1 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3 * gimplify.c (gimplify_init_constructor): Really never clear for an
4 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6 2018-06-19 Richard Biener <rguenther@suse.de>
8 PR tree-optimization/86179
9 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10 after failed recognition.
12 2018-06-18 Martin Sebor <msebor@redhat.com>
15 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
16 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
17 Handle integer subtraction.
18 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
19 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
21 2018-06-18 David Malcolm <dmalcolm@redhat.com>
23 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
24 param from rtx to rtx_insn *.
25 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
27 (frv_ifcvt_modify_insn): Likwise.
28 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
29 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
30 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
31 as_a <rtx_insn *> cast to local "unprotected_region" once
32 it's been established that it's not NULL or pc_rtx.
33 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
34 param "sethi" from rtx to rtx_insn *.
35 (nds32_group_float_insns): Likewise for param "insn".
36 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
38 (vax_output_int_subtract): Likewise.
39 * config/vax/vax.c (vax_output_int_add): Likewise for param
41 (vax_output_int_subtract): Likewise.
42 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
43 (emit_pattern_after): Likewise for param "after".
44 (emit_insn_after): Likewise.
45 (emit_jump_insn_after): Likewise.
46 (emit_call_insn_after): Likewise.
47 (emit_debug_insn_after): Likewise.
48 (emit_pattern_before): Likewise for param "before".
49 (emit_insn_before): Likewise.
50 (emit_jump_insn_before): Likewise.
51 * final.c (get_insn_template): Likewise for param "insn", removing
53 * output.h (get_insn_template): Likewise for 2nd param.
54 * rtl.h (emit_insn_before): Likewise.
55 (emit_jump_insn_before): Likewise.
56 (emit_debug_insn_before_noloc): Likewise.
57 (emit_insn_after): Likewise.
58 (emit_jump_insn_after): Likewise.
59 (emit_call_insn_after): Likewise.
60 (emit_debug_insn_after): Likewise.
61 (set_insn_deleted): Likewise for param.
63 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
66 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
67 floating point modes, so that IFmode is numerically greater than
68 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
69 to declare the ordering. This prevents IFmode from being
70 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
71 machine. Include rs6000-modes.h to share the fractional values
72 between genmodes* and the rest of the compiler.
76 * config/rs6000/rs6000-modes.h: New file.
77 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
78 meaning of rs6000_long_double_size so that 126..128 selects an
79 appropriate 128-bit floating point type.
80 (rs6000_option_override_internal): Likewise.
81 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
82 (TARGET_LONG_DOUBLE_128): Change the meaning of
83 rs6000_long_double_size so that 126..128 selects an appropriate
84 128-bit floating point type.
85 (LONG_DOUBLE_TYPE_SIZE): Update comment.
86 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
87 source and destination to match the standard usage.
88 (truncifkf2): Likewise.
89 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
90 ISA 2.07 to use an explicit clobber, instead of passing in a
92 (copysign<mode>3_soft): Likewise.
94 2018-06-18 David Malcolm <dmalcolm@redhat.com>
96 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
97 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
98 (vect_slp_analyze_instance_dependence): Likewise.
99 (vect_enhance_data_refs_alignment): Likewise.
100 (vect_analyze_data_refs_alignment): Likewise.
101 (vect_slp_analyze_and_verify_instance_alignment
102 (vect_analyze_data_ref_accesses): Likewise.
103 (vect_prune_runtime_alias_test_list): Likewise.
104 (vect_analyze_data_refs): Likewise.
105 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
106 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
107 (vect_analyze_scalar_cycles_1): Likewise.
108 (vect_get_loop_niters): Likewise.
109 (vect_analyze_loop_form_1): Likewise.
110 (vect_update_vf_for_slp): Likewise.
111 (vect_analyze_loop_operations): Likewise.
112 (vect_analyze_loop): Likewise.
113 (vectorizable_induction): Likewise.
114 (vect_transform_loop): Likewise.
115 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
116 * tree-vect-slp.c (vect_analyze_slp): Likewise.
117 (vect_make_slp_decision): Likewise.
118 (vect_detect_hybrid_slp): Likewise.
119 (vect_slp_analyze_operations): Likewise.
120 (vect_slp_bb): Likewise.
121 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
122 (vectorizable_bswap): Likewise.
123 (vectorizable_call): Likewise.
124 (vectorizable_simd_clone_call): Likewise.
125 (vectorizable_conversion): Likewise.
126 (vectorizable_assignment): Likewise.
127 (vectorizable_shift): Likewise.
128 (vectorizable_operation): Likewise.
129 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
131 2018-06-18 Martin Sebor <msebor@redhat.com>
133 PR tree-optimization/81384
134 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
135 * builtins.c (expand_builtin_strnlen): New function.
136 (expand_builtin): Call it.
137 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
138 * builtins.def (BUILT_IN_STRNLEN): New.
139 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
140 Warn for bounds in excess of maximum object size.
141 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
142 single-value ranges. Handle strnlen.
143 (handle_builtin_strlen): Handle strnlen.
144 (strlen_check_and_optimize_stmt): Same.
145 * doc/extend.texi (Other Builtins): Document strnlen.
147 2018-06-18 Maya Rashish <coypu@sdf.org>
149 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
150 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
151 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
153 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
155 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
157 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
159 * tree.c (escaped_string::escape): Replace cast to char * by
160 const_cast<char *> (unescaped).
162 2018-06-18 Nick Clifton <nickc@redhat.com>
165 * tree.c (escaped_string): New class. Converts an unescaped
166 string into its escaped equivalent.
167 (warn_deprecated_use): Use the new class to convert the
168 deprecation message, if present.
169 (test_escaped_strings): New self test.
170 (test_c_tests): Add test_escaped_strings.
171 * doc/extend.texi (deprecated): Add a note that the
172 deprecation message is affected by the -fmessage-length
173 option, and that control characters will be escaped.
174 (#pragma GCC error): Document this pragma.
175 (#pragma GCC warning): Likewise.
176 * doc/invoke.texi (-fmessage-length): Document this option's
177 effect on the #warning and #error preprocessor directives and
178 the deprecated attribute.
180 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
182 * tree.c (decl_value_expr_lookup): Revert latest change.
183 (decl_value_expr_insert): Likewise.
185 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
187 * gimplify.c (nonlocal_vlas): Delete.
188 (nonlocal_vla_vars): Likewise.
189 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
191 (gimplify_body): Do not create and destroy nonlocal_vlas.
192 * tree-nested.c: Include diagnostic.h.
193 (use_pointer_in_frame): Tweak.
194 (lookup_field_for_decl): Add assertion and declare the transformation.
195 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
196 internal error when the reference is in a wrong context. Do not
197 create a debug decl by default.
198 (note_nonlocal_block_vlas): Delete.
199 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
200 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
201 create a debug decl by default.
202 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
203 call is in a wrong context.
204 (fixup_vla_decls): New function.
205 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
206 debug variables were created.
207 * tree.c (decl_value_expr_lookup): Add checking assertion.
208 (decl_value_expr_insert): Likewise.
210 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
213 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
214 * tree-scalar-evolution.c (interpret_expr): Likewise.
215 (expression_expensive_p): Likewise.
216 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
217 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
218 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
219 (ssa_defined_by_minus_one_stmt_p): New.
221 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
224 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
225 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
226 * dojump.c (do_jump): Likewise.
227 * expr.c (expand_expr_real_2): Check operand type's sign.
228 * fold-const.c (const_unop): Handle ABSU_EXPR.
229 (fold_abs_const): Likewise.
230 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
231 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
232 (strip_sign_op_1): Likesise.
233 * match.pd: Add new pattern to generate ABSU_EXPR.
234 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
235 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
236 * tree-eh.c (operation_could_trap_helper_p): Likewise.
237 * tree-inline.c (estimate_operator_cost): Likewise.
238 * tree-pretty-print.c (dump_generic_node): Likewise.
239 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
240 * tree.def (ABSU_EXPR): New.
242 2018-06-16 Jakub Jelinek <jakub@redhat.com>
245 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
246 documented as preserved for backward compatibility only.
247 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
249 PR rtl-optimization/86108
250 * bb-reorder.c (create_forwarder_block): Renamed to ...
251 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
252 jump from new landing pad to the second part.
253 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
256 2018-06-15 Jakub Jelinek <jakub@redhat.com>
259 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
260 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
261 Only call store_expr for halves if the mode is the same.
264 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
265 Fix up comment formatting.
267 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
269 * typed-splay-tree.h (typed_splay_tree::remove): New function.
270 (typed_splay_tree::closure,
271 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
272 (typed_splay_tree::typed_splay_tree,
273 typed_splay_tree::operator =): Declared private.
274 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
275 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
276 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
277 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
278 typed_splay_tree::splay_tree_splay,
279 typed_splay_tree::splay_tree_foreach_helper,
280 typed_splay_tree::splay_tree_insert,
281 typed_splay_tree::splay_tree_remove,
282 typed_splay_tree::splay_tree_lookup,
283 typed_splay_tree::splay_tree_predecessor,
284 typed_splay_tree::splay_tree_successor,
285 typed_splay_tree::splay_tree_min,
286 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
287 (typed_splay_tree::root, typed_splay_tree::comp,
288 typed_splay_tree::delete_key,
289 typed_splay_tree::delete_value): New data members.
290 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
291 typed_splay_tree::remove.
293 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
295 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
296 -mginv and -mno-ginv to the assembler.
297 * config/mips/mips.opt (-mcrc): New option.
299 * doc/invoke.text (-mcrc): Document.
302 2018-06-15 Nick Clifton <nickc@redhat.com>
305 * tree.c (escaped_string): New class. Converts an unescaped
306 string into its escaped equivalent.
307 (warn_deprecated_use): Use the new class to convert the
308 deprecation message, if present.
309 (test_escaped_strings): New self test.
310 (test_c_tests): Add test_escaped_strings.
311 * doc/extend.texi (deprecated): Add a note that the
312 deprecation message is affected by the -fmessage-length
313 option, and that control characters will be escaped.
314 (#pragma GCC error): Document this pragma.
315 (#pragma GCC warning): Likewise.
316 * doc/invoke.texi (-fmessage-length): Document this option's
317 effect on the #warning and #error preprocessor directives and
318 the deprecated attribute.
320 2018-06-15 Richard Biener <rguenther@suse.de>
322 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
323 here, also noting vector size used.
324 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
325 size used in MSG_OPTIMIZED_LOCATIONS dump.
326 (pass_slp_vectorize::execute): Adjust.
328 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
331 * config/arc/arc.c (arc_return_address_register): Fix
334 2018-06-15 Richard Biener <rguenther@suse.de>
337 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
338 leave useless conversion stripping to force_gimple_operand_gsi.
339 (gimplify_build2): Likewise.
340 (gimplify_build1): Likewise.
342 2018-06-15 Richard Biener <rguenther@suse.de>
345 * tree-cfg.c (move_stmt_op): unshare invariant addresses
346 before adjusting their block.
348 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
350 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
351 multilibs for *-*-rtems*.
352 * config/riscv/t-rtems: New file.
354 2018-06-14 Jakub Jelinek <jakub@redhat.com>
357 * match.pd ((A +- CST1) +- CST2): Punt if last resort
358 unsigned_type_for returns NULL.
361 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
362 subregs of multi-word pseudos unless the float mode has word size.
364 2018-06-14 Richard Biener <rguenther@suse.de>
367 * tree-vect-generic.c (build_word_mode_vector_type): Remove
368 duplicate and harmful type_hash_canon.
369 * tree.c (type_hash_canon): Assert we didn't find ourselves.
371 2018-06-14 Richard Biener <rguenther@suse.de>
374 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
377 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
379 * config/rtems.h (STDINT_LONG32): Define.
381 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
382 Prachi Godbole <prachi.godbole@imgtec.com>
384 * config/mips/mips-cpus.def: Define P6600.
385 * config/mips/mips-tables.opt: Regenerate.
386 * config/mips/mips.c (mips_ucbranch_type): New enum.
387 (mips_rtx_cost_data): Add support for P6600.
388 (mips_issue_rate): Likewise.
389 (mips_multipass_dfa_lookahead): Likewise.
390 (mips_avoid_hazard): Likewise.
391 (mips_reorg_process_insns): Likewise.
392 (mips_classify_branch_p6600): New function.
393 * config/mips/mips.h (TUNE_P6600): New define.
394 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
395 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
396 * config/mips/mips.md: Include p6600.md.
397 (processor): Add p6600.
398 * config/mips/p6600.md: New file.
399 * doc/invoke.texi: Add p6600 to supported architectures.
401 2018-06-13 Martin Sebor <msebor@redhat.com>
403 PR tree-optimization/86114
404 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
406 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
408 2018-06-13 Richard Biener <rguenther@suse.de>
410 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
411 Properly set vector type of the intermediate stmt.
412 * tree-vect-stmts.c (vectorizable_operation): The destination
413 var always has vectype_out type.
415 2018-06-13 Jeff Law <law@redhat.com>
417 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
418 integer 0 for argument to print_rtl_with_bb.
419 (rl78_reorg): Likewise.
421 2018-06-13 David Malcolm <dmalcolm@redhat.com>
423 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
424 from rtx to rtx_insn *.
425 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
427 (add_sched_insns_for_speculation): Likewise for local "target",
428 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
429 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
430 from rtx_insn ** to rtx_code_label **.
431 (reorg_emit_nops): Likewise.
432 (c6x_reorg): Likewise for local "call_labels".
433 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
435 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
436 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
437 the loops over LABEL_REFS.
438 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
440 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
441 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
442 (split_branches): Strengthen local "olabel" from rtx to
443 rtx_insn *, adding a safe_as_a cast.
444 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
446 (add_insn_after): Likewise for first two params.
447 (add_insn_before): Likewise.
448 (remove_insn): Likewise for param.
449 (emit_pattern_before_noloc): Likewise for second and third params.
450 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
451 (emit_call_insn_before_noloc): Likewise.
452 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
454 (emit_barrier_before): Likewise.
455 (emit_label_before): Strengthen "label" param from "rtx" to
456 "rtx_code_label *". Strengthen "before" param from "rtx" to
458 (emit_insn_after_1): Strengthen "after" param from "rtx" to
460 (emit_pattern_after_noloc): Likewise.
461 (emit_insn_after_noloc): Likewise.
462 (emit_jump_insn_after_noloc): Likewise.
463 (emit_call_insn_after_noloc): Likewise.
464 (emit_debug_insn_after_noloc): Likewise.
465 (emit_barrier_after): Likewise.
466 (emit_label_after): Likewise for both params.
467 (emit_pattern_after_setloc): Likewise for "after" param. Convert
468 "loc" param from "int" to "location_t".
469 (emit_insn_after_setloc): Likewise.
470 (emit_jump_insn_after_setloc): Likewise.
471 (emit_call_insn_after_setloc): Likewise.
472 (emit_debug_insn_after_setloc): Likewise.
473 (emit_pattern_before_setloc): Likewise for "before" param. Convert
474 "loc" param from "int" to "location_t".
475 (emit_pattern_before): Convert NULL_RTX to NULL.
476 (emit_insn_before_setloc): Convert "loc" param from "int" to
478 (emit_jump_insn_before_setloc): Likewise.
479 (emit_call_insn_before_setloc): Likewise.
480 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
481 rtx_insn *. Convert "loc" param from "int" to "location_t".
482 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
483 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
484 Convert 3rd param from "int" to "location_t".
485 (emit_barrier_before, emit_barrier_after, next_real_insn):
486 Strengthen param from rtx to rtx_insn *.
487 (emit_label_before): Strengthen 1st param from "rtx" to
488 "rtx_code_label *". Strengthen 2nd param from "rtx" to
490 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
491 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
492 Strengthen 2nd param from "rtx" to "rtx_insn *".
493 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
494 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
495 Likewise. Convert 3rd param from "int" to "location_t".
496 (emit_label_after): Strengthen 1st param from "rtx" to
498 (next_real_insn, remove_insn): Strengthen param from "rtx" to
500 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
501 from "rtx" to "rtx_insn *".
503 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
505 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
506 bodies streamed in with -Q.
507 * dumpfile.c (dump_files): Add lto-stream-out dump file.
508 * dumpfile.h (tree_dump_index): Add lto_stream_out.
509 * gimple-streamer-out.c: Include gimple-pretty-print.h
510 (output_bb): Dump stmts streamed.
511 * lto-section-out.c: Include print-tree.h
512 (lto_begin_section): Dump sections created.
513 (lto_output_decl_index): Dump decl encoded.
514 * lto-streamer-out.c: Include print-tree.h
515 (create_output_block): Dump output block created.
516 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
517 (output_function): Dump function output.
518 (output_constructor): Dump constructor streamed.
519 (write_global_stream): Output indexes encoded.
520 (produce_asm_for_decls): Dump streams encoded.
521 * lto-streamer.c (streamer_dump_file): New global var.
522 * lto-streamer.h (streamer_dump_file): Declare.
523 * passes.c (ipa_write_summaries): Initialize streamer dump.
524 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
527 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
530 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
531 offsets for register save directives. Emit a second batch of save
532 directives, if need be, when the function accesses prior frames.
534 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
536 * config/arc/fpu.md (fmasf4): Force operand to register.
539 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
541 * config/arc/arc-protos.h (arc_pad_return): Remove.
542 * config/arc/arc.c (machine_function): Remove force_short_suffix
544 (arc_print_operand): Adjust printing of '&'.
545 (arc_verify_short): Remove conditional printing of short suffix.
546 (arc_final_prescan_insn): Remove reference to size_reason.
547 (pad_return): New function.
548 (arc_reorg): Call pad_return.
549 (arc_pad_return): Remove.
550 (arc_init_machine_status): Remove reference to force_short_suffix.
551 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
552 (attr length): When attribute iscompact is true force to 2
553 regardless; in the case of maybe check if we want to force the
554 instruction to have 4 bytes length.
555 (nopv): Change it to generate 4 byte long nop as well.
556 (blockage): New pattern.
557 (simple_return): Remove call to arc_pad_return.
558 (p_return_i): Likewise.
560 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
562 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
564 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
566 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
569 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
571 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
572 for ARC700 and ARCv2.
574 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
577 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
578 operands[2] instead of operands[1].
581 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
583 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
584 case, check whether the outer register overlaps an unallocatable
585 register, not just whether it fits the required class.
587 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
589 * poly-int.h (can_div_trunc_p): Add new overload in which all values
591 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
592 (memrefs_conflict_p): Likewise.
593 (init_alias_analysis): Likewise.
594 * cfgexpand.c (expand_debug_expr): Likewise.
595 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
596 * cse.c (fold_rtx): Likewise.
597 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
598 * expr.c (emit_block_move_hints): Likewise.
599 (clear_storage_hints, push_block, emit_push_insn): Likewise.
600 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
601 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
602 (emit_group_store): Likewise.
603 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
604 to read the PRE/POST_MODIFY increment.
605 * calls.c (store_one_arg): Use strip_offset.
606 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
608 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
610 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
611 (simplify_binary_operation_1): Extend CONST_INT handling to
613 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
614 than a HOST_WIDE_INT.
615 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
617 (adjust_mems, add_stores): Update accodingly.
618 (vt_canonicalize_addr): Track polynomial offsets.
619 (emit_note_insn_var_location): Likewise.
620 (vt_add_function_parameter): Likewise.
621 (vt_initialize): Likewise.
623 2018-06-12 Jeff Law <law@redhat.com>
625 * config.gcc (alpha*-*-freebsd*): Remove.
626 * config/alpha/freebsd.h: Remove.
628 2018-06-12 David Malcolm <dmalcolm@redhat.com>
631 * spellcheck-tree.c (levenshtein_distance): Rename to...
632 (get_edit_distance): ...this, and update for underlying renaming.
633 * spellcheck-tree.h (levenshtein_distance): Rename to...
634 (get_edit_distance): ...this.
635 * spellcheck.c (levenshtein_distance): Rename to...
636 (get_edit_distance): ...this. Convert from Levenshtein distance
637 to Damerau-Levenshtein distance by supporting transpositions of
638 adjacent characters. Rename "v1" to "v_next" and "v0" to
640 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
641 (selftest::test_edit_distance_unit_test_oneway): ...this, and
642 update for underlying renaming.
643 (selftest::levenshtein_distance_unit_test): Rename to...
644 (selftest::test_get_edit_distance_unit): ...this, and update for
646 (selftest::test_find_closest_string): Add example from PR 69968
647 where transposition helps
648 (selftest::test_metric_conditions): Update for renaming.
649 (selftest::test_metric_conditions): Likewise.
650 (selftest::spellcheck_c_tests): Likewise.
651 * spellcheck.h (levenshtein_distance): Rename both overloads to...
652 (get_edit_distance): ...this.
653 (best_match::consider): Update for renaming.
655 2018-06-12 Martin Sebor <msebor@redhat.com>
657 PR tree-optimization/85259
658 * builtins.c (compute_objsize): Handle constant offsets.
659 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
660 true iff a warning has been issued.
661 * gimple.h (gimple_nonartificial_location): New function.
662 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
663 gimple_nonartificial_location and handle -Wno-system-headers.
664 (handle_builtin_stxncpy): Same.
666 2018-06-12 Martin Sebor <msebor@redhat.com>
669 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
671 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
673 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
674 BUILTIN_VEC_XST entries for pointer to double and long long.
676 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
679 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
681 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
684 2018-06-12 Martin Liska <mliska@suse.cz>
686 * doc/options.texi: Document IntegerRange.
688 2018-06-12 Martin Liska <mliska@suse.cz>
690 * config/i386/i386.opt: Make MPX-related options as Deprecated.
691 * opt-functions.awk: Handle Deprecated flag.
692 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
694 (read_cmdline_option): Report warning for a deprecated option.
695 * opts.h (struct cl_option): Add new field cl_deprecated.
696 (CL_ERR_DEPRECATED): New.
698 2018-06-12 Martin Liska <mliska@suse.cz>
700 * doc/options.texi: Document Deprecated option flag.
702 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
704 * config/arc/arc-arch.h (arc_extras): New enum.
705 (arc_cpu_t):Add field extra.
706 (arc_cpu_types): Consider the extras.
707 * config/arc/arc-cpus.def: Add extras info.
708 * config/arc/arc-opts.h (processor_type): Consider extra field.
709 * config/arc/arc.c (arc_override_options): Handle extra field.
711 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
713 * config/arc/arc-arch.h: Update ARC_OPTX macro.
714 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
716 * config/arc/arc.c (arc_init): Update pic warning.
717 (irq_range): Update irq range parsing warnings.
718 (arc_override_options): Update various warning messages.
719 (arc_handle_aux_attribute): Likewise.
721 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
723 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
725 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
727 * doc/sourcebuild.texi: Document usage of line number 0 in verify
728 compiler messages directives.
730 2018-06-12 Matthew Fortune <mfortune@gmail.com>
732 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
733 * config/mips/mips-tables.opt: Regenerate.
734 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
736 * doc/invoke.texi: Document -march=i6500.
738 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
740 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
741 (i6400_gpmul): Add cpu_unit.
742 (i6400_gpdiv): Likewise.
743 (i6400_msa_add_d): Update reservations.
744 (i6400_msa_int_add) Likewise.
745 (i6400_msa_short_logic3) Likewise.
746 (i6400_msa_short_logic2) Likewise.
747 (i6400_msa_short_logic) Likewise.
748 (i6400_msa_move) Likewise.
749 (i6400_msa_cmp) Likewise.
750 (i6400_msa_short_float2) Likewise.
751 (i6400_msa_div_d) Likewise.
752 (i6400_msa_long_logic1) Likewise.
753 (i6400_msa_long_logic2) Likewise.
754 (i6400_msa_mult) Likewise.
755 (i6400_msa_long_float2) Likewise.
756 (i6400_msa_long_float4) Likewise.
757 (i6400_msa_long_float5) Likewise.
758 (i6400_msa_long_float8) Likewise.
759 (i6400_fpu_fadd): Include frint type.
760 (i6400_fpu_store): New define_insn_reservation.
761 (i6400_fpu_load): Likewise.
762 (i6400_fpu_move): Likewise.
763 (i6400_fpu_fcmp): Likewise.
764 (i6400_fpu_fmadd): Likewise.
765 (i6400_int_mult): Include imul3nc type and update reservation.
766 (i6400_int_div): Include idiv3 type and update reservation.
767 (i6400_int_load): Update to check type not move_type.
768 (i6400_int_store): Likewise.
769 (i6400_int_prefetch): Set zero latency.
771 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
773 * gcc.c: Document new %@{...} sequence.
774 (LINK_COMMAND_SPEC): Use it for the -L switches.
775 (cpp_unique_options): Use it for the -I switches.
776 (at_file_argbuf): New global variable.
777 (in_at_file): Likewise.
778 (alloc_args): Create at_file_argbuf.
779 (clear_args): Truncate at_file_argbuf.
780 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
781 (open_at_file): New function.
782 (close_at_file): Likewise.
783 (create_at_file): Delete.
784 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
787 (validate_switches_from_spec): Deal with %@{...} sequence.
788 (validate_switches): Likewise.
789 (driver::finalize): Call clear_args.
791 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
793 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
795 2018-06-11 Martin Sebor <msebor@redhat.com>
797 * doc/invoke.texi (-Wall): List -Wc++17-compat.
798 (Wno-class-memaccess): Add @opindex.
799 (Wno-templates, Wno-multiple-inheritance): Same.
800 (Wno-virtual-inheritance, Wno-namespaces): Same.
801 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
802 (Wno-format-overflow, Wno-format-truncation): Same.
803 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
804 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
805 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
806 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
807 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
808 (Wno-misspelled-isr): Same.
810 2018-06-11 Martin Sebor <msebor@redhat.com>
812 * PR tree-optimization/86083
813 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
815 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
817 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
819 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
822 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
823 on the correct operand.
824 (*movdi_internal64): Ditto.
826 2018-06-11 Martin Liska <mliska@suse.cz>
828 PR tree-optimization/86089
829 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
831 2018-06-11 Julia Koval <julia.koval@intel.com>
833 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
834 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
835 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
837 2018-06-11 Olivier Hainque <hainque@adacore.com>
839 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
840 for Ada with strict dwarf2.
842 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
845 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
848 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
850 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
852 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
854 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
857 2018-06-08 Martin Liska <mliska@suse.cz>
859 * tree-cfg.h (debug_function): Fix argument type to match
862 2018-06-08 Martin Liska <mliska@suse.cz>
864 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
865 Remove usage of MPX-related (and removed) fields.
866 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
868 2018-06-08 David Malcolm <dmalcolm@redhat.com>
870 * cfg.c (debug): Use TDF_NONE rather than 0.
871 * cfghooks.c (debug): Likewise.
872 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
873 (struct dump_option_value_info): Convert to...
874 (struct kv_pair): ...this template type.
875 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
877 (optinfo_verbosity_options): Likewise.
878 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
880 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
881 than int for "optgroup_flags" param.
882 (dump_generic_expr_loc): Use dump_flags_t rather than int for
884 (dump_dec): Likewise.
885 (dump_finish): Use TDF_NONE rather than 0.
886 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
887 rather than int for "optgroup_flags" param. Use TDF_NONE rather
888 than 0. Update for change to option_ptr.
889 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
890 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
891 0. Update for changes to optinfo_verbosity_options and
893 (opt_info_switch_p): Convert optgroup_flags from int to
895 (dump_basic_block): Use dump_flags_t rather than int
896 for "dump_kind" param.
897 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
898 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
899 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
900 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
901 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
902 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
903 TDF_NONE): Convert from macros to...
904 (enum dump_flag): ...this new enum.
905 (dump_flags_t): Update to use enum.
906 (operator|, operator&, operator~, operator|=, operator&=):
907 Implement for dump_flags_t.
908 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
909 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
910 Convert from macros to...
911 (enum optgroup_flag): ...this new enum.
912 (optgroup_flags_t): New typedef.
913 (operator|, operator|=): Implement for optgroup_flags_t.
914 (struct dump_file_info): Convert field "alt_flags" to
915 dump_flags_t. Convert field "optgroup_flags" to
917 (dump_basic_block): Use dump_flags_t rather than int for param.
918 (dump_generic_expr_loc): Likewise.
919 (dump_dec): Likewise.
920 (dump_register): Convert param "optgroup_flags" to
922 (opt_info_enable_passes): Likewise.
923 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
925 * gimple-pretty-print.c (debug): Likewise.
926 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
927 (merged_store_group::apply_stores): Likewise.
928 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
929 * gimple.c (verify_gimple_pp): Likewise.
930 * graphite-poly.c (print_pbb_body): Likewise.
931 * passes.c (pass_manager::register_one_dump_file): Convert
932 local "optgroup_flags" to optgroup_flags_t.
933 * print-tree.c (print_node): Use TDF_NONE rather than 0.
935 (debug_body): Likewise.
936 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
938 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
940 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
941 (convert_mult_to_fma): Likewise.
942 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
943 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
944 * tree-vect-data-refs.c (dump_lower_bound): Convert param
945 "dump_kind" to dump_flags_t.
947 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
949 * config/rs6000/rs6000.c (min, max): Delete.
951 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
953 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
954 -mabi=spe and -mabi=no-spe.
956 2018-06-08 Martin Liska <mliska@suse.cz>
958 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
959 where we expect an existing summary.
961 2018-06-08 Martin Liska <mliska@suse.cz>
963 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
964 * ipa-inline.h (estimate_edge_growth): Likewise.
966 2018-06-08 Martin Liska <mliska@suse.cz>
968 * cgraph.c (function_version_hasher::hash): Use
969 cgraph_node::get_uid ().
970 (function_version_hasher::equal):
971 * cgraph.h (cgraph_node::get_uid): New method.
972 * ipa-inline.c (update_caller_keys): Use
973 cgraph_node::get_uid ().
974 (update_callee_keys): Likewise.
975 * ipa-utils.c (searchc): Likewise.
976 (ipa_reduced_postorder): Likewise.
977 * lto-cgraph.c (input_node): Likewise.
978 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
979 * symbol-summary.h (symtab_insertion): Likewise.
980 (symtab_removal): Likewise.
981 (symtab_duplication): Likewise.
982 * tree-pretty-print.c (dump_function_header): Likewise.
983 * tree-sra.c (convert_callers_for_node): Likewise.
985 2018-06-08 Martin Liska <mliska@suse.cz>
987 * cgraph.c (symbol_table::create_edge): Always assign a new
989 (symbol_table::free_edge): Do not recycle numbers.
990 * cgraph.h (cgraph_edge::get): New method.
991 * symbol-summary.h (symtab_removal): Use it.
992 (symtab_duplication): Likewise.
993 (call_summary::hashable_uid): Remove.
995 2018-06-08 Martin Liska <mliska@suse.cz>
997 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
998 (initialize_growth_caches): Remove.
999 (free_growth_caches): Likewise.
1000 (do_estimate_edge_time): Use edge_growth_cache.
1001 (do_estimate_edge_size): Likewise.
1002 (do_estimate_edge_hints): Likewise.
1003 * ipa-inline.c (reset_edge_caches): Likewise.
1004 (recursive_inlining): Likewise.
1005 (inline_small_functions): Likewise.
1006 * ipa-inline.h (initialize_growth_caches): Remove.
1007 (estimate_edge_size): Likewise.
1008 (estimate_edge_time): Likewise.
1009 (estimate_edge_hints): Likewise.
1010 (reset_edge_growth_cache): Likewise.
1011 * symbol-summary.h (call_summary::remove): New method.
1013 2018-06-08 Martin Liska <mliska@suse.cz>
1015 * ipa-cp.c (class edge_clone_summary): New summary.
1016 (grow_edge_clone_vectors): Remove.
1017 (ipcp_edge_duplication_hook): Remove.
1018 (class edge_clone_summary_t): New call_summary class.
1019 (ipcp_edge_removal_hook): Remove.
1020 (edge_clone_summary_t::duplicate): New function.
1021 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1022 (create_specialized_node): Likewise.
1023 (ipcp_driver): Initialize edge_clone_summaries and do not
1026 2018-06-08 Martin Liska <mliska@suse.cz>
1028 * symbol-summary.h (get): New function.
1029 (call_summary::m_initialize_when_cloning): New class member.
1031 2018-06-08 Martin Liska <mliska@suse.cz>
1033 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1034 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1035 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1036 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1037 (remove_cgraph_node_from_order): Use the removed_nodes set.
1038 (do_per_function_toporder): Likwise.
1039 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1040 instead of summary_uid.
1041 (symtab_removal): Likewise.
1042 (symtab_duplication): Likewise.
1044 2018-06-08 Martin Liska <mliska@suse.cz>
1046 * ipa-cp.c (ipcp_store_bits_results): Use
1047 ipcp_transformation_sum.
1048 (ipcp_store_vr_results): Likewise.
1049 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1051 (ipcp_transformation_initialize): ... this.
1052 (ipa_set_node_agg_value_chain):
1053 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1054 (write_ipcp_transformation_info): Likewise.
1055 (read_ipcp_transformation_info): Likewise.
1056 (ipcp_update_bits): Likewise.
1057 (ipcp_update_vr): Likewise.
1058 (ipcp_transform_function): Likewise.
1059 * ipa-prop.h: Rename ipcp_transformation_summary to
1060 ipcp_transformation.
1061 (class ipcp_transformation_t): New function summary.
1062 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1063 (ipa_get_agg_replacements_for_node): Likewise.
1065 2018-06-08 Martin Liska <mliska@suse.cz>
1067 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1069 (class funct_state_summary_t): New function_summary class.
1070 (has_function_state): Remove.
1071 (get_function_state): Likewise.
1072 (set_function_state): Likewise.
1073 (add_new_function): Likewise.
1074 (funct_state_summary_t::insert): New function.
1075 (duplicate_node_data): Remove.
1076 (remove_node_data): Remove.
1077 (funct_state_summary_t::duplicate): New function.
1078 (register_hooks): Create new funct_state_summaries.
1079 (pure_const_generate_summary): Use it.
1080 (pure_const_write_summary): Likewise.
1081 (pure_const_read_summary): Likewise.
1082 (propagate_pure_const): Likewise.
1083 (propagate_nothrow): Likewise.
1084 (dump_malloc_lattice): Likewise.
1085 (propagate_malloc): Likewise.
1086 (execute): Do not register hooks, just remove summary
1088 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1091 2018-06-08 Martin Liska <mliska@suse.cz>
1093 * ipa-reference.c (remove_node_data): Remove.
1094 (duplicate_node_data): Likewise.
1095 (class ipa_ref_var_info_summary_t): New class.
1096 (class ipa_ref_opt_summary_t): Likewise.
1097 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1098 (get_reference_optimization_summary): Use
1099 ipa_ref_opt_sum_summaries.
1100 (set_reference_vars_info): Remove.
1101 (set_reference_optimization_summary): Likewise.
1102 (ipa_init): Create summaries.
1103 (init_function_info): Use function summary.
1104 (ipa_ref_opt_summary_t::duplicate): New function.
1105 (ipa_ref_opt_summary_t::remove): New function.
1106 (get_read_write_all_from_node): Fix GNU coding style.
1107 (propagate): Use function summary.
1108 (write_node_summary_p): Fix GNU coding style.
1109 (stream_out_bitmap): Likewise.
1110 (ipa_reference_read_optimization_summary): Use function summary.
1111 (ipa_reference_c_finalize): Do not release hooks.
1113 2018-06-08 Martin Liska <mliska@suse.cz>
1115 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1116 (analyze_function_body): Extract multiple calls of get_create.
1117 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1118 * ipa-inline.c (recursive_inlining): Use ::get method.
1119 * ipa-inline.h (estimate_edge_growth): Likewise.
1121 2018-06-08 Martin Liska <mliska@suse.cz>
1123 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1125 (hsa_function_summary::hsa_function_summary): Use the new enum
1127 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1128 * hsa-gen.c (hsa_get_host_function): Likewise.
1129 (get_brig_function_name): Likewise.
1130 * ipa-hsa.c (process_hsa_functions): Likewise.
1131 (ipa_hsa_write_summary): Likewise.
1132 * symbol-summary.h (symtab_duplication): Use ::get function/
1133 (get): New function.
1135 2018-06-08 Martin Liska <mliska@suse.cz>
1137 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1139 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1140 (hsa_register_kernel): Likewise.
1141 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1142 * hsa-gen.c (hsa_get_host_function): Likewise.
1143 (get_brig_function_name): Likewise.
1144 (generate_hsa): Likewise.
1145 (pass_gen_hsail::execute): Likewise.
1146 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1147 (devirtualization_time_bonus): Likewise.
1148 (ipcp_propagate_stage): Likewise.
1149 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1150 (edge_set_predicate): Likewise.
1151 (evaluate_conditions_for_known_args): Likewise.
1152 (evaluate_properties_for_edge): Likewise.
1153 (ipa_fn_summary::reset): Likewise.
1154 (ipa_fn_summary_t::duplicate): Likewise.
1155 (dump_ipa_call_summary): Likewise.
1156 (ipa_dump_fn_summary): Likewise.
1157 (analyze_function_body): Likewise.
1158 (compute_fn_summary): Likewise.
1159 (estimate_edge_devirt_benefit): Likewise.
1160 (estimate_edge_size_and_time): Likewise.
1161 (estimate_calls_size_and_time): Likewise.
1162 (estimate_node_size_and_time): Likewise.
1163 (inline_update_callee_summaries): Likewise.
1164 (remap_edge_change_prob): Likewise.
1165 (remap_edge_summaries): Likewise.
1166 (ipa_merge_fn_summary_after_inlining): Likewise.
1167 (ipa_update_overall_fn_summary): Likewise.
1168 (read_ipa_call_summary): Likewise.
1169 (inline_read_section): Likewise.
1170 (write_ipa_call_summary): Likewise.
1171 (ipa_fn_summary_write): Likewise.
1172 (ipa_free_fn_summary): Likewise.
1173 * ipa-hsa.c (process_hsa_functions): Likewise.
1174 (ipa_hsa_write_summary): Likewise.
1175 (ipa_hsa_read_section): Likewise.
1176 * ipa-icf.c (sem_function::merge): Likewise.
1177 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1178 (do_estimate_edge_time): Likewise.
1179 (estimate_size_after_inlining): Likewise.
1180 (estimate_growth): Likewise.
1181 (growth_likely_positive): Likewise.
1182 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1183 (inline_call): Likewise.
1184 * ipa-inline.c (caller_growth_limits): Likewise.
1185 (can_inline_edge_p): Likewise.
1186 (can_inline_edge_by_limits_p): Likewise.
1187 (compute_uninlined_call_time): Likewise.
1188 (compute_inlined_call_time): Likewise.
1189 (want_inline_small_function_p): Likewise.
1190 (edge_badness): Likewise.
1191 (update_caller_keys): Likewise.
1192 (update_callee_keys): Likewise.
1193 (recursive_inlining): Likewise.
1194 (inline_small_functions): Likewise.
1195 (inline_to_all_callers_1): Likewise.
1196 (dump_overall_stats): Likewise.
1197 (early_inline_small_functions): Likewise.
1198 (early_inliner): Likewise.
1199 * ipa-inline.h (estimate_edge_growth): Likewise.
1200 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1201 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1202 * ipa-prop.h (IPA_NODE_REF): Likewise.
1203 (IPA_EDGE_REF): Likewise.
1204 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1205 (propagate_malloc): Likewise.
1206 * ipa-split.c (execute_split_functions): Likewise.
1207 * symbol-summary.h: Rename get to get_create.
1209 (get_create): Likewise.
1210 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1212 2018-06-08 Martin Liska <mliska@suse.cz>
1214 * symbol-summary.h (release): Move definition out of class
1216 (symtab_removal): Likewise.
1217 (symtab_duplication): Likewise.
1219 2018-06-08 Martin Liska <mliska@suse.cz>
1221 * symbol-summary.h (function_summary): Move constructor
1222 implementation out of class declaration.
1223 (release): Likewise.
1224 (symtab_insertion): Likewise.
1225 (symtab_removal): Likewise.
1226 (symtab_duplication): Likewise.
1229 2018-06-08 Martin Liska <mliska@suse.cz>
1231 * Makefile.in: Remove support for MPX (macros, related functions,
1232 fields in cgraph_node, ...).
1233 * builtin-types.def (BT_BND): Likewise.
1234 (BT_FN_BND_CONST_PTR): Likewise.
1235 (BT_FN_CONST_PTR_BND): Likewise.
1236 (BT_FN_VOID_PTR_BND): Likewise.
1237 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1238 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1239 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1240 (expand_builtin_mempcpy_with_bounds): Likewise.
1241 (expand_builtin_memset_with_bounds): Likewise.
1242 (expand_builtin_memset_args): Likewise.
1243 (std_expand_builtin_va_start): Likewise.
1244 (expand_builtin): Likewise.
1245 (expand_builtin_with_bounds): Likewise.
1246 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1247 (DEF_LIB_BUILTIN_CHKP): Likewise.
1248 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1249 (DEF_CHKP_BUILTIN): Likewise.
1250 (BUILT_IN_MEMCPY): Likewise.
1251 (BUILT_IN_MEMMOVE): Likewise.
1252 (BUILT_IN_MEMPCPY): Likewise.
1253 (BUILT_IN_MEMSET): Likewise.
1254 (BUILT_IN_STPCPY): Likewise.
1255 (BUILT_IN_STRCAT): Likewise.
1256 (BUILT_IN_STRCHR): Likewise.
1257 (BUILT_IN_STRCPY): Likewise.
1258 (BUILT_IN_STRLEN): Likewise.
1259 (BUILT_IN_MEMCPY_CHK): Likewise.
1260 (BUILT_IN_MEMMOVE_CHK): Likewise.
1261 (BUILT_IN_MEMPCPY_CHK): Likewise.
1262 (BUILT_IN_MEMSET_CHK): Likewise.
1263 (BUILT_IN_STPCPY_CHK): Likewise.
1264 (BUILT_IN_STRCAT_CHK): Likewise.
1265 (BUILT_IN_STRCPY_CHK): Likewise.
1266 * calls.c (store_bounds): Likewise.
1267 (emit_call_1): Likewise.
1268 (special_function_p): Likewise.
1269 (maybe_warn_nonstring_arg): Likewise.
1270 (initialize_argument_information): Likewise.
1271 (finalize_must_preallocate): Likewise.
1272 (compute_argument_addresses): Likewise.
1273 (expand_call): Likewise.
1274 * cfgexpand.c (expand_call_stmt): Likewise.
1275 (expand_return): Likewise.
1276 (expand_gimple_stmt_1): Likewise.
1277 (pass_expand::execute): Likewise.
1278 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1279 (cgraph_node::remove): Likewise.
1280 (cgraph_node::dump): Likewise.
1281 (cgraph_node::verify_node): Likewise.
1282 * cgraph.h (chkp_function_instrumented_p): Likewise.
1283 (symtab_node::get_alias_target): Likewise.
1284 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1285 (cgraph_local_p): Likewise.
1286 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1287 (cgraph_edge::rebuild_references): Likewise.
1288 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1289 (walk_polymorphic_call_targets): Likewise.
1290 (cgraph_node::expand_thunk): Likewise.
1291 (symbol_table::output_weakrefs): Likewise.
1292 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1293 (ix86_handle_option): Likewise.
1294 * config/i386/constraints.md: Likewise.
1295 * config/i386/i386-builtin-types.def (BND): Likewise.
1299 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1300 (BDESC_FIRST): Likewise.
1302 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1303 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1304 * config/i386/i386.c (enum reg_class): Likewise.
1305 (ix86_target_string): Likewise.
1306 (ix86_option_override_internal): Likewise.
1307 (ix86_conditional_register_usage): Likewise.
1308 (ix86_valid_target_attribute_inner_p): Likewise.
1309 (ix86_set_indirect_branch_type): Likewise.
1310 (ix86_set_current_function): Likewise.
1311 (ix86_function_arg_regno_p): Likewise.
1312 (init_cumulative_args): Likewise.
1313 (ix86_function_arg_advance): Likewise.
1314 (ix86_function_arg): Likewise.
1315 (ix86_pass_by_reference): Likewise.
1316 (ix86_function_value_regno_p): Likewise.
1317 (ix86_function_value_1): Likewise.
1318 (ix86_function_value_bounds): Likewise.
1319 (ix86_return_in_memory): Likewise.
1320 (ix86_setup_incoming_vararg_bounds): Likewise.
1321 (ix86_va_start): Likewise.
1322 (indirect_thunk_need_prefix): Likewise.
1323 (print_reg): Likewise.
1324 (ix86_print_operand): Likewise.
1325 (ix86_expand_call): Likewise.
1326 (ix86_output_function_return): Likewise.
1327 (reg_encoded_number): Likewise.
1328 (BDESC_VERIFYS): Likewise.
1329 (ix86_init_mpx_builtins): Likewise.
1330 (ix86_init_builtins): Likewise.
1331 (ix86_emit_cmove): Likewise.
1332 (ix86_emit_move_max): Likewise.
1333 (ix86_expand_builtin): Likewise.
1334 (ix86_builtin_mpx_function): Likewise.
1335 (ix86_get_arg_address_for_bt): Likewise.
1336 (ix86_load_bounds): Likewise.
1337 (ix86_store_bounds): Likewise.
1338 (ix86_load_returned_bounds): Likewise.
1339 (ix86_store_returned_bounds): Likewise.
1340 (ix86_class_likely_spilled_p): Likewise.
1341 (ix86_hard_regno_mode_ok): Likewise.
1342 (x86_order_regs_for_local_alloc): Likewise.
1343 (ix86_mitigate_rop): Likewise.
1344 (ix86_bnd_prefixed_insn_p): Likewise.
1345 (ix86_mpx_bound_mode): Likewise.
1346 (ix86_make_bounds_constant): Likewise.
1347 (ix86_initialize_bounds): Likewise.
1348 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1349 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1350 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1351 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1352 (TARGET_CHKP_BOUND_MODE): Likewise.
1353 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1354 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1355 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1356 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1357 * config/i386/i386.h (TARGET_MPX): Likewise.
1358 (TARGET_MPX_P): Likewise.
1359 (VALID_BND_REG_MODE): Likewise.
1360 (FIRST_BND_REG): Likewise.
1361 (LAST_BND_REG): Likewise.
1362 (enum reg_class): Likewise.
1363 (BND_REG_P): Likewise.
1364 (BND_REGNO_P): Likewise.
1365 (BNDmode): Likewise.
1366 (ADJUST_INSN_LENGTH): Likewise.
1367 * config/i386/i386.md: Likewise.
1368 * config/i386/i386.opt: Likewise.
1369 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1370 (defined): Likewise.
1371 (LINK_MPX): Likewise.
1372 (MPX_SPEC): Likewise.
1373 (LIBMPX_SPEC): Likewise.
1374 (LIBMPXWRAPPERS_SPEC): Likewise.
1375 (CHKP_SPEC): Likewise.
1376 * config/i386/predicates.md: Likewise.
1377 * dbxout.c (dbxout_type): Likewise.
1378 * doc/extend.texi: Likewise.
1379 * doc/invoke.texi: Likewise.
1380 * doc/md.texi: Likewise.
1381 * doc/tm.texi: Likewise.
1382 * doc/tm.texi.in: Likewise.
1383 * dwarf2out.c (is_base_type): Likewise.
1384 (gen_formal_types_die): Likewise.
1385 (gen_subprogram_die): Likewise.
1386 (gen_type_die_with_usage): Likewise.
1387 (gen_decl_die): Likewise.
1388 (dwarf2out_late_global_decl): Likewise.
1389 * expr.c (expand_assignment): Likewise.
1390 (emit_storent_insn): Likewise.
1391 (store_expr_with_bounds): Likewise.
1392 (store_expr): Likewise.
1393 (expand_expr_real_1): Likewise.
1394 * expr.h (store_expr_with_bounds): Likewise.
1395 * function.c (use_register_for_decl): Likewise.
1396 (struct bounds_parm_data): Likewise.
1397 (assign_parms_augmented_arg_list): Likewise.
1398 (assign_parm_find_entry_rtl): Likewise.
1399 (assign_parm_is_stack_parm): Likewise.
1400 (assign_parm_load_bounds): Likewise.
1401 (assign_bounds): Likewise.
1402 (assign_parms): Likewise.
1403 (expand_function_start): Likewise.
1404 * gcc.c (CHKP_SPEC): Likewise.
1405 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1406 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1407 (wrestrict_dom_walker::check_call): Likewise.
1408 * gimple.c (gimple_build_call_from_tree): Likewise.
1409 * gimple.h (enum gf_mask): Likewise.
1410 (gimple_call_with_bounds_p): Likewise.
1411 (gimple_call_set_with_bounds): Likewise.
1412 * gimplify.c (gimplify_init_constructor): Likewise.
1413 * ipa-cp.c (initialize_node_lattices): Likewise.
1414 (propagate_constants_across_call): Likewise.
1415 (find_more_scalar_values_for_callers_subset): Likewise.
1416 * ipa-hsa.c (process_hsa_functions): Likewise.
1417 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1418 * ipa-icf.c (sem_function::merge): Likewise.
1419 * ipa-inline.c (early_inliner): Likewise.
1420 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1421 (warn_function_cold): Likewise.
1422 (propagate_pure_const): Likewise.
1423 * ipa-ref.h (enum GTY): Likewise.
1424 * ipa-split.c (find_retbnd): Likewise.
1425 (consider_split): Likewise.
1426 (split_function): Likewise.
1427 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1428 * ipa.c (walk_polymorphic_call_targets): Likewise.
1429 (symbol_table::remove_unreachable_nodes): Likewise.
1430 (process_references): Likewise.
1431 (cgraph_build_static_cdtor_1): Likewise.
1432 * lto-cgraph.c (lto_output_node): Likewise.
1433 (output_refs): Likewise.
1434 (compute_ltrans_boundary): Likewise.
1435 (input_overwrite_node): Likewise.
1436 (input_node): Likewise.
1437 (input_cgraph_1): Likewise.
1438 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1439 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1440 (class pass_chkp_instrumentation_passes): Likewise.
1441 (make_pass_chkp_instrumentation_passes): Likewise.
1442 * passes.def: Likewise.
1443 * rtl.h (struct GTY): Likewise.
1444 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1445 * stor-layout.c (layout_type): Likewise.
1446 * symtab.c: Likewise.
1447 * target.def: Likewise.
1448 * targhooks.c (default_chkp_bound_type): Likewise.
1449 (default_chkp_bound_mode): Likewise.
1450 (default_builtin_chkp_function): Likewise.
1451 (default_chkp_function_value_bounds): Likewise.
1452 (default_chkp_make_bounds_constant): Likewise.
1453 (default_chkp_initialize_bounds): Likewise.
1454 * targhooks.h (default_chkp_bound_type): Likewise.
1455 (default_chkp_bound_mode): Likewise.
1456 (default_builtin_chkp_function): Likewise.
1457 (default_chkp_function_value_bounds): Likewise.
1458 (default_chkp_make_bounds_constant): Likewise.
1459 (default_chkp_initialize_bounds): Likewise.
1460 * toplev.c (compile_file): Likewise.
1461 (process_options): Likewise.
1462 * tree-core.h (DEF_BUILTIN): Likewise.
1463 (DEF_BUILTIN_CHKP): Likewise.
1464 * tree-inline.c (declare_return_variable): Likewise.
1465 (remap_gimple_stmt): Likewise.
1466 (copy_bb): Likewise.
1467 (initialize_inlined_parameters): Likewise.
1468 (expand_call_inline): Likewise.
1469 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1470 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1471 (make_pass_ipa_chkp_produce_thunks): Likewise.
1472 (make_pass_chkp): Likewise.
1473 (make_pass_chkp_opt): Likewise.
1474 (make_pass_chkp_instrumentation_passes): Likewise.
1475 * tree-pretty-print.c (dump_generic_node): Likewise.
1476 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1477 * tree-ssa-dce.c (propagate_necessity): Likewise.
1478 (eliminate_unnecessary_stmts): Likewise.
1479 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1480 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1481 * tree-ssa-sccvn.h: Likewise.
1482 * tree-ssa-strlen.c (get_string_length): Likewise.
1483 (valid_builtin_call): Likewise.
1484 (adjust_last_stmt): Likewise.
1485 (handle_builtin_strchr): Likewise.
1486 (handle_builtin_strcpy): Likewise.
1487 (handle_builtin_stxncpy): Likewise.
1488 (handle_builtin_memcpy): Likewise.
1489 (handle_builtin_strcat): Likewise.
1490 (strlen_check_and_optimize_stmt): Likewise.
1491 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1492 * tree-streamer-in.c: Likewise.
1493 * tree-streamer.c (record_common_node): Likewise.
1494 * tree.c (tree_code_size): Likewise.
1495 (wide_int_to_tree_1): Likewise.
1496 (type_contains_placeholder_1): Likewise.
1497 (build_common_tree_nodes): Likewise.
1498 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1499 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1500 (POINTER_BOUNDS_P): Likewise.
1501 (BOUNDED_TYPE_P): Likewise.
1502 (BOUNDED_P): Likewise.
1503 (CALL_WITH_BOUNDS_P): Likewise.
1504 (pointer_bounds_type_node): Likewise.
1505 * value-prof.c (gimple_ic): Likewise.
1506 * var-tracking.c (vt_add_function_parameters): Likewise.
1507 * varasm.c (make_decl_rtl): Likewise.
1508 (assemble_start_function): Likewise.
1509 (output_constant): Likewise.
1510 (maybe_assemble_visibility): Likewise.
1511 * varpool.c (ctor_for_folding): Likewise.
1512 * chkp-builtins.def: Remove.
1513 * ipa-chkp.c: Remove.
1514 * ipa-chkp.h: Remove.
1515 * rtl-chkp.c: Remove.
1516 * rtl-chkp.h: Remove.
1517 * tree-chkp-opt.c: Remove.
1518 * tree-chkp.c: Remove.
1519 * tree-chkp.h: Remove.
1521 2018-06-07 Carl Love <cel@us.ibm.com>
1523 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1524 vextract_fp_from_shortl): Add BE support.
1526 2018-06-07 Paul Koning <ni1d@arrl.net>
1528 * compare-elim.c (try_merge_compare): Don't merge compare if
1529 address contains a side effect.
1530 (try_eliminate_compare): Likewise.
1532 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1534 * config.gcc: Support "tremont".
1535 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1536 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1538 * config/i386/i386.c (m_TREMONT): Define.
1539 (processor_target_table): Add "tremont".
1540 (PTA_TREMONT): Define.
1541 (ix86_lea_outperforms): Add TARGET_TREMONT.
1542 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1543 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1544 and M_INTEL_GOLDMONT_PLUS.
1545 (fold_builtin_cpu): Add "tremont".
1546 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1547 (ix86_option_override_internal): Add "tremont".
1548 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1549 (processor_type): Add PROCESSOR_TREMONT.
1550 * config/i386/x86-tune.def: Add m_TREMONT.
1551 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1553 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1555 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1556 symbol defined for msp430i* devices to be lower case.
1558 gcc/testsuite/gcc.target/msp430/
1559 * msp430i-device-symbol.c: New test.
1560 * msp430f-device-symbol.c: New test.
1561 * msp430.h: New test header file.
1563 2018-06-07 Richard Biener <rguenther@suse.de>
1565 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1566 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1567 Properly wrap signed arithmetic if overflow wraps.
1569 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1571 PR tree-optimization/69615
1572 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1573 of a cast from a same precision integral SSA_NAME in a bb dominated
1574 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1575 cast to utype if rhs2 has already a compatible type.
1577 2018-06-07 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/85935
1580 * graphite-scop-detection.c (find_params_in_bb): Analyze
1581 condition operands with respect to the correct loop. Assert
1582 the analysis doesn't fail.
1584 2018-06-04 Carl Love <cel@us.ibm.com>
1586 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1587 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1588 as it is slightly cheaper.
1589 (first_match_or_eos_index_<mode>):
1590 Calculate index using natural element order.
1591 (first_match_index_<mode>):
1592 Calculate index using natural element order.
1593 (first_match_or_eos_index_<mode>):
1594 Calculate index using natural order.
1595 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1596 for BE and LE modes.
1597 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1598 P9V_BUILTIN_VCLZLSBB_V16QI.
1599 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1602 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1604 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1605 indentation and line wrap for many prototypes. Add missing
1606 @smallexample directives around block of prototypes for vec_xl and
1609 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1611 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1612 track if we pass or return IEEE 128-bit floating point.
1613 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1614 C++ mangling that is compatible with GCC 8.1.
1615 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1616 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1617 floating point types.
1618 (rs6000_function_arg_advance_1): Likewise.
1619 (rs6000_mangle_type): Optionally generate mangled names that match
1620 what GCC 8.1 generated for IEEE 128-bit floating point types.
1621 (rs6000_globalize_decl_name): If we have an external function that
1622 passes or returns IEEE 128-bit types, generate a weak reference
1623 from the mangled name used in GCC 8.1 to the current mangled
1625 (rs6000_init_builtins): Make __ibm128 use the long double type if
1626 long double is IBM extended double. Make __float128 use the long
1627 double type if long double is IEEE 128-bit.
1630 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1631 macro for __ibm128 built-in functions.
1632 (PACK_IF): Add __ibm128 pack/unpack functions.
1633 (UNPACK_IF): Likewise.
1634 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1635 enable long double built-in functions if long double is IEEE
1636 128-bit floating point.
1637 (rs6000_invalid_builtin): Update long double built-in function
1639 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1640 functions, adjust the built-in function to use the long double
1641 built-in function if __ibm128 and long double are the same type.
1642 * doc/extend.texi (PowerPC builtins): Update documention for
1643 __builtin_{,un}pack_longdouble. Add documentation for
1644 __builtin_{,un}pack_ibm128.
1646 2018-06-06 Jim Wilson <jimw@sifive.com>
1648 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1649 (struct machine_function): New field interrupt_mode.
1650 (riscv_handle_type_attribute): New function. Add forward declaration.
1651 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1652 (riscv_expand_epilogue): Check interrupt_mode field.
1653 (riscv_set_current_function): Check interrupt attribute args and
1654 set interrupt_mode field.
1655 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1656 (riscv_sret, riscv_uret): New.
1657 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1658 new arguments to interrupt attribute.
1660 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1663 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1664 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1666 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1668 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1669 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1670 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1671 several redundant entries.
1673 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1675 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1676 type from "rtx" to "rtx_insn *".
1677 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1678 for local "call_insn", removing cast.
1679 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1681 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1683 PR tree-optimization/86066
1684 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1685 for BIT_INSERT_EXPR stores.
1687 2018-06-06 Richard Biener <rguenther@suse.de>
1689 PR tree-optimization/86062
1690 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1691 component refs ontop
1692 of to be offsetted base.
1694 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1696 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1697 to be static and remove check on interrupt attribute name.
1699 gcc/testsuite/gcc.target/msp430/
1700 * function-attributes-4.c: New test.
1701 * static-interrupts.c: New test.
1703 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1705 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1706 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1708 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1711 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1713 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1714 Remove second argument, change how error is called.
1715 (aarch64_layout_arg): Remove second argument from
1716 aarch64_err_no_fpadvsimd call.
1717 (aarch64_init_cumulative_args): Ditto.
1718 (aarch64_gimplify_va_arg_expr): Ditto.
1719 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1721 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1723 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1724 (*simple_return_indirect_internal<mode>): Rename from
1725 simple_return_indirect_internal. Use W mode iterator.
1726 (rstorssp): New expander.
1727 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1728 (clrssbsy): New expander.
1729 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1731 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1733 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1735 (cmse_check_pointed_object): Likewise.
1737 2018-06-05 Martin Liska <mliska@suse.cz>
1739 PR gcov-profile/47618
1740 * doc/invoke.texi: Document how -fprofile-dir format
1743 2018-06-05 Richard Biener <rguenther@suse.de>
1745 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1746 removal pretend DOM info isn't available so we do not update
1747 it and only remove edges, not dominated blocks. Actually free
1748 DOM info in case we removed something. Remove unreachable blocks.
1749 (mfb_keep_latches): Work with either DOM info or marked backedges.
1750 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1751 first. Mark backedges if DOM info isn't available.
1752 (Re-)compute DOM info after cleanup_control_flow_pre.
1754 2018-06-05 Richard Biener <rguenther@suse.de>
1756 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1757 (locus_discrim_hasher::hash): Adjust.
1758 (locus_discrim_hasher::equal): Likewise.
1759 (next_discriminator_for_locus): Work on line directly.
1760 (same_line_p): Pass in expanded locus1 as well.
1761 (assign_discriminators): Avoid redundant location expansions.
1763 2018-06-05 Richard Biener <rguenther@suse.de>
1765 PR tree-optimization/86046
1766 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1767 if required after clearing TREE_ADDRESSABLE.
1769 2018-06-05 Richard Biener <rguenther@suse.de>
1771 PR tree-optimization/86047
1772 * tree-ssa-loop.c (for_each_index): Glob handling of all
1773 decls and constants and really handle all of them.
1775 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1778 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1779 qualifier_void_pointer and qualifier_const_void_pointer.
1780 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1781 (arm_init_builtins): Handle the above.
1782 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1783 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1784 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1787 2018-06-05 Martin Liska <mliska@suse.cz>
1789 * auto-profile.c (read_autofdo_file): Do not use
1790 gcov_ctr_summary struct.
1791 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1792 * coverage.c (struct counts_entry): Likewise.
1793 (read_counts_file): Read just single summary entry.
1794 (get_coverage_counts): Use gcov_summary struct.
1795 * coverage.h (get_coverage_counts): Likewise.
1796 * gcov-dump.c (dump_working_sets): Likewise.
1797 (tag_summary): Dump just single summary.
1798 * gcov-io.c (gcov_write_summary): Write just histogram
1800 (gcov_read_summary): Read just single summary.
1801 (compute_working_sets): Use gcov_summary struct.
1802 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1803 of GCOV_COUNTERS_SUMMABLE.
1804 (GCOV_COUNTERS_SUMMABLE): Remove.
1805 (GCOV_FIRST_VALUE_COUNTER): Replace with
1806 GCOV_COUNTER_V_INTERVAL.
1807 (struct gcov_ctr_summary): Remove.
1808 (struct gcov_summary): Directly use fields of former
1810 (compute_working_sets): Use gcov_summary struct.
1811 * gcov.c (read_count_file): Do not use ctrs fields.
1812 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1814 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1816 * profile.c: Likewise.
1817 * profile.h: Likewise.
1819 2018-06-05 Martin Liska <mliska@suse.cz>
1821 PR gcov-profile/84846
1822 * gcov.c (output_lines): Print working directory only
1823 in intermediate format.
1825 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1827 * config/s390/s390-builtin-types.def: Add void function type.
1828 * config/s390/s390-builtins.def: Use the function type for the
1831 2018-06-04 Jim Wilson <jimw@sifive.com>
1833 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1835 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1836 and maybe_eh_return. Change regno to unsigned int. Use new args to
1837 handle EH_RETURN_DATA_REGNO registers properly.
1838 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1839 (riscv_expand_epilogue): Update comment. Change argument name and
1840 type. Update code to use new name and type. Pass new args to
1841 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1843 * config/riscv/riscv.md (NORMAL_RETURN): New.
1844 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1845 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1846 (eh_return): Call gen_eh_return_internal and emit barrier.
1847 (eh_return_internal): Call riscv_expand_epilogue.
1849 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1851 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1852 bit_insertion field and declare can_be_merged_into method.
1853 (merged_store_group::can_be_merged_into): New method.
1854 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1855 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1856 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1858 2018-06-04 Richard Biener <rguenther@suse.de>
1860 PR tree-optimization/85955
1861 * builtins.c (fold_builtin_sincos): Convert pointers to
1862 destination to appropriate type before dereferencing.
1864 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1866 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1868 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1870 * expr.c (expand_expr_real_1): Force the operand into memory if
1871 its TYPE_MODE is BLKmode and if there is no integer mode for
1872 the number of bits being extracted.
1874 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1878 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1879 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1881 2018-06-04 Richard Biener <rguenther@suse.de>
1883 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1884 (cleanup_tree_cfg_noloop): ... single caller. Do
1885 start_recording_case_labels later.
1887 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1889 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1890 to _IMMINTRIN_H_INCLUDED.
1891 * config/i386/pconfigintrin.h: Ditto.
1892 * config/i386/waitpkgintrin.h: Ditto.
1893 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1894 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1895 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1896 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1897 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1898 waitpkgintrin.h and cldemoteintrin.h.
1900 2018-06-04 Richard Biener <rguenther@suse.de>
1902 PR tree-optimization/86038
1903 * tracer.c (find_best_successor): Check probability for
1904 being initialized, bail out if not.
1906 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1909 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1910 of bits to ignore when comparing architectures.
1912 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1914 PR tree-optimization/69615
1915 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1916 maximum or minimum of the type, try to merge it also as if
1917 range1 is + [-, x - 1] or + [x + 1, -].
1920 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1922 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1924 PR tree-optimization/86034
1925 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1926 the unsigned bitfield type in a bit insertion sequence if it does not
1927 have a larger precision than the bitfield size.
1928 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1930 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1932 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1934 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1936 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1937 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1938 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1939 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1941 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1943 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1944 Disable -fdelete-null-pointer-checks for ELF toolchain.
1946 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1947 Kito Cheng <kito.cheng@gmail.com>
1949 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1950 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1951 (nds32*-*-*): Add glibc and uclibc conditions.
1952 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1953 (TARGET_EXCEPT_UNWIND_INFO): Define.
1954 * config/nds32/elf.h: New file.
1955 * config/nds32/linux.h: New file.
1956 * config/nds32/nds32-elf.opt: New file.
1957 * config/nds32/nds32-linux.opt: New file.
1958 * config/nds32/nds32-fp-as-gp.c
1959 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1960 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1962 (nds32_asm_file_end): Ditto.
1963 (nds32_print_operand): Ditto.
1964 (nds32_insert_attributes): Ditto.
1965 (nds32_init_libfuncs): New function.
1966 (TARGET_HAVE_TLS): Define.
1967 (TARGET_INIT_LIBFUNCS): Define.
1968 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1970 (TARGET_ELF): Apply different mcmodel setting.
1971 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1972 been migrated into elf.h and linux.h files.
1973 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1974 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1975 (mcmodel): The content has been migrated into nds32-elf.opt and
1976 nds32-linux.opt files.
1977 * config/nds32/t-elf: New file.
1978 * config/nds32/t-linux: New file.
1980 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1981 Shiva Chen <shiva0217@gmail.com>
1983 * config/nds32/constants.md (unspec_volatile_element): Add
1984 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1985 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1987 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1988 (make_pass_nds32_fp_as_gp): Declare.
1989 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1991 (nds32_asm_function_end_prologue): Remove unused asm output.
1992 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1993 (nds32_asm_file_start): Output necessary fp_as_gp information.
1994 (nds32_option_override): Adjust register usage.
1995 (nds32_expand_prologue): Consider fp_as_gp situation.
1996 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1997 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2000 (simple_return): Ditto.
2001 (omit_fp_begin): Output special directive for fp_as_gp.
2002 (omit_fp_end): Output special directive for fp_as_gp.
2003 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2004 mforbid-fp-as-gp): New options.
2006 2018-06-01 Mark Wielaard <mark@klomp.org>
2008 * dwarf2out.c (dwarf2out_finish): Remove generation of
2009 DW_AT_loclists_base.
2011 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2013 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2014 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2015 (struct merged_store_group): Add bit_insertion field.
2016 (dump_char_array): Use standard hexadecimal format.
2017 (merged_store_group::merged_store_group): Set bit_insertion to false.
2018 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2019 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2020 also print the mask in the dump file.
2021 (pass_store_merging::gate): Minor tweak.
2022 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2023 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2024 stores with INTEGER_CST stores.
2025 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2026 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2027 and use it throughout. Generate bit insertion sequences if need be.
2028 (pass_store_merging::process_store): Remove redundant condition.
2029 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2031 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2033 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2034 the 128-bit floating point types. Fix function comment.
2036 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2038 * config/aarch64/aarch64-simd.md
2039 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2041 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2044 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2046 PR tree-optimization/85989
2047 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2049 (backprop::intersect_uses): Check it when deciding whether this
2050 is a backedge reference.
2051 (backprop::process_block): Add each phi to m_visited_phis
2052 after visiting it, then clear it at the end.
2054 2018-06-01 Richard Biener <rguenther@suse.de>
2056 * tree-vectorizer.h (vect_dr_stmt): New function.
2057 (vect_get_load_cost): Adjust.
2058 (vect_get_store_cost): Likewise.
2059 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2060 Use vect_dr_stmt instead of DR_SMTT.
2061 (vect_record_base_alignments): Likewise.
2062 (vect_calculate_target_alignment): Likewise.
2063 (vect_compute_data_ref_alignment): Likewise and make static.
2064 (vect_update_misalignment_for_peel): Likewise.
2065 (vect_verify_datarefs_alignment): Likewise.
2066 (vector_alignment_reachable_p): Likewise.
2067 (vect_get_data_access_cost): Likewise. Pass down
2068 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2069 (vect_get_peeling_costs_all_drs): Likewise.
2070 (vect_peeling_hash_get_lowest_cost): Likewise.
2071 (vect_enhance_data_refs_alignment): Likewise.
2072 (vect_find_same_alignment_drs): Likewise.
2073 (vect_analyze_data_refs_alignment): Likewise.
2074 (vect_analyze_group_access_1): Likewise.
2075 (vect_analyze_group_access): Likewise.
2076 (vect_analyze_data_ref_access): Likewise.
2077 (vect_analyze_data_ref_accesses): Likewise.
2078 (vect_vfa_segment_size): Likewise.
2079 (vect_small_gap_p): Likewise.
2080 (vectorizable_with_step_bound_p): Likewise.
2081 (vect_prune_runtime_alias_test_list): Likewise.
2082 (vect_analyze_data_refs): Likewise.
2083 (vect_supportable_dr_alignment): Likewise.
2084 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2085 (vect_gen_prolog_loop_niters): Likewise.
2086 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2087 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2089 (vect_recog_mask_conversion_pattern): Likewise.
2090 (vect_try_gather_scatter_pattern): Likewise.
2091 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2092 to vect_get_store_cost.
2093 (vect_get_store_cost): Get stmt_info instead of DR.
2094 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2095 (vect_get_load_cost): Get stmt_info instead of DR.
2097 2018-06-01 Richard Biener <rguenther@suse.de>
2100 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2101 vars wrapped in MEM_REFs.
2103 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2105 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2106 Fix subreg tests so that we only return a choice between
2107 GENERAL_REGS and FP_REGS if the original classes included both.
2109 2018-06-01 Richard Biener <rguenther@suse.de>
2112 * tree-ssa-structalias.c (get_function_part_constraint):
2113 Handle NULL fi->decl.
2114 (find_func_aliases_for_call): Properly handle indirect
2115 fi from direct call.
2116 (find_func_clobbers): Likewise.
2117 (ipa_pta_execute): Likewise.
2118 (create_variable_info_for): For functions that are ifunc_resolver
2119 resolve to a varinfo that contains the result of the resolver call.
2120 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2123 2018-05-31 Michael Collison <michael.collison@arm.com>
2125 * config/aarch64/aarch64.md:
2126 (*fix_to_zero_extenddfdi2): New pattern.
2127 * gcc.target/aarch64/fix_extend1.c: New testcase.
2129 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2133 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2134 and BUILT_IN_STRNCMP_EQ.
2135 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2136 BUILT_IN_STRNCMP_EQ.
2137 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2138 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2139 (gimple_fold_builtin): Likewise.
2140 * tree-ssa-strlen.c (compute_string_length): New function.
2141 (determine_min_obsize): New function.
2142 (handle_builtin_string_cmp): New function to handle calls to
2143 string compare functions.
2144 (strlen_optimize_stmt): Add handling to builtin string compare
2146 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2147 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2148 * tree.c (build_common_builtin_nodes): Add new defines of
2149 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2151 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2154 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2155 functions with naked attribute.
2157 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2159 * config/i386/sse.md (avx_vec_concat<mode>):
2160 Substitute concat_tg_mode mode attribute with xtg_mode.
2161 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2162 (concat_tg_mode): Remove mode attribute.
2164 2018-05-31 Martin Sebor <msebor@redhat.com>
2167 * calls.c (alloc_max_size): Correct a logic error/typo.
2168 Treat excessive arguments as infinite. Warn for invalid arguments.
2169 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2171 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2174 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2175 and movx for Haswell.
2177 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2178 Cesar Philippidis <cesar@codesourcery.com>
2181 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2182 when emitting error on private/firstprivate reductions.
2183 * omp-low.c (lower_omp_target): Avoid reference-type processing
2184 on pointers for firstprivate clause.
2186 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2188 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2191 * config/aarch64/aarch64-simd.md
2192 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2193 (aarch64_ld1_x3_<mode>): Likewise
2194 (aarch64_st1x2<VALLDIF:mode>): Likewise
2195 (aarch64_st1_x2_<mode>): Likewise
2196 (aarch64_st1x3<VALLDIF:mode>): Likewise
2197 (aarch64_st1_x3_<mode>): Likewise
2198 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2199 (vld1_s8_x3): Likewise.
2200 (vld1_u16_x3): Likewise.
2201 (vld1_s16_x3): Likewise.
2202 (vld1_u32_x3): Likewise.
2203 (vld1_s32_x3): Likewise.
2204 (vld1_u64_x3): Likewise.
2205 (vld1_s64_x3): Likewise.
2206 (vld1_f16_x3): Likewise.
2207 (vld1_f32_x3): Likewise.
2208 (vld1_f64_x3): Likewise.
2209 (vld1_p8_x3): Likewise.
2210 (vld1_p16_x3): Likewise.
2211 (vld1_p64_x3): Likewise.
2212 (vld1q_u8_x3): Likewise.
2213 (vld1q_s8_x3): Likewise.
2214 (vld1q_u16_x3): Likewise.
2215 (vld1q_s16_x3): Likewise.
2216 (vld1q_u32_x3): Likewise.
2217 (vld1q_s32_x3): Likewise.
2218 (vld1q_u64_x3): Likewise.
2219 (vld1q_s64_x3): Likewise.
2220 (vld1q_f16_x3): Likewise.
2221 (vld1q_f32_x3): Likewise.
2222 (vld1q_f64_x3): Likewise.
2223 (vld1q_p8_x3): Likewise.
2224 (vld1q_p16_x3): Likewise.
2225 (vld1q_p64_x3): Likewise.
2226 (vst1_s64_x2): Likewise.
2227 (vst1_u64_x2): Likewise.
2228 (vst1_f64_x2): Likewise.
2229 (vst1_s8_x2): Likewise.
2230 (vst1_p8_x2): Likewise.
2231 (vst1_s16_x2): Likewise.
2232 (vst1_p16_x2): Likewise.
2233 (vst1_s32_x2): Likewise.
2234 (vst1_u8_x2): Likewise.
2235 (vst1_u16_x2): Likewise.
2236 (vst1_u32_x2): Likewise.
2237 (vst1_f16_x2): Likewise.
2238 (vst1_f32_x2): Likewise.
2239 (vst1_p64_x2): Likewise.
2240 (vst1q_s8_x2): Likewise.
2241 (vst1q_p8_x2): Likewise.
2242 (vst1q_s16_x2): Likewise.
2243 (vst1q_p16_x2): Likewise.
2244 (vst1q_s32_x2): Likewise.
2245 (vst1q_s64_x2): Likewise.
2246 (vst1q_u8_x2): Likewise.
2247 (vst1q_u16_x2): Likewise.
2248 (vst1q_u32_x2): Likewise.
2249 (vst1q_u64_x2): Likewise.
2250 (vst1q_f16_x2): Likewise.
2251 (vst1q_f32_x2): Likewise.
2252 (vst1q_f64_x2): Likewise.
2253 (vst1q_p64_x2): Likewise.
2254 (vst1_s64_x3): Likewise.
2255 (vst1_u64_x3): Likewise.
2256 (vst1_f64_x3): Likewise.
2257 (vst1_s8_x3): Likewise.
2258 (vst1_p8_x3): Likewise.
2259 (vst1_s16_x3): Likewise.
2260 (vst1_p16_x3): Likewise.
2261 (vst1_s32_x3): Likewise.
2262 (vst1_u8_x3): Likewise.
2263 (vst1_u16_x3): Likewise.
2264 (vst1_u32_x3): Likewise.
2265 (vst1_f16_x3): Likewise.
2266 (vst1_f32_x3): Likewise.
2267 (vst1_p64_x3): Likewise.
2268 (vst1q_s8_x3): Likewise.
2269 (vst1q_p8_x3): Likewise.
2270 (vst1q_s16_x3): Likewise.
2271 (vst1q_p16_x3): Likewise.
2272 (vst1q_s32_x3): Likewise.
2273 (vst1q_s64_x3): Likewise.
2274 (vst1q_u8_x3): Likewise.
2275 (vst1q_u16_x3): Likewise.
2276 (vst1q_u32_x3): Likewise.
2277 (vst1q_u64_x3): Likewise.
2278 (vst1q_f16_x3): Likewise.
2279 (vst1q_f32_x3): Likewise.
2280 (vst1q_f64_x3): Likewise.
2281 (vst1q_p64_x3): Likewise.
2283 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2285 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2286 user_label_prefix to name.
2288 * tree-core.h: Update comment about the format of NAME string
2289 passed to handler in attribute_spec.
2291 * config/msp430/msp430.md: Remove erroneous subreg expression from
2292 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2293 zero_extend{q,h}isi2.
2295 2018-05-30 Borislav Petkov <bp@suse.de>
2297 * doc/extend.texi: Document some architecture specific
2298 constraints and sort entries.
2300 2018-05-30 Martin Sebor <msebor@redhat.com>
2303 * builtins.c (expand_builtin_stpcpy_1): New function.
2304 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2305 only if the former succeeds.
2307 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2309 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2312 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2314 * doc/invoke.texi (-flinker-output): Document
2316 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2318 * passes.c (ipa_write_summaries): Only modify statements if body
2320 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2321 incrementally linking.
2322 (ipa_passes): Likewise.
2323 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2324 pass down resolution info.
2325 * common.opt (flag_incremental_link): Update info.
2326 * gcc.c (plugin specs): Turn flinker-output=* to
2327 -plugin-opt=-linker-output-known
2328 * toplev.c (compile_file): Also cut compilation when doing incremental
2330 * flag-types. (enum lto_partition_model): Add
2331 LTO_LINKER_OUTPUT_NOLTOREL.
2332 (invoke.texi): Add -flinker-output docs.
2333 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2334 link same way as WPA; do not stream in dead initializers.
2336 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2337 darf2out_register_external_decl): Support incremental link.
2339 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2341 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2343 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2345 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2346 it down to simple_object_copy_lto_debug_sections.
2347 (run_gcc): Determine incremental LTO link time and configure
2348 lto1 into non-wpa mode, disable renaming of debug sections.
2350 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2352 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2353 descriptions of various incorrectly documented functions.
2355 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2358 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2359 address check not strict.
2361 2018-05-30 Richard Biener <rguenther@suse.de>
2363 PR tree-optimization/85964
2364 * tracer.c (better_p): Drop initialized count check, we only
2365 call the function with initialized counts now.
2366 (find_best_successor): Do find a best edge if one
2367 has uninitialized count.
2368 (find_best_predecessor): Likewise. Do BB frequency check only
2369 if count is initialized.
2371 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2373 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2374 (aarch64_ldrstr_offset_compare): New.
2375 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2376 load/store orderings.
2377 (aarch64_gen_adjusted_ldpstp): Likewise.
2379 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2381 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2382 Check for subset of GENERAL_REGS and FP_REGS.
2383 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2386 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2388 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2389 and wi::to_poly_offset. Add the current offset and then check
2390 whether the sum fits, rather than using an unchecked addition of
2391 a checked term. Check for a shwi rather than a uhwi.
2392 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2393 (store_constructor): Use poly_int_tree_p.
2394 (expand_expr_real_1): Likewise.
2395 * function.c (assign_temp): Likewise.
2396 * fold-const.c (const_binop): Use poly_int_tree_p and
2398 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2400 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2401 to_poly_offset for MEM offsets.
2402 * ipa-icf.c (sem_variable::equals): Likewise.
2403 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2404 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2405 wi::to_poly_offset for BIT_FIELD_REF offsets.
2406 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2408 * var-tracking.c (emit_note_insn_var_location): Use
2409 tree_to_poly_uint64.
2411 2018-05-29 Jim Wilson <jimw@sifive.com>
2413 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2415 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2418 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2419 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2421 (sse4_1_round<mode>2): Use nonimmediate_operand
2422 for operand 1 predicate.
2424 2018-05-29 Martin Sebor <msebor@redhat.com>
2425 Richard Biener <rguenther@suse.de>
2428 * calls.c (get_size_range): Call determine_value_range instead
2429 of get_value_range..
2430 * tree-vrp.h (determine_value_range): Declared new function.
2431 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2433 2018-05-29 Richard Biener <rguenther@suse.de>
2435 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2436 sure to use non-pattern stmts for get_earlier_stmt arguments.
2437 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2438 called on pattern stmts.
2439 (get_later_stmt): Likewise.
2441 2018-05-29 Martin Liska <mliska@suse.cz>
2443 PR gcov-profile/85759
2444 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2447 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2449 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2451 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2455 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2456 VEC_PACK_FLOAT_EXPR): New tree codes.
2457 * tree-pretty-print.c (op_code_prio): Handle
2458 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2459 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2460 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2461 * tree-inline.c (estimate_operator_cost): Likewise.
2462 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2463 * fold-const.c (const_binop): Likewise.
2464 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2465 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2466 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2467 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2468 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2469 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2470 * expr.c (expand_expr_real_2): Likewise.
2471 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2472 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2473 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2475 * optabs.c (expand_widen_pattern_expr): For
2476 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2477 sign from result type rather than operand's type.
2478 (expand_binop_directly): For vec_packu_float_optab and
2479 vec_packs_float_optab allow result type to be different from operand's
2481 * optabs-tree.c (optab_for_tree_code): Handle
2482 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2483 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2484 * tree-vect-generic.c (expand_vector_operations_1): Handle
2485 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2486 VEC_PACK_FLOAT_EXPR.
2487 * tree-vect-stmts.c (supportable_widening_operation): Handle
2489 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2490 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2491 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2492 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2493 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2495 (vec_pack<floatprefix>_float_<mode>): New expander.
2496 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2498 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2499 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2500 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2501 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2502 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2504 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2505 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2506 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2507 VEC_PACK_FLOAT_EXPR): Document.
2509 2018-05-29 Richard Biener <rguenther@suse.de>
2511 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2513 (stmt_vec_info_vec): Make pointer.
2514 (init_stmt_vec_info_vec): Remove.
2515 (free_stmt_vec_info_vec): Likewise.
2516 (set_stmt_vec_info_vec): New function.
2517 (free_stmt_vec_infos): Likewise.
2518 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2519 (set_vinfo_for_stmt): Likewise.
2520 (get_earlier_stmt): Likewise.
2521 (get_later_stmt): Likewise.
2522 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2523 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2524 (vec_info::~vec_info): Free stmt_vec_infos.
2525 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2526 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2527 (pass_slp_vectorize::execute): Likewise.
2528 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2529 (free_stmt_vec_info_vec): Likewise.
2530 (set_stmt_vec_info_vec): New function.
2531 (free_stmt_vec_infos): Likewise.
2532 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2533 the global stmt_vec_info_vec.
2534 * tree-parloops.c (gather_scalar_reductions): Use
2535 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2538 2018-05-29 Richard Biener <rguenther@suse.de>
2540 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2542 2018-05-29 Martin Liska <mliska@suse.cz>
2543 David Malcolm <dmalcolm@redhat.com>
2545 * vec.c (test_reverse): New.
2546 (vec_c_tests): Add new test.
2547 * vec.h (vl_ptr>::reverse): New function.
2549 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2551 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2553 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2556 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2558 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2560 2018-05-28 Richard Biener <rguenther@suse.de>
2562 PR tree-optimization/85933
2563 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2564 look at stmts marked as vectorizable.
2566 2018-05-28 Richard Biener <rguenther@suse.de>
2568 PR tree-optimization/85934
2569 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2570 vector boolean check before scalar optimization.
2572 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2574 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2577 2018-05-28 Mark Wielaard <mark@klomp.org>
2579 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2580 if it is an expression containing a minus sign.
2582 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2584 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2586 2018-05-27 Paul Koning <ni1d@arrl.net>
2588 * config/pdp11/pdp11.md (truncsihi2): Remove.
2590 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2591 Chung-Ju Wu <jasonwucj@gmail.com>
2593 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2595 (unaligned_store_dw): Ditto.
2596 * config/nds32/nds32-memory-manipulation.c
2597 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2598 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2599 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2600 (emit_setmem_word_loop): Rename to ...
2601 (emit_setmem_doubleword_loop): ... this.
2602 (nds32_gen_dup_4_byte_to_word_value): New function.
2603 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2604 (nds32_expand_setmem_loop): Refine implementation.
2605 (nds32_expand_setmem_loop_v3m): Ditto.
2606 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2609 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2611 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2613 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2615 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2616 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2617 machine->attr_no_prologue_p.
2618 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2619 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2620 (nds32_expand_epilogue): Consider attr_naked_p.
2621 (nds32_expand_epilogue_v3pop): Likewise.
2622 (nds32_can_use_return_insn): Likewise.
2623 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2624 attr_no_prologue_p fields.
2625 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2627 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2630 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2632 * config/i386/sse.md
2633 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2635 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2637 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2639 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2641 (*<floatsuffix>floatv2div2sf2): Rename to ...
2642 (*float<floatunssuffix>v2div2sf2): ... this.
2643 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2644 (float<floatunssuffix>v2div2sf2_mask): ... this.
2645 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2646 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2647 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2649 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2651 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2653 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2655 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2657 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2659 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2660 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2661 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2662 gen_ufix_truncv8dfv8si2.
2663 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2664 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2665 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2666 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2667 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2668 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2669 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2670 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2672 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2676 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2678 2018-05-25 Jim Wilson <jimw@sifive.com>
2680 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2681 * config/riscv/riscv.c (struct machine_function): Add
2682 interrupt_handler_p and attribute_checked_p fields.
2683 (riscv_attribute_table): Add interrupt.
2684 (riscv_interrupt_type_p): New.
2685 (riscv_save_reg_p): Save extra regs for interrupt handler.
2686 (riscv_use_save_libcall): Return false for interrupt handler.
2687 (riscv_first_stack_step): Add forward declaration.
2688 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2689 for interrupt handler with large frame. Use it for saved reg list.
2690 (riscv_expand_prologue): Move flag_stack_usage_info support to
2691 eliminate duplication.
2692 (riscv_expand_epilogue): Generate mret for interrupt handler.
2693 (riscv_epilogue_uses): New.
2694 (riscv_can_use_return_insn): Return false for interrupt handler.
2695 (riscv_function_ok_for_sibcall): Likewise.
2696 (riscv_set_current_function): Add interrupt handler support.
2697 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2698 * config/riscv/riscv.md (UNSPECV_MRET): New.
2700 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2702 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2704 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2706 PR tree-optimization/85712
2707 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2708 this candidate has already been replaced in-situ by a copy.
2710 2018-05-25 Jason Merrill <jason@redhat.com>
2712 PR c++/80485 - inline function non-zero address.
2713 * symtab.c (nonzero_address): Check DECL_COMDAT.
2715 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2718 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2719 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2720 extension of SImode operation. Use const123_operand predicate.
2721 (*saddsi_1): Remove.
2722 (*saddl_se_1): Ditto.
2724 (*ssubl_se_1): Ditto.
2725 * config/alpha/predicates.md (const123_operand): New predicate.
2726 * config/alpha/constraints.md (P): Use IN_RANGE.
2728 2018-05-25 Richard Biener <rguenther@suse.de>
2730 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2732 (ref_maybe_used_by_stmt_p): Likewise.
2733 (stmt_may_clobber_ref_p): Likewise.
2734 (stmt_may_clobber_ref_p_1): Likewise.
2735 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2737 (ref_maybe_used_by_stmt_p): Likewise.
2738 (stmt_may_clobber_ref_p): Likewise.
2739 (stmt_may_clobber_ref_p_1): Likewise.
2740 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2741 the alias oracle to disambiguate DRs with stmts DR analysis
2743 (vect_analyze_data_refs): Do not give up on not analyzable
2744 DRs for BB vectorization. Remove code truncating the dataref
2747 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2750 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2751 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2752 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2754 2018-05-25 Richard Biener <rguenther@suse.de>
2756 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2757 function, combining stmt data ref gathering and fatal analysis
2759 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2760 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2761 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2762 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2763 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2765 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2767 PR tree-optimization/85720
2768 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2769 SCC if all partitions are builtins.
2770 (version_loop_by_alias_check): New parameter. Generate cancelable
2771 runtime alias check if all partitions are builtins.
2772 (distribute_loop): Update call to above function.
2774 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2776 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2777 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2778 (parm_default_def_partition_arg): Ditto.
2779 (set_parm_default_def_partition): Ditto.
2780 (get_parm_default_def_partitions): Ditto and make it static.
2781 (get_undefined_value_partitions): Ditto and make it static.
2782 (remove_ssa_form): Refactor call to init_var_map here.
2783 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2784 computation for loop region.
2785 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2786 (register_default_def): Delete.
2787 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2788 (parm_default_def_partition_arg): Ditto.
2789 (set_parm_default_def_partition): Ditto.
2790 (get_parm_default_def_partitions): Ditto and make it static.
2791 (get_undefined_value_partitions): Ditto and make it static.
2792 (coalesce_with_default, coalesce_with_default): Update comment.
2793 (create_coalesce_list_for_region): New func factored out from
2794 create_outofssa_var_map.
2795 (populate_coalesce_list_for_outofssa): New func factored out from
2796 create_outofssa_var_map and coalesce_ssa_name.
2797 (create_outofssa_var_map): Delete.
2798 (coalesce_ssa_name): Refactor to support live range computation.
2799 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2800 (get_parm_default_def_partitions): Delete.
2801 (get_undefined_value_partitions): Ditto.
2802 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2803 computation for loop region.
2804 (new_tree_live_info, loe_visit_block): Ditto.
2805 (live_worklist, set_var_live_on_entry): Ditto.
2806 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2807 * tree-ssa-live.h (struct _var_map): New fields.
2808 (init_var_map): Change decl.
2809 (region_contains_p): New.
2811 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2813 * tree-ssa-live.h (live_merge_and_clear): Delete.
2815 2018-05-25 Richard Biener <rguenther@suse.de>
2818 * tree-dump.c (dequeue_and_dump): Remove access to removed
2819 operand 2 of a SWITCH_EXPR.
2821 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2823 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2824 multiplication and division.
2825 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2826 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2827 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2828 (cond_udiv_optab, cond_umod_optab): New optabs.
2829 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2830 (IFN_COND_RDIV): New internal functions.
2831 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2832 TRUNC_MOD_EXPR and RDIV_EXPR.
2833 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2834 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2836 (SVE_INT_BINARY): Include mult.
2837 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2838 (optab, sve_int_op): Handle mult.
2839 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2841 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2842 for SVE_INT_BINARY_SD.
2844 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2846 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2847 (optab, sve_int_op): Handle div and udiv.
2848 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2849 for SVE_INT_BINARY_SD.
2850 (*<optab><mode>3): New insn for the same.
2852 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2854 * tree-vect-patterns.c: Include predict.h.
2855 (vect_recog_divmod_pattern): Restrict check for division support
2856 to when optimizing for size.
2858 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2860 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2861 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2862 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2863 (gimple_match_op::set_op): Likewise.
2864 (gimple_resimplify4): Declare.
2865 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2866 (expr::gen_transform): Likewise.
2867 (decision_tree::gen): Generate a simplification routine for 4 operands.
2868 * gimple-match-head.c (gimple_simplify): Add an overload for
2869 4 operands. In the top-level function, handle up to 4 call
2870 arguments and call gimple_resimplify4.
2871 (gimple_resimplify4): New function.
2872 (build_call_internal): Pass a fourth operand.
2873 (maybe_push_to_seq): Likewise.
2874 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2875 Fold VEC_COND_EXPRs of an operation and a default value into
2876 an IFN_COND_* function if possible.
2877 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2879 (SVE_COND_FP_BINARY): Include them.
2880 (optab, sve_fp_op): Handle them.
2881 (SVE_INT_BINARY_REV): New code iterator.
2882 (SVE_COND_FP_BINARY_REV): New int iterator.
2883 (commutative): New int attribute.
2884 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2886 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2888 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2889 (*cond_<optab><mode>): New patterns for reversed operands.
2891 2018-05-25 Richard Biener <rguenther@suse.de>
2893 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2894 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2895 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2896 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2897 * tree-vect-data-refs.c (everywhere): Adjust users.
2898 * tree-vect-loop.c (everywhere): Likewise.
2899 * tree-vect-slp.c (everywhere): Likewise.
2900 * tree-vect-stmts.c (everywhere): Likewise.
2901 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2903 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2905 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2907 (gcc_cv_as_section_exclude): ... this.
2908 Try Solaris as #exclude syntax.
2909 * configure: Regenerate.
2910 * config.in: Regenerate.
2911 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2913 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2914 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2916 * varasm.c (default_elf_asm_named_section): Don't check if
2917 HAVE_GAS_SECTION_EXCLUDE is defined.
2919 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2921 * doc/md.texi: Update the documentation of the cond_* optabs
2922 to mention the new final operand. Fix GET_MODE_NUNITS call.
2923 Describe the scalar case too.
2924 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2925 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2927 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2928 (get_conditional_internal_fn): Update comment.
2929 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2930 accumulator value as a final argument to conditional functions.
2931 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2932 a define_expand and add an "else" operand. Assert for now that
2933 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2934 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2935 (*cond_<optab><mode>): New patterns.
2936 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2937 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2938 (UNSPEC_COND_EOR): Delete.
2939 (optab): Remove associated mappings.
2940 (SVE_INT_BINARY): New code iterator.
2941 (sve_int_op): Remove int attribute and add "minus" to the code
2943 (SVE_COND_INT_OP): Delete.
2944 (SVE_COND_FP_OP): Rename to...
2945 (SVE_COND_FP_BINARY): ...this.
2947 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2949 * optabs.c (can_reuse_operands_p): New function.
2950 (maybe_legitimize_operands): Try to reuse the results for
2953 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2955 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2956 Add {q} suffix to insn mnemonic.
2958 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2960 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2961 (msp430_warn_func_return): New.
2963 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2965 * fold-const.c (tree_nonzero_bits): New function.
2966 * fold-const.h (tree_nonzero_bits): Likewise.
2967 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2968 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2970 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2974 * varasm.c (assemble_alias): Check ifunc_resolver only on
2977 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2980 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2981 when memory input operand is handled.
2983 2018-05-24 Luis Machado <luis.machado@linaro.org>
2985 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2987 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2989 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2991 * match.pd: Delay FMA folds until after vectorization.
2993 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2996 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2997 address check not strict.
2999 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3001 * gimple-match.h (gimple_match_op): New class.
3002 (mprts_hook): Replace parameters with a gimple_match_op *.
3003 (maybe_build_generic_op): Likewise.
3004 (gimple_simplified_result_is_gimple_val): Replace parameters with
3005 a const gimple_match_op *.
3006 (gimple_simplify): Replace code_helper * and tree * parameters with
3007 a gimple_match_op * parameter.
3008 (gimple_resimplify1): Replace code_helper *, tree and tree *
3009 parameters with a gimple_match_op * parameter.
3010 (gimple_resimplify2): Likewise.
3011 (gimple_resimplify3): Likewise.
3012 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3013 parameters with a gimple_match_op * parameter.
3014 * gimple-match-head.c (gimple_simplify): Change prototypes of
3015 auto-generated functions to take a gimple_match_op * instead of
3016 separate code_helper * and tree * parameters. Make the same
3017 change in the top-level overload and update calls to the
3018 gimple_resimplify routines. Update calls to the auto-generated
3019 functions and to maybe_push_res_to_seq in the publicly-facing
3020 operation-specific gimple_simplify overloads.
3021 (gimple_match_op::MAX_NUM_OPS): Define.
3022 (gimple_resimplify1): Replace rcode and ops with a single res_op
3023 parameter. Update call to gimple_simplify.
3024 (gimple_resimplify2): Likewise.
3025 (gimple_resimplify3): Likewise.
3026 (mprts_hook): Replace parameters with a gimple_match_op *.
3027 (maybe_build_generic_op): Likewise.
3028 (build_call_internal): Replace type, nargs and ops with
3029 a gimple_match_op *.
3030 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3031 with a single gimple_match_op *. Update calls to mprts_hook,
3032 build_call_internal and gimple_simplified_result_is_gimple_val.
3033 Factor out code that is common to the tree_code and combined_fn cases.
3034 * genmatch.c (expr::gen_transform): Replace tem_code and
3035 tem_ops with a gimple_match_op called tem_op. Update calls
3036 to the gimple_resimplify functions and maybe_push_res_to_seq.
3037 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3038 res_ops. Update call to the gimple_resimplify functions.
3039 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3040 (decision_tree::gen): Make the functions take a gimple_match_op *
3041 called res_op instead of separate res_code and res_ops parameters.
3042 Update call accordingly.
3043 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3044 and ops with a single res_op parameter. Update calls to
3045 maybe_build_generic_op and maybe_push_res_to_seq.
3046 (fold_stmt_1): Update calls to gimple_simplify and
3047 replace_stmt_with_simplification.
3048 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3049 and gimple_simplified_result_is_gimple_val.
3050 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3052 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3053 with a gimple_match_op *.
3054 (vn_nary_build_or_lookup): Likewise. Update call to
3055 vn_nary_build_or_lookup_1.
3056 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3057 gimple_match_op *. Update calls to the gimple_resimplify routines
3058 and to gimple_simplified_result_is_gimple_val.
3059 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3060 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3061 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3062 (visit_nary_op): Likewise.
3063 (visit_reference_op_load): Likewise.
3065 2018-05-23 Luis Machado <luis.machado@linaro.org>
3067 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3068 modifier for printing the step amount.
3070 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3073 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3076 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3078 * doc/sourcebuild.texi (Endianness): New subsubsection.
3080 2018-05-23 Luis Machado <luis.machado@linaro.org>
3082 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3083 <prefetch_dynamic_strides>: New const bool field.
3084 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3085 prefetch_dynamic_strides.
3086 (exynosm1_prefetch_tune): Likewise.
3087 (thunderxt88_prefetch_tune): Likewise.
3088 (thunderx_prefetch_tune): Likewise.
3089 (thunderx2t99_prefetch_tune): Likewise.
3090 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3092 (aarch64_override_options_internal): Update to set
3093 PARAM_PREFETCH_DYNAMIC_STRIDES.
3094 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3095 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3096 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3097 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3098 prefetch-dynamic-strides setting.
3100 2018-05-23 Luis Machado <luis.machado@linaro.org>
3102 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3103 <minimum_stride>: New const int field.
3104 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3105 minimum_stride field defaulting to -1.
3106 (exynosm1_prefetch_tune): Likewise.
3107 (thunderxt88_prefetch_tune): Likewise.
3108 (thunderx_prefetch_tune): Likewise.
3109 (thunderx2t99_prefetch_tune): Likewise.
3110 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3111 <default_opt_level>: Set to 3.
3112 (aarch64_override_options_internal): Update to set
3113 PARAM_PREFETCH_MINIMUM_STRIDE.
3114 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3115 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3116 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3117 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3118 stride is constant and is below the minimum stride threshold.
3120 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3122 * config/arm/arm-cpus.in (mode26): Delete.
3123 (armv4): Delete mode26 reference.
3124 * config/arm/arm.c (arm_configure_build_target): Delete use of
3127 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3129 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3131 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3132 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3134 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3135 (floatunsdidf2): Ditto.
3137 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3139 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3140 (fixuns_trunc<mode>si2_avx512f): Ditto.
3141 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3142 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3143 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3145 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3147 PR rtl-optimization/79985
3148 * df-scan.c (df_insn_refs_collect): Remove special case for
3149 global registers and asm statements.
3151 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3153 * extend.texi (Global Register Variables): Rewrite the bullet list.
3154 Note that the register is available for allocation. Note that access
3155 via inline asm must use constraints. Add note about async-signal
3156 handlers. Remove paragraph about automagic register selection.
3158 2018-05-23 Richard Biener <rguenther@suse.de>
3160 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3161 of fixed offset from memset VN.
3163 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3165 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3167 (alloc_cand_and_find_basis): Initialize first_interp field.
3168 (slsr_process_mul): Modify first_interp field.
3169 (slsr_process_add): Likewise.
3170 (slsr_process_cast): Modify first_interp field for each new
3172 (slsr_process_copy): Likewise.
3173 (dump_candidate): Dump first_interp field.
3174 (replace_mult_candidate): Process all interpretations, not just
3176 (replace_rhs_if_not_dup): Likewise.
3177 (replace_one_candidate): Likewise.
3179 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3181 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3183 (aarch64_needs_frame_chain): New function.
3184 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3186 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3189 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3190 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3191 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3192 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3193 as true for strict-align.
3194 (aarch64_can_inline_p): Perform checks even when callee has no
3195 attributes to check for strict alignment.
3196 * doc/extend.texi (AArch64 Function Attributes): Document
3198 * doc/invoke.texi: (AArch64 Options): Likewise.
3200 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3202 PR tree-optimization/85853
3203 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3204 the handling of the root of the node to...
3205 (vect_slp_analyze_node_operations_1): ...this new function,
3206 and run the whole thing with the child nodes' def types
3207 set according to their SLP node's def type.
3209 2018-05-23 Richard Biener <rguenther@suse.de>
3212 * tree-data-ref.c (create_runtime_alias_checks): Defer
3213 and ignore overflow warnings.
3215 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3217 PR tree-optimization/85822
3218 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3221 2018-05-23 Richard Biener <rguenther@suse.de>
3223 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3224 memset constants via native_interpret_expr.
3226 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3229 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3231 (cgraph_node::create_alias): Likewise.
3232 (cgraph_node::get_availability): Check ifunc_resolver instead
3233 of looking up ifunc attribute.
3234 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3235 * varasm.c (do_assemble_alias): Likewise.
3236 (assemble_alias): Likewise.
3237 (default_binds_local_p_3): Likewise.
3238 * cgraph.h (cgraph_node): Add ifunc_resolver.
3239 (cgraph_node::only_called_directly_or_aliased_p): Return false
3241 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3243 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3244 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3245 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3246 instead of looking up ifunc attribute.
3248 2018-05-22 Luis Machado <luis.machado@linaro.org>
3250 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3252 2018-05-22 Martin Sebor <msebor@redhat.com>
3255 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3256 only when expasion succeeds.
3257 (expand_builtin_strcmp): Same.
3258 (expand_builtin_strncmp): Same.
3260 2018-05-22 Martin Sebor <msebor@redhat.com>
3262 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3264 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3265 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3267 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3268 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3270 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3271 Add check for legitimate_address.
3272 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3273 (aarch64_swap_ldrstr_operands): New.
3274 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3277 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3278 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3280 * config/aarch64/aarch64.md: New patterns to generate stp
3282 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3283 single words and double words.
3284 (load_pair_sw, load_pair_dw): Likewise.
3285 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3287 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3289 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3290 for different mode ldpstp and add peephole for merged zero stores.
3292 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3294 (aarch64_gen_store_pair): Rename calls to match new patterns.
3295 (aarch64_gen_load_pair): Rename calls to match new patterns.
3296 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3297 (load_pair<DREG:mode><DREG2:mode>): ... This.
3298 (store_pair<mode>): Rename to...
3299 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3300 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3302 (V_INT_EQUIV): Handle SImode.
3303 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3306 2018-05-22 Martin Sebor <msebor@redhat.com>
3309 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3310 or ajust the presumed bound on an operation to avoid unnecessary
3313 2018-05-22 Martin Sebor <msebor@redhat.com>
3315 PR tree-optimization/85826
3316 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3317 assuming that a DECL necesarily has a constant size.
3319 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3322 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3324 2018-05-22 Richard Biener <rguenther@suse.de>
3326 PR tree-optimization/85834
3327 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3328 non-constant and non-zero memset arguments.
3330 2018-05-22 Martin Liska <mliska@suse.cz>
3333 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3335 2018-05-22 Richard Biener <rguenther@suse.de>
3337 PR tree-optimization/85863
3338 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3339 comparisons when vectype is specified.
3340 (vectorizable_condition): Do not specify vectype for
3341 vect_is_simple_cond when SLP vectorizing.
3343 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3346 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3347 define __ibm128 as long double.
3348 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3349 as a distinct type when IEEE 128-bit support is enabled.
3350 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3351 128-bit types to use the correct functions.
3352 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3353 convert between 128-bit floating point types that have different
3354 modes but the same representation, instead of using gen_lowpart to
3356 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3358 (IFKF_reg): New attributes to give the register constraints for
3360 (extend<mode>tf2_internal): New insns to mark an explicit
3361 conversion between 128-bit floating point types that have a
3362 different mode but share the same representation.
3364 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3366 PR tree-optimization/85814
3367 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3368 a null return from get_strinfo when unsharing the next
3369 strinfo in the chain.
3371 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3374 * varasm.c (weak_finish): Clean up weak_decls.
3376 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3378 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3379 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3380 UNSPEC_UADALP values.
3381 * config/aarch64/iterators.md (ABAL): New int iterator.
3384 (sur): Add mappings for the above.
3385 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3387 (aarch64_<sur>abal<mode>_4): Likewise.
3388 (aarch64_<sur>adalp<mode>_3): Likewise.
3389 (<sur>sadv16qi): New define_expand.
3391 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3393 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3394 (*movdf_internal): Ditto.
3395 (*rcpsf2_sse): Ditto.
3396 (*rsqrtsf2_sse): Ditto.
3397 (*sqrt<mode>2_sse): Ditto.
3399 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3401 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3403 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3404 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3405 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3406 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3408 * config/aarch64/arm_neon.h: Likewise.
3409 * config/aarch64/iterators.md (VQ_I): New.
3411 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3413 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3415 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3418 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3420 * config/nds32/constraints.md (S): New constraint.
3421 * config/nds32/nds32.md (call_internal): Use constraint S.
3422 (call_value_internal): Likewise.
3423 (sibcall_internal): Likewise.
3424 (sibcall_value_internal): Likewise.
3426 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3427 Chung-Ju Wu <jasonwucj@gmail.com>
3429 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3432 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3433 Chung-Ju Wu <jasonwucj@gmail.com>
3435 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3436 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3437 (nds32_rtx_costs_impl): Simplify.
3438 (nds32_address_cost_impl): Simplify.
3439 (nds32_init_rtx_costs): New function.
3440 (nds32_rtx_costs_speed_prefer): Likewise.
3441 (nds32_rtx_costs_size_prefer): Likewise.
3442 (nds32_address_cost_speed_prefer): Likewise.
3443 (nds32_address_cost_speed_fwprop): Likewise.
3444 (nds32_address_cost_size_prefer): Likewise.
3445 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3446 * config/nds32/nds32.c (nds32_option_override): Use
3447 nds32_init_rtx_costs function.
3449 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3451 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3452 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3453 (TARGET_PIPELINE_N8): Likewise.
3454 (TARGET_PIPELINE_N10): Likewise.
3455 (TARGET_PIPELINE_N13): Likewise.
3456 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3458 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3460 * config/nds32/nds32-fpu.md: Update copyright year.
3462 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3464 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3466 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3468 * config/nds32/nds32.c
3469 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3470 * config/nds32/nds32.opt (minline-asm-r15): New option.
3472 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3474 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3476 * config/nds32/nds32.md (abssi2): New pattern.
3478 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3480 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3481 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3482 Merge insn pattern from sse_cvtsi2ss<round_name> and
3483 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3484 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3485 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3486 using SWI48 mode iterator.
3487 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3488 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3489 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3490 pattern from sse_cvttss2si<round_saeonly_name>
3491 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3492 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3493 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3494 using SWI48 mode iterator.
3495 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3496 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3497 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3498 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3499 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3500 using SWI48 mode iterator.
3501 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3502 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3503 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3504 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3505 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3506 SWI48 mode iterator.
3507 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3508 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3509 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3510 pattern from sse_cvttsd2si<round_saeonly_name>
3511 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3513 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3515 * config/nds32/nds32-md-auxiliary.c
3516 (nds32_valid_smw_lwm_base_p): Refine.
3517 (nds32_output_smw_single_word): Refine.
3518 (nds32_output_smw_double_word): New.
3519 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3521 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3523 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3524 (nds32_output_stack_pop): Refine.
3525 (nds32_expand_unaligned_load): Refine.
3526 (nds32_expand_unaligned_store): Refine.
3528 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3529 Chung-Ju Wu <jasonwucj@gmail.com>
3531 * config/nds32/constants.md: Add TP_REGNUM constant.
3532 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3533 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3535 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3536 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3537 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3538 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3539 and PIC code generation.
3540 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3542 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3544 * config/nds32/nds32.md: Support TLS and PIC.
3545 * config/nds32/nds32.c: Support TLS and PIC.
3546 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3547 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3550 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3552 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3553 mode with E_ prefix.
3555 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3556 Chung-Ju Wu <jasonwucj@gmail.com>
3558 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3559 * config/nds32/nds32-md-auxiliary.c
3560 (symbolic_reference_mentioned_p): New.
3561 (nds32_legitimize_ict_address): New.
3562 (nds32_expand_ict_move): New.
3563 (nds32_indirect_call_referenced_p): New.
3564 (nds32_symbol_binds_local_p): Delete.
3565 (nds32_long_call_p): Modify.
3566 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3567 * config/nds32/nds32-protos.h
3568 (symbolic_reference_mentioned_p): Declare.
3569 (nds32_legitimize_ict_address): Declare.
3570 (nds32_expand_ict_move): Declare.
3571 (nds32_indirect_call_referenced_p): Declare.
3572 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3573 (nds32_relax_group): Use nds32_ict_const_p as condition.
3574 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3575 (nds32_asm_file_start): Output ict_model directive in asm code.
3576 (nds32_legitimate_address_p): Consider indirect call.
3577 (nds32_print_operand): Consider indirect call.
3578 (nds32_print_operand_address): Consider indirect call.
3579 (nds32_insert_attributes): Handle "indirect_call" attribute.
3580 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3581 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3582 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3583 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3584 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3585 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3586 (TARGET_ICT_MODEL_SMALL): Define.
3587 (TARGET_ICT_MODEL_LARGE): Define.
3588 * config/nds32/nds32.md (movsi): Consider ict model.
3589 (call, call_value): Consider ict model.
3590 (sibcall, sibcall_value): Consider ict model.
3591 * config/nds32/nds32.opt (mict-model): New option.
3592 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3595 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3596 Monk Chiang <sh.chiang04@gmail.com>
3597 Jim Wilson <jimw@sifive.com>
3599 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3600 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3601 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3602 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3603 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3604 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3605 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3606 compute save_libcall_adjustment properly.
3607 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3608 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3609 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3610 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3611 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3612 (ABI_SPEC): Handle mabi=ilp32e.
3613 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3614 (RVE): Add RVE mask.
3615 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3616 <-march>: Add rv32e as an example.
3618 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3621 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3622 (intra_create_variable_infos): Handle C++ constructors.
3624 2018-05-18 Martin Liska <mliska@suse.cz>
3626 * passes.def: Remove a redundant pass.
3628 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3631 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3633 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3635 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3637 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3639 (armv2, armv2a, armv3, armv3m): Delete architectures.
3640 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3641 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3642 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3644 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3645 (*mulsidi3adddi): Likewise.
3646 (mulsidi3): Likewise.
3647 (*mulsidi3_nov6): Likewise.
3648 (umulsidi3): Likewise.
3649 (umulsidi3_nov6): Likewise.
3650 (umaddsidi4): Likewise.
3651 (*umulsidi3adddi): Likewise.
3652 (smulsi3_highpart): Likewise.
3653 (*smulsi3_highpart_nov6): Likewise.
3654 (umulsi3_highpart): Likewise.
3655 (*umulsi3_highpart_nov6): Likewise.
3656 * config/arm/arm.h (arm_arch3m): Delete.
3657 * config/arm/arm.c (arm_arch3m): Delete.
3658 (arm_option_override_internal): Update armv3-related comment.
3659 (arm_configure_build_target): Delete use of isa_bit_mode32.
3660 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3661 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3662 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3664 (mulusa3): Likewise.
3665 * config/arm/arm-protos.h (arm_arch3m): Delete.
3666 * config/arm/arm-tables.opt: Regenerate.
3667 * config/arm/arm-tune.md: Likewise.
3668 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3669 deleted architectures.
3671 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3673 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3674 (armv5t, armv5te): New features.
3675 (ARMv5, ARMv5e): Delete fgroups.
3676 (ARMv5t, ARMv5te): Adjust for above changes.
3678 (armv5, armv5e): Delete arches.
3679 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3681 (*call_reg_arm): Likewise.
3682 (*call_value_reg_armv5): Likewise.
3683 (*call_value_reg_arm): Likewise.
3684 (*call_symbol): Likewise.
3685 (*call_value_symbol): Likewise.
3686 (*sibcall_insn): Likewise.
3687 (*sibcall_value_insn): Likewise.
3689 (prefetch): Likewise.
3690 (define_split and define_peephole2 dependent on arm_arch5):
3692 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3694 (TARGET_ARM_QBIT): Likewise.
3695 (TARGET_DSP_MULTIPLY): Likewise.
3696 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3697 (arm_arch5, arm_arch5e): Delete.
3698 (arm_arch5t, arm_arch5te): Declare.
3699 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3700 (arm_arch5t): Declare.
3701 (arm_option_reconfigure_globals): Update for the above.
3702 (arm_options_perform_arch_sanity_checks): Update comment, replace
3703 use of arm_arch5 with arm_arch5t.
3704 (use_return_insn): Likewise.
3705 (arm_emit_call_insn): Likewise.
3706 (output_return_instruction): Likewise.
3707 (arm_final_prescan_insn): Likewise.
3708 (arm_coproc_builtin_available): Likewise.
3709 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3710 arm_arch5e with arm_arch5t and arm_arch5te.
3711 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3712 (arm_arch5t, arm_arch5te): Declare.
3713 * config/arm/arm-tables.opt: Regenerate.
3714 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3715 * config/arm/t-multilib: Likewise.
3716 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3717 instead of arm_arch5.
3718 (*call_reg_thumb1): Likewise.
3719 (*call_value_reg_thumb1_v5): Likewise.
3720 (*call_value_reg_thumb1): Likewise.
3721 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3723 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3725 2018-05-18 Martin Liska <mliska@suse.cz>
3727 PR gcov-profile/84846
3728 * doc/gcov.texi: Document -t option of gcov tool.
3730 2018-05-18 Martin Liska <mliska@suse.cz>
3732 PR gcov-profile/84846
3733 * gcov.c (print_usage): Add new -t option.
3734 (process_args): Handle the option.
3735 (generate_results): Use stdout as output when requested by
3738 2018-05-18 Martin Liska <mliska@suse.cz>
3740 PR gcov-profile/84846
3741 * coverage.c (coverage_init): Write PWD to .gcno file.
3742 * doc/gcov.texi: Document how working directory is printed.
3743 * gcov-dump.c (dump_gcov_file): Print PWD.
3744 * gcov.c (output_intermediate_file): Likewise.
3745 (read_graph_file): Read PWD string.
3746 (output_lines): Print PWD.
3748 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3751 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3752 for retval and return false if all args to phi are zero.
3754 2018-05-18 Richard Biener <rguenther@suse.de>
3756 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3758 (evrp_dom_walker::before_dom_children): Call it.
3760 2018-05-18 Richard Biener <rguenther@suse.de>
3762 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3763 results when processing array refs with variable index.
3765 2018-05-18 Toon Moene <toon@moene.org>
3767 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3768 directly after that of -floop-interchange. Indicate that both
3769 options are enabled by default when specifying -O3.
3771 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3773 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3774 iterator. Delete separate integer-mode vec_set<mode> expander.
3775 (aarch64_simd_vec_setv2di): Delete.
3776 (vec_setv2di): Delete.
3777 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3778 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3779 the "w, r" alternative.
3781 2018-05-18 Martin Liska <mliska@suse.cz>
3783 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3784 * tree-pass.h (make_pass_lower_switch_O0): New function.
3785 * tree-switch-conversion.c (node_has_low_bound): Remove.
3786 (node_has_high_bound): Likewise.
3787 (node_is_bounded): Likewise.
3788 (class pass_lower_switch): Make it a template type and create
3790 (pass_lower_switch::execute): Add template argument.
3791 (make_pass_lower_switch): New function.
3792 (make_pass_lower_switch_O0): New function.
3793 (do_jump_if_equal): Remove.
3794 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3795 all the hard work to tree optimization passes.
3797 2018-05-18 Martin Liska <mliska@suse.cz>
3799 * dbgcnt.c (limit_low): Renamed from limit.
3800 (limit_high): New variable.
3801 (dbg_cnt_is_enabled): Check for upper limit.
3802 (dbg_cnt): Adjust dumping.
3803 (dbg_cnt_set_limit_by_index): Add new argument for high
3805 (dbg_cnt_set_limit_by_name): Likewise.
3806 (dbg_cnt_process_single_pair): Parse new format.
3807 (dbg_cnt_process_opt): Use strtok.
3808 (dbg_cnt_list_all_counters): Remove 'value' and add
3810 * doc/invoke.texi: Document changes.
3812 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3814 * doc/sourcebuild.texi (scalar_all_fma): Document.
3815 * tree.def (FMA_EXPR): Delete.
3816 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3817 * internal-fn.c (ternary_direct): New macro.
3818 (expand_ternary_optab_fn): Likewise.
3819 (direct_ternary_optab_supported_p): Likewise.
3820 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3821 * builtins.c (fold_builtin_fma): Delete.
3822 (fold_builtin_3): Don't call it.
3823 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3824 * expr.c (expand_expr_real_2): Likewise.
3825 * fold-const.c (operand_equal_p): Likewise.
3826 (fold_ternary_loc): Likewise.
3827 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3828 * gimple.c (DEFTREECODE): Likewise.
3829 * gimplify.c (gimplify_expr): Likewise.
3830 * optabs-tree.c (optab_for_tree_code): Likewise.
3831 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3832 * tree-eh.c (operation_could_trap_p): Likewise.
3833 (stmt_could_throw_1_p): Likewise.
3834 * tree-inline.c (estimate_operator_cost): Likewise.
3835 * tree-pretty-print.c (dump_generic_node): Likewise.
3836 (op_code_prio): Likewise.
3837 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3838 * tree-ssa-operands.c (get_expr_operands): Likewise.
3839 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3840 * fold-const-call.h (fold_fma): Delete.
3841 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3842 CFN_FNMA and CFN_FNMS.
3844 * genmatch.c (combined_fn): New enum.
3845 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3846 (commutative_op): New function.
3847 (commutate): Use it. Handle more than 2 operands.
3848 (dt_operand::gen_gimple_expr): Use commutative_op.
3849 (parser::parse_expr): Allow :c to be used with non-binary
3850 operators if the commutative operand is known.
3851 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3852 CFN_FMS, CFN_FNMA and CFN_FNMS.
3853 (backprop::process_assign_use): Remove FMA_EXPR handling.
3854 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3855 (gen_hsa_fma): New function.
3856 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3857 IFN_FNMA and IFN_FNMS.
3858 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3859 * gimple-fold.h (follow_all_ssa_edges): Declare.
3860 * gimple-fold.c (follow_all_ssa_edges): New function.
3861 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3862 gimple_build interface and use follow_all_ssa_edges to fold the result.
3863 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3864 instead of checking for optabs directly.
3865 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3866 rather than FMA_EXPRs.
3867 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3868 call to IFN_FMA instead of an FMA_EXPR.
3870 2018-05-17 Jim Wilson <jimw@sifive.com>
3872 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3873 sign extended promoted subreg, and the range does not have the sign bit
3874 set, then do a sign extend.
3876 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3877 test, check for sign extended subreg and/or constant operands, and
3878 do a sign extend in that case.
3880 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3882 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3883 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3885 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3886 Change logics_shift_reg to logics_shift_imm.
3887 (thunderx2t99_fp_loadpair_basic): Delete.
3888 (thunderx2t99_fp_storepair_basic): Delete.
3889 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3890 (thunderx2t99_asimd_polynomial): Delete.
3891 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3892 and neon_fp_mul_d_scalar_q.
3893 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3894 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3895 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3896 (thunderx2t99_asimd_lut): Add missing tbl types.
3897 (thunderx2t99_asimd_ext): Delete.
3898 (thunderx2t99_asimd_load1_1_mult): Delete.
3899 (thunderx2t99_asimd_load1_2_mult): Delete.
3900 (thunderx2t99_asimd_load1_ldp): New.
3901 (thunderx2t99_asimd_load1): New.
3902 (thunderx2t99_asimd_load2): Add missing *load2* types.
3903 (thunderx2t99_asimd_load3): New.
3904 (thunderx2t99_asimd_load4): New.
3905 (thunderx2t99_asimd_store1_1_mult): Delete.
3906 (thunderx2t99_asimd_store1_2_mult): Delete.
3907 (thunderx2t99_asimd_store2_mult): Delete.
3908 (thunderx2t99_asimd_store2_onelane): Delete.
3909 (thunderx2t99_asimd_store_stp): New.
3910 (thunderx2t99_asimd_store1): New.
3911 (thunderx2t99_asimd_store2): New.
3912 (thunderx2t99_asimd_store3): New.
3913 (thunderx2t99_asimd_store4): New.
3915 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3917 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3918 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3920 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3921 Segher Boessenkool <segher@kernel.crashing.org>
3924 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3927 2018-05-17 Richard Biener <rguenther@suse.de>
3929 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3930 for pruning loop and prune defs feeding only already visited PHIs.
3932 2018-05-17 Richard Biener <rguenther@suse.de>
3934 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3936 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3937 Richard Biener <rguenther@suse.de>
3939 PR tree-optimization/85793
3940 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3941 for VMAT_ELEMENTWISE.
3943 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3945 * internal-fn.h (lookup_internal_fn): Declare
3946 * internal-fn.c (lookup_internal_fn): New function.
3947 * gimple.c (gimple_build_call_from_tree): Handle calls to
3949 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3950 internal function names.
3951 * tree-pretty-print.c (dump_generic_node): Likewise.
3952 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3954 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3956 * gimple-fold.h (gimple_build): Make the function forms take
3957 combined_fn rather than built_in_function.
3958 (gimple_simplify): Likewise.
3959 * gimple-match-head.c (gimple_simplify): Likewise.
3960 * gimple-fold.c (gimple_build): Likewise.
3961 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3962 rather than gimple_build_call_internal.
3963 (get_initial_defs_for_reduction): Likewise.
3964 (vect_create_epilog_for_reduction): Likewise.
3965 (vectorizable_live_operation): Likewise.
3967 2018-05-17 Martin Liska <mliska@suse.cz>
3969 * gimple-ssa-sprintf.c (format_directive): Do not use
3970 space in between 'G_' and '('.
3972 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3975 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3976 even if the mask is not all ones.
3979 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3981 (ix86_gimple_fold_builtin): Likewise.
3984 * config/i386/i386.c: Include tree-vector-builder.h.
3985 (ix86_vector_shift_count): New function.
3986 (ix86_fold_builtin): Fold shift builtins by scalar count.
3987 (ix86_gimple_fold_builtin): Likewise.
3989 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3990 _mm512_setzero): New intrinsics.
3992 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3993 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3995 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3996 code generation for cases where splatting a value is not useful.
3997 * simplify-rtx.c (simplify_ternary_operation): Simplify
3998 vec_merge across a vec_duplicate and a paradoxical subreg forming
3999 a vector mode to a vec_concat.
4001 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4003 * config.gcc: Support "goldmont-plus".
4004 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4006 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4007 PROCESSOR_GOLDMONT_PLUS.
4008 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4009 (processor_target_table): Add "goldmont-plus".
4010 (PTA_GOLDMONT_PLUS): Define.
4011 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4012 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4013 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4014 (fold_builtin_cpu): Add "goldmont-plus".
4015 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4016 (ix86_option_override_internal): Add "goldmont-plus".
4017 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4018 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4019 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4020 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4022 2018-05-17 Richard Biener <rguenther@suse.de>
4024 PR tree-optimization/85757
4025 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4026 remove defs that only feed that PHI from further processing.
4028 2018-05-16 Jim Wilson <jimw@sifive.com>
4030 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4032 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4034 2018-05-16 Mark Wielaard <mark@klomp.org>
4036 * dwarf2out.c (count_index_strings): New function.
4037 (output_indirect_strings): Call count_index_strings and generate
4038 header for dwarf_version >= 5.
4040 2018-05-16 Mark Wielaard <mark@klomp.org>
4042 * dwarf2out.c (dwarf_FORM): New function.
4043 (set_indirect_string): Use dwarf_FORM.
4044 (reset_indirect_string): Likewise.
4045 (size_of_die): Likewise.
4046 (value_format): Likewise.
4047 (output_die): Likewise.
4048 (add_skeleton_AT_string): Likewise.
4049 (output_macinfo_op): Likewise.
4050 (index_string): Likewise.
4051 (output_index_string_offset): Likewise.
4052 (output_index_string): Likewise.
4053 (count_index_strings): Likewise.
4055 2018-05-16 Carl Love <cel@us.ibm.com>
4057 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4058 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4060 2018-05-16 Martin Jambor <mjambor@suse.cz>
4062 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4063 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4065 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4067 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4068 (fnma<mode>4): Likewise.
4069 (fms<mode>4): Likewise.
4070 (fnms<mode>4): Likewise.
4071 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4072 (aarch64_fnma<mode>4): Likewise.
4073 (aarch64_fms<mode>4): Likewise.
4074 (aarch64_fnms<mode>4): Likewise.
4075 (aarch64_fnmadd<mode>4): Likewise.
4077 2018-05-16 Jason Merrill <jason@redhat.com>
4079 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4081 2018-05-16 Richard Biener <rguenther@suse.de>
4083 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4084 (dump_stmt_cost): Declare.
4085 (add_stmt_cost): Dump cost we add.
4086 (add_stmt_costs): New function.
4087 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4089 (vect_analyze_stmt): Adjust prototype.
4090 (vectorizable_condition): Likewise.
4091 (vectorizable_live_operation): Likewise.
4092 (vectorizable_reduction): Likewise.
4093 (vectorizable_induction): Likewise.
4094 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4095 cost vector to pass to vectorizable_ and record afterwards.
4096 (vect_model_reduction_cost): Take cost vector argument and adjust.
4097 (vect_model_induction_cost): Likewise.
4098 (vectorizable_reduction): Likewise.
4099 (vectorizable_induction): Likewise.
4100 (vectorizable_live_operation): Likewise.
4101 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4102 SLP_TREE_NUMBER_OF_VEC_STMTS.
4103 (vect_analyze_slp_cost_1): Remove.
4104 (vect_analyze_slp_cost): Likewise.
4105 (vect_slp_analyze_node_operations): Take visited args and
4106 a target cost vector. Avoid processing already visited stmt sets.
4107 (vect_slp_analyze_operations): Use a local cost vector to gather
4108 costs and register those of non-discarded instances.
4109 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4110 (vect_schedule_slp_instance): Remove copying of
4111 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4113 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4114 adding cost. Record cost entry location.
4115 (vect_prologue_cost_for_slp_op): Function to compute cost of
4116 a constant or invariant generated for SLP vect in the prologue,
4117 split out from vect_analyze_slp_cost_1.
4118 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4119 (vect_model_promotion_demotion_cost): Likewise.
4120 (vect_model_store_cost): Likewise, make static.
4121 (vect_model_load_cost): Likewise.
4122 (vectorizable_bswap): Add cost vector arg and adjust.
4123 (vectorizable_call): Likewise.
4124 (vectorizable_simd_clone_call): Likewise.
4125 (vectorizable_conversion): Likewise.
4126 (vectorizable_assignment): Likewise.
4127 (vectorizable_shift): Likewise.
4128 (vectorizable_operation): Likewise.
4129 (vectorizable_store): Likewise.
4130 (vectorizable_load): Likewise.
4131 (vectorizable_condition): Likewise.
4132 (vectorizable_comparison): Likewise.
4133 (can_vectorize_live_stmts): Likewise.
4134 (vect_analyze_stmt): Likewise.
4135 (vect_transform_stmt): Adjust calls to vectorizable_*.
4136 * tree-vectorizer.c: Include gimple-pretty-print.h.
4137 (dump_stmt_cost): New function.
4139 2018-05-16 Richard Biener <rguenther@suse.de>
4141 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4142 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4143 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4144 (check_name): New callback.
4145 (dse_classify_store): Track cycles via a visited bitmap of PHI
4146 defs and simplify handling of in-loop and across loop dead stores
4147 and properly fail for loop-variant refs. Handle byte-tracking with
4148 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4151 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4153 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4154 (vect_get_mask_type_for_stmt): Likewise.
4155 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4157 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4158 to determine the statement's vector type and the vector type that
4159 should be used for calculating nunits. Deal with cases in which
4160 the type has to be deferred.
4161 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4162 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4163 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4164 (vect_determine_vf_for_stmt): New functions, split out from...
4165 (vect_determine_vectorization_factor): ...here.
4166 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4167 (vect_get_mask_type_for_stmt): New functions, split out from
4168 vect_determine_vectorization_factor.
4170 2018-05-16 Richard Biener <rguenther@suse.de>
4172 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4173 verify the [VEC_]COND_EXPR embedded comparison.
4175 2018-05-15 Martin Sebor <msebor@redhat.com>
4177 PR tree-optimization/85753
4178 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4179 RECORD_TYPE in addition to ARRAY_TYPE.
4181 2018-05-15 Martin Sebor <msebor@redhat.com>
4184 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4186 2018-05-15 Richard Biener <rguenther@suse.de>
4188 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4189 add by_clobber_p one. Change algorithm to collect all defs
4190 representing uses we need to walk and try reducing them to
4191 a single one before failing.
4192 (dse_dom_walker::dse_optimize_stmt): Adjust.
4194 2018-05-13 Mark Wielaard <mark@klomp.org>
4196 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4197 (size_of_loc_descr): Likewise.
4198 (output_loc_operands): Likewise.
4199 (output_loc_operands_raw): Likewise.
4200 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4201 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4202 (hash_loc_operands): Likewise.
4203 (compare_loc_operands): Likewise.
4205 2018-05-14 Mark Wielaard <mark@klomp.org>
4207 * dwarf2out.c (count_index_addrs): New function.
4208 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4210 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4212 PR tree-optimization/83648
4213 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4214 return value as malloc candidate.
4216 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4219 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4220 param as true in call to suggest_attribute.
4222 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4224 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4225 -mreadonly-in-sdata.
4227 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4229 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4231 (aarch64_crypto_aesd_fused): Likewise.
4233 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4235 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4236 (movsi_aarch64): Likewise.
4237 (load_pairsi): Likewise.
4238 (load_pairdi): Likewise.
4239 (store_pairsi): Likewise.
4240 (store_pairdi): Likewise.
4241 (load_pairsf): Likewise.
4242 (load_pairdf): Likewise.
4243 (store_pairsf): Likewise.
4244 (store_pairdf): Likewise.
4245 (zero_extend): Likewise.
4246 (trunc): Swap alternatives.
4247 (fcvt_target): Add '?' to prefer w over r.
4249 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4252 * config/i386/i386.md: Disallow non-commutative arithmetics in
4253 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4254 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4255 in the peephole2 before it.
4257 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4259 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4260 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4261 (ix86_handle_option): Handle -mcldemote.
4262 * config.gcc: New header.
4263 * config/i386/cldemoteintrin.h: New file.
4264 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4265 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4267 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4268 OPTION_MASK_ISA_CLDEMOTE.
4269 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4270 (ix86_valid_target_attribute_inner_p): Ditto.
4271 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4272 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4273 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4274 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4275 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4277 * config/i386/i386.opt: Add -mcldemote.
4278 * config/i386/x86intrin.h: New header.
4279 * doc/invoke.texi: Add -mcldemote.
4281 2018-05-14 Richard Biener <rguenther@suse.de>
4283 * doc/match-and-simplify.texi: Adjust :s documentation.
4285 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4287 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4288 intended memcpy size.
4289 (REORDER_45): Likewise.
4291 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4293 * sort.cc: New file.
4294 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4295 * vec.c (qsort_chk): Use gcc_qsort.
4296 * Makefile.in (OBJS-libcommon): Add sort.o.
4297 (build/sort.o): New target. Use it...
4298 (BUILD_RTL): ... here, and...
4299 (build/gencfn-macros): ... here, and...
4300 (build/genmatch): ... here.
4302 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4303 Chung-Ju Wu <jasonwucj@gmail.com>
4305 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4306 * config/nds32/nds32-graywolf.md: New file.
4307 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4308 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4310 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4311 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4312 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4313 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4314 * config/nds32/pipelines.md: Include n15 settings.
4316 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4317 Chung-Ju Wu <jasonwucj@gmail.com>
4319 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4320 * config/nds32/nds32-n13.md: New file.
4321 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4322 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4324 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4325 * config/nds32/nds32.md (pipeline_model): Add n13.
4326 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4327 * config/nds32/pipelines.md: Include n13 settings.
4329 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4330 Chung-Ju Wu <jasonwucj@gmail.com>
4332 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4333 * config/nds32/nds32-n10.md: New file.
4334 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4335 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4337 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4338 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4339 * config/nds32/nds32.md (pipeline_model): Add n10.
4340 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4341 * config/nds32/pipelines.md: Include n10 settings.
4343 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4344 Kito Cheng <kito.cheng@gmail.com>
4345 Chung-Ju Wu <jasonwucj@gmail.com>
4347 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4348 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4349 Add enum values for DSP extension instructions.
4350 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4352 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4353 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4355 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4356 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4357 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4358 * config/nds32/nds32-intrinsic.md: Likewise.
4359 * config/nds32/nds32_intrinsic.h: Likewise.
4360 * config/nds32/nds32-md-auxiliary.c: Likewise.
4361 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4362 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4363 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4364 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4365 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4366 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4367 TYPE_DMAC in switch statement.
4368 * config/nds32/nds32.c: New checking and implementation for DSP
4369 extension instructions.
4370 * config/nds32/nds32.h: Likewise.
4371 * config/nds32/nds32.md: Likewise.
4372 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4373 * config/nds32/predicates.md: Implement new predicates for DSP
4376 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4378 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4379 Reformat alternatives and attributes so it is easier to identify
4380 which constraints/attributes go with which instruction.
4381 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4382 (mov<mode>_softfloat32, FMOVE64): Likewise.
4383 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4384 (mov<mode>_softfloat64, FMOVE64): Likewise.
4386 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4388 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4390 (Basic PowerPC Built-in Functions): The new name of the
4391 subsection previously known as "PowerPC Built-in Functions".
4392 (Basic PowerPC Built-in Functions Available on all Configurations):
4394 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4395 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4396 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4397 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4399 2018-05-11 Martin Jambor <mjambor@suse.cz>
4402 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4405 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4408 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4410 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4412 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4413 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4414 (ix86_handle_option): Handle -mwaitpkg.
4415 * config.gcc: New header.
4416 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4417 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4418 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4420 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4421 OPTION_MASK_ISA_WAITPKG.
4422 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4423 (ix86_option_override_internal): Add PTA_WAITPKG.
4424 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4425 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4426 IX86_BUILTIN_TPAUSE.
4427 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4428 __builtin_ia32_umwait and __builtin_ia32_tpause.
4429 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4430 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4431 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4432 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4433 UNSPECV_TPAUSE): New.
4434 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4435 * config/i386/i386.opt: Add -mwaitpkg.
4436 * config/i386/waitpkgintrin.h: New file.
4437 * config/i386/x86intrin.h: New header.
4438 * doc/invoke.texi: Add -mwaitpkg.
4440 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4443 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4445 (cortex-m0): Use armv6s-m isa.
4446 (cortex-m0plus): Likewise.
4447 (cortex-m1): Likewise.
4448 (cortex-m0.small-multiply): Likewise.
4449 (cortex-m0plus.small-multiply): Likewise.
4450 (cortex-m1.small-multiply): Likewise.
4452 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4453 Jakub Jelinek <jakub@redhat.com>
4455 PR tree-optimization/85692
4456 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4457 source permute as well.
4459 2018-05-11 Martin Liska <mliska@suse.cz>
4462 * doc/extend.texi: Document LLVM style format for no_sanitize
4465 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4467 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4468 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4469 (mode_supports_vsx_dform_quad): Likewise.
4470 (mode_supports_vmx_dform): Move these functions to be next to the
4471 other mode_supports functions.
4472 (mode_supports_dq_form): Likewise.
4473 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4474 mode_supports_dq_form.
4475 (reg_offset_addressing_ok_p): Likewise.
4476 (offsettable_ok_by_alignment): Likewise.
4477 (rs6000_legitimate_offset_address_p): Likewise.
4478 (legitimate_lo_sum_address_p): Likewise.
4479 (rs6000_legitimize_address): Likewise.
4480 (rs6000_legitimize_reload_address): Likewise.
4481 (rs6000_secondary_reload_inner): Likewise.
4482 (rs6000_preferred_reload_class): Likewise.
4483 (rs6000_output_move_128bit): Likewise.
4485 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4487 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4488 Generate SImode target register for null target.
4489 <case IX86_BUILTIN_XGETBV>: Ditto.
4490 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4491 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4493 2018-05-10 Carl Love <cel@us.ibm.com>
4495 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4496 dcbtt and dcbtstt if operands[2] is 0.
4498 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4501 * config/i386/sse.md (usadv64qi): New expander.
4503 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4505 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4506 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4507 -maltivec=be support.
4508 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4509 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4510 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4511 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4512 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4513 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4514 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4515 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4516 altivec_vsumsws): Adjust.
4517 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4518 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4519 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4521 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4522 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4523 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4524 (altivec_lve<VI_char>x): Delete expand.
4525 (*altivec_lve<VI_char>x_internal): Rename to...
4526 (altivec_lve<VI_char>x): ... this.
4527 (altivec_lvxl_<mode>): Delete expand.
4528 (*altivec_lvxl_<mode>_internal): Rename to ...
4529 (altivec_lvxl_<mode>): ... this.
4530 (altivec_stvxl_<mode>): Delete expand.
4531 (*altivec_stvxl_<mode>_internal): Rename to ...
4532 (altivec_stvxl_<mode>): ... this.
4533 (altivec_stve<VI_char>x): Delete expand.
4534 (*altivec_stve<VI_char>x_internal): Rename to ...
4535 (altivec_stve<VI_char>x): ... this.
4536 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4537 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4538 reduc_plus_scal_<mode>): Adjust.
4539 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4541 (rs6000_cpu_cpp_builtins): Adjust.
4542 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4543 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4544 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4545 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4546 -maltivec=be support.
4547 (rs6000_split_vec_extract_var): Adjust.
4548 (rs6000_split_v4si_init): Adjust.
4549 (swap_selector_for_mode): Delete.
4550 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4551 altivec_expand_stvex_be): Delete.
4552 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4553 -maltivec=be support.
4554 (rs6000_gimple_fold_builtin): Ditto.
4555 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4557 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4558 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4559 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4560 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4561 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4562 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4563 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4564 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4565 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4566 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4567 anonymous split): Adjust.
4568 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4569 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4571 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4573 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4574 when --with-gxx-include-dir is also specified.
4575 * configure: Regenerate.
4577 2018-05-09 Jim Wilson <jimw@sifive.com>
4580 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4581 * config/riscv/t-withmultilib: New.
4582 * config/riscv/withmultilib.h: New.
4583 * doc/install.texi: Document RISC-V --with-multilib-list support.
4585 2018-05-09 Richard Biener <rguenther@suse.de>
4587 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4589 (vect_bb_vectorization_profitable_p): Adjust. Compute
4590 actual scalar cost using the cost vector and the add_stmt_cost
4593 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4595 PR rtl-optimization/85645
4596 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4597 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4599 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4601 PR rtl-optimization/85645
4602 * shrink-wrap.c (spread_components): Return a boolean saying if
4603 anything was changed.
4604 (try_shrink_wrapping_separate): Iterate spread_components until
4605 nothing changes anymore.
4607 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4609 PR rtl-optimization/85645
4610 * regrename.c (build_def_use): Also kill the chains that include the
4611 destination of a REG_CFA_REGISTER note.
4613 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4615 PR rtl-optimization/85645
4616 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4617 insn that has a REG_CFA_REGISTER note.
4619 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4621 * cfgexpand.c (expand_clobber): New function.
4622 (expand_gimple_stmt_1): Use it.
4623 * tree-vect-stmts.c (vect_clobber_variable): New function,
4625 (vectorizable_simd_clone_call): ...here.
4626 (vectorizable_store): Emit a clobber either side of an
4627 IFN_STORE_LANES sequence.
4628 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4630 2018-05-09 Tom de Vries <tom@codesourcery.com>
4633 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4634 (define_insn "trap_if_false"): Add exit after trap.
4636 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4638 PR rtl-optimization/85638
4639 * bb-reorder.c: Include common/common-target.h.
4640 (create_forwarder_block): New function extracted from...
4641 (fix_up_crossing_landing_pad): ...here. Rename into...
4642 (dw2_fix_up_crossing_landing_pad): ...this.
4643 (sjlj_fix_up_crossing_landing_pad): New function.
4644 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4645 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4646 from both partitions and exit the loop after one iteration.
4648 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4651 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4653 (Basic PowerPC Built-in Functions): The new name of the
4654 subsection previously known as "PowerPC Built-in Functions".
4655 (Basic PowerPC Built-in Functions Available on all Configurations):
4657 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4659 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4661 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4663 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4666 2018-05-08 Jim Wilson <jimw@sifive.com>
4668 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4669 (LD_EMUL_SUFFIX): New.
4670 (LINK_SPEC): Use it.
4672 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4674 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4676 (Basic PowerPC Built-in Functions): The new name of the
4677 subsection previously known as "PowerPC Built-in Functions".
4678 (Basic PowerPC Built-in Functions Available on all Configurations):
4680 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4682 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4684 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4686 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4689 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4692 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4693 after cmpelim optimization.
4695 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4697 * config.gcc: Support "goldmont".
4698 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4699 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4701 * config/i386/i386.c (m_GOLDMONT): Define.
4702 (processor_target_table): Add "goldmont".
4703 (PTA_GOLDMONT): Define.
4704 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4705 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4706 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4707 (fold_builtin_cpu): Add "goldmont".
4708 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4709 (ix86_option_override_internal): Add "goldmont".
4710 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4711 (processor_type): Add PROCESSOR_GOLDMONT.
4712 * config/i386/i386.md: Add CPU "glm".
4713 * config/i386/glm.md: New file.
4714 * config/i386/x86-tune.def: Add m_GOLDMONT.
4715 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4717 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4720 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4722 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4723 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4724 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4727 * config/i386/i386.c (ix86_fold_builtin): Handle
4728 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4731 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4732 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4734 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4737 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4738 (check_arch): Likewise.
4739 (check_fpu): Return the result rather than printing it.
4740 (end arch): Fix operator precedence.
4741 (end cpu): Likewise.
4742 (END): Print the result from check_fpu.
4744 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4745 Alan Hayward <alan.hayward@arm.com>
4746 David Sherwood <david.sherwood@arm.com>
4748 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4749 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4750 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4751 (*fcmuo<mode>_and): New patterns.
4753 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4755 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4756 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4757 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4758 (cmp_op, sve_imm_con): New code attributes.
4759 (SVE_COND_INT_CMP, imm_con): Delete.
4760 (cmp_op): Remove above unspecs from int attribute.
4761 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4763 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4764 comparison-specific unspecs.
4765 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4766 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4767 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4768 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4769 (*vec_fcm<cmp_op><mode>): Rename to...
4770 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4771 (*vec_fcmuo<mode>): Rename to...
4772 (*fcmuo<mode>): ...this and adjust likewise.
4773 (*pred_fcm<cmp_op><mode>): New pattern.
4774 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4775 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4777 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4779 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4780 (aarch64_emit_sve_predicated_cond): New function.
4781 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4782 (aarch64_emit_unspec_cond_or): Replace with...
4783 (aarch64_emit_sve_or_conds): ...this new function. Use
4784 aarch64_emit_sve_ptrue_op for the individual comparisons and
4785 aarch64_emit_binop to OR them together.
4786 (aarch64_emit_inverted_unspec_cond): Replace with...
4787 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4788 aarch64_emit_sve_ptrue_op for the comparison and
4789 aarch64_emit_unop to invert the result.
4790 (aarch64_expand_sve_vec_cmp_float): Update after the above
4791 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4793 2018-05-07 Nathan Sidwell <nathan@acm.org>
4795 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4796 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4797 (Backwards Compatibility): Likewise.
4799 2018-05-07 Luis Machado <luis.machado@linaro.org>
4803 2018-05-07 Luis Machado <luis.machado@linaro.org>
4805 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4806 <prefetch_dynamic_strides>: New const bool field.
4807 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4808 prefetch_dynamic_strides.
4809 (exynosm1_prefetch_tune): Likewise.
4810 (thunderxt88_prefetch_tune): Likewise.
4811 (thunderx_prefetch_tune): Likewise.
4812 (thunderx2t99_prefetch_tune): Likewise.
4813 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4815 (aarch64_override_options_internal): Update to set
4816 PARAM_PREFETCH_DYNAMIC_STRIDES.
4817 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4818 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4819 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4820 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4821 prefetch-dynamic-strides setting.
4823 2018-05-07 Luis Machado <luis.machado@linaro.org>
4825 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4826 <minimum_stride>: New const int field.
4827 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4828 minimum_stride field.
4829 (exynosm1_prefetch_tune): Likewise.
4830 (thunderxt88_prefetch_tune): Likewise.
4831 (thunderx_prefetch_tune): Likewise.
4832 (thunderx2t99_prefetch_tune): Likewise.
4833 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4834 (aarch64_override_options_internal): Update to set
4835 PARAM_PREFETCH_MINIMUM_STRIDE.
4836 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4837 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4838 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4839 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4840 stride is constant and is below the minimum stride threshold.
4842 2018-05-07 Luis Machado <luis.machado@linaro.org>
4844 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4847 2018-05-07 Luis Machado <luis.machado@linaro.org>
4849 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4850 <prefetch_dynamic_strides>: New const bool field.
4851 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4852 prefetch_dynamic_strides.
4853 (exynosm1_prefetch_tune): Likewise.
4854 (thunderxt88_prefetch_tune): Likewise.
4855 (thunderx_prefetch_tune): Likewise.
4856 (thunderx2t99_prefetch_tune): Likewise.
4857 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4859 (aarch64_override_options_internal): Update to set
4860 PARAM_PREFETCH_DYNAMIC_STRIDES.
4861 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4862 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4863 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4864 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4865 prefetch-dynamic-strides setting.
4867 2018-05-07 Luis Machado <luis.machado@linaro.org>
4869 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4870 <minimum_stride>: New const int field.
4871 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4872 minimum_stride field.
4873 (exynosm1_prefetch_tune): Likewise.
4874 (thunderxt88_prefetch_tune): Likewise.
4875 (thunderx_prefetch_tune): Likewise.
4876 (thunderx2t99_prefetch_tune): Likewise.
4877 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4878 (aarch64_override_options_internal): Update to set
4879 PARAM_PREFETCH_MINIMUM_STRIDE.
4880 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4881 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4882 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4883 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4884 stride is constant and is below the minimum stride threshold.
4886 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4889 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4890 the type is addressable. Don't force op into register if it has
4893 2018-05-05 Roland McGrath <mcgrathr@google.com>
4896 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4897 any section for which we don't know a specific type it should have,
4898 regardless of name. Previously this was done only for the exact
4899 names ".init_array", ".fini_array", and ".preinit_array".
4900 (default_elf_asm_named_section): Add comment about
4901 relationship with default_section_type_flags and SECTION_NOTYPE.
4902 (get_section): Don't consider it a type conflict if one side has
4903 SECTION_NOTYPE and the other doesn't, as long as neither has the
4904 SECTION_BSS et al used in the default_section_type_flags logic.
4906 2018-05-05 Tom de Vries <tom@codesourcery.com>
4909 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4910 (workaround_barsyncs): New function.
4911 (nvptx_reorg): Use workaround_barsyncs.
4912 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4913 (define_expand "nvptx_membar_cta"): New define_expand.
4914 (define_insn "*nvptx_membar_cta"): New insn.
4916 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4918 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4919 To improve optimization opportunities.
4920 * builtin-types.def: The new needed builtin types for the above.
4922 2018-05-04 Richard Biener <rguenther@suse.de>
4924 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4925 * gimple-ssa-store-merging.c
4926 (imm_store_chain_info::output_merged_store): Remove redundant create,
4927 release split_store vector contents on failure.
4928 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4929 scalar stmt vector on cache hit.
4931 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4933 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4935 * config.gcc (powerpc-xilinx-eabi*): Remove.
4936 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4938 (fusion_addis_mem_combo_load): Ditto.
4939 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4941 (rs6000_cpu_cpp_builtins): Ditto.
4942 * config/rs6000/rs6000-linux.c
4943 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4944 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4945 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4947 (rs6000_setup_reg_addr_masks): Ditto.
4948 (rs6000_init_hard_regno_mode_ok): Ditto.
4949 (rs6000_option_override_internal): Ditto.
4950 (legitimate_lo_sum_address_p): Ditto.
4951 (rs6000_legitimize_address): Ditto.
4952 (rs6000_legitimize_reload_address): Ditto.
4953 (rs6000_legitimate_address_p): Ditto.
4954 (abi_v4_pass_in_fpr): Ditto.
4955 (setup_incoming_varargs): Ditto.
4956 (rs6000_gimplify_va_arg): Ditto.
4957 (rs6000_split_multireg_move): Ditto.
4958 (rs6000_savres_strategy): Ditto.
4959 (rs6000_emit_prologue_components): Ditto.
4960 (rs6000_emit_epilogue_components): Ditto.
4961 (rs6000_emit_prologue): Ditto.
4962 (rs6000_emit_epilogue): Ditto.
4963 (rs6000_elf_file_end): Ditto.
4964 (rs6000_function_value): Ditto.
4965 (rs6000_libcall_value): Ditto.
4966 * config/rs6000/rs6000.h: Ditto.
4967 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4968 (TARGET_MINMAX): ... this. New.
4969 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4970 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4971 (*movsi_internal1_single): Delete.
4972 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4973 mfpu=, mxilinx-fpu): Delete.
4974 * config/rs6000/singlefp.h: Delete.
4975 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4976 * config/rs6000/t-rs6000: Ditto.
4977 * config/rs6000/t-xilinx: Delete.
4978 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4979 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4980 (VStype_simple): Delete.
4981 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4982 * config/rs6000/xfpu.h: Delete.
4983 * config/rs6000/xfpu.md: Delete.
4984 * config/rs6000/xilinx.h: Delete.
4985 * config/rs6000/xilinx.opt: Delete.
4986 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4987 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4989 2018-05-04 Tom de Vries <tom@codesourcery.com>
4992 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4995 2018-05-04 Richard Biener <rguenther@suse.de>
4998 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4999 (expand_complex_div_wide): Likewise.
5000 (expand_complex_operations_1): Likewise.
5001 (expand_complex_libcall): Preserve EH info of the original stmt.
5002 (tree_lower_complex): Handle removed blocks.
5003 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5004 on complex multiplication and division libcall builtins.
5006 2018-05-04 Richard Biener <rguenther@suse.de>
5009 * fold-const.c (negate_expr_p): Restrict negation of operand
5010 zero of a division to when we know that can happen without
5012 (fold_negate_expr_1): Likewise.
5014 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5017 * real.h (real_nextafter): Declare.
5018 * real.c (real_nextafter): New function.
5019 * fold-const-call.c (fold_const_nextafter): New function.
5020 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5021 CASE_CFN_NEXTTOWARD.
5022 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5023 even when arg1_mode is different from arg0_mode.
5025 2018-05-03 Nathan Sidwell <nathan@acm.org>
5027 * doc/extend.texi (Deprecated Features): Remove
5029 (Backwards Compatibility): Likewise.
5030 * doc/invoke.texi (C++ Language Options): Likewise.
5031 (C++ Dialect Options): Likewise.
5033 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5036 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5037 _mm512_mask_mullox_epi64): New intrinsics.
5039 2018-05-03 Tom de Vries <tom@codesourcery.com>
5042 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5043 dump files): Add offload-tree.
5045 2018-05-03 Richard Biener <rguenther@suse.de>
5047 PR tree-optimization/85615
5048 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5049 to loops not nested in BBs loop father to avoid creating multi-entry
5052 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5054 PR tree-optimization/70291
5055 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5056 arguments. Change return type to tree. Emit libcall as a new
5057 statement rather than replacing existing one when inplace_p is true.
5058 (expand_complex_multiplication_components): New function.
5059 (expand_complex_multiplication): Expand floating-point complex
5060 multiplication using the above.
5061 (expand_complex_division): Rename inner_type parameter to type.
5062 Update expand_complex_libcall call-site.
5063 (expand_complex_operations_1): Update expand_complex_multiplication
5064 and expand_complex_division call-sites.
5066 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5069 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5070 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5071 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5072 the highest significant bit of the shift count mask is clear. In
5073 check whether and[sq]i3 is needed verify that all significant bits
5074 of the shift count other than the highest are set.
5076 2018-05-02 Tom de Vries <tom@codesourcery.com>
5079 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5080 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5081 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5082 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5083 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5084 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5085 __builtin_goacc_parlevel_size.
5087 2018-05-02 Richard Biener <rguenther@suse.de>
5089 PR tree-optimization/85597
5090 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5091 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5094 2018-05-02 Tom de Vries <tom@codesourcery.com>
5097 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5098 dump files): Add ltrans-tree.
5100 2018-05-02 Tom de Vries <tom@codesourcery.com>
5103 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5104 dump files): Add wpa-ipa.
5106 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5108 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5109 powerpc*-*-linux*paired* target.
5110 * config/rs6000/750cl.h: Delete.
5111 * config/rs6000/paired.h: Delete.
5112 * config/rs6000/paired.md: Delete.
5113 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5115 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5116 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5117 comment. Remove paired float support.
5118 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5119 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5121 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5122 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5124 * config/rs6000/rs6000.c: Remove paired float support.
5125 (paired_expand_vector_init, paired_expand_vector_move,
5126 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5127 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5128 paired_expand_builtin, paired_expand_predicate_builtin,
5129 paired_init_builtins): Delete.
5130 * config/rs6000/rs6000.h: Remove paired float support.
5131 * config/rs6000/rs6000.md: Remove paired float support.
5132 (move_from_CR_ov_bit): Delete.
5133 * config/rs6000/rs6000.opt (mpaired): Delete.
5134 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5135 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5137 2018-05-02 Richard Biener <rguenther@suse.de>
5140 * gimplify.c (gimplify_save_expr): When in SSA form allow
5141 SAVE_EXPRs to compute to SSA vars.
5143 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5146 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5147 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5148 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5149 clobber operands[2], instead use a new pseudo. Formatting fixes.
5151 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5153 PR tree-optimization/85586
5154 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5155 exit early for statements in the same group if the accesses are
5158 2018-05-02 Tom de Vries <tom@codesourcery.com>
5161 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5164 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5166 PR tree-optimization/85143
5167 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5169 2018-05-01 Tom de Vries <tom@codesourcery.com>
5172 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5173 not found" error message.
5175 2018-05-01 Tom de Vries <tom@codesourcery.com>
5178 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5179 * vec.c (test_ordered_remove_if): New function.
5180 (vec_c_tests): Call test_ordered_remove_if.
5181 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5182 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5183 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5184 VEC_ORDERED_REMOVE_IF.
5186 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5188 PR tree-optimization/82665
5189 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5190 pointer subtraction where arguments come from a memchr call.
5192 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5194 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5195 --push-state --as-needed and --pop-state instead of --as-needed and
5196 --no-as-needed if ld supports it.
5197 * configure: Regenerated.
5200 * doc/install.texi2html: Replace _002d with - and _002a with * in
5201 generated html files using sed.
5203 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5206 * gcc-rich-location.c (blank_line_before_p): New function.
5207 (use_new_line): New function.
5208 (gcc_rich_location::add_fixit_insert_formatted): New function.
5209 * gcc-rich-location.h
5210 (gcc_rich_location::add_fixit_insert_formatted): New function.
5212 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5214 * selftest.c (assert_streq): Rename "expected" and "actual" to
5215 "val1" and "val2". Extend NULL-handling to cover both inputs
5216 symmetrically, while still requiring both to be non-NULL for a pass.
5217 * selftest.h (assert_streq): Rename "expected" and "actual" to
5219 (ASSERT_EQ): Likewise.
5220 (ASSERT_EQ_AT): Likewise.
5221 (ASSERT_KNOWN_EQ): Likewise.
5222 (ASSERT_KNOWN_EQ_AT): Likewise.
5223 (ASSERT_NE): Likewise.
5224 (ASSERT_MAYBE_NE): Likewise.
5225 (ASSERT_MAYBE_NE_AT): Likewise.
5226 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5227 the assertion to pass.
5228 (ASSERT_STREQ_AT): Likewise.
5230 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5232 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5233 interaction with -pie.
5235 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5237 * selftest.h: Fix alphabetization of per-source-file selftest
5240 2018-04-30 Jason Merrill <jason@redhat.com>
5242 PR c++/61982 - dead stores to destroyed objects.
5243 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5246 2018-04-30 Jason Merrill <jason@redhat.com>
5248 * tree.c (build_clobber): New.
5249 * tree.h: Declare it.
5250 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5252 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5254 * diagnostic-show-locus.c (layout::layout): Update for
5255 location_get_source_line returning a char_span.
5256 (struct char_span): Move to input.h.
5257 (struct correction): Update for fields in char_span becoming
5259 (struct source_line): Update for location_get_source_line
5260 returning a char_span.
5261 (layout::print_line): Likewise.
5262 * edit-context.c (edited_file::print_content): Likewise.
5263 (edited_file::print_diff_hunk): Likewise.
5264 (edited_file::print_run_of_changed_lines): Likewise.
5265 (edited_file::get_num_lines): Likewise.
5266 (edited_line::edited_line): Likewise.
5267 * final.c (asm_show_source): Likewise.
5268 * input.c (location_get_source_line): Convert return type
5269 from const char * to char_span, losing the final "line_len"
5271 (dump_location_info): Update for the above.
5272 (get_substring_ranges_for_loc): Likewise. Use a char_span
5273 when handling the literal within the line.
5274 (test_reading_source_line): Update for location_get_source_line
5275 returning a char_span.
5276 * input.h (class char_span): Move here from
5277 diagnostic-show-locus.c, converting from a struct to a class.
5278 Make data members private.
5279 (char_span::operator bool): New.
5280 (char_span::length): New.
5281 (char_span::get_buffer): New.
5282 (char_span::operator[]): New.
5283 (char_span::subspan): Make const.
5284 (char_span::xstrdup): New.
5285 (location_get_source_line): Convert return type from const char *
5286 to char_span, losing the final "line_size" param.
5288 2018-04-30 Jan Hubicka <jh@suse.cz>
5290 * lto-wrapper.c (ltrans_priorities): New static var.
5291 (cmp_priority): New.
5292 (run_gcc): Read priorities and if doing parallel build order
5293 the Makefile by them.
5295 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5297 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5299 2018-04-30 Richard Biener <rguenther@suse.de>
5301 * tree-cfg.c (verify_address): Remove base argument, add
5302 flag whether to check TREE_ADDRESSABLE and do that.
5303 (verify_expr): Remove.
5304 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5305 (verify_gimple_assign_single): Likewise.
5306 (verify_gimple_switch): Likewise.
5307 (verify_expr_location_1): Dereference tp once. Add (disabled)
5308 piece from verify_expr.
5309 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5311 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5313 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5315 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5317 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5318 (small_data_pattern): Likewise.
5319 (arc_rewrite_small_data): Likewise.
5320 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5321 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5322 (get_symbol_alignment): New function.
5323 (legitimate_small_data_address_p): Likewise.
5324 (legitimate_scaled_address): Update, call
5325 legitimate_small_data_address_p.
5326 (output_sdata): New static variable.
5327 (arc_print_operand): Update how we handle small data operands.
5328 (arc_print_operand_address): Likewise.
5329 (arc_legitimate_address_p): Update, use
5330 legitimate_small_data_address_p.
5331 (arc_rewrite_small_data_p): Remove.
5332 (arc_rewrite_small_data_1): Likewise.
5333 (arc_rewrite_small_data): Likewise.
5334 (small_data_pattern): Likewise.
5335 (compact_sda_memory_operand): Update to use
5336 legitimate_small_data_address_p and get_symbol_alignment.
5337 (prepare_move_operands): Don't rewite sdata pattern.
5338 (prepare_extend_operands): Remove.
5339 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5341 (zero_extendqisi2): Likewise.
5342 (zero_extendhisi2): Likewise.
5343 (extendqihi2): Likewise.
5344 (extendqisi2): Likewise.
5345 (extendhisi2): Likewise.
5349 * config/arc/constraints.md (Usd): Change it to memory constraint.
5351 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5353 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5354 as source of std instructions.
5355 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5356 allow 6-bit constants as source for store instructions.
5357 (movdi_insn): Update instruction pattern to allow 6-bit constants
5358 as source for store instructions.
5360 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5362 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5364 2018-04-30 Nathan Sidwell <nathan@acm.org>
5365 Sandra Loosemore <sandra@codesourcery.com>
5367 * dumpfile.c (dump_open): Allow '-' for stdout.
5368 * doc/invoke.texi (Developer Options): Document dump filename
5369 determination early. Document stdin/stdout selection.
5371 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5373 Microblaze Target: PIC data text relative
5375 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5376 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5378 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5379 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5380 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5381 New addressing mode for data-text relative position indepenedent code.
5382 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5383 'ADDRESS_SYMBOLIC_TXT_REL'.
5384 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5385 (microblaze_legitimate_pic_operand): Exclude function calls from
5386 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5387 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5389 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5390 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5391 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5392 for 'address + offset'.
5393 (microblaze_expand_prologue): Add new function prologue call for
5395 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5396 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5397 table in case of TARGET_PIC_DATA_TEXT_REL.
5398 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5399 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5400 Add new macros 'UNSPEC_TEXT',
5401 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5402 + exclude function calls from 'UNSPEC_PLT' in case of data text
5404 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5405 new target hook for generating address diff vector tables in case of
5407 * doc/tm.texi : Regenerate.
5408 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5409 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5410 of addr diff vector generation.
5411 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5412 target hook definition.
5413 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5414 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5415 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5416 option for data text relative.
5418 2018-04-30 Richard Biener <rguenther@suse.de>
5420 * tree-chrec.h (evolution_function_is_constant_p): Remove
5422 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5424 2018-04-30 Richard Biener <rguenther@suse.de>
5427 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5429 2018-04-30 Richard Biener <rguenther@suse.de>
5431 PR tree-optimization/28364
5432 PR tree-optimization/85275
5433 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5434 copying first exit test.
5436 2018-04-28 Mark Wielaard <mark@klomp.org>
5438 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5440 (dwarf_AT): Handle DW_AT_addr_base.
5441 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5443 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5446 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5447 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5448 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5449 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5451 2018-04-28 Richard Biener <rguenther@suse.de>
5453 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5454 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5455 to reflect use. Only add interesting stmts.
5457 2018-04-27 Martin Jambor <mjambor@suse.cz>
5460 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5461 the jump function allows for passing through aggregate values.
5463 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5465 * input.h (in_system_header_at): Convert from macro to inline
5467 (from_macro_expansion_at): Likewise.
5468 (from_macro_definition_at): Likewise.
5470 2018-04-27 Jeff Law <law@redhat.com>
5472 * config.gcc: Mark tile* targets as deprecated/obsolete.
5474 2018-04-27 Richard Biener <rguenther@suse.de>
5476 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5479 2018-04-27 Richard Biener <rguenther@suse.de>
5481 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5483 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5485 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5486 with Yd constraint. Set "preferred_for_speed" attribute from
5487 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5489 (*movdi_internal): Ditto.
5490 (movti_interunit splitters): Remove
5491 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5492 (movdi_interunit splitters): Ditto.
5493 * config/i386/constraints.md (Ye): Remove.
5494 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5496 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5499 * config/aarch64/constraints.md (Usg): Limit to 31.
5502 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5504 PR tree-optimization/85529
5505 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5506 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5507 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5508 zero extension or masking of the MSB bit.
5509 (optimize_range_tests): Add FIRST_BB argument, pass it through
5510 to optimize_range_tests_var_bound.
5511 (maybe_optimize_range_tests, reassociate_bb): Adjust
5512 optimize_range_tests callers.
5514 2018-04-26 Richard Biener <rguenther@suse.de>
5515 Jakub Jelinek <jakub@redhat.com>
5517 * cgraph.h (symbol_table): Just declare debug method here.
5518 * symtab.c (symbol_table::debug): Define.
5520 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5522 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5524 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5526 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5527 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5528 (*movdi_internal): Substitute Yi and Yj constraint with x
5529 and Ym and Yn constraint with y constraint. Update "isa"
5530 attribute and set "preferred_for_speed" attribute from
5531 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5532 (*movsi_internal): Ditto.
5533 (*movdf_internal): Ditto.
5534 (*movsf_internal): Ditto.
5535 (*zero_extendsidi2): Ditto.
5536 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5537 (sse2_loadld): Ditto.
5538 (*vec_extract<ssevecmodelower>_0): Ditto.
5539 (*vec_extractv4si_0_zext_sse4): Ditto.
5540 (vec_concatv2di): Ditto.
5541 (*vec_dup<mode>): Ditto.
5542 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5543 * config/i386/constraints.md (Yi): Remove.
5548 2018-04-26 Nathan Sidwell <nathan@acm.org>
5550 * dumpfile.c (dump_open): New.
5551 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5552 (dump_finish): Detect stdio/stderr by value not name.
5554 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5556 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5558 2018-04-26 Tom de Vries <tom@codesourcery.com>
5561 * config/nvptx/nvptx.c (verify_neutering_jumps)
5562 (verify_neutering_labels): New function
5563 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5565 2018-04-26 Tom de Vries <tom@codesourcery.com>
5568 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5569 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5572 2018-04-26 Richard Biener <rguenther@suse.de>
5573 Tom de Vries <tom@codesourcery.com>
5576 * lto-streamer-out.c (output_function): Fixup loops if required to match
5577 discovery done in the reader.
5579 2018-04-26 Richard Biener <rguenther@suse.de>
5581 PR tree-optimization/85116
5582 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5583 have a loop exit from the single latch predecessor. Remove
5584 case of header with just condition.
5585 (ch_base::copy_headers): Exclude infinite loops from any
5587 (pass_ch::execute): Record exits.
5589 2018-04-26 Richard Biener <rguenther@suse.de>
5591 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5592 prologue cost vector and pass it to vect_get_load_cost.
5593 (vect_get_peeling_costs_all_drs): Likewise.
5594 (vect_peeling_hash_get_lowest_cost): Likewise.
5595 (vect_enhance_data_refs_alignment): Likewise.
5597 2018-04-26 Richard Biener <rguenther@suse.de>
5600 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5601 checking of integer<->pointer conversions.
5602 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5603 sign-/zero-extending pointer types.
5604 (expand_omp_for_static_chunk): Likewise.
5606 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5607 Jean Lee <xiaoyur347@gmail.com>
5609 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5610 (TARGET_ASAN_SHADOW_OFFSET): Define.
5611 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5612 true for -fsanitize=address.
5614 2018-04-25 Mark Wielaard <mark@klomp.org>
5616 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5619 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5621 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5622 than "alu", remove explicit "memory" and "imm_disp" attributes.
5623 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5626 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5627 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5628 gen_lowpart_no_emit.
5630 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5633 * config/i386/i386.c (ix86_expand_builtin): Change memory
5634 operand to XI, extend p0 to Pmode.
5635 * config/i386/i386.md: Change unspec volatile and operand
5636 1 mode to XI, change operand 0 mode to P.
5638 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5640 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5641 GET_MODE_MASK before any checking.
5642 (nds32_can_use_bset_p): Likewise.
5643 (nds32_can_use_btgl_p): Likewise.
5645 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5647 * config/nds32/nds32-doubleword.md: New define_split pattern for
5648 illegal register number.
5650 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5652 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5654 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5656 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5658 2018-04-25 Richard Biener <rguenther@suse.de>
5660 * lto-streamer.h (LTO_major_version): Bump to 8.
5662 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5664 * BASE-VER: Set to 9.0.0.
5666 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5668 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5669 in __abskf2 and __powikf2.
5671 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5674 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5675 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5676 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5677 Use the above on operand 2. Reindent.
5678 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5680 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5683 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5685 * config/i386/i386.opt (mcet): Removed.
5686 * doc/install.texi: Remove -mcet documentation.
5687 * doc/invoke.texi: Likewise.
5689 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5692 * doc/install.texi: Remove -mcet from bootstrap-cet.
5694 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5697 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5698 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5702 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5703 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5704 containing a CONST_VECTOR.
5706 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5708 * doc/install.texi: Update newlib dependency for nvptx.
5710 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5713 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5714 instead of INTVAL when shifting x left.
5716 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5718 PR tree-optimization/85478
5719 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5720 vect_grouped_store_supported for single element vectors.
5722 2018-04-24 Richard Biener <rguenther@suse.de>
5725 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5726 load cost increase to the case of non-constant step.
5728 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5731 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5732 destination if any_malformed_asm.
5734 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5737 * expr.c (store_field): In the bitfield case, if the value comes from
5738 a function call and is returned in registers by means of a PARALLEL,
5739 do not change the mode of the temporary unless BLKmode and VOIDmode.
5741 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5743 PR rtl-optimization/85423
5744 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5745 dependencies to debug insns when the previous insn is non-debug.
5747 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5749 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5750 enums into a single definition.
5751 (fls): Fix predicates and printing.
5754 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5756 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5757 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5758 and short u6 immediate.
5759 (check_if_valid_sleep_operand): Remove.
5760 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5762 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5764 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5765 flag_always_save_lp condition.
5766 * config/nds32/nds32.opt (malways-save-lp): New option.
5768 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5770 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5771 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5772 * config/nds32/nds32.h
5773 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5774 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5776 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5778 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5779 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5781 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5782 Chung-Ju Wu <jasonwucj@gmail.com>
5784 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5786 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5787 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5789 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5791 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5793 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5795 * config/nds32/nds32-protos.h (nds32_data_alignment,
5796 nds32_local_alignment): Declare.
5797 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5798 nds32_local_alignment): New functions.
5799 (TARGET_CONSTANT_ALIGNMENT): Define.
5800 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5802 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5804 * config/nds32/nds32.c
5805 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5806 (TARGET_MODES_TIEABLE_P): Likewise.
5808 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5810 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5813 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5814 Chung-Ju Wu <jasonwucj@gmail.com>
5816 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5817 for unaligned access.
5818 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5820 * config/nds32/nds32-intrinsic.md: Likewise.
5821 * config/nds32/nds32_intrinsic.h: Likewise.
5822 * config/nds32/nds32.h (nds32_builtins): Likewise.
5823 * config/nds32/nds32.opt (munaligned-access): New option.
5824 * config/nds32/nds32.c (nds32_asm_file_start): Display
5825 flag_unaligned_access status.
5827 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5829 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5830 -mno-relax is present.
5831 * config/riscv/linux.h (LINK_SPEC): Ditto.
5833 2018-04-20 Martin Sebor <msebor@redhat.com>
5836 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5838 (gimple_fold_builtin_stxcpy_chk): Same.
5839 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5841 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5844 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5845 __powikf2 when long double is IEEE 128-bit.
5847 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5849 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5850 step to make sure stack always aligned.
5852 2018-04-20 Carl Love <cel@us.ibm.com>
5855 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5856 size check for arg0.
5858 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5859 Tom de Vries <tom@codesourcery.com>
5862 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5863 Emit insns for calls too.
5864 (nvptx_find_par): Always look for worker-level predecessor insn.
5865 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5867 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5868 (nvptx_process_pars): Propagate frames for calls.
5870 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5873 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5875 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5876 (ix86_handle_option): Don't handle OPT_mibt.
5877 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5879 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5881 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5882 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5883 (ix86_target_macros): Define __CET__ with flag_cf_protection
5884 for -fcf-protection.
5885 * config/i386/i386.c (isa2_opts): Remove -mibt.
5886 * config/i386/i386.h (TARGET_IBT): Removed.
5887 (TARGET_IBT_P): Likewise.
5888 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5889 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5890 * config/i386/i386.opt (mcet): Update help message.
5893 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5894 -mcet as an alias for -mshstk.
5896 2018-04-20 Richard Biener <rguenther@suse.de>
5899 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5900 complexity by forcing a single use of the multiply operand.
5902 2018-04-20 Martin Jambor <mjambor@suse.cz>
5905 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5906 recursion dependency to only apply to non-clones.
5908 2018-04-20 Martin Jambor <mjambor@suse.cz>
5911 * ipa-cp.c (create_specialized_node): Check that clones of
5912 self-recursive edges exist during IPA-CP.
5914 2018-04-19 Toon Moene <toon@moene.org>
5916 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5919 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5921 PR tree-optimization/85467
5922 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5923 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5924 VECTOR_CST element to type.
5926 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5929 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5930 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5931 (builtin_longjmp): Likewise.
5932 (save_stack_nonlocal): New pattern.
5933 (restore_stack_nonlocal): Likewise.
5935 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5938 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5939 Replace ASM_OUTPUT_LABEL with fprintf.
5941 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5944 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5945 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5946 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5947 define __IBT__ and __SHSTK__ for -fcf-protection.
5948 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5950 (ix86_trampoline_init): Likewise.
5951 (x86_output_mi_thunk): Likewise.
5952 (ix86_notrack_prefixed_insn_p): Likewise.
5953 (ix86_option_override_internal): Don't disallow -fcf-protection.
5954 * config/i386/i386.md (rdssp<mode>): Also enable for
5956 (incssp<mode>): Likewise.
5957 (nop_endbr): Likewise.
5958 * config/i386/i386.opt (mcet): Change help message to built-in
5962 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5963 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5964 enable CET built-in functions.
5966 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5968 * common/config/i386/i386-common.c
5969 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5970 OPTION_MASK_ISA_MOVDIRI_UNSET,
5971 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5972 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5973 * config.gcc (movdirintrin.h): New header.
5974 * config/i386/cpuid.h (bit_MOVDIRI,
5975 bit_MOVDIR64B): New bits.
5976 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5978 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5979 (VOID, PVOID, PCVOID)): New function types.
5980 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5981 __builtin_ia32_directstoreu_u64,
5982 __builtin_ia32_movdir64b): New builtins.
5983 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5984 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5986 (ix86_valid_target_attribute_inner_p): Ditto.
5987 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5988 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5989 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5990 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5991 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5992 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5993 (movdiri<mode>, movdir64b_<mode>): New.
5994 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5995 * config/i386/immintrin.h: Include movdirintrin.h.
5996 * config/i386/movdirintrin.h: New file.
5997 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5999 2018-04-19 Richard Biener <rguenther@suse.de>
6002 * cfg.c (clear_bb_flags): When loop state says we have
6003 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6005 2018-04-19 Richard Biener <rguenther@suse.de>
6007 PR tree-optimization/84737
6008 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6009 copying restrict info.
6010 (vect_setup_realignment): Use it.
6011 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6012 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6013 the first DR to all generated stores.
6014 (vectorizable_load): Likewise for loads.
6016 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6018 PR tree-optimization/85446
6019 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6020 the integral and pointer types to have the same precision.
6022 * doc/install.texi: Document --disable-cet being the default and
6025 2018-04-18 Martin Liska <mliska@suse.cz>
6027 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6030 2018-04-18 Martin Liska <mliska@suse.cz>
6033 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6036 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6037 arguments if they are comparable.
6039 2018-04-18 Martin Liska <mliska@suse.cz>
6042 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6045 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6048 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6051 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6052 ENDBR after calling __morestack.
6054 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6057 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6058 by using gcc_base_ver to generate a gcc_driver_version, and use
6059 it when generating GCC_DRIVER_NAME.
6060 * configure: Regenerate.
6062 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6065 * config.gcc: Obsolete powerpc*-*-*spe*.
6067 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6070 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6071 (stabstr_D): Change type of unum from unsigned int to
6072 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6075 2018-04-17 Jim Wilson <jimw@sifive.com>
6078 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6079 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6080 Set arg_pointer_offset after using pretend_args_size.
6082 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6084 PR rtl-optimization/85431
6085 * dse.c (record_store): Ignore zero width stores.
6088 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6089 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6090 __builtin_stack_restore rather than after it.
6091 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6092 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6093 argument instead of virtual_dynamic_stack_rtx.
6095 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6097 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6099 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6100 Add note to error message to explain internal mapping of overloaded
6101 built-in function name to non-overloaded built-in function name.
6102 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6105 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6108 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6109 where the inputs overlap with the output.
6111 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6113 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6114 (=v, v) alternative and explicit "memory" attribute.
6115 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6116 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6118 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6119 "sselog1" type instead of "sselog".
6120 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6121 "sselog". Remove explicit "memory" attribute.
6122 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6123 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6125 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6126 "sselog1" type instead of "sselog". Remove explicit "memory"
6128 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6129 use "sselog1" type instead of "sselog". Remove explicit "memory"
6131 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6132 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6134 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6135 "sselog1" type instead of "sselog". Remove explicit "memory"
6137 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6138 use "sselog1" type instead of "sselog". Remove explicit "memory"
6142 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6145 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6148 2018-04-17 Martin Jambor <mjambor@suse.cz>
6151 * ipa-cp.c (create_specialized_node): Call
6152 expand_all_artificial_thunks if necessary.
6154 2018-04-17 Martin Liska <mliska@suse.cz>
6157 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6158 in message, remote space in between '_G' and '('.
6160 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6163 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6164 avx512f_vmcmp<mode>3<round_saeonly_name>,
6165 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6166 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6167 avx512f_rndscale<mode><round_saeonly_name>,
6168 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6169 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6170 Use %<iptr>2 instead of %2 for -masm=intel.
6171 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6172 avx512f_vcvttss2usi<round_saeonly_name>,
6173 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6175 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6176 avx512f_vcvttsd2usi<round_saeonly_name>,
6177 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6178 Use %q1 instead of %1 for -masm=intel.
6179 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6180 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6181 of %3 for -masm=intel.
6182 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6184 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6186 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6188 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6189 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6191 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6193 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6194 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6195 %g1 and one with %0 and %1.
6196 (avx512er_vmrcp28<mode><round_saeonly_name>,
6197 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6199 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6200 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6201 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6202 of %0 and %{%4%} for -masm=intel.
6203 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6204 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6205 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6206 order of %0 and %{%5%}%{z%} for -masm=intel.
6208 2018-04-17 Jan Hubicka <jh@suse.cz>
6211 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6213 2018-04-17 Martin Liska <mliska@suse.cz>
6216 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6217 for target_clone error message. Make default implementation
6218 clone to be a local declaration.
6219 (separate_attrs): Add new argument and check for an empty
6221 (expand_target_clones): Handle it.
6222 (ipa_target_clone): Make redirection just for target_clones
6225 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6226 Tom de Vries <tom@codesourcery.com>
6229 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6230 tiled basic blocks without omp continue statements.
6232 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6235 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6236 vec_extract expression as having side effects to make sure it gets
6239 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6242 * config/i386/i386.c (get_builtin_code_for_version): Check
6245 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6248 * gcc/config.gcc: Support "skylake".
6249 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6251 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6252 (processor_target_table): Add "skylake".
6253 (ix86_option_override_internal): Add "skylake".
6254 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6255 PROCESSOR_CANNONLAKE.
6256 (get_builtin_code_for_version): Fix priority for
6257 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6258 PROCESSOR_SKYLAKE-AVX512.
6259 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6260 (processor_type): Add PROCESSOR_SKYLAKE.
6262 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6263 Jason Merrill <jason@redhat.com>
6266 * convert.c (convert_to_integer_1): Use direct recursion for
6267 enumeral types and types with a precision less than the number
6268 of bits in their mode.
6270 2018-04-16 Julia Koval <julia.koval@intel.com>
6273 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6274 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6276 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6279 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6280 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6281 and -mno-direct-move.
6283 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6286 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6287 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6288 Ensure negative shifts result in {0}.
6290 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6292 PR rtl-optimization/79916
6293 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6294 regs (if any) to define how to gnerate SD moves when LRA is in
6297 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6299 PR rtl-optimization/85393
6300 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6301 * except.c (expand_dw2_landing_pad_for_region): Make static.
6302 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6303 a label and unconditional jump to old_bb, rather than
6304 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6307 PR rtl-optimization/85376
6308 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6309 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6310 instead of a specific value.
6312 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6313 Bin Cheng <bin.cheng@arm.com>
6315 PR tree-optimization/82965
6316 PR tree-optimization/83991
6317 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6318 by_profile_only parameter.
6319 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6320 information if the loop was predicted to iterate too many times.
6321 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6323 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6326 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6329 2018-04-13 Martin Liska <mliska@suse.cz>
6330 Jakub Jelinek <jakub@redhat.com>
6333 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6334 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6335 * builtins.c (expand_builtin_memory_copy_args): Use
6336 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6337 handle dest_addr == pc_rtx.
6339 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6342 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6343 asked to not generate direct moves.
6344 (fix_trunc<mode>si2_stfiwx): Similar.
6345 (fix_trunc<mode>si2_internal): Similar.
6347 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6350 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6351 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6352 lookup if dest in some wider mode is known to be const0_rtx and
6353 if so, record permanent equivalence for it to be ZERO_EXTEND of
6354 the narrower mode destination.
6356 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6358 * lto-streamer-out.c (output_function): Revert 259346.
6359 * omp-expand.c (expand_oacc_for): Likewise.
6361 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6363 PR rtl-optimization/85354
6364 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6365 * sel-sched.c (sel_global_init): ... here.
6367 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6370 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6371 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6372 mode for PE-COFF targets.
6373 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6374 (i386_pe_asm_lto_end): Likewise.
6375 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6376 (TARGET_ASM_LTO_END): Likewise.
6377 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6378 (i386_pe_asm_lto_start): New function.
6379 (i386_pe_asm_lto_end): Likewise.
6381 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6382 Richard Biener <rguenther@suse.de>
6385 * lto-streamer-out.c (output_function): Fix CFG loop state before
6387 * omp-expand.c (expand_oacc_for): Handle calls to internal
6388 functions like regular functions.
6390 2018-04-12 Richard Biener <rguenther@suse.de>
6393 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6394 for the early LTO debug to properly generate references to it
6395 during DIE emission. Do not re-use that for the skeleton for
6397 (dwarf2out_early_finish): Likewise.
6399 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6402 * config/i386/sse.md
6403 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6404 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6405 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6406 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6407 and output is a reg, avoid creating invalid lowpart subreg, but
6408 instead split into a 512-bit move. Don't split if not AVX512VL,
6409 input is xmm16+ reg and output is a mem.
6410 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6411 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6412 xmm16+ reg and output is a mem.
6414 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6416 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6417 also for flag_dwarf2_cfi_asm.
6419 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6421 PR rtl-optimization/85342
6422 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6423 a bool scalar var inside of the loop instead. Don't try to update
6424 recog_data.operand after failed apply_change_group.
6426 2018-04-12 Tom de Vries <tom@codesourcery.com>
6429 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6430 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6431 array with flexible array member as array without given dimension.
6432 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6433 argument for undefined param to true.
6435 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6438 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6439 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6440 from PowerPC section.
6441 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6442 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6443 help text that is too long.
6444 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6445 help text that is too long.
6446 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6447 help text that is too long.
6449 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6451 * config/alpha/alpha.md (stack_probe_internal): Rename
6452 from "probe_stack". Update all callers.
6454 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6456 PR rtl-optimization/84566
6457 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6458 sched_macro_fuse_insns.
6460 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6463 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6464 (compute_block_dependences): ... from here.
6466 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6468 PR tree-optimization/85331
6469 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6470 from int to HOST_WIDE_INT.
6472 2018-04-11 Martin Jambor <mjambor@suse.cz>
6475 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6476 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6477 not the same as the source val.
6478 (cgraph_edge_brings_value_p): New parameter.
6479 (gather_edges_for_value): Pass destination value to
6480 cgraph_edge_brings_value_p.
6481 (perhaps_add_new_callers): Likewise.
6482 (get_info_about_necessary_edges): Likewise and exclude values brought
6483 only by self-recursive edges.
6484 (create_specialized_node): Redirect only clones of self-calling edges.
6485 (+self_recursive_pass_through_p): New function.
6486 (find_more_scalar_values_for_callers_subset): Use it.
6487 (find_aggregate_values_for_callers_subset): Likewise.
6488 (known_aggs_to_agg_replacement_list): Removed.
6489 (decide_whether_version_node): Re-calculate known constants for all
6490 remaining context clones.
6492 2018-04-11 Richard Biener <rguenther@suse.de>
6495 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6496 from early DWARF output.
6497 (dwarf2out_early_finish): Output line info unconditionally into
6498 early DWARF and add reference to it.
6500 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6503 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6504 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6505 other than V2DFmode using iptr mode attribute.
6506 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6508 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6510 PR rtl-optimization/84659
6511 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6513 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6516 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6518 (output_loc_list): Pass address of a dummy size variable even in the
6519 locview handling loop.
6520 (index_location_lists): Add comment on why skip_loc_list_entry can't
6523 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6526 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6529 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6532 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6533 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6534 and -mstring-compare-inline-limit.
6536 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6539 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6540 for stack clash protection in a register whenever we need it to be in
6543 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6545 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6546 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6548 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6551 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6553 (mlong-double-): Ditto.
6554 * config/rs6000/sysv4.opt (msdata=): Ditto.
6555 (mtls-size=): Ditto.
6557 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6559 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6560 erroneous entries for
6561 "vector int vec_ldl (int, long int *)", and
6562 "vector unsigned int vec_ldl (int, unsigned long int *)".
6563 Add comments and entries for
6564 "vector bool char vec_ldl (int, bool char *)",
6565 "vector bool short vec_ldl (int, bool short *)",
6566 "vector bool int vec_ldl (int, bool int *)",
6567 "vector bool long long vec_ldl (int, bool long long *)",
6568 "vector pixel vec_ldl (int, pixel *)",
6569 "vector long long vec_ldl (int, long long *)",
6570 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6571 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6572 type tree bool_long_long_type_node and correct definition of
6573 bool_V2DI_type_node to make reference to this new type tree.
6574 (rs6000_mangle_type): Replace erroneous reference to
6575 bool_long_type_node with bool_long_long_type_node.
6576 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6577 comments to emphasize sign distinctions for char and int types and
6578 replace RS6000_BTI_bool_long constant with
6579 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6580 use of RS6000_BTI_pixel.
6581 (bool_long_type_node): Remove this macro definition.
6582 (bool_long_long_type_node): New macro definition
6584 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6586 PR rtl-optimization/85300
6587 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6588 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6589 simplify_unary_operation fails.
6591 2018-04-10 Martin Liska <mliska@suse.cz>
6593 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6594 cgraph_edge and ipa_ref.
6596 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6600 * config/i386/sse.md
6601 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6602 computation of the VEC_MERGE selector from mask.
6603 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6604 Fix decoding of the VEC_MERGE selector into mask.
6606 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6608 PR tree-optimization/85286
6609 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6611 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6613 * final.c (final_1): Set insn_last_address as well as
6614 insn_current_address.
6616 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6619 * explow.c (emit_stack_probe): Call validize_mem on memory location
6620 before passing it to gen_probe_stack. Create address operand and
6621 legitimize it for the probe_stack_address case.
6623 2018-04-09 Jan Hubicka <jh@suse.cz>
6626 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6627 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6628 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6630 (free_lang_data): Rebuild type inheritance graph.
6632 2018-04-09 Martin Sebor <msebor@redhat.com>
6634 * invoke.texi (-finline-small-functions): Mention other optimization
6636 (-findirect-inlining, -fpartial-inlining): Same.
6637 (-finline-functions-called-once): Same.
6638 (-freorder-blocks-and-partition): Same.
6640 2018-04-09 Jan Hubicka <jh@suse.cz>
6643 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6644 jumps; choose last target that matches the criteria (i.e.
6645 no partition changes for non-crossing jumps).
6646 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6647 support for redirecting crossing jumps to non-crossing.
6649 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6651 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6652 also for naked functions.
6654 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6656 * config/arc/arc.md (add_shift): New pattern.
6657 (add_shift2): Likewise.
6658 (sub_shift): Likewise.
6659 (sub_shift_cmp0_noout): Likewise.
6660 (compare_si_ashiftsi): Likewise.
6661 (xbfu_cmp0_noout): New combine pattern.
6662 (xbfu_cmp0"): Likewise.
6663 (movsi_set_cc_insn): Place the predicable variant first.
6664 (commutative_binary_cmp0_noout): Remove clobber.
6665 (commutative_binary_cmp0): New pattern.
6666 (noncommutative_binary_cmp0): Likewise.
6667 (noncommutative_binary_cmp0_noout): Likewise.
6668 (noncommutative_binary_comparison_result_used): Removed.
6669 (rsub_cmp0): New pattern.
6670 (rsub_cmp0_noout): Likewise.
6671 (extzvsi): Changed, keep only meaningful variants.
6672 (SQH, SEZ): New iterators.
6673 (SQH_postfix): New mode attribute.
6674 (SEZ_prefix): New code attribute.
6675 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6676 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6677 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6679 (noncommutative_operator): Check the availability of barrel
6682 2018-04-09 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/85284
6685 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6686 Only use the niter constraining form of simple_iv when the exit
6689 2018-04-09 Tom de Vries <tom@codesourcery.com>
6692 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6693 (define_expand "*memory_barrier"): New define_expand.
6694 (define_insn "memory_barrier"): New insn.
6696 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6698 PR rtl-optimization/80463
6699 PR rtl-optimization/83972
6700 PR rtl-optimization/83480
6702 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6703 correct producer for the insn.
6704 (tidy_control_flow): Fixup seqnos in case of debug insns.
6706 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6708 PR rtl-optimization/83913
6710 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6711 different sched-times when merging exprs.
6713 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6715 PR rtl-optimization/83962
6717 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6718 tidy_fallthru_edge and tidy_control_flow.
6720 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6722 PR rtl-optimization/83530
6724 * sel-sched.c (force_next_insn): New global variable.
6725 (remove_insn_for_debug): When force_next_insn is true, also leave only
6726 next insn in the ready list.
6727 (sel_sched_region): When the region wasn't scheduled, make another pass
6728 over it with force_next_insn set to 1.
6730 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6732 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6734 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6735 for interrupt control.
6736 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6737 functions for interrupt control.
6738 * config/nds32/nds32-intrinsic.md: Likewise.
6739 * config/nds32/nds32_intrinsic.h: Likewise.
6740 * config/nds32/nds32.h (nds32_builtins): Likewise.
6742 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6744 * config/nds32/nds32.c (nds32_init_machine_status,
6745 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6746 strict_aligned_p field.
6747 (nds32_expand_to_rtl_hook): New function.
6748 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6749 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6751 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6752 Chung-Ju Wu <jasonwucj@gmail.com>
6754 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6755 * config/nds32/nds32-n7.md: New file.
6756 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6757 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6759 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6760 * config/nds32/nds32.md (pipeline_model): Add n7.
6761 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6762 * config/nds32/pipelines.md: Include n7 settings.
6764 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6765 Chung-Ju Wu <jasonwucj@gmail.com>
6767 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6768 * config/nds32/nds32-e8.md: New file.
6769 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6770 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6772 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6773 * config/nds32/nds32.md (pipeline_model): Add e8.
6774 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6775 * config/nds32/pipelines.md: Include e8 settings.
6777 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6778 Chung-Ju Wu <jasonwucj@gmail.com>
6780 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6781 * config/nds32/nds32-n8.md: New file.
6782 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6783 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6785 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6786 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6787 * config/nds32/nds32.md (pipeline_model): Add n8.
6788 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6789 * config/nds32/pipelines.md: Include n8 settings.
6791 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6792 Chung-Ju Wu <jasonwucj@gmail.com>
6794 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6795 * config/nds32/nds32-n9-2r1w.md: New file.
6796 * config/nds32/nds32-n9-3r2w.md: New file.
6797 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6798 nds32_register_ports): New or modify for cpu n9.
6799 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6801 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6802 * config/nds32/nds32-utils.c: New file.
6803 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6804 TARGET_MUL_SLOW): Define.
6805 * config/nds32/nds32.md (pipeline_model): New attribute.
6806 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6807 New options that support cpu n9.
6808 * config/nds32/pipelines.md: Include n9 settings.
6809 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6811 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6813 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6814 information if necessary.
6815 (output_cond_branch_compare_zero): Likewise.
6816 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6817 (nds32_target_alignment): Refine for alignment.
6818 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6819 (FUNCTION_BOUNDARY): Modify.
6820 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6822 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6824 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6826 * config/nds32/constants.md (unspec_volatile_element): Add values for
6827 TLB operation and data prefetch.
6828 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6829 functions for TLB operation and data prefetch.
6830 * config/nds32/nds32-intrinsic.md: Likewise.
6831 * config/nds32/nds32_intrinsic.h: Likewise.
6832 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6833 (nds32_print_operand): Likewise.
6834 * config/nds32/nds32.h (nds32_builtins): Likewise.
6836 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6837 Andrew Pinski <pinsika@gcc.gnu.org>
6840 * match.pd: Use constant_boolean_node of correct type instead of
6841 boolean_true_node or boolean_false_node for simplifying
6842 pointer comparisons to zero.
6844 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6846 PR tree-optimization/80021
6847 * tree.c (verify_type_variant): Make error call in verify_variant_match
6848 translatable and remove final full stop.
6850 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6852 * config/nds32/constants.md (unspec_volatile_element): Add
6853 UNSPEC_VOLATILE_EH_RETURN.
6854 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6855 nds32_output_stack_pop): Support dwarf exception handling process.
6856 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6857 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6858 exception handling process.
6859 (nds32_compute_stack_frame): Likewise.
6860 (nds32_return_addr_rtx): Likewise.
6861 (nds32_initial_elimination_offset): Likewise.
6862 (nds32_expand_prologue): Likewise.
6863 (nds32_expand_epilogue): Likewise.
6864 (nds32_dynamic_chain_address): New function.
6865 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6867 (DYNAMIC_CHAIN_ADDRESS): Define.
6868 (EH_RETURN_DATA_REGNO): Define.
6869 (EH_RETURN_STACKADJ_RTX): Define.
6870 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6871 patterns for dwarf exception handling.
6873 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6875 * config/nds32/nds32.h: Clean up obsolete macros.
6877 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6879 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6880 Add enum values for particular instructions.
6881 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6882 particular intrinsic functions.
6883 * config/nds32/nds32-intrinsic.md: Likewise.
6884 * config/nds32/nds32_intrinsic.h: Likewise.
6885 * config/nds32/nds32.h (nds32_builtins): Likewise.
6886 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6887 (btst, ave): New patterns for particular instructions.
6889 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6891 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6892 Add enum values for atomic load/store and memory sync.
6893 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6895 * config/nds32/nds32-intrinsic.md: Likewise.
6896 * config/nds32/nds32_intrinsic.h: Likewise.
6897 * config/nds32/nds32.h (nds32_builtins): Likewise.
6899 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6901 PR tree-optimization/85257
6902 * fold-const.c (native_encode_vector): If not all elts could fit
6903 and off is -1, return 0 rather than offset.
6904 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6905 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6906 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6907 adjust buffer in native_interpret_expr call.
6909 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6911 * config/nds32/constants.md (unspec_volatile_element): Add cache
6912 control enum values.
6913 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6914 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6915 * config/nds32/nds32.c (nds32_cctl_names): New.
6916 (nds32_print_operand): Handle cache control register names.
6917 * config/nds32/nds32.h (nds32_builtins): New enum values.
6918 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6920 * config/nds32/nds32.md (type): Add mmu.
6921 * config/nds32/pipelines.md (simple_insn): Add mmu.
6923 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6925 * config/nds32/nds32.md (type): Remove call.
6926 * config/nds32/pipelines.md (simple_insn): Likewise.
6928 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6930 * config/nds32/constants.md (unspec_volatile_element): Add
6931 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6932 UNSPEC_VOLATILE_FMFCFG.
6933 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6934 description for fmfcfg and fmfcsr.
6935 (bdesc_1arg): Add fmtcsr.
6936 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6937 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6938 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6939 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6940 unspec_fmfcfg): New patterns.
6941 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6942 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6943 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6944 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6945 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6946 __nds32__fmfcfg): Define.
6948 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6950 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6951 intrinsic register names.
6952 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6953 intrinsic register enum values and macros.
6955 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6957 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6958 for load/store addressing form.
6959 (nds32_print_operand_address): Likewise.
6961 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6964 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6965 based on LABEL_REF. Remove useless assertion.
6966 (pic_address_needs_scratch): Fix formatting.
6967 (sparc_legitimize_pic_address): Minor tweaks.
6968 (sparc_delegitimize_address): Adjust assertion accordingly.
6969 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6970 into symbolic_operand.
6971 (movsi_high_pic_label_ref): Likewise.
6972 (movsi_lo_sum_pic_label_ref): Likewise.
6973 (movdi_pic_label_ref): Likewise.
6974 (movdi_high_pic_label_ref): Likewise.
6975 (movdi_lo_sum_pic_label_ref): Likewise.
6977 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6979 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6980 custom LIB_SPEC setup.
6982 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6983 Kito Cheng <kito.cheng@gmail.com>
6985 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6986 * config/riscv/freebsd.h: New.
6988 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6990 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6991 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6994 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6995 Kito Cheng <kito.cheng@gmail.com>
6997 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6998 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6999 * config/nds32/nds32-protos.h (nds32_output_call,
7000 nds32_output_return): Declare.
7001 * config/nds32/nds32.md: Refine all the call and return patterns.
7003 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7006 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7007 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7009 PR rtl-optimization/84872
7010 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7011 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7014 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7016 * expr.c (copy_blkmode_to_reg): Revert 254862.
7017 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7019 2018-04-06 Richard Biener <rguenther@suse.de>
7022 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7023 after seeing a component reference with an adjacent field. Treat
7024 refs to arrays at struct end of external decls similar to
7025 refs to unconstrained commons.
7027 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7030 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7031 look through SAVE_EXPRs with non-side-effects argument. Adjust
7033 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7036 2018-04-06 Richard Biener <rguenther@suse.de>
7039 * alias.c (find_base_term): New wrapper around find_base_term
7040 unwinding CSELIB_VAL_PTR changes.
7041 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7044 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7046 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7048 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7049 constant definitions.
7050 ("nop"): lr 0,0 -> nopr r0
7051 ("nop_lr0", "nop_lr1"): New insn definitions.
7053 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7055 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7056 NDS32_V3PUSH_AVAILABLE_P macro.
7058 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7059 Chung-Ju Wu <jasonwucj@gmail.com>
7061 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7062 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7063 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7064 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7065 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7066 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7067 * config/nds32/constraints.md: New constraints and checking for hard
7068 float configuration.
7069 * config/nds32/iterators.md: New mode iterator and attribute for hard
7070 float configuration.
7071 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7073 * config/nds32/nds32-fpu.md: New file.
7074 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7075 deal with hard float code generation.
7076 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7078 (abi_type, float_reg_number): New enum type.
7079 * config/nds32/nds32-predicates.c: New predicates for hard float.
7080 * config/nds32/nds32-protos.h: Declare functions for hard float.
7081 * config/nds32/nds32.c: Implementation for hard float configuration.
7082 * config/nds32/nds32.h: Definitions for hard float configuration.
7083 * config/nds32/nds32.md: Include hard float machine description and
7084 modify patterns for hard float configuration.
7085 * config/nds32/nds32.opt: New options for hard float configuration.
7086 * config/nds32/predicates.md: New predicates for hard float
7089 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7091 * common/config/nds32/nds32-common.c
7092 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7094 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7097 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7098 CONSTRUCTOR_ELT (ctor, ...)->value.
7100 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7103 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7105 2018-04-05 Tom de Vries <tom@codesourcery.com>
7108 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7111 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7112 Kito Cheng <kito.cheng@gmail.com>
7114 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7115 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7116 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7117 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7119 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7120 Kito Cheng <kito.cheng@gmail.com>
7122 * config/nds32/constraints.md (Ufe): New memory constraint.
7123 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7124 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7125 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7127 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7128 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7130 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7132 * config/nds32/nds32.md: Use optimize_size in the condition for
7133 alu-shift instructions.
7135 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7137 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7139 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7141 * config/nds32/nds32.md (negsi2): Refine pattern.
7143 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7144 Chung-Ju Wu <jasonwucj@gmail.com>
7146 * config/nds32/iterators.md (shift_rotate): New code iterator.
7147 (shift): New code attribute.
7148 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7149 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7150 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7151 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7152 bit-wise operations.
7153 (andsi3, *andsi3): Ditto.
7154 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7155 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7156 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7157 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7158 nds32_ior_operand, nds32_xor_operand): New predicates.
7160 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7162 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7163 (addsi3, subsi3): ... this.
7165 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7167 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7169 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7171 * config/nds32/nds32.md: Adjust indention.
7173 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7175 * config/nds32/nds32.md (feature): New attribute.
7177 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7179 * config/nds32/nds32.md (subtype): New attribute.
7181 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7184 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7185 expansion to perform a bitwise AND of the argument followed by a
7186 boolean negation of the result.
7188 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7190 PR rtl-optimization/84878
7191 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7192 the basic block. Assert the use reference is not artificial and that
7193 it has an associated insn.
7195 2018-04-04 Michael Matz <matz@suse.de>
7197 * builtins.c (compute_objsize): Pass correct operand
7198 to array_at_struct_end_p.
7200 2018-04-04 Richard Biener <rguenther@suse.de>
7203 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7204 from contexts for DINFO_LEVEL_TERSE and below.
7206 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7208 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7209 resiter_operand condition.
7210 * config/nds32/nds32.md (*move<mode>): Ditto.
7212 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7213 Monk Chiang <sh.chiang04@gmail.com>
7215 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7217 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7221 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7222 Kito Cheng <kito.cheng@gmail.com>
7224 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7225 nds32_cond_code_str, output_cond_branch,
7226 output_cond_branch_compare_zero, nds32_expand_cbranch,
7227 nds32_expand_cstore, nds32_expand_movcc,
7228 nds32_output_cbranchsi4_equality_zero,
7229 nds32_output_cbranchsi4_equality_reg,
7230 nds32_output_cbranchsi4_equality_reg_or_const_int,
7231 nds32_output_cbranchsi4_greater_less_zero: New functions.
7232 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7233 nds32_expand_cstore, nds32_expand_movcc,
7234 nds32_output_cbranchsi4_equality_zero,
7235 nds32_output_cbranchsi4_equality_reg,
7236 nds32_output_cbranchsi4_equality_reg_or_const_int,
7237 nds32_output_cbranchsi4_greater_less_zero): Declare.
7238 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7239 nds32_rimm11s_operand): New predicates.
7240 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7241 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7244 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7246 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7247 * config/nds32/nds32.md: Ditto.
7248 * config/nds32/pipelines.md: Ditto.
7250 2018-04-04 Richard Biener <rguenther@suse.de>
7252 PR tree-optimization/85168
7253 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7254 propagating abnormals.
7256 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7258 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7260 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7261 Kito Cheng <kito.cheng@gmail.com>
7263 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7264 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7265 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7266 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7267 * config/nds32/nds32.md (sibcall_internal): New.
7268 (sibcall_register): Remove.
7269 (sibcall_immediate): Remove.
7270 (sibcall_value_internal): New.
7271 (sibcall_value_register): Remove.
7272 (sibcall_value_immediate): Remove.
7273 * config/nds32/predicates.md (nds32_general_register_operand): New.
7274 (nds32_call_address_operand): New.
7276 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7278 PR rtl-optimization/85167
7279 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7280 bb_defs if *split_p, instead preinitialize it to NULL.
7282 PR tree-optimization/85156
7283 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7284 evaluating the argument multiple times.
7286 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7288 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7290 (_mm_cvtpd_ps): Likewise.
7291 (_mm_cvttpd_epi32): Likewise.
7292 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7293 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7294 vector, pixel, and bool following altivec.h include.
7296 2018-04-03 Martin Sebor <msebor@redhat.com>
7298 * doc/extend.texi (Common Function Attributes): Clarify.
7299 (const attribute): Likewise.
7300 (pure attribute): Likewise.
7302 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7305 * config/i386/i386.c (ix86_expand_vector_set): Use
7306 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7308 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7310 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7311 instructions when changing rounding bits to preserve precision bits
7312 in the x87 control word.
7314 2018-04-03 Martin Liska <mliska@suse.cz>
7316 PR tree-optimization/82491
7317 * rtl.h (strip_offset_and_add): Replace += suboffset with
7318 poly_uint64 () + suboffset.
7320 2018-03-29 Martin Liska <mliska@suse.cz>
7321 Martin Jambor <mjambor@suse.cz>
7324 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7325 param_type is not an integral or pointer type.
7327 2018-04-03 Richard Biener <rguenther@suse.de>
7329 * sese.h (recompute_all_dominators): Remove.
7331 2018-04-02 Martin Sebor <msebor@redhat.com>
7333 * doc/invoke.texi (-Wrestrict): Fix typos.
7335 2018-04-02 Jim Wilson <jimw@sifive.com>
7337 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7338 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7339 (<optab>di3, <optab>si3_extend): Likewise.
7340 (<optab>si3_mask, <optab>si3_mask_1): New.
7341 (<optab>di3_mask, <optab>di3_mask_1): New.
7342 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7343 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7344 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7346 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7348 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7351 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7353 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7354 (nds32_canonicalize_comparison): New function.
7356 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7357 Kito Cheng <kito.cheng@gmail.com>
7358 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7360 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7361 * config/nds32/constants.md (unspec_volatile_element): Add
7362 UNSPEC_VOLATILE_RELAX_GROUP.
7363 * config/nds32/nds32-relax-opt.c: New file.
7364 * config/nds32/nds32-predicates.c
7365 (nds32_symbol_load_store_p): New function.
7366 * config/nds32/nds32-protos.h
7367 (nds32_symbol_load_store_p): Declare function.
7368 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7369 * config/nds32/nds32.c
7370 (nds32_register_pass): New function to register pass.
7371 (nds32_register_passes): New function to register passes.
7372 * config/nds32/nds32.md (relax_group): New pattern.
7373 * config/nds32/nds32.opt (mrelax-hint): New option.
7374 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7376 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7378 * config/nds32/t-nds32: Modify files dependency.
7380 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7382 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7383 (PROFILE_HOOK): Define its implementation.
7385 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7387 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7388 type and 32-bit size.
7390 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7393 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7394 (V_128_256): New mode iterator.
7395 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7396 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7397 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7399 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7400 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7402 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7405 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7406 NaN inputs correctly.
7408 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7411 * config/rs6000/vsx.md (??r): New mode attribute.
7412 (*vsx_mov<mode>_64bit): Use it.
7413 (*vsx_mov<mode>_32bit): Likewise.
7415 2018-03-30 Martin Sebor <msebor@redhat.com>
7417 PR tree-optimization/84818
7418 * builtins.c (check_access): Use warning_n.
7420 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7423 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7425 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7428 2018-03-30 Julia Koval <julia.koval@intel.com>
7431 * x86-tune.def (movx, partial_reg_dependency): Enable for
7434 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7437 * lra-constraints.c (process_alt_operands): Move setting
7438 this_alternative_matches below.
7440 2018-03-29 Martin Liska <mliska@suse.cz>
7443 * doc/invoke.texi: Document how LTO works with debug info.
7444 Describe auto-load support of binutils. Mention 'x86-64'
7445 as valid option value of -march option.
7447 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7449 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7452 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7453 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7454 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7457 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7460 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7461 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7462 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7463 for RS6000_BTM_POWERPC64.
7464 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7465 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7466 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7471 2018-03-28 Carl Love <cel@us.ibm.com>
7474 2017-09-27 Carl Love <cel@us.ibm.com>
7476 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7477 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7478 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7481 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7483 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7484 instead of __vector bool.
7485 (_mm_max_pu8): Likewise.
7486 (_mm_min_pi16): Likewise.
7488 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7491 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7492 (DIVWEUO): Likewise.
7494 (DIVDEUO): Likewise.
7495 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7496 DIVWEUO and DIVDEUO.
7497 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7498 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7499 (div_extend): Likewise.
7500 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7502 (__builtin_divweuo): Likewise.
7503 (__builtin_divdeo): Likewise.
7504 (__builtin_divdeuo): Likewise.
7506 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7509 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7510 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7512 PR tree-optimization/82004
7513 * gimple-match-head.c (optimize_pow_to_exp): New function.
7514 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7515 Don't fold to exp if optimize_pow_to_exp is false.
7517 2018-03-28 Martin Liska <mliska@suse.cz>
7520 * calls.c (initialize_argument_information): Fix trailing space.
7521 * common.opt: Fix typo and provide better explanation for
7522 -fsanitize-coverage option.
7523 * config/i386/i386.opt: Fix typo.
7525 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7526 Martin Liska <mliska@suse.cz>
7529 * gimplify.c (asan_poison_variable): Don't do the check for
7530 gimplify_omp_ctxp here.
7531 (gimplify_decl_expr): Do it here.
7532 (gimplify_target_expr): Likewise.
7534 2018-03-28 Martin Liska <mliska@suse.cz>
7537 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7538 chkp_type_bounds_count if MPX is not enabled.
7540 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7542 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7544 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7547 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7548 function to create the function decl for complex long double
7549 multiply and divide for -mabi=ieeelongdouble.
7550 (init_float128_ieee): Call it.
7552 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7555 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7556 -fcf-protection=branch -mibt.
7557 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7559 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7562 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7563 arm_word_relocations.
7565 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7568 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7569 extern array declarations.
7571 2018-03-27 Richard Biener <rguenther@suse.de>
7574 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7575 explicit single_use checks.
7577 2018-03-27 Richard Biener <rguenther@suse.de>
7579 PR tree-optimization/85082
7580 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7583 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7585 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7586 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7587 Turn on fasynchronous-unwind-tables and funwind-tables.
7589 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7592 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7593 (*bmi_blsr_<mode>_ccz): Ditto.
7595 2018-03-26 Tom de Vries <tom@codesourcery.com>
7597 PR tree-optimization/85063
7598 * omp-general.c (offloading_function_p): New function. Factor out
7600 * omp-offload.c (pass_omp_target_link::gate): ... here.
7601 * omp-general.h (offloading_function_p): Declare.
7602 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7603 with attribute omp declare target for offloading functions.
7605 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7607 PR tree-optimization/84005
7608 * tree-data-ref.h (get_base_for_alignment): Declare.
7609 * tree-data-ref.c (get_base_for_alignment_1): New function.
7610 (get_base_for_alignment): Likewise.
7611 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7612 get_base_for_alignment to find a suitable base object, instead
7613 of always using drb->base_address.
7615 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7618 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7619 known size by default.
7621 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7624 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7625 and non BLKmode operands.
7627 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7630 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7631 Clean up attributes.
7633 2018-03-23 Richard Biener <rguenther@suse.de>
7636 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7637 we are going to emit early debug for LTO.
7639 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7642 * function.c (match_asm_constraints_1): Don't optimize if input
7643 doesn't satisfy general_operand predicate for output's mode.
7646 * alias.c (write_dependence_p): Don't require for x_canonicalized
7647 non-VOIDmode if x has VOIDmode.
7650 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7651 just don't try to optimize it rather than assert it never happens.
7653 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7655 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7656 macro expansions for definition of ST_INTERNAL_<mode> and
7657 LD_INTERNAL_<mode> builtins.
7658 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7660 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7662 (altivec_expand_st_builtin): Likewise.
7663 (altivec_expand_builtin): Remove calls to deleted functions.
7664 (rs6000_address_for_altivec): Delete this function.
7665 * config/rs6000/vector.md: Remove expands for
7666 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7668 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7671 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7672 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7673 re-computing once computed.
7674 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7675 (arm_init_machine_status): Initialize
7676 machine->static_chain_stack_bytes.
7678 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7681 * doc/extend.texi: Add four new prototypes for vec_ld.
7682 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7683 definitions for more logical presentation.
7684 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7685 entries for V1TI variants of __builtin_altivec_ld builtin.
7686 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7687 handling of V1TI variant of LVX icode pattern.
7688 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7689 (rs6000_gimple_fold_builtin): Likewise.
7690 (altivec_init_builtins): Add code to define
7691 __builtin_altivec_lvx_v1ti function.
7693 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7696 * function.c (match_asm_constraints_1): Don't do the optimization
7697 if input isn't a REG, SUBREG, MEM or constant.
7699 2018-03-22 Tom de Vries <tom@codesourcery.com>
7701 PR tree-optimization/84956
7702 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7703 bb_has_abnormal_pred.
7705 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7708 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7709 DECL_INITIAL (decl) to decl at the end.
7710 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7713 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7715 * doc/extend.texi (__builtin_tgmath): Document when complex
7716 integer types are treated as _Complex _Float64.
7718 2018-03-21 Tom de Vries <tom@codesourcery.com>
7720 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7722 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7724 PR tree-optimization/84960
7725 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7726 if it is ENTRY block, move them into single succ of ENTRY in that case.
7728 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7730 PR tree-optimization/84811
7731 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7732 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7733 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7734 (known_subrange_p): Update accordingly. Cast each value involved
7735 in the size comparison, rather than casting the result of the
7738 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7740 PR tree-optimization/84982
7741 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7742 by flipping the least significant bit rather than all bits from
7743 bitpos to bitpos + bitsize - 1.
7745 2018-03-21 Nathan Sidwell <nathan@acm.org>
7747 * doc/extend.texi (Deprecated Features): Remove mention of
7748 long-deleted deprecations.
7750 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7753 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7754 * configure: Regenerate.
7756 2018-03-21 Tom de Vries <tom@codesourcery.com>
7758 PR tree-optimization/83126
7759 * tree-parloops.c (num_phis): New function.
7760 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7762 2018-03-21 Nathan Sidwell <nathan@acm.org>
7764 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7765 mention anon-struct/union members and trailing attributes.
7767 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7769 PR tree-optimization/84969
7770 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7771 builtin memset partitions if they set different rhs values.
7773 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7775 PR rtl-optimization/84989
7776 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7777 VEC_DUPLICATE with scalar result mode.
7779 2018-03-21 Martin Liska <mliska@suse.cz>
7782 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7783 not intended return statement.
7785 2018-03-21 Martin Liska <mliska@suse.cz>
7788 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7789 (chkp_find_bound_slots_1): Limit number of iterations.
7791 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7794 * Minor grammar fixes for x86 options.
7796 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7799 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7800 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7802 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7805 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7806 (altivec_lvx_<mode>_1op): Likewise.
7807 (altivec_stvx_<mode>_2op): Likewise.
7808 (altivec_stvx_<mode>_1op): Likewise.
7809 (altivec_lvx_<VM2:mode>): New define_expand.
7810 (altivec_stvx_<VM2:mode>): Likewise.
7811 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7812 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7813 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7814 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7815 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7816 (rs6000_gen_lvx): Likewise.
7817 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7818 (altivec_expand_stv_builtin): Likewise.
7819 (altivec_expand_builtin): Likewise.
7820 * config/rs6000/vector.md: Likewise.
7822 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7825 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7828 2018-03-20 Richard Biener <rguenther@suse.de>
7831 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7832 sign-conversions as zero, fall back to standard scalar_stmt
7835 2018-03-20 Martin Liska <mliska@suse.cz>
7838 * predict.c (rebuild_frequencies): Handle case when we have
7839 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7841 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7844 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7845 flag_section_anchors.
7846 * varasm.c (use_blocks_for_decl_p): Remove hack for
7847 dw2_force_const_mem.
7850 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7852 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7853 be created, use lowpart_subreg of operands[0] rather than operands[0]
7855 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7856 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7857 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7858 and n constraint instead of aarch64_shift_imm_di and Usd.
7859 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7860 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7862 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7865 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7866 to favor GPR over NEON registers.
7867 (<shift>di3_neon): Likewise.
7869 2018-03-20 Tom de Vries <tom@codesourcery.com>
7872 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7873 (nvptx_process_pars): Emit bar.sync asap and alap.
7875 2018-03-20 Tom de Vries <tom@codesourcery.com>
7878 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7879 seen_label if seen_label is already set.
7881 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7884 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7885 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7886 Use 1U instead of 1. Formatting fixes.
7889 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7890 instead of TREE_TYPE (s1) for the return value.
7892 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7894 PR tree-optimization/84946
7895 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7896 bitsize + bitsize in poly_uint64 rather than poly_int64.
7899 * dwarf2asm.c: Include fold-const.c.
7900 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7901 of decl rather than decl itself.
7903 PR rtl-optimization/84643
7904 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7906 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7909 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7910 calling assemble_variable.
7912 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7915 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7916 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7918 2018-03-19 Jim Wilson <jimw@sifive.com>
7921 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7922 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7923 (riscv_first_stack_step): Likewise.
7924 (riscv_option_override): Use STACK_BOUNDARY instead of
7926 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7928 (BIGGEST_ALIGNMENT): Set to 128.
7929 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7930 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7933 2018-03-19 Richard Biener <rguenther@suse.de>
7935 PR tree-optimization/84933
7936 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7937 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7939 2018-03-19 Richard Biener <rguenther@suse.de>
7941 PR tree-optimization/84859
7942 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7943 (cond_if_else_store_replacement): Perform sinking operation on
7944 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7945 Generalize what a BB with a single eligible store is.
7947 2018-03-19 Richard Biener <rguenther@suse.de>
7949 PR tree-optimization/84929
7950 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7951 chrec_is_positive against non-chrec arg.
7953 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7956 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7958 2018-03-18 Martin Liska <mliska@suse.cz>
7960 PR rtl-optimization/84635
7961 * regrename.c (build_def_use): Use matches_mode only when
7964 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7966 PR tree-optimization/84913
7967 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7968 vectorize chains of COND_EXPRs.
7970 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7972 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7974 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7976 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7978 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7980 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7982 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7983 Kito Cheng <kito.cheng@gmail.com>
7985 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7986 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7987 (nds32_adjust_reg_alloc_order): New function.
7988 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7990 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7992 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7993 nds32_print_operand, nds32_print_operand_address): Use
7994 HOST_WIDE_INT_PRINT_DEC instead.
7996 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7998 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8000 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8003 * config/i386/i386.c (initial_ix86_tune_features,
8004 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8006 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8007 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8008 rather than 1u << ix86_tune. Formatting fix.
8009 (ix86_option_override_internal): Change ix86_arch_mask from
8010 unsigned int to unsigned HOST_WIDE_INT, initialize to
8011 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8012 (ix86_function_specific_restore): Likewise.
8014 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8017 * postreload.c (reload_combine_recognize_pattern): Perform
8018 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8019 truncate_int_for_mode the result for the destination's mode.
8022 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8023 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8025 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8028 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8031 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8032 regno_allocno_class_array and sorted_pseudos.
8033 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8034 insns where regno is used.
8036 2018-03-16 Martin Liska <mliska@suse.cz>
8039 * multiple_target.c (create_dispatcher_calls): Redirect
8040 reference in the symbol table.
8042 2018-03-16 Martin Liska <mliska@suse.cz>
8045 * multiple_target.c (create_dispatcher_calls): Redirect also
8048 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8052 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8053 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8054 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8056 2018-03-16 Julia Koval <julia.koval@intel.com>
8058 * doc/invoke.texi (Skylake Server): Add CLWB.
8059 Cannonlake): Remove CLWB.
8061 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8063 PR tree-optimization/84841
8064 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8066 (FLOAT_ONE_CONST_TYPE): Define.
8067 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8068 (sort_by_operand_rank): Put entries with higher constant_type last
8069 rather than first to match comments.
8071 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8073 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8076 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8079 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8080 check and use address of target always.
8082 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8085 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8086 (indirect_thunk_bnd_needed): Likewise.
8087 (indirect_thunks_used): Likewise.
8088 (indirect_thunks_bnd_used): Likewise.
8089 (indirect_return_needed): New.
8090 (indirect_return_bnd_needed): Likewise.
8091 (output_indirect_thunk_function): Add a bool argument for
8093 (output_indirect_thunk_function): Don't generate alias for
8094 function return thunk.
8095 (ix86_code_end): Call output_indirect_thunk_function to generate
8096 function return thunks.
8097 (ix86_output_function_return): Set indirect_return_bnd_needed
8098 and indirect_return_needed instead of indirect_thunk_bnd_needed
8099 and indirect_thunk_needed.
8101 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8103 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8104 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8105 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8107 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8108 Paul Hua <paul.hua.gm@gmail.com>
8111 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8113 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8115 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8116 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8119 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8122 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8123 instead of GET_MODE_SIZE when comparing Units.
8125 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8128 * varasm.c (hash_section): Return an unchangeble hash value
8129 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8130 Return !aarch64_can_use_per_function_literal_pools_p ().
8132 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8135 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8136 rather than address of cmode as last argument to prepare_cmp_insn.
8138 2018-03-15 Julia Koval <julia.koval@intel.com>
8140 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8141 F_AVX512VNNI, F_AVX512BITALG): New.
8143 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8146 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8147 insn for floating-point loads and stores.
8149 2018-03-14 Carl Love <cel@us.ibm.com>
8151 * config/rs6000/rs6000-c.c: Add macro definitions for
8152 ALTIVEC_BUILTIN_VEC_PERMXOR.
8153 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8154 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8155 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8156 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8158 * config/doc/extend.texi: Add prototypes for vec_permxor.
8160 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8163 * diagnostic-show-locus.c (class layout_point): Convert m_line
8164 from int to linenum_type.
8165 (line_span::comparator): Use linenum "compare" function when
8166 comparing line numbers.
8167 (test_line_span): New function.
8168 (layout_range::contains_point): Convert param "row" from int to
8170 (layout_range::intersects_line_p): Likewise.
8171 (layout::will_show_line_p): Likewise.
8172 (layout::print_source_line): Likewise.
8173 (layout::should_print_annotation_line_p): Likewise.
8174 (layout::print_annotation_line): Likewise.
8175 (layout::print_leading_fixits): Likewise.
8176 (layout::annotation_line_showed_range_p): Likewise.
8177 (struct line_corrections): Likewise for field m_row.
8178 (line_corrections::line_corrections): Likewise for param "row".
8179 (layout::print_trailing_fixits): Likewise.
8180 (layout::get_state_at_point): Likewise.
8181 (layout::get_x_bound_for_row): Likewise.
8182 (layout::print_line): Likewise.
8183 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8184 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8185 * input.c (selftest::test_linenum_comparisons): New function.
8186 (selftest::input_c_tests): Call it.
8187 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8188 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8189 * selftest.h (ASSERT_GT): New macro.
8190 (ASSERT_GT_AT): New macro.
8191 (ASSERT_LT): New macro.
8192 (ASSERT_LT_AT): New macro.
8194 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8196 PR rtl-optimization/84780
8197 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8199 2018-03-14 Martin Liska <mliska@suse.cz>
8201 * tree.c (record_node_allocation_statistics): Use
8202 get_stats_node_kind.
8203 (get_stats_node_kind): New function extracted from
8204 record_node_allocation_statistics.
8205 (free_node): Use get_stats_node_kind.
8207 2018-03-14 Richard Biener <rguenther@suse.de>
8209 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8210 that the value-set of ANTIC_IN doesn't grow.
8213 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8215 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8216 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8218 2018-03-14 Julia Koval <julia.koval@intel.com>
8220 * config.gcc (icelake-client, icelake-server): New.
8222 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8223 (initial_ix86_arch_features): Ditto.
8224 (PTA_SKYLAKE): Add SGX.
8225 (PTA_ICELAKE): Remove.
8226 (PTA_ICELAKE_CLIENT): New.
8227 (PTA_ICELAKE_SERVER): New.
8228 (ix86_option_override_internal): Split up icelake on icelake client and
8230 (get_builtin_code_for_version): Ditto.
8231 (fold_builtin_cpu): Ditto.
8232 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8233 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8234 * config/i386/i386.h (processor_type): Ditto.
8235 * doc/invoke.texi: Ditto.
8237 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8240 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8241 INTEGER_CST offset, add it together with bitpos / 8 and
8242 sign extend based on POINTER_SIZE.
8246 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8249 * config/i386/constraints.md (Yc): New register constraint.
8250 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8251 Use Yc constraint for alternative 2 of operand 0. Remove
8252 preferred_for_speed attribute.
8254 2018-03-14 Richard Biener <rguenther@suse.de>
8256 PR tree-optimization/84830
8257 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8258 with the old one to avoid oscillations.
8260 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8263 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8265 (assign_by_spills): Return a flag of reload assignment failure.
8266 Do not process the reload assignment failures. Do not spill other
8267 reload pseudos if they has the same reg class. Update n if
8269 (lra_assign): Add a return arg. Set up from the result of
8270 assign_by_spills call.
8271 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8272 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8273 usage_insns if it is not NULL.
8274 (spill_hard_reg_in_range): New function.
8275 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8276 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8277 function prototypes.
8278 (lra_assign): Change prototype.
8279 * lra.c (lra): Add code to deal with fails by splitting hard reg
8282 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8284 * config/riscv/riscv.opt (mrelax): New option.
8285 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8286 "norelax" when riscv_mrelax is disabled.
8287 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8289 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8292 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8293 reassociation for int modes.
8295 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8297 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8298 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8300 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8301 * config/aarch64/aarch64-sve.md
8302 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8303 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8304 (*extend<mode><Vwide>2): Rename to...
8305 (aarch64_sve_extend<mode><Vwide>2): ...this.
8306 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8307 renaming the old pattern to...
8308 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8310 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8311 define_expand, renaming the old pattern to...
8312 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8313 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8314 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8315 account when deciding which SVE instruction the optab should use.
8316 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8318 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8320 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8321 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8322 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8323 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8324 (tlsdesc_small_<mode>): Turn a define_expand and use
8325 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8326 (tlsdesc_small_advsimd_<mode>): ...this.
8327 (tlsdesc_small_sve_<mode>): New pattern.
8329 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8331 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8332 (UNSPEC_UMUL_HIGHPART): New constants.
8333 (MUL_HIGHPART): New int iteraor.
8334 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8335 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8337 (*<su>mul<mode>3_highpart): New define_insn.
8339 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8342 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8345 2018-03-13 Martin Liska <mliska@suse.cz>
8348 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8350 (sem_item_optimizer::~sem_item_optimizer): Release it.
8351 (sem_item_optimizer::merge_classes): Register variable aliases.
8352 (sem_item_optimizer::fixup_pt_set): New function.
8353 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8354 * ipa-icf.h: Declare new variables and functions.
8356 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8359 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8360 integer_pow2p@2 and test integer_pow2p in condition.
8361 (A < 0 ? C : 0): Similarly for @1.
8364 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8365 characters starting at p contain '\0' character, don't look beyond
8369 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8370 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8373 * reg-stack.c (change_stack): Change update_end var from int to
8374 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8375 also call set_block_for_insn on the newly added insns and rescan.
8378 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8379 on the last operand.
8382 * tree.c (stabilize_reference_1): Return save_expr (e) for
8383 STATEMENT_LIST even if it doesn't have side-effects.
8385 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8387 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8389 2018-03-12 Renlin Li <renlin.li@arm.com>
8391 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8392 aarch64_output_scalar_simd_mov_immediate.
8394 2018-03-12 Martin Sebor <msebor@redhat.com>
8396 PR tree-optimization/83456
8397 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8398 for perfectly overlapping calls to memcpy.
8399 (gimple_fold_builtin_memory_chk): Same.
8400 (gimple_fold_builtin_strcpy): Handle no-warning.
8401 (gimple_fold_builtin_stxcpy_chk): Same.
8402 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8404 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8406 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8407 parameter. Use it for SFmode.
8408 (rs6000_function_arg_advance_1): Adjust.
8409 (rs6000_function_arg): Adjust.
8410 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8412 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8414 PR rtl-optimization/84169
8415 PR rtl-optimization/84780
8416 * combine.c (can_combine_p): Check for a 2-insn combination whether
8417 the destination register is used between the two insns, too.
8419 2018-03-12 Richard Biener <rguenther@suse.de>
8421 PR tree-optimization/84803
8422 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8423 for refs DR analysis didn't process.
8425 2018-03-12 Richard Biener <rguenther@suse.de>
8427 PR tree-optimization/84777
8428 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8429 force-vectorize loops ignore whether we are optimizing for size.
8431 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8433 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8434 (TARGET_MD_ASM_ADJUST): Define.
8436 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8437 Kito Cheng <kito.cheng@gmail.com>
8438 Chung-Ju Wu <jasonwucj@gmail.com>
8440 * config/nds32/nds32.c (nds32_compute_stack_frame,
8441 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8442 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8443 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8444 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8445 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8446 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8447 * config/nds32/nds32.md (prologue, epilogue): Use macro
8448 NDS32_V3PUSH_AVAILABLE_P to do checking.
8450 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8453 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8454 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8455 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8456 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8457 addition of most attributes on !orig_type_die or the attribute not
8458 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8460 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8461 Chung-Ju Wu <jasonwucj@gmail.com>
8463 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8465 * config/nds32/nds32.opt (mvh): New option.
8467 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8468 Chung-Ju Wu <jasonwucj@gmail.com>
8470 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8472 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8473 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8476 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8477 Chung-Ju Wu <jasonwucj@gmail.com>
8479 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8481 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8482 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8484 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8485 Kito Cheng <kito.cheng@gmail.com>
8486 Chung-Ju Wu <jasonwucj@gmail.com>
8488 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8489 UNSPEC_FFMISM and UNSPEC_FLMISM.
8490 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8491 for ffb, ffmism and flmism.
8492 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8493 (unspec_ffmism): Ditto.
8494 (unspec_flmism): Ditto.
8495 (nds32_expand_builtin_impl): Check if string extension is available.
8496 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8497 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8499 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8502 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8505 * lra-assigns.c (assign_by_spills): Return a flag of reload
8506 assignment failure. Do not process the reload assignment
8507 failures. Do not spill other reload pseudos if they has the same
8509 (lra_assign): Add a return arg. Set up from the result of
8510 assign_by_spills call.
8511 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8512 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8513 usage_insns if it is not NULL.
8514 (spill_hard_reg_in_range): New function.
8515 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8516 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8517 function prototypes.
8518 (lra_assign): Change prototype.
8519 * lra.c (lra): Add code to deal with fails by splitting hard reg
8522 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8525 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8527 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8530 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8531 (dw_val_node): Add val_symbolic_view.
8532 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8533 (symview_upper_bound): New.
8534 (new_line_info_table): Initialize symviews_since_reset.
8535 (dwarf2out_source_line): Count symviews_since_reset and set
8536 symview_upper_bound.
8537 (dw_val_equal_p): Handle symview.
8538 (add_AT_symview): New.
8539 (print_dw_val): Handle symview.
8540 (attr_checksum, attr_checksum_ordered): Likewise.
8541 (same_dw_val_p, size_of_die): Likewise.
8542 (value_format, output_die): Likewise.
8543 (add_high_low_attributes): Use add_AT_symview for entry_view.
8544 (dwarf2out_finish): Reset symview_upper_bound, clear
8547 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8550 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8551 Add strict argument and use it.
8552 (rs6000_split_multireg_move): Update for new strict argument.
8553 (mem_operand_gpr): Disallow all non-offsettable addresses.
8554 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8556 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8559 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8560 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8561 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8564 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8565 decl, use remap_type if we want to use the type.
8567 2018-03-09 Martin Sebor <msebor@redhat.com>
8569 PR tree-optimization/84526
8570 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8572 (builtin_access::generic_overlap): Be prepared to handle non-array
8575 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8577 PR rtl-optimization/84682
8578 * lra-constraints.c (process_address_1): Check is_address flag
8579 for address constraints.
8580 (process_alt_operands): Likewise.
8581 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8582 preprocess_constraints.
8583 * recog.h (preprocess_constraints): Add oploc parameter.
8585 * recog.c (preprocess_constraints): Test address_operand for
8586 CT_ADDRESS constraints.
8588 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8591 * lra-assigns.c (assign_by_spills): Return a flag of reload
8592 assignment failure. Do not process the reload assignment
8593 failures. Do not spill other reload pseudos if they has the same
8595 (lra_assign): Add a return arg. Set up from the result of
8596 assign_by_spills call.
8597 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8598 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8599 usage_insns if it is not NULL.
8600 (spill_hard_reg_in_range): New function.
8601 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8602 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8603 function prototypes.
8604 (lra_assign): Change prototype.
8605 * lra.c (lra): Add code to deal with fails by splitting hard reg
8608 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8611 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8612 Accept complain bool parameter. Only emit errors if it is true.
8613 (arm_parse_cpu_option_name): Likewise.
8614 (arm_target_thumb_only): Adjust callers of the above.
8615 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8616 prototype to take a default true bool parameter.
8617 (arm_parse_arch_option_name): Likewise.
8619 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8620 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8624 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8625 * configure: Regenerate.
8626 * configure.ac ("linker --version-script option"): New.
8627 ("linker soname option"): New.
8629 2018-03-09 Richard Biener <rguenther@suse.de>
8631 PR tree-optimization/84775
8632 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8633 immediate uses of predicate stmts and mark them modified.
8636 PR tree-optimization/84178
8637 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8639 (version_loop_for_if_conversion): Delay update_ssa call.
8640 (tree_if_conversion): Delay update_ssa until after predicate
8643 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8646 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8647 when the function accesses prior frames.
8649 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8652 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8653 gen_llsym, otherwise call maybe_gen_llsym.
8656 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8657 has ',' character inside of it.
8659 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8662 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8663 as clobbering CC_REGNUM.
8665 2018-03-08 Richard Biener <rguenther@suse.de>
8668 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8669 (follow_copies_to_constant): Do not follow SSA names registered
8672 2018-03-08 Richard Biener <rguenther@suse.de>
8674 PR tree-optimization/84178
8675 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8677 (version_loop_for_if_conversion): Delay update_ssa call.
8678 (tree_if_conversion): Delay update_ssa until after predicate
8681 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8683 PR tree-optimization/84178
8684 * tree-if-conv.c (release_bb_predicate): Remove the
8685 the assertion that the stmts have NULL use_ops.
8686 Discard the statements, asserting that they haven't
8687 yet been added to a BB.
8689 2018-03-08 Richard Biener <rguenther@suse.de>
8691 PR tree-optimization/84746
8692 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8693 (phi_translate): Pass in destination ANTIC_OUT set.
8694 (phi_translate_1): Likewise. For a simplified result lookup
8695 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8696 (phi_translate_set): Adjust.
8697 (do_pre_regular_insertion): Likewise.
8698 (do_pre_partial_partial_insertion): Likewise.
8700 2018-03-08 Martin Liska <mliska@suse.cz>
8702 PR gcov-profile/84735
8703 * doc/gcov.texi: Document usage of profile files.
8704 * gcov-io.h: Document changes in the format.
8706 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8710 * dwarf2out.c (struct dw_line_info_table): Update comments for
8712 (FORCE_RESET_NEXT_VIEW): New.
8713 (FORCE_RESETTING_VIEW_P): New.
8714 (RESETTING_VIEW_P): Check for -1 too.
8715 (ZERO_VIEW_P): Likewise.
8716 (new_line_info_table): Force-reset next view.
8717 (dwarf2out_begin_function): Likewise.
8718 (dwarf2out_source_line): Simplify zero_view_p initialization.
8719 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8720 view directly. Omit view when omitting .loc at line 0.
8722 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8724 PR tree-optimization/84740
8725 * tree-switch-conversion.c (process_switch): Call build_constructors
8726 only if info.phi_count is non-zero.
8728 PR tree-optimization/84739
8729 * tree-tailcall.c (find_tail_calls): Check call arguments against
8730 DECL_ARGUMENTS (current_function_decl) rather than
8731 DECL_ARGUMENTS (func) when checking for tail recursion.
8733 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8735 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8736 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8737 Volker Reichelt's entry and add entries for people that perform
8738 GCC fuzzy testing and report numerous bugs.
8740 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8743 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8744 readonly data in sdata, if that is disabled.
8745 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8746 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8747 -mreadonly-in-sdata option.
8749 2018-03-07 Martin Sebor <msebor@redhat.com>
8751 PR tree-optimization/84468
8752 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8753 basic block when looking for nul assignment.
8755 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8758 * except.h (output_function_exception_table): Adjust prototype.
8759 * except.c (output_function_exception_table): Remove FNNAME parameter
8760 and add SECTION parameter. Ouput one part of the table at a time.
8761 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8762 the first part of the exception table and emit unwind directives.
8763 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8764 (i386_pe_seh_cold_init): Likewise.
8765 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8766 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8767 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8768 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8769 * config/i386/winnt.c: Include except.h.
8770 (struct seh_frame_state): Add reg_offset, after_prologue and
8771 in_cold_section fields.
8772 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8773 (i386_pe_seh_cold_init): New function.
8774 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8775 to seh->in_cold_section.
8776 (seh_emit_push): Record the offset of the push.
8777 (seh_emit_save): Record the offet of the save.
8778 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8779 Test seh->after_prologue to disregard the epilogue.
8780 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8781 (i386_pe_end_cold_function): New function.
8783 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8786 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8787 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8790 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8792 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8793 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8796 * multiple_target.c: Include tree-inline.h and intl.h.
8797 (expand_target_clones): Diagnose and fail if node->definition and
8798 !tree_versionable_function_p (node->decl).
8800 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8802 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8804 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8805 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8806 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8808 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8811 * combine.c (try_combine): Use reg_or_subregno instead of handling
8812 just paradoxical SUBREGs and REGs.
8814 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8816 * config/arc/arc.c (arc_finalize_pic): Remove function.
8817 (arc_must_save_register): We use single base PIC register, remove
8818 checks to save/restore the PIC register.
8819 (arc_expand_prologue): Likewise.
8820 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8822 (arc_verify_short): Likewise.
8823 (arc_attr_type): Likewise.
8824 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8825 (walk_stores): Likewise.
8826 (arc_address_cost): Make it static.
8827 (arc_verify_short): Likewise.
8828 (branch_dest): Likewise.
8829 (arc_attr_type): Likewise.
8830 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8831 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8832 (arc_final_prescan_insn): Remove inserting the nops due to
8833 hardware hazards. It is done in reorg step.
8834 (insn_length_variant_t): Remove.
8835 (insn_length_parameters_t): Likewise.
8836 (arc_insn_length_parameters): Likewise.
8837 (arc_get_insn_variants): Likewise.
8838 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8840 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8843 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8846 PR tree-optimization/84687
8847 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8849 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8851 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8853 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8854 Rename to ppc_speculation_barrier.
8855 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8856 __builtin_ppc_speculation_barrier.
8858 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8861 * combine.c (combine_simplify_rtx): Don't try to simplify if
8862 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8865 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8867 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8868 to 32 bytes when compiling for POWER9.
8870 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8873 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8874 regparm >= 3 with no arg reg available also for calls with
8875 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8878 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8880 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8882 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8885 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8887 2018-03-05 Richard Biener <rguenther@suse.de>
8889 PR tree-optimization/84486
8890 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8891 When inserting a __builtin_assume_aligned call set the LHS
8892 SSA name alignment info accordingly.
8894 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8896 PR tree-optimization/84114
8897 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8898 Avoid reassociation of FLOAT_MODE addition.
8900 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8902 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8903 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8904 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8905 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8906 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8907 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8908 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8910 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8911 __builtin_ia32_wbinvd): New builtins.
8912 (SPECIAL_ARGS2): New.
8913 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8914 (SPECIAL_ARGS2): New.
8915 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8916 (ix86_valid_target_attribute_inner_p): Ditto.
8917 (ix86_init_mmx_sse_builtins): Add special_args2.
8918 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8919 TARGET_WBNOINVD_P): New.
8920 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8921 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8922 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8923 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8924 * config/i386/pconfigintrin.h: New file.
8925 * config/i386/wbnoinvdintrin.h: Ditto.
8926 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8928 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8930 2018-03-05 Richard Biener <rguenther@suse.de>
8932 PR tree-optimization/84670
8933 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8935 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8936 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8937 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8938 doesn't grow if all successors (recursively) were visited at least
8941 2018-03-05 Richard Biener <rguenther@suse.de>
8943 PR tree-optimization/84650
8944 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8945 if executed in the loop pipeline.
8947 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8949 * doc/configfiles.texi (Configuration Files): Move info about
8950 conditionalizing $target-protos.h to...
8951 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8952 differs from $target-protos.h.
8954 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8955 Chung-Ju Wu <jasonwucj@gmail.com>
8957 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8958 * config/nds32/nds32-multiple.md (setmemsi): Define.
8959 * config/nds32/nds32-memory-manipulation.c
8960 (nds32_gen_dup_4_byte_to_word_value): New.
8961 (emit_setmem_word_loop): New.
8962 (emit_setmem_byte_loop): New.
8963 (nds32_expand_setmem_loop): New.
8964 (nds32_expand_setmem_loop_v3m): New.
8965 (nds32_expand_setmem_unroll): New.
8966 (nds32_expand_setmem): New.
8968 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8969 Chung-Ju Wu <jasonwucj@gmail.com>
8971 * config/nds32/nds32-memory-manipulation.c
8972 (nds32_emit_load_store): New.
8973 (nds32_emit_post_inc_load_store): New.
8974 (nds32_emit_mem_move): New.
8975 (nds32_emit_mem_move_block): New.
8976 (nds32_expand_movmemsi_loop_unknown_size): New.
8977 (nds32_expand_movmemsi_loop_known_size): New.
8978 (nds32_expand_movmemsi_loop): New.
8979 (nds32_expand_movmemsi_unroll): New.
8980 (nds32_expand_movmemqi): Rename ...
8981 (nds32_expand_movmemsi): ... to this.
8982 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8983 (movmemsi): ... to this.
8984 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8985 (nds32_expand_movmemsi): ... to this.
8987 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8988 Monk Chiang <sh.chiang04@gmail.com>
8989 Chung-Ju Wu <jasonwucj@gmail.com>
8991 * config/nds32/nds32-protos.h
8992 (nds32_expand_load_multiple): New arguments.
8993 (nds32_expand_store_multiple): Ditto.
8994 (nds32_valid_multiple_load_store): Rename ...
8995 (nds32_valid_multiple_load_store_p): ... to this.
8996 * config/nds32/nds32-memory-manipulation.c
8997 (nds32_expand_load_multiple): Refine implementation.
8998 (nds32_expand_store_multiple): Ditto.
8999 * config/nds32/nds32-multiple.md
9000 (load_multiple): Update nds32_expand_load_multiple interface.
9001 (store_multiple): Update nds32_expand_store_multiple interface.
9002 * config/nds32/nds32-predicates.c
9003 (nds32_valid_multiple_load_store): Rename ...
9004 (nds32_valid_multiple_load_store_p): ... to this and refine
9006 * config/nds32/predicates.md
9007 (nds32_load_multiple_and_update_address_operation): New predicate.
9008 (nds32_store_multiple_and_update_address_operation): New predicate.
9010 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9011 Chung-Ju Wu <jasonwucj@gmail.com>
9013 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9014 (combo): New attribute.
9015 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9017 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9019 * config/nds32/nds32.opt: Change -mcmodel= default value.
9021 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9022 Monk Chiang <sh.chiang04@gmail.com>
9023 Chung-Ju Wu <jasonwucj@gmail.com>
9025 * config/nds32/constants.md (unspec_element): New enum.
9026 * config/nds32/constraints.md (Umw): New constraint.
9027 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9028 * config/nds32/nds32-intrinsic.md: Likewise.
9029 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9030 (nds32_valid_smw_lwm_base_p): New.
9031 (nds32_output_smw_single_word): New.
9032 (nds32_output_lmw_single_word): New.
9033 (nds32_expand_unaligned_load): New.
9034 (nds32_expand_unaligned_store): New.
9035 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9036 (nds32_output_smw_single_word): Declare.
9037 (nds32_output_lmw_single_word): Declare.
9038 (nds32_expand_unaligned_load): Declare.
9039 (nds32_expand_unaligned_store): Declare.
9040 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9041 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9042 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9043 NDS32_BUILTIN_UASTORE_DW.
9044 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9047 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9048 Kito Cheng <kito.cheng@gmail.com>
9049 Chung-Ju Wu <jasonwucj@gmail.com>
9051 * config/nds32/nds32-intrinsic.c
9052 (nds32_expand_builtin_null_ftype_reg): Delete.
9053 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9054 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9055 (nds32_read_argument): New.
9056 (nds32_legitimize_target): Ditto.
9057 (nds32_legitimize_argument): Ditto.
9058 (nds32_check_constant_argument): Ditto.
9059 (nds32_expand_unop_builtin): Ditto.
9060 (nds32_expand_unopimm_builtin): Ditto.
9061 (nds32_expand_binop_builtin): Ditto.
9062 (nds32_builtin_decl_impl): Ditto.
9063 (builtin_description): Ditto.
9064 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9065 (nds32_init_builtins_impl): Ditto.
9066 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9067 (nds32_builtin_decl): New.
9068 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9069 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9071 2018-03-02 Jeff Law <law@redhat.com>
9073 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9074 (redundant_insn, fill_simple_delay_slots): Likewise.
9075 (fill_slots_from_thread): Likewise.
9076 * resource.c (mark_referenced_resources): Likewise.
9077 (mark_set_resources, find_dead_or_set_registers): Likewise.
9079 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9081 * substring-locations.h (format_warning_va): Formatting fix for
9083 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9085 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9086 * substring-locations.c: Include intl.h.
9087 (format_warning_va): Turned into small wrapper around
9088 format_warning_n_va, renamed to ...
9089 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9090 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9092 (format_warning_at_substring_n): New function.
9093 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9094 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9095 format_warning_at_substring with just a shorter name instead of
9096 const function pointer.
9097 (fmtwarn_n): New function.
9098 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9099 appropriate, get rid of all the fmtstr temporaries, move conditionals
9100 with G_() wrapped string literals directly into fmtwarn arguments,
9101 cast dir.len to (int), formatting fixes.
9103 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9105 * doc/invoke.texi: Remove "Cilk Plus" references.
9107 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9108 Richard Biener <rguenther@suse.de>
9111 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9112 for error or warning attributes if CALL_FROM_THUNK_P is set.
9115 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9118 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9119 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9122 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9123 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9125 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9126 instead of -1U in last predictors element's probability member.
9128 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9131 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9132 arguments if they are comparable.
9134 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9136 PR tree-optimization/84634
9137 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9138 masks and masked_loop_p with a single loop_masks, making sure it's
9139 null for bb vectorization.
9141 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9143 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9144 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9145 loop->force_vectorize to check whether there is no alias.
9147 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9150 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9152 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9154 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9155 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9156 instead of a loop around prev_real_insn.
9157 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9161 * config/i386/i386.c (ix86_print_operand): Use conditional
9162 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9165 2018-03-02 Richard Biener <rguenther@suse.de>
9167 PR tree-optimization/84427
9168 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9169 (bitmap_set_subtract_values): Rewrite to handle multiple
9172 (prune_clobbered_mems): Likewise.
9173 (phi_translate): Take edge instead of pred/phiblock.
9174 (phi_translate_1): Likewise.
9175 (phi_translate_set): Likewise. Insert all translated
9176 exprs for a value into the set, keeping possibly multiple
9177 expressions per value.
9178 (compute_antic_aux): Adjust for phi_translate changes.
9179 When intersecting union the expressions and prune those
9180 not in the final value set, keeping possibly multiple
9181 expressions per value. Do not use value-insertion
9182 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9183 all expressions. Add verification that the value-sets
9184 only shrink during iteration.
9185 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9186 (do_pre_regular_insertion): Likewise.
9187 (do_pre_partial_partial_insertion): Likewise.
9189 2018-03-02 Richard Biener <rguenther@suse.de>
9192 * config/darwin.c (saved_debug_info_level): New static global.
9193 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9194 (darwin_asm_lto_end): Restore debug info generation settings.
9196 2018-03-01 Martin Liska <mliska@suse.cz>
9199 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9202 2018-03-01 Richard Biener <rguenther@suse.de>
9205 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9206 type annotation with LTO when debug was disabled at compile-time.
9208 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9210 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9212 (mips_final_postscan_insn): Likewise.
9214 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9216 PR rtl-optimization/84528
9217 * alias.c (init_alias_target): Add commentary.
9218 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9219 a unique base value if the frame pointer is not eliminated
9220 to the stack pointer.
9222 2018-03-01 Tom de Vries <tom@codesourcery.com>
9224 PR rtl-optimization/83327
9225 * lra-int.h (hard_regs_spilled_into): Declare.
9226 * lra.c (hard_regs_spilled_into): Define.
9227 (init_reg_info): Init hard_regs_spilled_into.
9228 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9229 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9230 (process_bb_lives): Handle hard_regs_spilled_into.
9231 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9232 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9234 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9236 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9237 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9238 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9239 * config/rs6000/aix72.h: New file.
9241 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9243 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9244 instead of warning_at with conditional singular and plural messages
9248 * stor-layout.c (update_alignment_for_field): For
9249 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9250 && !DECL_PACKED (field), do the alignment update, just use
9251 only desired_align instead of MAX (type_align, desired_align)
9253 (place_field): Don't do known_align < desired_align handling
9254 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9255 is non-NULL, instead do it after rli->prev_field handling and
9256 only if not within a bitfield word. For DECL_PACKED (field)
9257 use type_align of BITS_PER_UNIT.
9259 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9261 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9262 superfluous parentheses and trailing spaces.
9264 2018-02-28 Richard Biener <rguenther@suse.de>
9266 PR tree-optimization/84584
9267 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9268 SCoPs with fake exit edge.
9270 2018-02-28 Martin Liska <mliska@suse.cz>
9273 * timevar.c (timer::print): Fix format to properly print 100%
9276 2018-02-28 Richard Biener <rguenther@suse.de>
9279 * genmatch.c (capture_info::walk_match): Do not mark
9280 captured expressions without operands as expr_p given
9281 they act more like predicates and should be subject to
9282 "lost tail" side-effect preserving.
9284 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9286 PR rtl-optimization/81611
9287 * auto-inc-dec.c (attempt_change): Move dead note from
9288 mem_insn if it's the next use of regno
9289 (find_address): Take address use of reg holding
9290 non-incremented value. Add parm to limit search to the named
9292 (merge_in_block): Attempt to use a mem insn that is the next
9293 use of the original regno.
9295 2018-02-27 Martin Sebor <msebor@redhat.com>
9298 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9299 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9301 2018-02-27 Martin Sebor <msebor@redhat.com>
9303 PR translation/84207
9304 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9305 n argument to unsigned HOST_WIDE_INT.
9306 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9307 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9308 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9309 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9311 2018-02-27 Richard Biener <rguenther@suse.de>
9313 PR tree-optimization/84512
9314 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9315 Do not use the estimate returned from record_stmt_cost for
9316 the scalar iteration cost but sum properly using add_stmt_cost.
9318 2018-02-27 Richard Biener <rguenther@suse.de>
9320 PR tree-optimization/84466
9321 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9322 Adjust last change to less strictly validate use operands.
9324 2018-02-27 Martin Liska <mliska@suse.cz>
9326 PR gcov-profile/84548
9327 * gcov.c (process_file): Allow partial overlap and consider it
9328 also as group functions.
9329 (output_lines): Properly calculate range of lines for a group.
9331 2018-02-27 Martin Liska <mliska@suse.cz>
9333 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9334 'ggc' suffixes. Change first column width.
9335 (timer::print): Fix formatting of the column.
9337 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9339 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9340 preserve inline entry blocks for the sake of debug inline
9341 entry point markers alone.
9342 (remove_unused_locals): Suggest in comments a better place to
9343 force the preservation of inline entry blocks that are
9344 otherwise unused, but do not preserve them.
9346 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9348 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9350 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9353 * config/i386/constraints.md (Bs): Replace
9354 ix86_indirect_branch_register with
9355 TARGET_INDIRECT_BRANCH_REGISTER.
9357 * config/i386/i386.md (indirect_jump): Likewise.
9358 (tablejump): Likewise.
9359 (*sibcall_memory): Likewise.
9360 (*sibcall_value_memory): Likewise.
9361 Peepholes of indirect call and jump via memory: Likewise.
9362 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9363 (*sibcall_value_GOT_32): Likewise.
9364 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9365 (GOT_memory_operand): Likewise.
9366 (call_insn_operand): Likewise.
9367 (sibcall_insn_operand): Likewise.
9368 (GOT32_symbol_operand): Likewise.
9369 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9371 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9373 PR rtl-optimization/83496
9374 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9375 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9376 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9377 redundant insn, if any.
9378 (relax_delay_slots): Likewise.
9379 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9381 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9383 PR tree-optimization/83965
9384 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9385 that grouped statements are part of a reduction chain. Return
9386 true if the statement is not marked as a reduction itself but
9388 (vect_recog_dot_prod_pattern): Don't check whether the statement
9389 is part of a group here.
9390 (vect_recog_sad_pattern): Likewise.
9391 (vect_recog_widen_sum_pattern): Likewise.
9393 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9396 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9398 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9401 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9403 (ix86_output_indirect_function_return): New prototype.
9404 (ix86_split_simple_return_pop_internal): Likewise.
9405 * config/i386/i386.c (indirect_return_via_cx): New.
9406 (indirect_return_via_cx_bnd): Likewise.
9407 (indirect_thunk_name): Handle return va CX_REG.
9408 (output_indirect_thunk_function): Create alias for
9409 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9410 (ix86_output_indirect_jmp): Remove the bool argument.
9411 (ix86_output_indirect_function_return): New function.
9412 (ix86_split_simple_return_pop_internal): Likewise.
9413 * config/i386/i386.md (*indirect_jump): Don't pass false
9414 to ix86_output_indirect_jmp.
9415 (*tablejump_1): Likewise.
9416 (simple_return_pop_internal): Change it to define_insn_and_split.
9417 Call ix86_split_simple_return_pop_internal to split it for
9419 (simple_return_indirect_internal): Call
9420 ix86_output_indirect_function_return instead of
9421 ix86_output_indirect_jmp.
9423 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9426 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9427 memset and value initialization afterwards.
9429 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9431 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9433 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9436 * common/config/aarch64/aarch64-common.c
9437 (aarch_option_optimization_table[]): Switch
9438 off fomit-frame-pointer
9440 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9441 Chung-Ju Wu <jasonwucj@gmail.com>
9443 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9445 (store_multiple): Ditto.
9447 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9449 * config.gcc: Add --with-cpu support for nds32 target.
9450 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9451 * config/nds32/nds32.opt: Add -mcpu= option.
9453 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9455 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9456 isel=yes): Warn for these deprecated options.
9458 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9460 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9461 ISA_2_5_MASKS_EMBEDDED.
9463 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9465 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9466 p->max as pointers rather than using iterative_hash_expr.
9468 2018-02-23 Carl Love <cel@us.ibm.com>
9470 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9471 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9473 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9474 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9475 P8V_BUILTIN_VEC_VUNSIGNED2.
9477 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9480 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9481 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9483 * lra-constraints.c (curr_insn_transform): Set up
9484 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9486 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9487 * lra-eliminations.c (spill_pseudos): Ditto.
9488 (process_insn_for_elimination): Ditto.
9489 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9490 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9491 LRA_NON_CLOBBERED_ALT.
9493 2018-02-22 Martin Sebor <msebor@redhat.com>
9495 PR tree-optimization/84480
9496 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9497 to maybe_diag_stxncpy_trunc. Call it.
9498 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9499 from gimple_fold_builtin_strcpy. Print inlining stack.
9500 (handle_builtin_stxncpy): Print inlining stack.
9501 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9503 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9506 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9507 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9508 and -fcheck-pointer-bounds are used together.
9509 (indirect_thunk_prefix): New enum.
9510 (indirect_thunk_need_prefix): New function.
9511 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9512 "_nt" instead of "_bnd" for NOTRACK prefix.
9513 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9514 (output_indirect_thunk_function): Likewise.
9516 (ix86_code_end): Update output_indirect_thunk_function calls.
9517 (ix86_output_indirect_branch_via_reg): Replace
9518 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9519 (ix86_output_indirect_branch_via_push): Likewise.
9520 (ix86_output_function_return): Likewise.
9521 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9522 incompatible with -fcf-protection=branch and
9523 -fcheck-pointer-bounds.
9525 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9528 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9529 Change gcc_assert call to output_operand_lossage.
9531 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9533 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9535 2018-02-22 DJ Delorie <dj@redhat.com>
9536 Sebastian Perta <sebastian.perta@renesas.com>
9537 Oleg Endo <olegendo@gcc.gnu.org>
9539 * config/rx/rx.c (rx_rtx_costs): New function.
9540 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9542 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9544 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9546 2018-02-22 Martin Liska <mliska@suse.cz>
9549 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9550 Add "native" as a possible value.
9552 2018-02-22 Martin Liska <mliska@suse.cz>
9555 * config/i386/i386.c (ix86_option_override_internal):
9556 Add "native" as a possible value for -march and -mtune.
9558 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9561 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9562 to all type variants.
9564 PR tree-optimization/84503
9565 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9566 width as info->bitpos + info->bitsize - start.
9567 (merged_store_group::merge_overlapping): Simplify width computation.
9568 (check_no_overlap): New function.
9569 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9570 start + width and last_order of the group, fail if check_no_overlap
9572 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9573 to group if check_no_overlap fails.
9575 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9577 * config/rs6000/altivec.md: Delete contraint arguments to
9578 define_expand, define_split, and define_peephole2, and in
9579 define_insn_and_split if always unused.
9580 * config/rs6000/darwin.md: Ditto.
9581 * config/rs6000/dfp.md: Ditto.
9582 * config/rs6000/rs6000.md: Ditto.
9583 * config/rs6000/sync.md: Ditto.
9584 * config/rs6000/vector.md: Ditto.
9585 * config/rs6000/vsx.md: Ditto.
9587 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9589 * config/rs6000/altivec.md: Write output control strings as braced
9590 blocks instead of double-quoted strings.
9591 * config/rs6000/darwin.md: Ditto.
9592 * config/rs6000/rs6000.md: Ditto.
9593 * config/rs6000/vector.md: Ditto.
9594 * config/rs6000/vsx.md: Ditto.
9596 2018-02-21 Jason Merrill <jason@redhat.com>
9598 PR c++/84314 - ICE with templates and fastcall attribute.
9599 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9601 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9603 * ipa-cp.c (determine_versionability): Fix comment typos.
9605 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9608 * ipa-cp.c (determine_versionability): Do not version functions caling
9611 2018-02-21 Martin Liska <mliska@suse.cz>
9614 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9615 Add "native" as a possible value.
9616 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9617 the macro when native cpu detection is available.
9619 2018-02-21 Martin Liska <mliska@suse.cz>
9622 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9623 Add "native" as a possible value.
9624 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9625 when native cpu detection is available.
9627 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9628 Martin Sebor <msebor@redhat.com>
9630 PR tree-optimization/84478
9631 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9633 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9634 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9635 support which is conservatively correct, for 2 only stay conservative
9636 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9637 argument to the 2 argument get_range_strlen, adjust 6 arg
9638 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9640 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9641 (gimple_fold_builtin_strlen): Pass true as last argument to
9644 2018-02-20 Martin Sebor <msebor@redhat.com>
9647 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9648 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9649 (builtin_memref::builtin_memref): Factor out parts into
9650 set_base_and_offset and call it.
9652 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9655 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9656 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9657 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9658 search at the associated MODE_INT.
9660 2018-02-20 Jeff Law <law@redhat.com>
9663 PR tree-optimization/81592
9665 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9666 for range data rather than using global data.
9667 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9668 range data rather than using global data.
9669 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9670 pass it to children as needed.
9671 (struct directive::fmtresult): Similarly.
9672 (struct directive::set_width): Similarly.
9673 (struct directive::set_precision): Similarly.
9674 (format_integer, format_directive, parse_directive): Similarly.
9675 (format_none): Accept unnamed vr_values parameter.
9676 (format_percent, format_floating, format_character): Similarly.
9677 (format_string, format_plain): Similarly.
9678 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9679 the EVRP range analyzer for range data rather than using global data.
9680 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9681 gimple-ssa-evrp-analyze.h
9682 (class sprintf_dom_walker): Add after_dom_children member function.
9683 Add evrp_range_analyzer member.
9684 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9685 range analyzer as needed.
9686 (sprintf_dom_walker::after_dom_children): New member function.
9687 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9689 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9690 (evrp_range_analyzer::pop_to_marker): Likewise.
9692 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9694 PR tree-optimization/84419
9695 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9696 with the required type if its current type is compatible but
9699 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9702 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9703 after vectorization.
9705 2018-02-20 Martin Liska <mliska@suse.cz>
9708 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9709 possible values if we don't have a hint.
9711 2018-02-20 Martin Liska <mliska@suse.cz>
9715 * final.c (shorten_branches): Build align_tab array with one
9717 * opts.c (finish_options): Add alignment option limit check.
9718 (MAX_CODE_ALIGN): Likewise.
9719 (MAX_CODE_ALIGN_VALUE): Likewise.
9720 * doc/invoke.texi: Document maximum allowed option value for
9721 all -falign-* options.
9723 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9726 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9727 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9728 * var-tracking.c (emit_note_insn_var_location): Remove all references
9729 to NOTE_INSN_CALL_ARG_LOCATION.
9730 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9731 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9732 Use copy_rtx_if_shared.
9733 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9734 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9735 (dwarf2out_var_location): Remove handling of
9736 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9738 * final.c (final_scan_insn): Remove all references to
9739 NOTE_INSN_CALL_ARG_LOCATION.
9740 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9741 before dumping final insns.
9742 * except.c (emit_note_eh_region_end): Remove all references to
9743 NOTE_INSN_CALL_ARG_LOCATION.
9744 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9745 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9746 * config/arc/arc.c (hwloop_optimize): Likewise.
9747 * config/arm/arm.c (create_fix_barrier): Likewise.
9748 * config/s390/s390.c (s390_chunkify_start): Likewise.
9749 * config/sh/sh.c (find_barrier): Likewise.
9750 * config/i386/i386.c (rest_of_insert_endbranch,
9751 ix86_seh_fixup_eh_fallthru): Likewise.
9752 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9753 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9754 * config/frv/frv.c (frv_function_prologue): Likewise.
9755 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9757 (note_outside_basic_block_p): Remove all references to
9758 NOTE_INSN_CALL_ARG_LOCATION.
9759 * gengtype.c (adjust_field_rtx_def): Likewise.
9760 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9762 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9763 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9766 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9769 PR tree-optimization/84452
9770 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9771 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9774 2018-02-19 Martin Liska <mliska@suse.cz>
9777 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9779 2018-02-19 Martin Liska <mliska@suse.cz>
9780 Richard Sandiford <richard.sandiford@linaro.org>
9782 PR tree-optimization/82491
9783 * gimple-fold.c (get_base_constructor): Make earlier bail out
9786 2018-02-19 Carl Love <cel@us.ibm.com>
9788 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9789 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9791 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9792 P8V_BUILTIN_VEC_NEG.
9794 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9796 * config/rl78/rl78.md (movdf): New define expand.
9798 2018-02-19 Martin Liska <mliska@suse.cz>
9801 * doc/invoke.texi: Fix typo.
9802 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9804 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9806 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9807 handle rs6000_single_float and rs6000_double_float specially for
9810 2018-02-16 Jeff Law <law@redhat.com>
9812 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9815 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9818 * ipa-inline.c (inline_small_functions): Fix a typo.
9820 2018-02-16 Nathan Sidwell <nathan@acm.org>
9822 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9824 2018-02-16 Carl Love <cel@us.ibm.com>
9826 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9827 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9828 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9829 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9830 expansion to P8V_BUILTIN_VEC_FLOAT2.
9832 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9834 PR rtl-optimization/70023
9835 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9836 src_regno into account.
9838 2018-02-16 Carl Love <cel@us.ibm.com>
9840 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9841 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9842 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9843 * config/rs6000/rs6000.c: Remove case statements for
9844 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9845 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9846 and P9V_BUILTIN_VEC_VINSERT4B.
9847 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9848 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9849 * config/rs6000/vsx.md:
9850 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9853 2018-02-16 Carl Love <cel@us.ibm.com>
9855 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9857 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9859 * config/rs6000/rs6000-c.c: Add the definitions for
9860 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9861 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9862 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9863 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9864 definition for insert4b and define insn *insert3b_internal.
9865 * doc/extend.texi: Add documentation for vec_extract4b.
9867 2018-02-16 Nathan Sidwell <nathan@acm.org>
9869 * doc/extend.texi (Backwards Compatibility): Mention friend
9870 injection. Note for-scope is deprecated.
9871 * doc/invoke.texi (-ffriend-injection): Deprecate.
9873 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9875 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9876 that moved to I2, also allow destinations that are a paradoxical
9877 subreg (instead of a normal reg).
9879 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9882 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9885 2018-02-16 Richard Biener <rguenther@suse.de>
9887 PR tree-optimization/84037
9888 PR tree-optimization/84016
9890 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9891 Adjust vec_construct for the fact we need additional higher latency
9892 128bit inserts for AVX256 and AVX512 vector builds.
9893 (ix86_add_stmt_cost): Scale vector construction cost for
9896 2018-02-16 Richard Biener <rguenther@suse.de>
9898 PR tree-optimization/84417
9899 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9900 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9901 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9903 2018-02-16 Martin Liska <mliska@suse.cz>
9906 * internal-fn.def (ASAN_CHECK): Set proper flags.
9907 (ASAN_MARK): Likewise.
9909 2018-02-16 Julia Koval <julia.koval@intel.com>
9911 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9912 from PTA_CANNONLAKE.
9914 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9917 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9918 Use ++iter rather than iter++ for std::list iterators.
9919 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9920 defer deleting them until all nodes in the forest are processed. Do
9921 free even leaf nodes. Change to_process into auto_vec.
9924 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9925 * vec.h (vec_default_construct): Use memset instead of placement new
9926 if BROKEN_VALUE_INITIALIZATION is defined.
9927 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9928 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9931 PR rtl-optimization/83723
9932 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9933 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9934 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9936 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9938 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9940 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9942 PR rtl-optimization/81443
9943 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9944 from inner REGs to paradoxical SUBREGs.
9946 2018-02-16 Richard Biener <rguenther@suse.de>
9948 PR tree-optimization/84399
9949 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9950 For operands we can analyze at their definition make sure we can
9951 analyze them at each use as well.
9953 2018-02-16 Richard Biener <rguenther@suse.de>
9955 PR tree-optimization/84190
9956 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9957 volatile accesses if the decl isn't volatile.
9959 2018-02-15 Jason Merrill <jason@redhat.com>
9961 PR c++/84314 - ICE with templates and fastcall attribute.
9962 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9963 TYPE_CANONICAL on an existing type.
9965 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9967 PR tree-optimization/84383
9968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9969 dstoff nor call operand_equal_p if dstbase is NULL.
9971 PR tree-optimization/84334
9972 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9973 also a CONSTANT_CLASS_P, punt.
9975 2018-02-14 Jim Wilson <jimw@sifive.com>
9977 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9978 first SMALL_OPERAND check. New local min_second_step. Move assert
9979 to where locals are set. Add TARGET_RVC support.
9980 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9982 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9984 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9986 2018-02-14 Martin Sebor <msebor@redhat.com>
9988 PR tree-optimization/83698
9989 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9990 arrays constrain the offset range to their bounds.
9991 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9992 (builtin_access::overlap): Avoid setting the size of overlap if it's
9994 (maybe_diag_overlap): Also consider arrays when deciding what values
9995 of offsets to include in diagnostics.
9997 2018-02-14 Martin Sebor <msebor@redhat.com>
10000 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10001 that correspond to the kind of a declaration.
10003 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10006 * config/pa/pa.md: Load address of PIC label using the linkage table
10007 if the label is nonlocal.
10009 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10011 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10012 warning message if user requests -maltivec=be.
10013 * doc/invoke.texi: Document deprecation of -maltivec=be.
10015 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10018 * config/rs6000/rs6000-c.c: Update definitions for
10019 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10020 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10022 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10025 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10026 add _get_ssp intrinsics. Remove argument from
10027 __builtin_ia32_rdssp[d|q].
10028 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10029 * config/i386/i386-builtin.def: Remove argument from
10030 __builtin_ia32_rdssp[d|q].
10031 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10032 ix86_expand_special_args_builtin for _rdssp[d|q].
10033 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10034 Clear register before usage.
10035 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10036 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10038 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10040 PR tree-optimization/84357
10041 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10042 operand 1 of an ARRAY_REF too.
10044 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10047 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10048 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10050 (set_of_reg): New struct.
10051 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10052 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10053 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10055 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10056 Split into bitclr, bitset, bitinvert patterns if appropriate.
10057 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10058 use rx_fuse_in_memory_bitop.
10059 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10060 to named insn, correct maximum insn length.
10062 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10065 * machmode.def: Define a complex mode for PARTIAL_INT.
10066 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10068 * doc/rtl.texi: Document CSPImode.
10069 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10071 (msp430_hard_regno_nregs_with_padding): Likewise.
10073 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10076 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10078 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10080 PR rtl-optimization/84169
10081 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10082 we generated a parallel as new i3 and we split that to new i2 and i3
10083 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10084 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10085 those to i2, not i1. Partially rewrite this scan code.
10087 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10090 * stor-layout.c (place_field): For variable length fields, adjust
10091 offset_align afterwards not just based on the field's alignment,
10092 but also on the size.
10094 PR middle-end/84309
10095 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10096 of exps and logs in the use_exp2 case.
10098 2018-02-13 Jeff Law <law@redhat.com>
10100 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10101 entry for "vector".
10103 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10106 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10110 * common.opt (gas-loc-support, gas-locview-support): New.
10111 (ginline-points, ginternal-reset-location-views): New.
10112 * doc/invoke.texi: Document them. Use @itemx where intended.
10113 (gvariable-location-views): Adjust.
10114 * target.def (reset_location_view): New.
10115 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10116 (TARGET_RESET_LOCATION_VIEW): New.
10117 * doc/tm.texi: Rebuilt.
10118 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10119 (dwarf2out_default_as_locview_support): New.
10120 (output_asm_line_debug_info): Use option variables.
10121 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10122 (output_loc_list): Likewise.
10123 (add_high_low_attributes): Check option variables.
10124 Don't output entry view attribute in strict mode.
10125 (gen_inlined_subroutine_die): Check option variables.
10126 (dwarf2out_inline_entry): Likewise.
10127 (init_sections_and_labels): Likewise.
10128 (dwarf2out_early_finish): Likewise.
10129 (maybe_reset_location_view): New, from...
10130 (dwarf2out_var_location): ... here. Call it.
10131 * debug.h (dwarf2out_default_as_loc_support): Declare.
10132 (dwarf2out_default_as_locview_support): Declare.
10133 * hooks.c (hook_int_rtx_insn_0): New.
10134 * hooks.h (hook_int_rtx_insn_0): Declare.
10135 * toplev.c (process_options): Take -gas-loc-support and
10136 -gas-locview-support from dwarf2out. Enable
10137 -gvariable-location-views by default only with locview
10138 assembler support. Enable -ginternal-reset-location-views by
10139 default only if the target defines the corresponding hook.
10140 Enable -ginline-points by default if location views are
10141 enabled; force it disabled if statement frontiers are
10143 * tree-inline.c (expand_call_inline): Check option variables.
10144 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10146 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10148 PR tree-optimization/84321
10149 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10150 handling. Also check whether the anti-range contains any values
10151 that satisfy the mask; switch to a VR_RANGE if not.
10153 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10156 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10158 2018-02-13 Martin Jambor <mjambor@suse.cz>
10161 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10162 of call statements, also set location of a load to a temporary.
10164 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10166 * config/rl78/rl78.c (add_vector_labels): New function.
10167 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10168 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10169 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10170 which checks that no arguments are passed.
10171 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10172 * doc/extend.texi: Documentation for the new attribute.
10174 2018-02-13 Andreas Schwab <schwab@suse.de>
10176 * config/riscv/linux.h (CPP_SPEC): Define.
10178 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10181 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10182 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10183 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10184 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10185 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10186 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10187 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10188 temporarily for AES and PCLMUL builtins.
10190 PR tree-optimization/84339
10191 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10192 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10195 PR middle-end/84309
10196 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10197 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10198 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10200 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10202 * omp-simd-clone.h: New file.
10203 * omp-simd-clone.c: Include omp-simd-clone.h.
10204 (expand_simd_clones): No longer static.
10205 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10206 cgraph.h and omp-simd-clone.h.
10207 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10208 (vect_recog_widen_shift_pattern): Formatting fix.
10209 (vect_pattern_recog_1): Don't check optab for calls.
10212 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10213 operands[2] into a REG before using gen_lowpart on it.
10215 2018-02-12 Jeff Law <law@redhat.com>
10218 * config/sh/sh.c (find_barrier): Consider a sibling call
10221 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10222 successfully back substituting a reg.
10224 2018-02-12 Richard Biener <rguenther@suse.de>
10226 PR tree-optimization/84037
10227 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10228 parameter, move visited init to caller.
10229 (vect_slp_analyze_operations): Separate cost from validity
10230 check, initialize visited once for all instances.
10231 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10233 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10235 (vect_model_promotion_demotion_cost): Likewise.
10236 (vectorizable_bswap): Guard cost modeling with !slp_node
10237 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10239 (vectorizable_call): Likewise.
10240 (vectorizable_conversion): Likewise.
10241 (vectorizable_assignment): Likewise.
10242 (vectorizable_shift): Likewise.
10243 (vectorizable_operation): Likewise.
10244 (vectorizable_store): Likewise.
10245 (vectorizable_load): Likewise.
10246 (vectorizable_condition): Likewise.
10247 (vectorizable_comparison): Likewise.
10249 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10252 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10253 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10254 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10256 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10258 PR middle-end/83665
10259 * params.def (inline-min-speedup): Increase from 8 to 15.
10260 (max-inline-insns-auto): Decrease from 40 to 30.
10261 * ipa-split.c (consider_split): Add some buffer for function to
10262 be considered inlining candidate.
10263 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10266 2018-02-12 Richard Biener <rguenther@suse.de>
10268 PR tree-optimization/84037
10269 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10270 matched stmts if we cannot swap the non-matched ones.
10272 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10274 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10275 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10276 _mm_maskz_scalef_round_ss): New intrinsics.
10277 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10278 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10279 __builtin_ia32_scalefss_round): Remove.
10280 (__builtin_ia32_scalefsd_mask_round,
10281 __builtin_ia32_scalefss_mask_round): New intrinsics.
10282 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10283 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10284 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10285 "<round_constraint>")): Changed to ...
10286 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10287 "<round_scalar_constraint>")): ... this.
10288 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10289 %0, %1, %2<round_op3>}"): Changed to ...
10290 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10291 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10292 %2<round_scalar_mask_op3>}"): ... this.
10293 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10295 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10297 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10298 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10299 (_mm_maskz_sqrt_round_ss): New intrinsics.
10300 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10301 (__builtin_ia32_sqrtsd_mask_round)
10302 (__builtin_ia32_sqrtss_mask_round): New builtins.
10303 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10304 (__builtin_ia32_sqrtss_round): Remove.
10305 (__builtin_ia32_sqrtsd_mask_round)
10306 (__builtin_ia32_sqrtss_mask_round): New builtins.
10307 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10308 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10309 ((match_operand:VF_128 1 "vector_operand"
10310 "xBm,<round_constraint>")): Changed to ...
10311 ((match_operand:VF_128 1 "vector_operand"
10312 "xBm,<round_scalar_constraint>")): ... this.
10313 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10314 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10315 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10316 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10317 %<iptr>1<round_scalar_mask_op3>}): ... this.
10318 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10319 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10321 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10324 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10325 Cast vec_cmpeq result to correct type.
10326 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10327 Cast vec_cmpgt result to correct type.
10329 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10331 * final.c (final_scan_insn_1): Renamed from...
10332 (final_scan_insn): ... this. New wrapper, to recover
10333 seen from the outermost call in recursive ones.
10334 * config/sparc/sparc.c (output_return): Drop seen from call.
10335 (output_sibcall): Likewise.
10336 * config/visium/visium.c (output_branch): Likewise.
10338 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10340 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10343 2018-02-10 Alan Modra <amodra@gmail.com>
10346 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10347 Specify LR as an input.
10349 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10352 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10353 remove_member_access_dummy_vars): New functions.
10354 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10355 lower_omp_1, execute_lower_omp): Use them.
10357 PR rtl-optimization/84308
10358 * shrink-wrap.c (spread_components): Release todo vector.
10360 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10362 PR rtl-optimization/57193
10363 * ira-color.c (struct allocno_color_data): Add member
10364 conflict_allocno_hard_prefs.
10365 (update_conflict_allocno_hard_prefs): New.
10366 (bucket_allocno_compare_func): Add a preference based on
10367 conflict_allocno_hard_prefs.
10368 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10369 (color_allocnos): Remove a dead code. Initiate
10370 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10372 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10375 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10376 constraint from =wa to wa. Avoid a subreg on the output operand,
10377 instead use a pseudo and subreg it in a move.
10378 (p9_xxbrd_<mode>): Changed to ...
10379 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10380 (p9_xxbrd_v2df): New expander.
10381 (p9_xxbrw_<mode>): Changed to ...
10382 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10383 (p9_xxbrw_v4sf): New expander.
10385 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10387 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10389 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10392 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10393 multiply in 32-bit mode.
10394 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10395 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10398 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10400 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10401 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10402 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10403 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10405 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10408 * dwarf2out.c (is_trivial_indirect_ref): New function.
10409 (dwarf2out_late_global_decl): Do not generate a location
10410 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10411 and that are not defined in the current unit.
10413 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10415 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10416 instead of a libcall for UNORDERED.
10418 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10421 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10422 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10424 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10427 * config/s390/s390.c (s390_set_current_function): Invoke
10428 s390_indirect_branch_settings also if fndecl didn't change.
10430 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10432 * config/rs6000/rs6000.md (blockage): Set length to zero.
10434 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10436 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10438 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10441 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10442 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10446 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10447 PARALLEL incoming that failed vt_get_decl_and_offset check.
10449 PR middle-end/84237
10450 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10451 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10453 (get_variable_section): For decls in named .bss* sections pass true as
10454 second argument to bss_initializer_p.
10456 2018-02-09 Marek Polacek <polacek@redhat.com>
10457 Jakub Jelinek <jakub@redhat.com>
10460 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10461 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10462 Sync some changes from cxx_fold_indirect_ref.
10464 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10466 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10468 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10469 (BLOCK_INLINE_ENTRY_LABEL): New.
10470 (dwarf2out_var_location): Disregard inline entry markers.
10471 (inline_entry_data): New struct.
10472 (inline_entry_data_hasher): New hashtable type.
10473 (inline_entry_data_hasher::hash): New.
10474 (inline_entry_data_hasher::equal): New.
10475 (inline_entry_data_table): New variable.
10476 (add_high_low_attributes): Add DW_AT_entry_pc and
10477 DW_AT_GNU_entry_view attributes if a pending entry is found
10478 in inline_entry_data_table. Add old entry_pc attribute only
10479 if debug nonbinding markers are disabled.
10480 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10481 markers are enabled.
10482 (block_within_block_p, dwarf2out_inline_entry): New.
10483 (dwarf2out_finish): Check that no entries remained in
10484 inline_entry_data_table.
10485 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10486 (final_scan_insn, notice_source_line): Likewise.
10487 (rest_of_clean_state): Skip inline entry markers.
10488 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10490 * gimple.c (gimple_build_debug_inline_entry): New.
10491 * gimple.h (enum gimple_debug_subcode): Add
10492 GIMPLE_DEBUG_INLINE_ENTRY.
10493 (gimple_build_debug_inline_entry): Declare.
10494 (gimple_debug_inline_entry_p): New.
10495 (gimple_debug_nonbind_marker_p): Adjust.
10496 * insn-notes.def (INLINE_ENTRY): New.
10497 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10498 inline entry marker notes.
10499 (print_insn): Likewise.
10500 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10501 (INSN_DEBUG_MARKER_KIND): Likewise.
10502 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10503 * tree-inline.c (expand_call_inline): Build and insert
10504 debug_inline_entry stmt.
10505 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10506 inline entry blocks early, if nonbind markers are enabled.
10507 (dump_scope_block): Dump fragment info.
10508 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10509 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10510 (gimple_build_debug_inline_entry): New.
10511 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10512 Enable/disable inline entry points too.
10513 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10514 (DEBUG_INSN): Describe inline entry markers.
10516 * common.opt (gvariable-location-views): New.
10517 (gvariable-location-views=incompat5): New.
10518 * config.in: Rebuilt.
10519 * configure: Rebuilt.
10520 * configure.ac: Test assembler for view support.
10521 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10522 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10523 * dwarf2out.c (var_loc_view): New typedef.
10524 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10525 (dwarf2out_locviews_in_attribute): New.
10526 (dwarf2out_locviews_in_loclist): New.
10527 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10528 (enum dw_line_info_opcode): Add LI_adv_address.
10529 (struct dw_line_info_table): Add view.
10530 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10531 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10532 (zero_view_p): New variable.
10533 (ZERO_VIEW_P): New macro.
10534 (output_asm_line_debug_info): New.
10535 (struct var_loc_node): Add view.
10536 (add_AT_view_list, AT_loc_list): New.
10537 (add_var_loc_to_decl): Add view param. Test it against last.
10538 (new_loc_list): Add view params. Record them.
10539 (AT_loc_list_ptr): Handle loc and view lists.
10540 (view_list_to_loc_list_val_node): New.
10541 (print_dw_val): Handle dw_val_class_view_list.
10542 (size_of_die): Likewise.
10543 (value_format): Likewise.
10544 (loc_list_has_views): New.
10545 (gen_llsym): Set vl_symbol too.
10546 (maybe_gen_llsym, skip_loc_list_entry): New.
10547 (dwarf2out_maybe_output_loclist_view_pair): New.
10548 (output_loc_list): Output view list or entries too.
10549 (output_view_list_offset): New.
10550 (output_die): Handle dw_val_class_view_list.
10551 (output_dwarf_version): New.
10552 (output_compilation_unit_header): Use it.
10553 (output_skeleton_debug_sections): Likewise.
10554 (output_rnglists, output_line_info): Likewise.
10555 (output_pubnames, output_aranges): Update version comments.
10556 (output_one_line_info_table): Output view numbers in asm comments.
10557 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10558 Call maybe_gen_llsym.
10559 (loc_list_from_tree_1): Adjust.
10560 (add_AT_location_description): Create view list attribute if
10561 needed, check it's absent otherwise.
10562 (convert_cfa_to_fb_loc_list): Adjust.
10563 (maybe_emit_file): Call output_asm_line_debug_info for test.
10564 (dwarf2out_var_location): Reset views as needed. Precompute
10565 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10566 attribute. Set view.
10567 (new_line_info_table): Reset next view.
10568 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10569 (dwarf2out_source_line): Likewise. Output view resets and labels to
10570 the assembler, or select appropriate line info opcodes.
10571 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10572 (optimize_string_length): Catch it. Adjust.
10573 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10574 dw_val_class_view_list, and remove it if no longer needed.
10575 (hash_loc_list): Hash view numbers.
10576 (loc_list_hasher::equal): Compare them.
10577 (optimize_location_lists): Check whether a view list symbol is
10578 needed, and whether the locview attribute is present, and
10579 whether they match. Remove the locview attribute if no longer
10581 (index_location_lists): Call skip_loc_list_entry for test.
10582 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10583 Use output_dwarf_version.
10584 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10585 (struct dw_val_node): Add val_view_list.
10586 * final.c (SEEN_NEXT_VIEW): New.
10587 (set_next_view_needed): New.
10588 (clear_next_view_needed): New.
10589 (maybe_output_next_view): New.
10590 (final_start_function): Rename to...
10591 (final_start_function_1): ... this. Take pointer to FIRST,
10592 add SEEN parameter. Emit param bindings in the initial view.
10593 (final_start_function): Reintroduce SEEN-less interface.
10594 (final): Rename to...
10595 (final_1): ... this. Take SEEN parameter. Output final pending
10596 next view at the end.
10597 (final): Reintroduce seen-less interface.
10598 (final_scan_insn): Output pending next view before switching
10599 sections or ending a block. Mark the next view as needed when
10600 outputting variable locations. Notify debug backend of section
10601 changes, and of location view changes.
10602 (rest_of_handle_final): Adjust.
10603 * toplev.c (process_options): Autodetect value for debug variable
10604 location views option. Warn on incompat5 without -gdwarf-5.
10605 * doc/invoke.texi (gvariable-location-views): New.
10606 (gvariable-location-views=incompat5): New.
10607 (gno-variable-location-views): New.
10609 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10611 PR tree-optimization/84136
10612 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10613 that the result of find_edge is non-NULL.
10615 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10618 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10619 storing integer register in SImode. Fix cost of 256 and 512
10620 byte aligned SSE register store.
10622 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10624 * config/i386/i386.c (ix86_multiplication_cost): Fix
10625 multiplication cost for TARGET_AVX512DQ.
10627 2018-02-08 Marek Polacek <polacek@redhat.com>
10629 PR tree-optimization/84238
10630 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10633 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10635 PR tree-optimization/84265
10636 * tree-vect-stmts.c (vectorizable_store): Don't treat
10637 VMAT_CONTIGUOUS accesses as grouped.
10638 (vectorizable_load): Likewise.
10640 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10642 PR tree-optimization/81635
10643 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10644 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10645 (test_round_for_mask): New functions.
10646 (wide_int_cc_tests): Call test_round_for_mask.
10647 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10648 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10649 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10650 range returned by get_range_info.
10652 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10655 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10656 * symtab.c: Include builtins.h
10657 (symtab_node::output_to_lto_symbol_table_p): Move here
10658 from lto-streamer-out.c:output_symbol_p.
10659 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10660 (output_symbol_p): Move all logic to symtab.c
10661 (produce_symtab): Update.
10663 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10665 * config/s390/s390-opts.h (enum indirect_branch): Define.
10666 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10667 (s390_indirect_branch_via_thunk)
10668 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10669 (enum s390_indirect_branch_type): Define.
10670 * config/s390/s390.c (struct s390_frame_layout, struct
10671 machine_function): Remove.
10672 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10673 (indirect_branch_table_label_no, indirect_branch_table_name):
10675 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10676 (enum s390_indirect_branch_option): Define.
10677 (s390_return_addr_from_memory): New function.
10678 (s390_handle_string_attribute): New function.
10679 (s390_attribute_table): Add new attribute handler.
10680 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10681 (s390_indirect_branch_via_thunk): New function.
10682 (s390_indirect_branch_via_inline_thunk): New function.
10683 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10684 sibling call optimization for non z10 compiles.
10685 (s390_emit_call): Force indirect branch target to be a single
10686 register. Add r1 clobber for non-z10 compiles.
10687 (s390_emit_epilogue): Emit return jump via return_use expander.
10688 (s390_reorg): Handle JUMP_INSNs as execute targets.
10689 (s390_option_override_internal): Perform validity checks for the
10690 new command line options.
10691 (s390_indirect_branch_attrvalue): New function.
10692 (s390_indirect_branch_settings): New function.
10693 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10694 (s390_output_indirect_thunk_function): New function.
10695 (s390_code_end): Implement target hook.
10696 (s390_case_values_threshold): Implement target hook.
10697 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10699 * config/s390/s390.h (struct s390_frame_layout)
10700 (struct machine_function): Move here from s390.c.
10701 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10702 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10703 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10704 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10705 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10706 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10707 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10708 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10709 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10710 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10711 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10712 (mnemonic attribute): Add values which aren't recognized
10714 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10715 pattern for branch conversion. Fix mnemonic attribute.
10716 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10717 indirect branch via thunk if requested.
10718 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10719 ("*indirect_jump"): Disable for branch conversion using out of
10721 ("indirect_jump_via_thunk<mode>_z10")
10722 ("indirect_jump_via_thunk<mode>")
10723 ("indirect_jump_via_inlinethunk<mode>_z10")
10724 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10725 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10726 ("casesi_jump_via_inlinethunk<mode>_z10")
10727 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10728 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10729 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10730 ("*indirect2_jump"): Disable for branch conversion.
10731 ("casesi_jump"): Turn into expander and expand patterns for branch
10733 ("return_use"): New expander.
10734 ("*return"): Emit return via thunk and rename it to ...
10735 ("*return<mode>"): ... this one.
10736 * config/s390/s390.opt: Add new options and and enum for the
10739 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10741 * lra-constraints.c (match_reload): Unconditionally use
10742 gen_lowpart_SUBREG, rather than selecting between that
10743 and equivalent gen_rtx_SUBREG code.
10745 2018-02-08 Richard Biener <rguenther@suse.de>
10747 PR tree-optimization/84233
10748 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10749 changed flag instead of boguously re-using phi_inserted.
10751 2018-02-08 Martin Jambor <mjambor@suse.cz>
10753 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10754 static local variables.
10756 2018-02-08 Richard Biener <rguenther@suse.de>
10758 PR tree-optimization/84278
10759 * tree-vect-stmts.c (vectorizable_store): When looking for
10760 smaller vector types to perform grouped strided loads/stores
10761 make sure the mode is supported by the target.
10762 (vectorizable_load): Likewise.
10764 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10766 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10767 Increase LDP/STP opportunities by adding adjacent callee-saves.
10769 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10771 PR rtl-optimization/84068
10772 PR rtl-optimization/83459
10773 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10775 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10777 PR tree-optimization/84224
10778 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10779 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10780 non-zero arguments.
10782 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10785 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10786 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10787 count, remove one USE.
10789 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10791 * doc/install.texi (Configuration): Document the
10792 --with-long-double-format={ibm,ieee} PowerPC configuration
10796 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10797 Convert from define_expand to be define_insn_and_split. Rework
10798 float/double/_Float128 conversions to QI/HI/SImode to work with
10799 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10800 conversions to QI/HImode types did a store and then a load to
10801 truncate the value. For conversions to VSX registers, don't split
10802 the insn, instead emit the code directly. Use the code iterator
10803 any_fix to combine signed and unsigned conversions.
10804 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10805 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10806 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10807 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10808 (fix_<mode>di2_hw): Likewise.
10809 (fixuns_<mode>di2_hw): Likewise.
10810 (fix_<mode>si2_hw): Likewise.
10811 (fixuns_<mode>si2_hw): Likewise.
10812 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10813 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10814 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10815 fix<uns>_trunc<SFDF:mode>si2_p8.
10816 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10817 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10818 (fix<uns>_<mode>_mem): Likewise.
10819 (fctiw<u>z_<mode>_mem): Likewise.
10820 (fix<uns>_<mode>_mem): Likewise.
10821 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10822 the register allocator from doing a direct move to the GPRs to do
10823 a store, and instead use the ISA 3.0 store byte/half-word from
10824 vector register instruction. For IEEE 128-bit floating point,
10825 also optimize stores of 32-bit ints.
10826 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10828 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10830 * genextract.c (push_pathstr_operand): New function to support
10832 (walk_rtx): Call push_pathstr_operand.
10833 (print_path): Support [a-zA-Z].
10835 2018-02-07 Richard Biener <rguenther@suse.de>
10837 PR tree-optimization/84037
10838 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10839 (cse_and_gimplify_to_preheader): Declare.
10840 (vect_get_place_in_interleaving_chain): Likewise.
10841 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10843 (_loop_vec_info::~_loop_vec_info): Delete it.
10844 (cse_and_gimplify_to_preheader): New function.
10845 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10846 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10847 (vectorizable_load): Likewise. For grouped stores always base
10848 the IV on the first element.
10849 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10850 condition before gimplifying.
10852 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10854 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10855 *DIV_EXPR and *MOD_EXPR.
10857 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10860 * config/i386/i386.c (ix86_option_override_internal): Mask out
10861 the CF_SET bit when checking -fcf-protection.
10863 2018-02-07 Tom de Vries <tom@codesourcery.com>
10866 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10869 2018-02-07 Richard Biener <rguenther@suse.de>
10871 PR tree-optimization/84204
10872 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10875 PR tree-optimization/84205
10876 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10877 special-case isl_ast_op_zdiv_r.
10879 PR tree-optimization/84223
10880 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10881 Only add conditions from within the region.
10882 (gather_bbs::after_dom_children): Adjust.
10884 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10887 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10888 * config/avr/avr.md: Only post-reload split REG-REG moves if
10889 either register is GENERAL_REG_P.
10891 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10893 PR tree-optimization/84235
10894 * tree-ssa-scopedtables.c
10895 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10896 if the subtraction is performed in floating point type where NaNs are
10897 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10898 build 1. Formatting fix.
10900 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10903 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10904 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10905 and skip it regardless of bb boundaries. Use CALL_P macro,
10906 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10907 unnecessarily, formatting fix.
10909 2018-02-06 Michael Collison <michael.collison@arm.com>
10911 * config/arm/thumb2.md:
10912 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10913 (*thumb_mov_notscc): Ditto.
10915 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10918 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10919 unsigned_fix, not "s".
10921 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10923 * configure.ac (gcc_fn_eh_frame_ro): New function.
10924 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10925 correct .eh_frame permissions.
10926 * configure: Regenerate.
10928 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10930 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10931 irrelevant options.
10933 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10935 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10936 Display warning message for -mno-speculate-indirect-jumps.
10938 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10940 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10942 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10944 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10946 PR tree-optimization/84225
10947 * tree-eh.c (find_trapping_overflow): Only call
10948 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10950 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10953 * config/i386/i386.c: Reimplement the check of possible options
10954 -mibt/-mshstk conbination. Change error messages.
10955 * doc/invoke.texi: Fix a typo: remove extra '='.
10957 2018-02-06 Marek Polacek <polacek@redhat.com>
10959 PR tree-optimization/84228
10960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10962 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10965 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10966 emitted arch directives.
10967 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10968 __ARM_FEATURE_COPROC before changing architectures.
10970 2018-02-06 Richard Biener <rguenther@suse.de>
10972 * config/i386/i386.c (print_reg): Fix typo.
10973 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10975 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10977 * configure: Regenerate.
10979 2018-02-05 Martin Sebor <msebor@redhat.com>
10981 PR tree-optimization/83369
10982 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10985 2018-02-05 Martin Liska <mliska@suse.cz>
10987 * doc/invoke.texi: Cherry-pick upstream r323995.
10989 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10991 * ira.c (ira_init_register_move_cost): Adjust comment.
10993 2018-02-05 Martin Liska <mliska@suse.cz>
10995 PR gcov-profile/84137
10996 * doc/gcov.texi: Fix typo in documentation.
10998 2018-02-05 Martin Liska <mliska@suse.cz>
11000 PR gcov-profile/83879
11001 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11002 using dlopen functionality.
11004 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11006 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11007 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11008 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11009 _mm_maskz_range_round_ss): New intrinsics.
11010 (__builtin_ia32_rangesd128_round)
11011 (__builtin_ia32_rangess128_round): Remove.
11012 (__builtin_ia32_rangesd128_mask_round,
11013 __builtin_ia32_rangess128_mask_round): New builtins.
11014 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11015 __builtin_ia32_rangess128_round): Remove.
11016 (__builtin_ia32_rangesd128_mask_round,
11017 __builtin_ia32_rangess128_mask_round): New builtins.
11018 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11019 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11020 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11021 "<round_saeonly_constraint>")): Changed to ...
11022 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11023 "<round_saeonly_scalar_constraint>")): ... this.
11024 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11025 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11026 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11027 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11028 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11030 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11032 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11034 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11035 Remove all values except native, 8540 and 8548.
11037 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11039 * config/i386/i386.c (ix86_output_function_return): Pass
11040 INVALID_REGNUM, instead of -1, as invalid register number to
11041 indirect_thunk_name and output_indirect_thunk.
11043 2018-02-02 Julia Koval <julia.koval@intel.com>
11045 * config.gcc: Add -march=icelake.
11046 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11047 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11048 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11049 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11050 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11051 (processor_target_table): Add icelake.
11052 (ix86_option_override_internal): Handle new PTAs.
11053 (get_builtin_code_for_version): Handle icelake.
11054 (M_INTEL_COREI7_ICELAKE): New.
11055 (fold_builtin_cpu): Handle icelake.
11056 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11057 * doc/invoke.texi: Add -march=icelake.
11059 2018-02-02 Julia Koval <julia.koval@intel.com>
11061 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11062 to wide_int_bitmask.
11063 * wide-int-bitmask.h: New.
11065 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11068 * config/i386/i386.md: Replace Pmode with word_mode in
11069 builtin_setjmp_setup and builtin_longjmp to support x32.
11071 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11075 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11077 (rs6000_supported_cpu_names): New static variable.
11078 (linux_cpu_translation_table): Likewise.
11079 (elf_platform) <cpu>: Define new static variable and use it.
11080 Translate kernel AT_PLATFORM name to canonical name if needed.
11081 Error if platform name is unknown.
11083 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11086 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11088 2018-02-01 Jeff Law <law@redhat.com>
11091 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11092 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11093 the scratch if RELEASE_VIA_POP is false.
11094 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11095 If we have to save a temporary register, decrement SIZE appropriately.
11096 Pass new arguments to release_scratch_register_on_entry.
11097 (ix86_adjust_stack_and_probe): Likewise.
11098 (ix86_emit_probe_stack_range): Pass new arguments to
11099 release_scratch_register_on_entry.
11101 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11103 PR rtl-optimization/84157
11104 * combine.c (change_zero_ext): Use REG_P predicate in
11105 front of HARD_REGISTER_P predicate.
11107 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11109 * config/avr/avr.c (avr_option_override): Move disabling of
11110 -fdelete-null-pointer-checks to...
11111 * common/config/avr/avr-common.c (avr_option_optimization_table):
11114 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11116 PR tree-optimization/81635
11117 * tree-data-ref.c (split_constant_offset_1): For types that
11118 wrap on overflow, try to use range info to prove that wrapping
11121 2018-02-01 Renlin Li <renlin.li@arm.com>
11124 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11125 TAILCALL_ADDR_REGS.
11126 (aarch64_register_move_cost): Likewise.
11127 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11128 TAILCALL_ADDR_REGS.
11129 (REG_CLASS_NAMES): Likewise.
11130 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11131 TAILCALL_ADDR_REGS. Remove IP registers.
11132 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11134 2018-02-01 Richard Biener <rguenther@suse.de>
11136 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11137 for specifying RPO order and allow NULL for that.
11138 * domwalk.c (dom_walker::dom_walker): Likewise.
11139 (dom_walker::walk): Handle NULL RPO order.
11140 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11142 (rewrite_update_dom_walker): Likewise.
11143 (mark_def_dom_walker): Likewise.
11145 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11147 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11148 (aarch64_maybe_expand_sve_subreg_move): Declare.
11149 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11150 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11152 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11153 that are semantically a reverse operation.
11154 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11155 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11156 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11158 (aarch64_can_change_mode_class): For big-endian, forbid changes
11159 between two SVE modes if they have different element sizes.
11161 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11163 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11164 the TImode handling for big-endian targets.
11166 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11168 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11169 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11171 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11172 Remove BSWAP handing for big-endian targets and use the form of
11173 LD1RQ appropariate for the mode.
11175 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11177 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11178 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11179 duplicated element.
11181 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11184 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11185 check for operands that need to go through aarch64_sve_reload_be.
11187 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11189 PR tree-optimization/81661
11190 PR tree-optimization/84117
11191 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11192 * tree-eh.c: Include gimplify.h.
11193 (find_trapping_overflow, replace_trapping_overflow,
11194 rewrite_to_non_trapping_overflow): New functions.
11195 * tree-vect-loop.c: Include tree-eh.h.
11196 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11197 * tree-data-ref.c: Include tree-eh.h.
11198 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11200 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11202 PR rtl-optimization/84123
11203 * combine.c (change_zero_ext): Check if hard register satisfies
11204 can_change_dest_mode before calling gen_lowpart_SUBREG.
11206 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11209 * ira.c (ira_init_register_move_cost): Remove assert.
11211 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11213 PR rtl-optimization/84071
11214 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11215 * doc/tm.texi: Regenerate.
11217 2018-01-31 Richard Biener <rguenther@suse.de>
11219 PR tree-optimization/84132
11220 * tree-data-ref.c (analyze_miv_subscript): Properly
11221 check whether evolution_function_is_affine_multivariate_p
11222 before calling gcd_of_steps_may_divide_p.
11224 2018-01-31 Julia Koval <julia.koval@intel.com>
11227 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11228 * config/i386/i386.md (rdpid_rex64) New.
11229 (rdpid): Make 32bit only.
11231 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11234 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11235 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11237 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11240 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11242 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11244 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11246 PR rtl-optimization/84071
11247 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11248 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11250 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11252 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11253 (arc_attribute_table): Add 'aux' attribute.
11254 (arc_in_small_data_p): Consider aux like variables.
11255 (arc_is_aux_reg_p): New function.
11256 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11257 (arc_get_aux_arg): New function.
11258 (prepare_move_operands): Handle aux-register access.
11259 (arc_handle_aux_attribute): New function.
11260 * doc/extend.texi (ARC Variable attributes): Add subsection.
11262 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11264 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11265 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11266 (arc_attribute_table): Add 'uncached' attribute.
11267 (arc_print_operand): Print '.di' flag for uncached memory
11269 (arc_in_small_data_p): Do not consider for small data the uncached
11271 (arc_is_uncached_mem_p): New function.
11272 * config/arc/predicates.md (compact_store_memory_operand): Check
11273 for uncached memory accesses.
11274 (nonvol_nonimm_operand): Likewise.
11275 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11277 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11280 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11281 falign-loops=): Add Optimization flag.
11283 2018-01-30 Jeff Law <law@redhat.com>
11286 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11287 INT_REGISTERS_SAVED. Check it prior to calling
11288 get_scratch_register_on_entry.
11289 (ix86_adjust_stack_and_probe): Similarly.
11290 (ix86_emit_probe_stack_range): Similarly.
11291 (ix86_expand_prologue): Corresponding changes.
11293 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11296 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11297 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11299 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11302 * lra-constraints.c (curr_insn_transform): Process AND in the
11305 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11307 PR rtl-optimization/83986
11308 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11309 dependence against last_pending_memory_flush in addition to
11310 pending_jump_insns.
11312 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11314 PR tree-optimization/81611
11315 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11318 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11321 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11324 2018-01-30 Richard Biener <rguenther@suse.de>
11325 Jakub Jelinek <jakub@redhat.com>
11327 PR tree-optimization/84111
11328 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11329 inner loops added during recursion, as they don't have up-to-date
11332 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11335 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11336 (can_inline_edge_by_limits_p): ... here.
11337 (can_early_inline_edge_p, check_callers,
11338 update_caller_keys, update_callee_keys, recursive_inlining,
11339 add_new_edges_to_heap, speculation_useful_p,
11340 inline_small_functions,
11341 inline_small_functions, flatten_function,
11342 inline_to_all_callers_1): Update.
11344 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11346 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11349 2018-01-30 Richard Biener <rguenther@suse.de>
11351 PR tree-optimization/83008
11352 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11353 invariant and constant vector uses in stmts when they need
11354 more than one stmt.
11356 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11359 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11360 * configure: Regenerate.
11362 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11364 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11366 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11367 Use gen_rtx_REG rather than gen_lowpart.
11369 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11371 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11372 rather than 0 when creating partial subregs.
11374 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11376 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11379 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11382 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11383 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11384 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11385 flags. This restores the settings used before the 2017-07-24.
11386 Turning off pre increment/decrement/modify allows IVOPTS to
11387 optimize DF/SF loops where the index is an int.
11389 2018-01-29 Richard Biener <rguenther@suse.de>
11390 Kelvin Nilsen <kelvin@gcc.gnu.org>
11393 * tree-vect-stmts.c (vectorizable_call): Don't call
11394 targetm.vectorize_builtin_md_vectorized_function if callee is
11397 2018-01-22 Carl Love <cel@us.ibm.com>
11399 * doc/extend.tex: Fix typo in second arg in
11400 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11402 2018-01-29 Richard Biener <rguenther@suse.de>
11404 PR tree-optimization/84086
11405 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11406 (flush_ssaname_freelist): When SSA names were released reset
11407 the SCEV hash table.
11409 2018-01-29 Richard Biener <rguenther@suse.de>
11411 PR tree-optimization/84057
11412 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11413 removed paths when removing edges.
11415 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11417 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11418 -mfunction-return=@var{choice}.
11420 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11422 PR diagnostic/84034
11423 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11424 Handle CR like TAB.
11425 (layout::print_source_line): Likewise.
11426 (test_get_line_width_without_trailing_whitespace): Add test cases.
11428 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11430 PR middle-end/84040
11431 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11434 2018-01-26 Jim Wilson <jimw@sifive.com>
11436 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11438 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11441 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11443 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11444 and CMP + SUB-immediate -> SUBS.
11446 2018-01-26 Martin Sebor <msebor@redhat.com>
11448 PR tree-optimization/83896
11449 * tree-ssa-strlen.c (get_string_len): Rename...
11450 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11451 Avoid assuming length is constant.
11452 (handle_char_store): Use HOST_WIDE_INT for string length.
11454 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11457 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11458 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11460 2018-01-26 Richard Biener <rguenther@suse.de>
11462 PR rtl-optimization/84003
11463 * dse.c (record_store): Only record redundant stores when
11464 the earlier store aliases at least all accesses the later one does.
11466 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11468 PR rtl-optimization/83985
11469 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11470 REG_CFA_RESTORE insns.
11471 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11472 REG_CFA_RESTORE insns here.
11475 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11476 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11478 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11480 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11481 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11482 (arc_init): Likewise.
11483 (arc_override_options): Likewise.
11484 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11486 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11488 * config/arc/arc.h (TARGET_DBNZ): Define.
11489 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11490 properly set the tune attribute.
11491 (dbnz): Use TARGET_DBNZ guard.
11492 * config/arc/arc.opt (mtune): Add core3 option.
11494 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11496 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11497 recognize new pic like addresses.
11498 (arc_delegitimize_address): Clean up.
11500 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11502 * config/arc/arc-arches.def: Option mrf16 valid for all
11504 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11505 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11506 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11507 * config/arc/arc-tables.opt: Regenerate.
11508 * config/arc/arc.c (arc_conditional_register_usage): Handle
11509 reduced register file case.
11510 (arc_file_start): Set must have build attributes.
11511 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11512 mrf16 option value.
11513 * config/arc/arc.opt (mrf16): Add new option.
11514 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11515 * config/arc/genmultilib.awk: Handle new mrf16 option.
11516 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11517 * config/arc/t-multilib: Regenerate.
11518 * doc/invoke.texi (ARC Options): Document mrf16 option.
11520 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11522 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11523 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11524 (arc_attribute_table): Add 'secure_call' attribute.
11525 (arc_print_operand): Print secure call operand.
11526 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11528 (arc_is_secure_call_p): New function. * config/arc/arc.md
11529 (call_i): Add support for sjli instruction.
11530 (call_value_i): Likewise.
11531 * config/arc/constraints.md (Csc): New constraint.
11533 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11534 John Eric Martin <John.Martin@emmicro-us.com>
11536 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11537 * config/arc/arc.c (_arc_jli_section): New struct.
11538 (arc_jli_section): New type.
11539 (rc_jli_sections): New static variable.
11540 (arc_handle_jli_attribute): New function.
11541 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11542 (arc_file_end): New function.
11543 (TARGET_ASM_FILE_END): Define.
11544 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11545 (arc_add_jli_section): New function.
11546 (jli_call_scan): Likewise.
11547 (arc_reorg): Call jli_call_scan.
11548 (arc_output_addsi): Remove 'S' from printing asm operand.
11549 (arc_is_jli_call_p): New function.
11550 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11552 (movhi_insn): Likewise.
11553 (movsi_insn): Likewise.
11554 (movsi_set_cc_insn): Likewise.
11555 (loadqi_update): Likewise.
11556 (load_zeroextendqisi_update): Likewise.
11557 (load_signextendqisi_update): Likewise.
11558 (loadhi_update): Likewise.
11559 (load_zeroextendhisi_update): Likewise.
11560 (load_signextendhisi_update): Likewise.
11561 (loadsi_update): Likewise.
11562 (loadsf_update): Likewise.
11563 (movsicc_insn): Likewise.
11564 (bset_insn): Likewise.
11565 (bxor_insn): Likewise.
11566 (bclr_insn): Likewise.
11567 (bmsk_insn): Likewise.
11568 (bicsi3_insn): Likewise.
11569 (cmpsi_cc_c_insn): Likewise.
11570 (movsi_ne): Likewise.
11571 (movsi_cond_exec): Likewise.
11572 (clrsbsi2): Likewise.
11573 (norm_f): Likewise.
11583 (call_i): Remove 'S' asm letter, add jli instruction.
11584 (call_value_i): Likewise.
11585 * config/arc/arc.op (mjli-always): New option.
11586 * config/arc/constraints.md (Cji): New constraint.
11587 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11589 (subsf3_fpx): Likewise.
11590 (mulsf3_fpx): Likewise.
11591 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11593 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11594 function attrbutes.
11595 * doc/invoke.texi (ARC): Document mjli-always option.
11597 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11599 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11600 avoid addition with 0 and use incw and decw where possible.
11602 2018-01-26 Richard Biener <rguenther@suse.de>
11604 PR tree-optimization/81082
11605 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11606 association if it requires casting to unsigned.
11607 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11608 from fold_plusminus_mult_expr to catch important cases late when
11609 range info is available.
11611 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11613 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11614 * configure.ac (hidden_linkonce): New test.
11615 * configure: Regenerate.
11616 * config.in: Regenerate.
11618 2018-01-26 Julia Koval <julia.koval@intel.com>
11620 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11621 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11622 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11623 _mm_mask_bitshuffle_epi64_mask): Fix type.
11624 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11625 USI_FTYPE_V4DI_V4DI_USI): Remove.
11626 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11627 __builtin_ia32_vpshufbitqmb256_mask,
11628 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11629 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11630 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11632 2018-01-26 Alan Modra <amodra@gmail.com>
11635 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11636 UNSPEC_VBPERMQ. Sort other unspecs.
11638 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11640 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11642 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11644 PR middle-end/83055
11645 * predict.c (drop_profile): Do not push/pop cfun; update also
11647 (handle_missing_profiles): Fix logic looking for zero profiles.
11649 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11651 PR middle-end/83977
11652 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11653 on functions with #pragma omp declare simd or functions with simd
11655 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11656 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11657 Remove trailing \n from warning_at calls.
11659 2018-01-25 Tom de Vries <tom@codesourcery.com>
11662 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11663 for neutered workers.
11665 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11668 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11669 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11671 2018-01-24 Jeff Law <law@redhat.com>
11674 * i386.c (get_probe_interval): Move to earlier point.
11675 (ix86_compute_frame_layout): If -fstack-clash-protection and
11676 the frame is larger than the probe interval, then use pushes
11677 to save registers rather than reg->mem moves.
11678 (ix86_expand_prologue): Remove conditional for int_registers_saved
11681 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11684 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11685 min/max for never referenced object.
11687 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11689 PR middle-end/83977
11690 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11692 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11693 attributes from DECL_ATTRIBUTES (decl) without affecting
11694 DECL_ATTRIBUTES (current_function_decl).
11695 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11696 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11698 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11700 PR tree-optimization/83979
11701 * fold-const.c (fold_comparison): Use constant_boolean_node
11702 instead of boolean_{true,false}_node.
11704 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11706 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11709 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11711 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11712 Simplify the clause that sets the length attribute.
11713 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11714 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11715 clause that sets the length attribute.
11716 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11718 2018-01-24 Tom de Vries <tom@codesourcery.com>
11721 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11722 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11723 Add strict parameter.
11724 (prevent_branch_around_nothing): Insert dummy insn between branch to
11725 label and label with no ptx insn inbetween.
11726 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11728 2018-01-24 Tom de Vries <tom@codesourcery.com>
11731 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11732 for neutered threads in warp.
11733 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11735 2018-01-24 Richard Biener <rguenther@suse.de>
11737 PR tree-optimization/83176
11738 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11741 2018-01-24 Richard Biener <rguenther@suse.de>
11743 PR tree-optimization/82819
11744 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11745 code generating pluses that are no-ops in the target precision.
11747 2018-01-24 Richard Biener <rguenther@suse.de>
11749 PR middle-end/84000
11750 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11752 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11754 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11755 to merge probabilities.
11756 * predict.c (probably_never_executed): Also mark as cold functions
11757 with global 0 profile and guessed local profile.
11758 * profile-count.c (profile_probability::combine_with_count): New
11760 * profile-count.h (profile_probability::operator*,
11761 profile_probability::operator*=, profile_probability::operator/,
11762 profile_probability::operator/=): Reduce precision to adjusted
11763 and set value to guessed on contradictory divisions.
11764 (profile_probability::combine_with_freq): Remove.
11765 (profile_probability::combine_wiht_count): Declare.
11766 (profile_count::force_nonzero):: Set to adjusted.
11767 (profile_count::probability_in):: Set quality to adjusted.
11768 * tree-ssa-tail-merge.c (replace_block_by): Use
11769 combine_with_count.
11771 2018-01-23 Andrew Waterman <andrew@sifive.com>
11772 Jim Wilson <jimw@sifive.com>
11774 * config/riscv/riscv.c (riscv_stack_boundary): New.
11775 (riscv_option_override): Set riscv_stack_boundary. Handle
11776 riscv_preferred_stack_boundary_arg.
11777 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11778 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11779 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11780 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11781 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11782 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11784 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11787 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11788 of struct ix86_frame.
11789 (ix86_expand_epilogue): Likewise. Add a local variable for
11790 the reg_save_offset field in struct ix86_frame.
11792 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11794 PR tree-optimization/82604
11795 * tree-loop-distribution.c (enum partition_kind): New enum item
11796 PKIND_PARTIAL_MEMSET.
11797 (partition_builtin_p): Support above new enum item.
11798 (generate_code_for_partition): Ditto.
11799 (compute_access_range): Differentiate cases that equality can be
11800 proven at all loops, the innermost loops or no loops.
11801 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11802 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11803 (finalize_partitions, distribute_loop): Don't fuse partition of
11804 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11805 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11806 parloop is enabled.
11808 2018-01-23 Martin Liska <mliska@suse.cz>
11810 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11811 order to ignore the predictor.
11812 (PRED_POLYMORPHIC_CALL): Likewise.
11813 (PRED_RECURSIVE_CALL): Likewise.
11815 2018-01-23 Martin Liska <mliska@suse.cz>
11817 * tree-profile.c (tree_profiling): Print function header to
11818 aware reader which function we are working on.
11819 * value-prof.c (gimple_find_values_to_profile): Do not print
11820 not interesting value histograms.
11822 2018-01-23 Martin Liska <mliska@suse.cz>
11824 * profile-count.h (enum profile_quality): Add
11825 profile_uninitialized as the first value. Do not number values
11826 as they are zero based.
11827 (profile_count::verify): Update sanity check.
11828 (profile_probability::verify): Likewise.
11830 2018-01-23 Nathan Sidwell <nathan@acm.org>
11832 * doc/invoke.texi (ffor-scope): Deprecate.
11834 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11836 PR tree-optimization/83510
11837 * domwalk.c (set_all_edges_as_executable): New function.
11838 (dom_walker::dom_walker): Convert bool param
11839 "skip_unreachable_blocks" to enum reachability. Move setup of
11840 edge flags to set_all_edges_as_executable and only do it when
11841 reachability is REACHABLE_BLOCKS.
11842 * domwalk.h (enum dom_walker::reachability): New enum.
11843 (dom_walker::dom_walker): Convert bool param
11844 "skip_unreachable_blocks" to enum reachability.
11845 (set_all_edges_as_executable): New decl.
11846 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11847 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11849 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11850 but converting true to REACHABLE_BLOCKS.
11851 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11853 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11854 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11855 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11857 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11858 if check_all_array_refs will be called.
11860 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11862 * tree.c (selftest::test_location_wrappers): Add more test
11865 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11867 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11868 (selftest::test_bit_in_range): Likewise.
11870 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11873 * doc/sourcebuild.texi (vect_float): Say that the selector
11874 only describes the situation when -funsafe-math-optimizations is on.
11875 (vect_float_strict): Document.
11877 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11879 PR tree-optimization/83965
11880 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11881 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11882 instead of checking only for a reduction.
11883 (vect_recog_widen_sum_pattern): Likewise.
11885 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11887 * predict.c (probably_never_executed): Only use precise profile info.
11888 (compute_function_frequency): Skip after inlining hack since we now
11889 have quality checking.
11891 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11893 * profile-count.h (profile_probability::very_unlikely,
11894 profile_probability::unlikely, profile_probability::even): Set
11895 precision to guessed.
11897 2018-01-23 Richard Biener <rguenther@suse.de>
11899 PR tree-optimization/83963
11900 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11901 Properly terminate dominator walk when crossing the exit edge not
11902 when visiting its source block.
11904 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11907 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11908 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11910 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11912 PR tree-optimization/83957
11913 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11914 semicolon after for body surrounded by braces.
11916 PR tree-optimization/83081
11917 * profile-count.h (profile_probability::split): New method.
11918 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11919 Use profile_probability::split.
11920 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11921 when splitting a single conditional jump into 2.
11923 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11925 PR tree-optimization/69452
11926 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11929 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11931 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11932 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11933 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11937 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11939 * config/rl78/rl78.md (movdi): New define_expand.
11940 * config/rl78/rl78.c (rl78_split_movdi): New function.
11942 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11945 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11947 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11948 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11949 128-bit to produce an UNSPEC move to get the double word with the
11950 signbit and then a shift directly to do signbit.
11951 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11952 implementation with a new version that just does either a direct
11953 move or a regular move. Move memory interface to separate insns.
11954 Move insns so they are next to the expander.
11955 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11956 with signbit move. Split big and little endian case.
11957 (signbit<mode>2_dm_mem_le): Likewise.
11958 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11959 (signbit<mode>2_dm2): Likewise.
11961 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11963 * config/rl78/rl78.md (anddi3): New define_expand.
11965 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11967 * config/rl78/rl78.md (umindi3): New define_expand.
11969 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11971 * config/rl78/rl78.md (smindi3): New define_expand.
11973 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11975 * config/rl78/rl78.md (smaxdi3): New define_expand.
11977 2018-01-22 Carl Love <cel@us.ibm.com>
11979 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11980 LVX_V1TI): Add macro expansion.
11981 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11982 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11983 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11984 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11985 Change check to determine if the instruction is a byte reversing
11986 entry. Fix typo in comment.
11987 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11988 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11989 Add def_builtin calls for new builtins.
11990 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11991 Add define_insn expansion.
11993 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11995 * config/rl78/rl78.md (umaxdi3): New define_expand.
11997 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11999 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12000 for non-QImode registers.
12002 2018-01-22 Richard Biener <rguenther@suse.de>
12004 PR tree-optimization/83963
12005 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12006 including the loop exit block.
12007 (scop_detection::merge_sese): Likewise.
12008 (scop_detection::add_scop): Do it here instead.
12010 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12012 * doc/sourcebuild.texi (arm_softfloat): Document.
12014 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12017 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12018 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12019 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12021 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12022 David Edelsohn <dje.gcc@gmail.com>
12025 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12026 Change "crset eq" to "crset 2".
12027 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12028 (*call_indirect_aix<mode>_nospec): Likewise.
12029 (*call_value_indirect_aix<mode>_nospec): Likewise.
12030 (*call_indirect_elfv2<mode>_nospec): Likewise.
12031 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12032 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12033 change assembly output from . to $.
12034 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12035 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12036 (*tablejump<mode>_internal1_nospec): Likewise.
12038 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12041 * config/sh/sh_optimize_sett_clrt.cc:
12042 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12044 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12046 PR tree-optimization/83940
12047 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12048 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12049 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12050 use it to pass back the definition type.
12051 (vect_check_store_rhs): Likewise rhs_dt_out.
12052 (vect_build_gather_load_calls): Add a mask_dt argument and use
12053 it instead of a call to vect_is_simple_use.
12054 (vectorizable_store): Update calls to vect_check_load_store_mask
12055 and vect_check_store_rhs. Use the dt returned by the latter instead
12056 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12057 instead of calls to vect_is_simple_use. Pass the scalar rather
12058 than the vector operand to vect_is_simple_use when handling
12059 second and subsequent copies of an rhs value.
12060 (vectorizable_load): Update calls to vect_check_load_store_mask
12061 and vect_build_gather_load_calls. Use the cached mask_dt and
12062 gs_info.offset_dt instead of calls to vect_is_simple_use.
12064 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12066 PR middle-end/83945
12067 * tree-emutls.c: Include gimplify.h.
12068 (lower_emutls_2): New function.
12069 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12070 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12071 it before further processing.
12074 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12075 UINTVAL (trueop1) instead of INTVAL (op1).
12077 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12081 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12082 INCOMING_FRAME_SP_OFFSET if not defined.
12083 (scan_trace): Add ENTRY argument. If true and
12084 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12085 emit a note to adjust the CFA offset.
12086 (create_cfi_notes): Adjust scan_trace callers.
12087 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12088 INCOMING_FRAME_SP_OFFSET in the CIE.
12089 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12090 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12092 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12093 * doc/tm.texi: Regenerated.
12095 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12097 PR rtl-optimization/83147
12098 * lra-constraints.c (remove_inheritance_pseudos): Use
12099 lra_substitute_pseudo_within_insn.
12101 2018-01-19 Tom de Vries <tom@codesourcery.com>
12102 Cesar Philippidis <cesar@codesourcery.com>
12105 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12107 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12110 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12111 spaces for function labels.
12113 2018-01-19 Martin Liska <mliska@suse.cz>
12115 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12116 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12117 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12118 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12119 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12120 (PRED_CONST_RETURN): Change from 69 to 65.
12121 (PRED_NULL_RETURN): Change from 91 to 71.
12122 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12123 (PRED_LOOP_GUARD): Change from 66 to 73.
12125 2018-01-19 Martin Liska <mliska@suse.cz>
12127 * predict.c (predict_insn_def): Add new assert.
12128 (struct branch_predictor): Change type to signed integer.
12129 (test_prediction_value_range): Amend test to cover
12130 PROB_UNINITIALIZED.
12131 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12132 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12133 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12134 (PRED_LOOP_IV_COMPARE): Likewise.
12135 * predict.h (PROB_UNINITIALIZED): Define new constant.
12137 2018-01-19 Martin Liska <mliska@suse.cz>
12139 * predict.c (dump_prediction): Add new format for
12140 analyze_brprob.py script which is enabled with -details
12142 * profile-count.h (precise_p): New function.
12144 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12146 PR tree-optimization/83922
12147 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12148 there are no statements that need masking.
12149 (vect_active_double_reduction_p): New function.
12150 (vect_analyze_loop_operations): Use it when handling phis that
12151 are not in the loop header.
12153 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12155 PR tree-optimization/83914
12156 * tree-vect-loop.c (vectorizable_induction): Don't convert
12157 init_expr or apply the peeling adjustment for inductions
12158 that are nested within the vectorized loop.
12160 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12162 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12165 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12169 * function.h (gimplify_parameters): Add gimple_seq * argument.
12170 * function.c: Include gimple.h and options.h.
12171 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12172 for the added local temporaries if needed.
12173 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12174 if there are any parameter cleanups, wrap whole body into a
12175 try/finally with the cleanups.
12177 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12180 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12181 Use GET_MODE_CLASS for scalar floating point.
12183 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12187 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12188 Fix call of call_cgraph_insertion_hooks.
12190 2018-01-18 Martin Sebor <msebor@redhat.com>
12192 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12194 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12197 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12200 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12203 * common.opt: (-ffile-prefix-map): New option.
12204 * opts.c (common_handle_option): Defer it.
12205 * opts-global.c (handle_common_deferred_options): Handle it.
12206 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12207 * file-prefix-map.h: New file.
12208 (remap_debug_filename, add_debug_prefix_map): ...here.
12209 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12210 * final.c (debug_prefix_map, add_debug_prefix_map
12211 remap_debug_filename): Move to...
12212 * file-prefix-map.c: New file.
12213 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12214 generalize, get rid of alloca(), use strrchr() instead of strchr().
12215 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12216 Implement in terms of add_prefix_map().
12217 (remap_macro_filename, remap_debug_filename): Implement in term of
12219 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12220 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12221 * dbxout.c: Include file-prefix-map.h.
12222 * varasm.c: Likewise.
12223 * vmsdbgout.c: Likewise.
12224 * xcoffout.c: Likewise.
12225 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12226 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12227 * doc/invoke.texi (-ffile-prefix-map): Document.
12228 (-fdebug-prefix-map): Update description.
12230 2018-01-18 Martin Liska <mliska@suse.cz>
12232 * config/i386/i386.c (indirect_thunk_name): Document that also
12234 (output_indirect_thunk): Document why both instructions
12235 (pause and lfence) are generated.
12237 2018-01-18 Richard Biener <rguenther@suse.de>
12239 PR tree-optimization/83887
12240 * graphite-scop-detection.c
12241 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12242 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12243 (scop_detection::merge_sese): Re-implement with a flood-fill
12244 algorithm that properly finds a SESE region if it exists.
12246 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12249 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12250 pointer_diff optimizations use view_convert instead of convert.
12252 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12254 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12255 Generate different code for -mno-speculate-indirect-jumps.
12256 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12257 (*call_indirect_aix<mode>): Disable for
12258 -mno-speculate-indirect-jumps.
12259 (*call_indirect_aix<mode>_nospec): New define_insn.
12260 (*call_value_indirect_aix<mode>): Disable for
12261 -mno-speculate-indirect-jumps.
12262 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12263 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12264 -mno-speculate-indirect-jumps.
12265 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12267 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12269 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12270 long double type, set the flags for noting the default long double
12271 type, even if we don't pass or return a long double type.
12273 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12276 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12279 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12281 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12282 support for merge[hl].
12283 (fold_mergehl_helper): New helper function.
12284 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12285 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12286 (altivec_vmrglw_direct): Add xxmrglw insn.
12288 2018-01-17 Andrew Waterman <andrew@sifive.com>
12290 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12291 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12293 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12296 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12297 call the lto_location_cache before reading the
12298 DECL_SOURCE_LOCATION of the types.
12300 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12301 Richard Sandiford <richard.sandiford@linaro.org>
12303 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12304 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12305 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12306 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12307 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12309 * config/aarch64/constraints.md (aarch64_movti_operand):
12311 * config/aarch64/predicates.md (Uti): Add new constraint.
12313 2018-01-17 Carl Love <cel@us.ibm.com>
12315 * config/rs6000/vsx.md (define_expand xl_len_r,
12316 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12317 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12319 (define_expand, define_insn): Move the shift left from the
12320 define_insn to the define_expand for lxvl and stxvl instructions.
12321 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12322 and XL_LEN_R definitions to PURE.
12324 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12326 * config/i386/i386.c (indirect_thunk_name): Declare regno
12327 as unsigned int. Compare regno with INVALID_REGNUM.
12328 (output_indirect_thunk): Ditto.
12329 (output_indirect_thunk_function): Ditto.
12330 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12331 in the call to output_indirect_thunk_function.
12333 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12335 PR middle-end/83884
12336 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12337 rather than the size of inner_type to determine the stack slot size
12338 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12340 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12343 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12346 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12348 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12349 endian Linux systems to optionally enable multilibs for selecting
12350 the long double type if the user configured an explicit type.
12351 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12352 have no long double multilibs if not defined.
12353 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12354 warn if the user used -mabi={ieee,ibm}longdouble and we built
12355 multilibs for long double.
12356 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12357 appropriate multilib option.
12358 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12360 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12361 for building long double multilibs.
12362 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12364 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12366 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12369 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12371 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12374 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12377 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12380 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12382 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12383 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12385 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12387 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12388 different rtl trees depending on TARGET_64BIT.
12389 (rs6000_gen_lvx): Likewise.
12391 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12393 * config/visium/visium.md (nop): Tweak comment.
12394 (hazard_nop): Likewise.
12396 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12398 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12399 -mspeculate-indirect-jumps.
12400 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12401 for -mno-speculate-indirect-jumps.
12402 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12403 (*call_value_indirect_elfv2<mode>): Disable for
12404 -mno-speculate-indirect-jumps.
12405 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12406 (indirect_jump): Emit different RTL for
12407 -mno-speculate-indirect-jumps.
12408 (*indirect_jump<mode>): Disable for
12409 -mno-speculate-indirect-jumps.
12410 (*indirect_jump<mode>_nospec): New define_insn.
12411 (tablejump): Emit different RTL for
12412 -mno-speculate-indirect-jumps.
12413 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12414 (tablejumpsi_nospec): New define_expand.
12415 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12416 (tablejumpdi_nospec): New define_expand.
12417 (*tablejump<mode>_internal1): Disable for
12418 -mno-speculate-indirect-jumps.
12419 (*tablejump<mode>_internal1_nospec): New define_insn.
12420 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12423 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12425 * caller-save.c (insert_save): Drop unnecessary parameter. All
12428 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12429 Richard Biener <rguenth@suse.de>
12432 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12433 return early, inline manually is_gimple_sizepos. Make sure if we
12434 call gimplify_expr we don't end up with a gimple constant.
12435 * tree.c (variably_modified_type_p): Don't return true for
12436 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12437 * gimplify.h (is_gimple_sizepos): Remove.
12439 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12441 PR tree-optimization/83857
12442 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12443 vectorizable_live_operation for pure SLP statements.
12444 (vectorizable_live_operation): Handle PHIs.
12446 2018-01-16 Richard Biener <rguenther@suse.de>
12448 PR tree-optimization/83867
12449 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12450 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12452 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12455 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12456 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12457 If off is not INTEGER_CST, issue a may not be aligned warning
12458 rather than isn't aligned. Use isn%'t rather than isn't.
12459 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12461 <case MULT_EXPR>: Improve the case when bottom and one of the
12462 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12463 operand, in that case check if the other operand is multiple of
12464 bottom divided by the INTEGER_CST operand.
12466 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12469 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12470 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12471 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12472 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12473 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12474 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12475 (pa_function_arg_size): New function.
12477 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12479 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12480 in a separate statement.
12482 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12484 PR tree-optimization/83847
12485 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12486 group gathers and scatters.
12488 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12490 PR rtl-optimization/86620
12491 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12493 PR rtl-optimization/83213
12494 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12495 to last if both are JUMP_INSNs.
12497 PR tree-optimization/83843
12498 * gimple-ssa-store-merging.c
12499 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12500 store_immediate_info for bswap/nop orig_stores.
12502 2018-01-15 Andrew Waterman <andrew@sifive.com>
12504 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12506 <UDIV>: Increase cost if !TARGET_DIV.
12508 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12510 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12511 (define_attr "cr_logical_3op"): New.
12512 (cceq_ior_compare): Adjust.
12513 (cceq_ior_compare_complement): Adjust.
12514 (*cceq_rev_compare): Adjust.
12515 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12516 (is_cracked_insn): Adjust.
12517 (insn_must_be_first_in_group): Adjust.
12518 * config/rs6000/40x.md: Adjust.
12519 * config/rs6000/440.md: Adjust.
12520 * config/rs6000/476.md: Adjust.
12521 * config/rs6000/601.md: Adjust.
12522 * config/rs6000/603.md: Adjust.
12523 * config/rs6000/6xx.md: Adjust.
12524 * config/rs6000/7450.md: Adjust.
12525 * config/rs6000/7xx.md: Adjust.
12526 * config/rs6000/8540.md: Adjust.
12527 * config/rs6000/cell.md: Adjust.
12528 * config/rs6000/e300c2c3.md: Adjust.
12529 * config/rs6000/e500mc.md: Adjust.
12530 * config/rs6000/e500mc64.md: Adjust.
12531 * config/rs6000/e5500.md: Adjust.
12532 * config/rs6000/e6500.md: Adjust.
12533 * config/rs6000/mpc.md: Adjust.
12534 * config/rs6000/power4.md: Adjust.
12535 * config/rs6000/power5.md: Adjust.
12536 * config/rs6000/power6.md: Adjust.
12537 * config/rs6000/power7.md: Adjust.
12538 * config/rs6000/power8.md: Adjust.
12539 * config/rs6000/power9.md: Adjust.
12540 * config/rs6000/rs64.md: Adjust.
12541 * config/rs6000/titan.md: Adjust.
12543 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12545 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12546 ix86_indirect_branch_register logic.
12548 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12550 * config/i386/constraints.md (Bs): Update
12551 ix86_indirect_branch_register check. Don't check
12552 ix86_indirect_branch_register with GOT_memory_operand.
12554 * config/i386/predicates.md (GOT_memory_operand): Don't check
12555 ix86_indirect_branch_register here.
12556 (GOT32_symbol_operand): Likewise.
12558 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12560 * config/i386/predicates.md (constant_call_address_operand):
12561 Rewrite ix86_indirect_branch_register logic.
12562 (sibcall_insn_operand): Likewise.
12564 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12566 * config/i386/constraints.md (Bs): Replace
12567 ix86_indirect_branch_thunk_register with
12568 ix86_indirect_branch_register.
12570 * config/i386/i386.md (indirect_jump): Likewise.
12571 (tablejump): Likewise.
12572 (*sibcall_memory): Likewise.
12573 (*sibcall_value_memory): Likewise.
12574 Peepholes of indirect call and jump via memory: Likewise.
12575 * config/i386/i386.opt: Likewise.
12576 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12577 (GOT_memory_operand): Likewise.
12578 (call_insn_operand): Likewise.
12579 (sibcall_insn_operand): Likewise.
12580 (GOT32_symbol_operand): Likewise.
12582 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12584 PR middle-end/83837
12585 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12586 type rather than type addr's type points to.
12587 (expand_omp_atomic_mutex): Likewise.
12588 (expand_omp_atomic): Likewise.
12590 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12593 * config/i386/i386.c (output_indirect_thunk_function): Use
12594 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12595 for __x86_return_thunk.
12597 2018-01-15 Richard Biener <rguenther@suse.de>
12599 PR middle-end/83850
12600 * expmed.c (extract_bit_field_1): Fix typo.
12602 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12605 * config/arm/iterators.md (VF): New mode iterator.
12606 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12607 Remove integer-related logic from pattern.
12608 (neon_vabd<mode>_3): Likewise.
12610 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12612 PR middle-end/82694
12613 * common.opt (fstrict-overflow): No longer an alias.
12614 (fwrapv-pointer): New option.
12615 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12616 also for pointer types based on flag_wrapv_pointer.
12617 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12618 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12619 opts->x_flag_wrapv got set.
12620 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12621 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12622 POINTER_TYPE_OVERFLOW_UNDEFINED.
12623 * match.pd: Likewise in address comparison pattern.
12624 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12626 2018-01-15 Richard Biener <rguenther@suse.de>
12629 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12630 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12631 Reset type names to their identifier if their TYPE_DECL doesn't
12632 have linkage (and thus is used for ODR and devirt).
12633 (save_debug_info_for_decl): Remove.
12634 (save_debug_info_for_type): Likewise.
12635 (add_tree_to_fld_list): Adjust.
12636 * tree-pretty-print.c (dump_generic_node): Make dumping of
12637 type names more robust.
12639 2018-01-15 Richard Biener <rguenther@suse.de>
12641 * BASE-VER: Bump to 8.0.1.
12643 2018-01-14 Martin Sebor <msebor@redhat.com>
12646 * builtins.c (check_access): Avoid warning when the no-warning bit
12649 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12651 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12652 * ira-color (allocno_hard_regs_compare): Likewise.
12654 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12657 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12658 Use .pushsection/.popsection.
12660 2018-01-14 Martin Sebor <msebor@redhat.com>
12663 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12665 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12667 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12668 entry from extra_headers.
12669 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12670 extra_headers, make the list bitwise identical to the i?86-*-* one.
12672 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12674 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12675 -mcmodel=large with -mindirect-branch=thunk,
12676 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12677 -mfunction-return=thunk-extern.
12678 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12679 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12680 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12682 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12684 * config/i386/i386.c (print_reg): Print the name of the full
12685 integer register without '%'.
12686 (ix86_print_operand): Handle 'V'.
12687 * doc/extend.texi: Document 'V' modifier.
12689 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12691 * config/i386/constraints.md (Bs): Disallow memory operand for
12692 -mindirect-branch-register.
12694 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12695 (GOT_memory_operand): Likewise.
12696 (call_insn_operand): Likewise.
12697 (sibcall_insn_operand): Likewise.
12698 (GOT32_symbol_operand): Likewise.
12699 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12700 for -mindirect-branch-register.
12701 (tablejump): Likewise.
12702 (*sibcall_memory): Likewise.
12703 (*sibcall_value_memory): Likewise.
12704 Disallow peepholes of indirect call and jump via memory for
12705 -mindirect-branch-register.
12706 (*call_pop): Replace m with Bw.
12707 (*call_value_pop): Likewise.
12708 (*sibcall_pop_memory): Replace m with Bs.
12709 * config/i386/i386.opt (mindirect-branch-register): New option.
12710 * doc/invoke.texi: Document -mindirect-branch-register option.
12712 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12714 * config/i386/i386-protos.h (ix86_output_function_return): New.
12715 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12716 set function_return_type.
12717 (indirect_thunk_name): Add ret_p to indicate thunk for function
12719 (output_indirect_thunk_function): Pass false to
12720 indirect_thunk_name.
12721 (ix86_output_indirect_branch_via_reg): Likewise.
12722 (ix86_output_indirect_branch_via_push): Likewise.
12723 (output_indirect_thunk_function): Create alias for function
12724 return thunk if regno < 0.
12725 (ix86_output_function_return): New function.
12726 (ix86_handle_fndecl_attribute): Handle function_return.
12727 (ix86_attribute_table): Add function_return.
12728 * config/i386/i386.h (machine_function): Add
12729 function_return_type.
12730 * config/i386/i386.md (simple_return_internal): Use
12731 ix86_output_function_return.
12732 (simple_return_internal_long): Likewise.
12733 * config/i386/i386.opt (mfunction-return=): New option.
12734 (indirect_branch): Mention -mfunction-return=.
12735 * doc/extend.texi: Document function_return function attribute.
12736 * doc/invoke.texi: Document -mfunction-return= option.
12738 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12740 * config/i386/i386-opts.h (indirect_branch): New.
12741 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12742 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12743 with local indirect jump when converting indirect call and jump.
12744 (ix86_set_indirect_branch_type): New.
12745 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12746 (indirectlabelno): New.
12747 (indirect_thunk_needed): Likewise.
12748 (indirect_thunk_bnd_needed): Likewise.
12749 (indirect_thunks_used): Likewise.
12750 (indirect_thunks_bnd_used): Likewise.
12751 (INDIRECT_LABEL): Likewise.
12752 (indirect_thunk_name): Likewise.
12753 (output_indirect_thunk): Likewise.
12754 (output_indirect_thunk_function): Likewise.
12755 (ix86_output_indirect_branch_via_reg): Likewise.
12756 (ix86_output_indirect_branch_via_push): Likewise.
12757 (ix86_output_indirect_branch): Likewise.
12758 (ix86_output_indirect_jmp): Likewise.
12759 (ix86_code_end): Call output_indirect_thunk_function if needed.
12760 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12762 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12763 (ix86_attribute_table): Add indirect_branch.
12764 * config/i386/i386.h (machine_function): Add indirect_branch_type
12765 and has_local_indirect_jump.
12766 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12768 (tablejump): Likewise.
12769 (*indirect_jump): Use ix86_output_indirect_jmp.
12770 (*tablejump_1): Likewise.
12771 (simple_return_indirect_internal): Likewise.
12772 * config/i386/i386.opt (mindirect-branch=): New option.
12773 (indirect_branch): New.
12776 (thunk-inline): Likewise.
12777 (thunk-extern): Likewise.
12778 * doc/extend.texi: Document indirect_branch function attribute.
12779 * doc/invoke.texi: Document -mindirect-branch= option.
12781 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12784 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12786 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12788 * ipa-inline.c (want_inline_small_function_p): Return false if
12789 inlining has already failed with CIF_FINAL_ERROR.
12790 (update_caller_keys): Call want_inline_small_function_p before
12792 (update_callee_keys): Likewise.
12794 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12796 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12798 (rs6000_quadword_masked_address_p): Likewise.
12799 (quad_aligned_load_p): Likewise.
12800 (quad_aligned_store_p): Likewise.
12801 (const_load_sequence_p): Add comment to describe the outer-most loop.
12802 (mimic_memory_attributes_and_flags): New function.
12803 (rs6000_gen_stvx): Likewise.
12804 (replace_swapped_aligned_store): Likewise.
12805 (rs6000_gen_lvx): Likewise.
12806 (replace_swapped_aligned_load): Likewise.
12807 (replace_swapped_load_constant): Capitalize argument name in
12808 comment describing this function.
12809 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12810 and stores that access quad-word aligned addresses and replace
12811 with stvx or lvx instructions when appropriate.
12812 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12813 New function prototype.
12814 (rs6000_quadword_masked_address_p): Likewise.
12815 (rs6000_gen_lvx): Likewise.
12816 (rs6000_gen_stvx): Likewise.
12817 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12818 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12819 when memory address is aligned.
12820 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12821 this split to select lvx instruction when memory address is aligned.
12822 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12823 instruction when memory address is aligned.
12824 (*vsx_le_perm_load_v16qi): Likewise.
12825 (four unnamed splitters): Modify to select the stvx instruction
12826 when memory is aligned.
12828 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12830 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12831 which appears in the queue multiple times.
12833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12834 Alan Hayward <alan.hayward@arm.com>
12835 David Sherwood <david.sherwood@arm.com>
12837 * tree-vectorizer.h (vec_lower_bound): New structure.
12838 (_loop_vec_info): Add check_nonzero and lower_bounds.
12839 (LOOP_VINFO_CHECK_NONZERO): New macro.
12840 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12841 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12842 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12843 fields. Make seg_len the distance travelled, not including the
12845 (dr_direction_indicator): Declare.
12846 (dr_zero_step_indicator): Likewise.
12847 (dr_known_forward_stride_p): Likewise.
12848 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12850 (runtime_alias_check_p): Allow runtime alias checks with
12852 (operator ==): Compare access_size and align.
12853 (prune_runtime_alias_test_list): Rework for new distinction between
12854 the access_size and seg_len.
12855 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12857 (get_segment_min_max): New function.
12858 (create_intersect_range_checks): Use it.
12859 (dr_step_indicator): New function.
12860 (dr_direction_indicator): Likewise.
12861 (dr_zero_step_indicator): Likewise.
12862 (dr_known_forward_stride_p): Likewise.
12863 * tree-loop-distribution.c (data_ref_segment_size): Return
12864 DR_STEP * (niters - 1).
12865 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12867 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12868 (vect_preserves_scalar_order_p): New function, split out from...
12869 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12870 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12871 (vect_vfa_access_size): New function.
12872 (vect_vfa_align): Likewise.
12873 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12874 (dump_lower_bound): New function.
12875 (vect_check_lower_bound): Likewise.
12876 (vect_small_gap_p): Likewise.
12877 (vectorizable_with_step_bound_p): Likewise.
12878 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12879 depencies if the vectorization factor is 1. Convert the checks
12880 for nonzero steps into checks on the bounds of DR_STEP. Try using
12881 a bunds check for variable steps if the minimum required step is
12882 relatively small. Update calls to the dr_with_seg_len
12883 constructor and to vect_compile_time_alias.
12884 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12886 (vect_loop_versioning): Call it.
12887 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12889 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12892 Alan Hayward <alan.hayward@arm.com>
12893 David Sherwood <david.sherwood@arm.com>
12895 * doc/sourcebuild.texi (vect_scatter_store): Document.
12896 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12898 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12900 * genopinit.c (main): Add supports_vec_scatter_store and
12901 supports_vec_scatter_store_cached to target_optabs.
12902 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12903 IFN_MASK_SCATTER_STORE.
12904 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12906 * internal-fn.h (internal_store_fn_p): Declare.
12907 (internal_fn_stored_value_index): Likewise.
12908 * internal-fn.c (scatter_store_direct): New macro.
12909 (expand_scatter_store_optab_fn): New function.
12910 (direct_scatter_store_optab_supported_p): New macro.
12911 (internal_store_fn_p): New function.
12912 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12913 IFN_MASK_SCATTER_STORE.
12914 (internal_fn_mask_index): Likewise.
12915 (internal_fn_stored_value_index): New function.
12916 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12917 for scatter stores.
12918 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12919 * optabs-query.c (supports_vec_scatter_store_p): New function.
12920 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12921 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12922 true for scatter stores.
12923 (vect_gather_scatter_fn_p): Handle scatter stores too.
12924 (vect_check_gather_scatter): Consider using scatter stores if
12925 supports_vec_scatter_store_p.
12926 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12927 scatter stores too.
12928 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12929 internal_fn_stored_value_index.
12930 (check_load_store_masking): Handle scatter stores too.
12931 (vect_get_store_rhs): Make public.
12932 (vectorizable_call): Use internal_store_fn_p.
12933 (vectorizable_store): Handle scatter store internal functions.
12934 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12935 when deciding whether the end of the group has been reached.
12936 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12937 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12938 (mask_scatter_store<mode>): New insns.
12940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12941 Alan Hayward <alan.hayward@arm.com>
12942 David Sherwood <david.sherwood@arm.com>
12944 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12945 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12946 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12948 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12949 Use vect_truncate_gather_scatter_offset if we can't treat the
12950 operation as a normal gather load or scatter store.
12951 (get_group_load_store_type): Take the gather_scatter_info
12952 as argument. Try using a gather load or scatter store for
12953 single-element groups.
12954 (get_load_store_type): Update calls to get_group_load_store_type
12955 and vect_use_strided_gather_scatters_p.
12957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12958 Alan Hayward <alan.hayward@arm.com>
12959 David Sherwood <david.sherwood@arm.com>
12961 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12962 optional tree argument.
12963 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12965 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12966 but continue to use the current value as a fallback.
12967 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12968 to compare the updates.
12969 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12970 (get_load_store_type): Use it when handling a strided access.
12971 (vect_get_strided_load_store_ops): New function.
12972 (vect_get_data_ptr_increment): Likewise.
12973 (vectorizable_load): Handle strided gather loads. Always pass
12974 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * doc/md.texi (gather_load@var{m}): Document.
12981 (mask_gather_load@var{m}): Likewise.
12982 * genopinit.c (main): Add supports_vec_gather_load and
12983 supports_vec_gather_load_cached to target_optabs.
12984 * optabs-tree.c (init_tree_optimization_optabs): Use
12985 ggc_cleared_alloc to allocate target_optabs.
12986 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12987 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12989 * internal-fn.h (internal_load_fn_p): Declare.
12990 (internal_gather_scatter_fn_p): Likewise.
12991 (internal_fn_mask_index): Likewise.
12992 (internal_gather_scatter_fn_supported_p): Likewise.
12993 * internal-fn.c (gather_load_direct): New macro.
12994 (expand_gather_load_optab_fn): New function.
12995 (direct_gather_load_optab_supported_p): New macro.
12996 (direct_internal_fn_optab): New function.
12997 (internal_load_fn_p): Likewise.
12998 (internal_gather_scatter_fn_p): Likewise.
12999 (internal_fn_mask_index): Likewise.
13000 (internal_gather_scatter_fn_supported_p): Likewise.
13001 * optabs-query.c (supports_at_least_one_mode_p): New function.
13002 (supports_vec_gather_load_p): Likewise.
13003 * optabs-query.h (supports_vec_gather_load_p): Declare.
13004 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13005 and memory_type field.
13006 (NUM_PATTERNS): Bump to 15.
13007 * tree-vect-data-refs.c: Include internal-fn.h.
13008 (vect_gather_scatter_fn_p): New function.
13009 (vect_describe_gather_scatter_call): Likewise.
13010 (vect_check_gather_scatter): Try using internal functions for
13011 gather loads. Recognize existing calls to a gather load function.
13012 (vect_analyze_data_refs): Consider using gather loads if
13013 supports_vec_gather_load_p.
13014 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13015 (vect_get_gather_scatter_offset_type): Likewise.
13016 (vect_convert_mask_for_vectype): Likewise.
13017 (vect_add_conversion_to_patterm): Likewise.
13018 (vect_try_gather_scatter_pattern): Likewise.
13019 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13020 (vect_vect_recog_func_ptrs): Add it.
13021 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13022 internal_fn_mask_index and internal_gather_scatter_fn_p.
13023 (check_load_store_masking): Take the gather_scatter_info as an
13024 argument and handle gather loads.
13025 (vect_get_gather_scatter_ops): New function.
13026 (vectorizable_call): Check internal_load_fn_p.
13027 (vectorizable_load): Likewise. Handle gather load internal
13029 (vectorizable_store): Update call to check_load_store_masking.
13030 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13031 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13032 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13033 (aarch64_gather_scale_operand_d): New predicates.
13034 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13035 (mask_gather_load<mode>): New insns.
13037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13038 Alan Hayward <alan.hayward@arm.com>
13039 David Sherwood <david.sherwood@arm.com>
13041 * optabs.def (fold_left_plus_optab): New optab.
13042 * doc/md.texi (fold_left_plus_@var{m}): Document.
13043 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13044 * internal-fn.c (fold_left_direct): Define.
13045 (expand_fold_left_optab_fn): Likewise.
13046 (direct_fold_left_optab_supported_p): Likewise.
13047 * fold-const-call.c (fold_const_fold_left): New function.
13048 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13049 * tree-parloops.c (valid_reduction_p): New function.
13050 (gather_scalar_reductions): Use it.
13051 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13052 (vect_finish_replace_stmt): Declare.
13053 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13054 (needs_fold_left_reduction_p): New function, split out from...
13055 (vect_is_simple_reduction): ...here. Accept reductions that
13056 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13057 (vect_force_simple_reduction): Also store the reduction type in
13058 the assignment's STMT_VINFO_REDUC_TYPE.
13059 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13060 (merge_with_identity): New function.
13061 (vect_expand_fold_left): Likewise.
13062 (vectorize_fold_left_reduction): Likewise.
13063 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13064 scalar phi in place for it. Check for target support and reject
13065 cases that would reassociate the operation. Defer the transform
13066 phase to vectorize_fold_left_reduction.
13067 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13068 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13069 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13073 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13074 call to ifc_temp_var.
13076 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13077 Alan Hayward <alan.hayward@arm.com>
13078 David Sherwood <david.sherwood@arm.com>
13080 * target.def (legitimize_address_displacement): Take the original
13081 offset as a poly_int.
13082 * targhooks.h (default_legitimize_address_displacement): Update
13084 * targhooks.c (default_legitimize_address_displacement): Likewise.
13085 * doc/tm.texi: Regenerate.
13086 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13087 as an argument, moving assert of ad->disp == ad->disp_term to...
13088 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13089 Try calling targetm.legitimize_address_displacement before expanding
13090 the address rather than afterwards, and adjust for the new interface.
13091 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13092 Match the new hook interface. Handle SVE addresses.
13093 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13094 new hook interface.
13096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13098 * Makefile.in (OBJS): Add early-remat.o.
13099 * target.def (select_early_remat_modes): New hook.
13100 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13101 * doc/tm.texi: Regenerate.
13102 * targhooks.h (default_select_early_remat_modes): Declare.
13103 * targhooks.c (default_select_early_remat_modes): New function.
13104 * timevar.def (TV_EARLY_REMAT): New timevar.
13105 * passes.def (pass_early_remat): New pass.
13106 * tree-pass.h (make_pass_early_remat): Declare.
13107 * early-remat.c: New file.
13108 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13110 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13113 Alan Hayward <alan.hayward@arm.com>
13114 David Sherwood <david.sherwood@arm.com>
13116 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13117 vfm1 with a bound_epilog parameter.
13118 (vect_do_peeling): Update calls accordingly, and move the prologue
13119 call earlier in the function. Treat the base bound_epilog as 0 for
13120 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13121 this base when peeling for gaps.
13122 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13123 with fully-masked loops.
13124 (vect_estimate_min_profitable_iters): Handle the single peeled
13125 iteration in that case.
13127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13128 Alan Hayward <alan.hayward@arm.com>
13129 David Sherwood <david.sherwood@arm.com>
13131 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13132 single-element interleaving even if the size is not a power of 2.
13133 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13134 accesses for single-element interleaving if the group size is
13137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13138 Alan Hayward <alan.hayward@arm.com>
13139 David Sherwood <david.sherwood@arm.com>
13141 * doc/md.texi (fold_extract_last_@var{m}): Document.
13142 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13143 * optabs.def (fold_extract_last_optab): New optab.
13144 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13145 * internal-fn.c (fold_extract_direct): New macro.
13146 (expand_fold_extract_optab_fn): Likewise.
13147 (direct_fold_extract_optab_supported_p): Likewise.
13148 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13149 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13150 EXTRACT_LAST_REDUCTION.
13151 (get_initial_def_for_reduction): Do not create an initial vector
13152 for EXTRACT_LAST_REDUCTION reductions.
13153 (vectorizable_reduction): Leave the scalar phi in place for
13154 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13155 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13156 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13157 transform phase to vectorizable_condition.
13158 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13160 (vect_finish_stmt_generation): ...here.
13161 (vect_finish_replace_stmt): New function.
13162 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13163 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13165 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13168 Alan Hayward <alan.hayward@arm.com>
13169 David Sherwood <david.sherwood@arm.com>
13171 * doc/md.texi (extract_last_@var{m}): Document.
13172 * optabs.def (extract_last_optab): New optab.
13173 * internal-fn.def (EXTRACT_LAST): New internal function.
13174 * internal-fn.c (cond_unary_direct): New macro.
13175 (expand_cond_unary_optab_fn): Likewise.
13176 (direct_cond_unary_optab_supported_p): Likewise.
13177 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13178 loops using EXTRACT_LAST.
13179 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13180 (extract_last_<mode>): ...this optab.
13181 (vec_extract<mode><Vel>): Update accordingly.
13183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13184 Alan Hayward <alan.hayward@arm.com>
13185 David Sherwood <david.sherwood@arm.com>
13187 * target.def (empty_mask_is_expensive): New hook.
13188 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13189 * doc/tm.texi: Regenerate.
13190 * targhooks.h (default_empty_mask_is_expensive): Declare.
13191 * targhooks.c (default_empty_mask_is_expensive): New function.
13192 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13193 if the target says that empty masks are expensive.
13194 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13196 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13203 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13204 (vect_use_loop_mask_for_alignment_p): New function.
13205 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13206 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13207 niters_skip argument. Make sure that the first niters_skip elements
13208 of the first iteration are inactive.
13209 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13210 Update call to vect_set_loop_masks_directly.
13211 (get_misalign_in_elems): New function, split out from...
13212 (vect_gen_prolog_loop_niters): ...here.
13213 (vect_update_init_of_dr): Take a code argument that specifies whether
13214 the adjustment should be added or subtracted.
13215 (vect_update_init_of_drs): Likewise.
13216 (vect_prepare_for_masked_peels): New function.
13217 (vect_do_peeling): Skip prologue peeling if we're using a mask
13218 instead. Update call to vect_update_inits_of_drs.
13219 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13221 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13222 alignment. Do not include the number of peeled iterations in
13223 the minimum threshold in that case.
13224 (vectorizable_induction): Adjust the start value down by
13225 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13226 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13227 Take the number of skipped iterations into account when calculating
13229 * tree-vect-stmts.c (vect_gen_while_not): New function.
13231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13232 Alan Hayward <alan.hayward@arm.com>
13233 David Sherwood <david.sherwood@arm.com>
13235 * doc/sourcebuild.texi (vect_fully_masked): Document.
13236 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13237 default value to 0.
13238 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13240 (vect_analyze_loop_2): ...here. Don't check the vectorization
13241 factor against the number of loop iterations if the loop is
13244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13245 Alan Hayward <alan.hayward@arm.com>
13246 David Sherwood <david.sherwood@arm.com>
13248 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13249 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13250 (dump_groups): Update accordingly.
13251 (iv_use::mem_type): New member variable.
13252 (address_p): New function.
13253 (record_use): Add a mem_type argument and initialize the new
13255 (record_group_use): Add a mem_type argument. Use address_p.
13256 Remove obsolete null checks of base_object. Update call to record_use.
13257 (find_interesting_uses_op): Update call to record_group_use.
13258 (find_interesting_uses_cond): Likewise.
13259 (find_interesting_uses_address): Likewise.
13260 (get_mem_type_for_internal_fn): New function.
13261 (find_address_like_use): Likewise.
13262 (find_interesting_uses_stmt): Try find_address_like_use before
13263 calling find_interesting_uses_op.
13264 (addr_offset_valid_p): Use the iv mem_type field as the type
13265 of the addressed memory.
13266 (add_autoinc_candidates): Likewise.
13267 (get_address_cost): Likewise.
13268 (split_small_address_groups_p): Use address_p.
13269 (split_address_groups): Likewise.
13270 (add_iv_candidate_for_use): Likewise.
13271 (autoinc_possible_for_pair): Likewise.
13272 (rewrite_groups): Likewise.
13273 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13274 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13275 (get_alias_ptr_type_for_ptr_address): New function.
13276 (rewrite_use_address): Rewrite address uses in calls that were
13277 identified by find_address_like_use.
13279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13280 Alan Hayward <alan.hayward@arm.com>
13281 David Sherwood <david.sherwood@arm.com>
13283 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13285 * gimple-expr.h (is_gimple_addressable: Likewise.
13286 * gimple-expr.c (is_gimple_address): Likewise.
13287 * internal-fn.c (expand_call_mem_ref): New function.
13288 (expand_mask_load_optab_fn): Use it.
13289 (expand_mask_store_optab_fn): Likewise.
13291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13292 Alan Hayward <alan.hayward@arm.com>
13293 David Sherwood <david.sherwood@arm.com>
13295 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13296 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13297 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13298 (cond_umax@var{mode}): Document.
13299 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13300 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13301 (cond_umin_optab, cond_umax_optab): New optabs.
13302 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13303 (COND_IOR, COND_XOR): New internal functions.
13304 * internal-fn.h (get_conditional_internal_fn): Declare.
13305 * internal-fn.c (cond_binary_direct): New macro.
13306 (expand_cond_binary_optab_fn): Likewise.
13307 (direct_cond_binary_optab_supported_p): Likewise.
13308 (get_conditional_internal_fn): New function.
13309 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13310 Cope with reduction statements that are vectorized as calls rather
13312 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13313 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13314 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13315 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13316 (UNSPEC_COND_EOR): New unspecs.
13317 (optab): Add mappings for them.
13318 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13319 (sve_int_op, sve_fp_op): New int attributes.
13321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13322 Alan Hayward <alan.hayward@arm.com>
13323 David Sherwood <david.sherwood@arm.com>
13325 * optabs.def (while_ult_optab): New optab.
13326 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13327 * internal-fn.def (WHILE_ULT): New internal function.
13328 * internal-fn.h (direct_internal_fn_supported_p): New override
13329 that takes two types as argument.
13330 * internal-fn.c (while_direct): New macro.
13331 (expand_while_optab_fn): New function.
13332 (convert_optab_supported_p): Likewise.
13333 (direct_while_optab_supported_p): New macro.
13334 * wide-int.h (wi::udiv_ceil): New function.
13335 * tree-vectorizer.h (rgroup_masks): New structure.
13336 (vec_loop_masks): New typedef.
13337 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13338 and fully_masked_p.
13339 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13340 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13341 (vect_max_vf): New function.
13342 (slpeel_make_loop_iterate_ntimes): Delete.
13343 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13344 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13345 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13346 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13347 internal-fn.h, stor-layout.h and optabs-query.h.
13348 (vect_set_loop_mask): New function.
13349 (add_preheader_seq): Likewise.
13350 (add_header_seq): Likewise.
13351 (interleave_supported_p): Likewise.
13352 (vect_maybe_permute_loop_masks): Likewise.
13353 (vect_set_loop_masks_directly): Likewise.
13354 (vect_set_loop_condition_masked): Likewise.
13355 (vect_set_loop_condition_unmasked): New function, split out from
13356 slpeel_make_loop_iterate_ntimes.
13357 (slpeel_make_loop_iterate_ntimes): Rename to..
13358 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13359 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13360 (vect_do_peeling): Update call accordingly.
13361 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13363 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13364 mask_compare_type, can_fully_mask_p and fully_masked_p.
13365 (release_vec_loop_masks): New function.
13366 (_loop_vec_info): Use it to free the loop masks.
13367 (can_produce_all_loop_masks_p): New function.
13368 (vect_get_max_nscalars_per_iter): Likewise.
13369 (vect_verify_full_masking): Likewise.
13370 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13371 retries, and free the mask rgroups before retrying. Check loop-wide
13372 reasons for disallowing fully-masked loops. Make the final decision
13373 about whether use a fully-masked loop or not.
13374 (vect_estimate_min_profitable_iters): Do not assume that peeling
13375 for the number of iterations will be needed for fully-masked loops.
13376 (vectorizable_reduction): Disable fully-masked loops.
13377 (vectorizable_live_operation): Likewise.
13378 (vect_halve_mask_nunits): New function.
13379 (vect_double_mask_nunits): Likewise.
13380 (vect_record_loop_mask): Likewise.
13381 (vect_get_loop_mask): Likewise.
13382 (vect_transform_loop): Handle the case in which the final loop
13383 iteration might handle a partial vector. Call vect_set_loop_condition
13384 instead of slpeel_make_loop_iterate_ntimes.
13385 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13386 (check_load_store_masking): New function.
13387 (prepare_load_store_mask): Likewise.
13388 (vectorizable_store): Handle fully-masked loops.
13389 (vectorizable_load): Likewise.
13390 (supportable_widening_operation): Use vect_halve_mask_nunits for
13392 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13393 (vect_gen_while): New function.
13394 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13395 (aarch64_uqdec<mode>): New insn.
13397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13398 Alan Hayward <alan.hayward@arm.com>
13399 David Sherwood <david.sherwood@arm.com>
13401 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13402 (reduc_xor_scal_optab): New optabs.
13403 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13404 (reduc_xor_scal_@var{m}): Document.
13405 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13406 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13407 internal functions.
13408 * fold-const-call.c (fold_const_call): Handle them.
13409 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13410 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13411 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13412 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13413 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13414 (UNSPEC_XORV): New unspecs.
13415 (optab): Add entries for them.
13416 (BITWISEV): New int iterator.
13417 (bit_reduc_op): New int attributes.
13419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13420 Alan Hayward <alan.hayward@arm.com>
13421 David Sherwood <david.sherwood@arm.com>
13423 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13424 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13425 * optabs.def (vec_shl_insert_optab): New optab.
13426 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13427 (duplicate_and_interleave): Likewise.
13428 * tree-vect-loop.c: Include internal-fn.h.
13429 (neutral_op_for_slp_reduction): New function, split out from
13430 get_initial_defs_for_reduction.
13431 (get_initial_def_for_reduction): Handle option 2 for variable-length
13432 vectors by loading the neutral value into a vector and then shifting
13433 the initial value into element 0.
13434 (get_initial_defs_for_reduction): Replace the code argument with
13435 the neutral value calculated by neutral_op_for_slp_reduction.
13436 Use gimple_build_vector for constant-length vectors.
13437 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13438 but the first group_size elements have a neutral value.
13439 Use duplicate_and_interleave otherwise.
13440 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13441 Update call to get_initial_defs_for_reduction. Handle SLP
13442 reductions for variable-length vectors by creating one vector
13443 result for each scalar result, with the elements associated
13444 with other scalar results stubbed out with the neutral value.
13445 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13446 Require IFN_VEC_SHL_INSERT for double reductions on
13447 variable-length vectors, or SLP reductions that have
13448 a neutral value. Require can_duplicate_and_interleave_p
13449 support for variable-length unchained SLP reductions if there
13450 is no neutral value, such as for MIN/MAX reductions. Also require
13451 the number of vector elements to be a multiple of the number of
13452 SLP statements when doing variable-length unchained SLP reductions.
13453 Update call to vect_create_epilog_for_reduction.
13454 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13455 and remove initial values.
13456 (duplicate_and_interleave): Make public.
13457 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13458 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13461 Alan Hayward <alan.hayward@arm.com>
13462 David Sherwood <david.sherwood@arm.com>
13464 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13465 (can_duplicate_and_interleave_p): New function.
13466 (vect_get_and_check_slp_defs): Take the vector of statements
13467 rather than just the current one. Remove excess parentheses.
13468 Restriction rejectinon of vect_constant_def and vect_external_def
13469 for variable-length vectors to boolean types, or types for which
13470 can_duplicate_and_interleave_p is false.
13471 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13472 (duplicate_and_interleave): New function.
13473 (vect_get_constant_vectors): Use gimple_build_vector for
13474 constant-length vectors and suitable variable-length constant
13475 vectors. Use duplicate_and_interleave for other variable-length
13476 vectors. Don't defer the update when inserting new statements.
13478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13479 Alan Hayward <alan.hayward@arm.com>
13480 David Sherwood <david.sherwood@arm.com>
13482 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13483 min_profitable_iters doesn't go negative.
13485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13486 Alan Hayward <alan.hayward@arm.com>
13487 David Sherwood <david.sherwood@arm.com>
13489 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13490 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13491 * optabs.def (vec_mask_load_lanes_optab): New optab.
13492 (vec_mask_store_lanes_optab): Likewise.
13493 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13494 (MASK_STORE_LANES): Likewise.
13495 * internal-fn.c (mask_load_lanes_direct): New macro.
13496 (mask_store_lanes_direct): Likewise.
13497 (expand_mask_load_optab_fn): Handle masked operations.
13498 (expand_mask_load_lanes_optab_fn): New macro.
13499 (expand_mask_store_optab_fn): Handle masked operations.
13500 (expand_mask_store_lanes_optab_fn): New macro.
13501 (direct_mask_load_lanes_optab_supported_p): Likewise.
13502 (direct_mask_store_lanes_optab_supported_p): Likewise.
13503 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13505 (vect_load_lanes_supported): Likewise.
13506 * tree-vect-data-refs.c (strip_conversion): New function.
13507 (can_group_stmts_p): Likewise.
13508 (vect_analyze_data_ref_accesses): Use it instead of checking
13509 for a pair of assignments.
13510 (vect_store_lanes_supported): Take a masked_p parameter.
13511 (vect_load_lanes_supported): Likewise.
13512 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13513 vect_store_lanes_supported and vect_load_lanes_supported.
13514 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13515 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13516 parameter. Don't allow gaps for masked accesses.
13517 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13518 and vect_load_lanes_supported.
13519 (get_load_store_type): Take a masked_p parameter and update
13520 call to get_group_load_store_type.
13521 (vectorizable_store): Update call to get_load_store_type.
13522 Handle IFN_MASK_STORE_LANES.
13523 (vectorizable_load): Update call to get_load_store_type.
13524 Handle IFN_MASK_LOAD_LANES.
13526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13527 Alan Hayward <alan.hayward@arm.com>
13528 David Sherwood <david.sherwood@arm.com>
13530 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13532 * config/aarch64/aarch64-protos.h
13533 (aarch64_sve_struct_memory_operand_p): Declare.
13534 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13535 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13536 (VPRED, vpred): Handle SVE structure modes.
13537 * config/aarch64/constraints.md (Utx): New constraint.
13538 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13539 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13540 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13541 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13542 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13543 structure modes. Split into pieces after RA.
13544 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13545 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13547 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13548 SVE structure modes.
13549 (aarch64_classify_address): Likewise.
13550 (sizetochar): Move earlier in file.
13551 (aarch64_print_operand): Handle SVE register lists.
13552 (aarch64_array_mode): New function.
13553 (aarch64_sve_struct_memory_operand_p): Likewise.
13554 (TARGET_ARRAY_MODE): Redefine.
13556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13557 Alan Hayward <alan.hayward@arm.com>
13558 David Sherwood <david.sherwood@arm.com>
13560 * target.def (array_mode): New target hook.
13561 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13562 * doc/tm.texi: Regenerate.
13563 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13564 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13565 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13566 targetm.array_mode.
13567 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13571 Alan Hayward <alan.hayward@arm.com>
13572 David Sherwood <david.sherwood@arm.com>
13574 * fold-const.c (fold_binary_loc): Check the argument types
13575 rather than the result type when testing for a vector operation.
13577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13579 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13580 * doc/tm.texi: Regenerate.
13582 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13583 Alan Hayward <alan.hayward@arm.com>
13584 David Sherwood <david.sherwood@arm.com>
13586 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13587 (sve): Document new AArch64 extension.
13588 * doc/md.texi (w): Extend the description of the AArch64
13589 constraint to include SVE vectors.
13590 (Upl, Upa): Document new AArch64 predicate constraints.
13591 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13593 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13594 (msve-vector-bits=): New option.
13595 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13596 SVE when these are disabled.
13597 (sve): New extension.
13598 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13599 modes. Adjust their number of units based on aarch64_sve_vg.
13600 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13601 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13602 aarch64_addr_query_type.
13603 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13604 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13605 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13606 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13607 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13608 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13609 (aarch64_simd_imm_zero_p): Delete.
13610 (aarch64_check_zero_based_sve_index_immediate): Declare.
13611 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13612 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13613 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13614 (aarch64_sve_float_mul_immediate_p): Likewise.
13615 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13616 rather than an rtx.
13617 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13618 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13619 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13620 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13621 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13622 (aarch64_regmode_natural_size): Likewise.
13623 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13624 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13626 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13627 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13628 for VG and the SVE predicate registers.
13629 (V_ALIASES): Add a "z"-prefixed alias.
13630 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13631 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13632 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13633 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13634 (REG_CLASS_NAMES): Add entries for them.
13635 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13636 and the predicate registers.
13637 (aarch64_sve_vg): Declare.
13638 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13639 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13640 (REGMODE_NATURAL_SIZE): Define.
13641 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13643 * config/aarch64/aarch64.c: Include cfgrtl.h.
13644 (simd_immediate_info): Add a constructor for series vectors,
13645 and an associated step field.
13646 (aarch64_sve_vg): New variable.
13647 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13648 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13649 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13650 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13651 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13652 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13653 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13654 (aarch64_get_mask_mode): New functions.
13655 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13656 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13657 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13658 predicate modes and predicate registers. Explicitly restrict
13659 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13660 to store a vector mode if it is recognized by
13661 aarch64_classify_vector_mode.
13662 (aarch64_regmode_natural_size): New function.
13663 (aarch64_hard_regno_caller_save_mode): Return the original mode
13665 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13666 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13667 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13668 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13670 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13671 does not overlap dest if the function is frame-related. Handle
13673 (aarch64_split_add_offset): New function.
13674 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13675 them aarch64_add_offset.
13676 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13677 and update call to aarch64_sub_sp.
13678 (aarch64_add_cfa_expression): New function.
13679 (aarch64_expand_prologue): Pass extra temporary registers to the
13680 functions above. Handle the case in which we need to emit new
13681 DW_CFA_expressions for registers that were originally saved
13682 relative to the stack pointer, but now have to be expressed
13683 relative to the frame pointer.
13684 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13686 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13687 IP0 and IP1 values for SVE frames.
13688 (aarch64_expand_vec_series): New function.
13689 (aarch64_expand_sve_widened_duplicate): Likewise.
13690 (aarch64_expand_sve_const_vector): Likewise.
13691 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13692 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13693 into the register, rather than emitting a SET directly.
13694 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13695 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13696 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13697 (offset_9bit_signed_scaled_p): New functions.
13698 (aarch64_replicate_bitmask_imm): New function.
13699 (aarch64_bitmask_imm): Use it.
13700 (aarch64_cannot_force_const_mem): Reject expressions involving
13701 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13702 (aarch64_classify_index): Handle SVE indices, by requiring
13703 a plain register index with a scale that matches the element size.
13704 (aarch64_classify_address): Handle SVE addresses. Assert that
13705 the mode of the address is VOIDmode or an integer mode.
13706 Update call to aarch64_classify_symbol.
13707 (aarch64_classify_symbolic_expression): Update call to
13708 aarch64_classify_symbol.
13709 (aarch64_const_vec_all_in_range_p): New function.
13710 (aarch64_print_vector_float_operand): Likewise.
13711 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13712 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13713 and the FP immediates 1.0 and 0.5.
13714 (aarch64_print_address_internal): Handle SVE addresses.
13715 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13716 (aarch64_regno_regclass): Handle predicate registers.
13717 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13719 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13720 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13721 (aarch64_convert_sve_vector_bits): New function.
13722 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13723 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13724 rather than an rtx.
13725 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13726 Handle SVE vector and predicate modes. Accept VL-based constants
13727 that need only one temporary register, and VL offsets that require
13728 no temporary registers.
13729 (aarch64_conditional_register_usage): Mark the predicate registers
13730 as fixed if SVE isn't available.
13731 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13732 Return true for SVE vector and predicate modes.
13733 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13734 rather than an unsigned int. Handle SVE modes.
13735 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13737 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13739 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13740 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13741 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13742 (aarch64_sve_float_mul_immediate_p): New functions.
13743 (aarch64_sve_valid_immediate): New function.
13744 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13745 Explicitly reject structure modes. Check for INDEX constants.
13746 Handle PTRUE and PFALSE constants.
13747 (aarch64_check_zero_based_sve_index_immediate): New function.
13748 (aarch64_simd_imm_zero_p): Delete.
13749 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13750 vector modes. Accept constants in the range of CNT[BHWD].
13751 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13752 ask for an Advanced SIMD mode.
13753 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13754 (aarch64_simd_vector_alignment): Handle SVE predicates.
13755 (aarch64_vectorize_preferred_vector_alignment): New function.
13756 (aarch64_simd_vector_alignment_reachable): Use it instead of
13758 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13759 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13761 (MAX_VECT_LEN): Delete.
13762 (expand_vec_perm_d): Add a vec_flags field.
13763 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13764 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13765 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13767 (aarch64_evpc_rev): Rename to...
13768 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13769 (aarch64_evpc_rev_global): New function.
13770 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13771 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13773 (aarch64_evpc_sve_tbl): New function.
13774 (aarch64_expand_vec_perm_const_1): Update after rename of
13775 aarch64_evpc_rev. Handle SVE permutes too, trying
13776 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13777 than aarch64_evpc_tbl.
13778 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13779 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13780 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13781 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13782 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13783 (aarch64_expand_sve_vcond): New functions.
13784 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13785 of aarch64_vector_mode_p.
13786 (aarch64_dwarf_poly_indeterminate_value): New function.
13787 (aarch64_compute_pressure_classes): Likewise.
13788 (aarch64_can_change_mode_class): Likewise.
13789 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13790 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13791 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13792 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13793 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13794 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13795 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13796 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13798 (Dn, Dl, Dr): Accept const as well as const_vector.
13799 (Dz): Likewise. Compare against CONST0_RTX.
13800 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13801 of "vector" where appropriate.
13802 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13803 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13804 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13805 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13806 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13807 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13808 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13809 (v_int_equiv): Extend to SVE modes.
13810 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13812 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13813 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13814 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13815 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13816 (SVE_COND_FP_CMP): New int iterators.
13817 (perm_hilo): Handle the new unpack unspecs.
13818 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13820 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13821 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13822 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13823 (aarch64_equality_operator, aarch64_constant_vector_operand)
13824 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13825 (aarch64_sve_nonimmediate_operand): Likewise.
13826 (aarch64_sve_general_operand): Likewise.
13827 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13828 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13829 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13830 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13831 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13832 (aarch64_sve_float_arith_immediate): Likewise.
13833 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13834 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13835 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13836 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13837 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13838 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13839 (aarch64_sve_float_arith_operand): Likewise.
13840 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13841 (aarch64_sve_float_mul_operand): Likewise.
13842 (aarch64_sve_vec_perm_operand): Likewise.
13843 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13844 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13845 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13846 as well as const_vector.
13847 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13848 in file. Use CONST0_RTX and CONSTM1_RTX.
13849 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13850 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13851 Use aarch64_simd_imm_zero.
13852 * config/aarch64/aarch64-sve.md: New file.
13853 * config/aarch64/aarch64.md: Include it.
13854 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13855 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13856 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13857 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13858 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13859 (sve): New attribute.
13860 (enabled): Disable instructions with the sve attribute unless
13862 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13863 aarch64_expand_mov_immediate.
13864 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13865 CNT[BHSD] immediates.
13866 (movti): Split CONST_POLY_INT moves into two halves.
13867 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13868 Split additions that need a temporary here if the destination
13869 is the stack pointer.
13870 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13871 (*add<mode>3_poly_1): New instruction.
13872 (set_clobber_cc): New expander.
13874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13876 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13877 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13878 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13879 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13880 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13881 Change innermode from fixed_mode_size to machine_mode.
13882 (simplify_subreg): Update call accordingly. Handle a constant-sized
13883 subreg of a variable-length CONST_VECTOR.
13885 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13886 Alan Hayward <alan.hayward@arm.com>
13887 David Sherwood <david.sherwood@arm.com>
13889 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13890 (add_offset_to_base): New function, split out from...
13891 (create_mem_ref): ...here. When handling a scale other than 1,
13892 check first whether the address is valid without the offset.
13893 Add it into the base if so, leaving the index and scale as-is.
13895 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13898 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13899 fold_for_warn before checking if arg2 is INTEGER_CST.
13901 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13903 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13904 (store_multiple_operation): Delete.
13905 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13906 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13907 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13908 guarded by TARGET_STRING.
13909 (rs6000_output_load_multiple): Delete.
13910 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13911 OPTION_MASK_STRING / TARGET_STRING handling.
13912 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13913 (const rs6000_opt_masks) <"string">: Change mask to 0.
13914 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13915 (MASK_STRING): Delete.
13916 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13918 (load_multiple): Delete.
13925 (store_multiple): Delete.
13932 (movmemsi_8reg): Delete.
13933 (corresponding unnamed define_insn): Delete.
13934 (movmemsi_6reg): Delete.
13935 (corresponding unnamed define_insn): Delete.
13936 (movmemsi_4reg): Delete.
13937 (corresponding unnamed define_insn): Delete.
13938 (movmemsi_2reg): Delete.
13939 (corresponding unnamed define_insn): Delete.
13940 (movmemsi_1reg): Delete.
13941 (corresponding unnamed define_insn): Delete.
13942 * config/rs6000/rs6000.opt (mno-string): New.
13943 (mstring): Replace by deprecation warning stub.
13944 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13946 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13948 * regrename.c (regrename_do_replace): If replacing the same
13949 reg multiple times, try to reuse last created gen_raw_REG.
13952 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13953 main to workaround a bug in GDB.
13955 2018-01-12 Tom de Vries <tom@codesourcery.com>
13958 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13960 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13962 PR rtl-optimization/80481
13963 * ira-color.c (get_cap_member): New function.
13964 (allocnos_conflict_by_live_ranges_p): Use it.
13965 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13966 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13968 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13971 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13972 (*saddl_se_1): Ditto.
13973 (*ssubsi_1): Ditto.
13974 (*ssubl_se_1): Ditto.
13976 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13978 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13979 rather than wi::to_widest for DR_INITs.
13980 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13981 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13982 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13984 (vect_analyze_group_access_1): Note that here.
13986 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13988 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13989 polynomial type sizes.
13991 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13993 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13994 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13995 (gimple_add_tmp_var): Likewise.
13997 2018-01-12 Martin Liska <mliska@suse.cz>
13999 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14000 (gimple_alloc_sizes): Likewise.
14001 (dump_gimple_statistics): Use PRIu64 in printf format.
14002 * gimple.h: Change uint64_t to int.
14004 2018-01-12 Martin Liska <mliska@suse.cz>
14006 * tree-core.h: Use uint64_t instead of int.
14007 * tree.c (tree_node_counts): Likewise.
14008 (tree_node_sizes): Likewise.
14009 (dump_tree_statistics): Use PRIu64 in printf format.
14011 2018-01-12 Martin Liska <mliska@suse.cz>
14013 * Makefile.in: As qsort_chk is implemented in vec.c, add
14014 vec.o to linkage of gencfn-macros.
14015 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14016 passing the info to record_node_allocation_statistics.
14017 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14019 * ggc-common.c (struct ggc_usage): Add operator== and use
14020 it in operator< and compare function.
14021 * mem-stats.h (struct mem_usage): Likewise.
14022 * vec.c (struct vec_usage): Remove operator< and compare
14023 function. Can be simply inherited.
14025 2018-01-12 Martin Jambor <mjambor@suse.cz>
14028 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14029 * tree-ssa-math-opts.c: Include domwalk.h.
14030 (convert_mult_to_fma_1): New function.
14031 (fma_transformation_info): New type.
14032 (fma_deferring_state): Likewise.
14033 (cancel_fma_deferring): New function.
14034 (result_of_phi): Likewise.
14035 (last_fma_candidate_feeds_initial_phi): Likewise.
14036 (convert_mult_to_fma): Added deferring logic, split actual
14037 transformation to convert_mult_to_fma_1.
14038 (math_opts_dom_walker): New type.
14039 (math_opts_dom_walker::after_dom_children): New method, body moved
14040 here from pass_optimize_widening_mul::execute, added deferring logic
14042 (pass_optimize_widening_mul::execute): Moved most of code to
14043 math_opts_dom_walker::after_dom_children.
14044 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14045 * config/i386/i386.c (ix86_option_override_internal): Added
14046 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14048 2018-01-12 Richard Biener <rguenther@suse.de>
14051 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14052 inline instance vars.
14054 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14057 * config/rx/rx.c (rx_is_restricted_memory_address):
14058 Handle SUBREG case.
14060 2018-01-12 Richard Biener <rguenther@suse.de>
14062 PR tree-optimization/80846
14063 * target.def (split_reduction): New target hook.
14064 * targhooks.c (default_split_reduction): New function.
14065 * targhooks.h (default_split_reduction): Declare.
14066 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14067 target requests first reduce vectors by combining low and high
14069 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14070 (get_vectype_for_scalar_type_and_size): Export.
14071 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14072 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14073 * doc/tm.texi: Regenerate.
14074 * config/i386/i386.c (ix86_split_reduction): Implement
14075 TARGET_VECTORIZE_SPLIT_REDUCTION.
14077 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14080 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14081 in PIC mode except for TARGET_VXWORKS_RTP.
14082 * config/sparc/sparc.c: Include cfgrtl.h.
14083 (TARGET_INIT_PIC_REG): Define.
14084 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14085 (sparc_pic_register_p): New predicate.
14086 (sparc_legitimate_address_p): Use it.
14087 (sparc_legitimize_pic_address): Likewise.
14088 (sparc_delegitimize_address): Likewise.
14089 (sparc_mode_dependent_address_p): Likewise.
14090 (gen_load_pcrel_sym): Remove 4th parameter.
14091 (load_got_register): Adjust call to above. Remove obsolete stuff.
14092 (sparc_expand_prologue): Do not call load_got_register here.
14093 (sparc_flat_expand_prologue): Likewise.
14094 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14095 (sparc_use_pseudo_pic_reg): New function.
14096 (sparc_init_pic_reg): Likewise.
14097 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14098 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14100 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14102 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14103 Add item for branch_cost.
14105 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14107 PR rtl-optimization/83565
14108 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14109 not extend the result to a larger mode for rotate operations.
14110 (num_sign_bit_copies1): Likewise.
14112 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14115 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14117 Use values-Xc.o for -pedantic.
14118 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14120 2018-01-12 Martin Liska <mliska@suse.cz>
14123 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14125 (possible_polymorphic_call_targets): Use it.
14126 (ipa_devirt): Likewise.
14128 2018-01-12 Martin Liska <mliska@suse.cz>
14130 * profile-count.h (enum profile_quality): Use 0 as invalid
14131 enum value of profile_quality.
14133 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14135 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14136 -mext-string options.
14138 2018-01-12 Richard Biener <rguenther@suse.de>
14140 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14141 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14142 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14144 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14146 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14148 * configure.ac (--with-long-double-format): Add support for the
14149 configuration option to change the default long double format on
14151 * config.gcc (powerpc*-linux*-*): Likewise.
14152 * configure: Regenerate.
14153 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14154 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14155 used without modification.
14157 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14159 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14160 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14161 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14162 MISC_BUILTIN_SPEC_BARRIER.
14163 (rs6000_init_builtins): Likewise.
14164 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14166 (speculation_barrier): New define_insn.
14167 * doc/extend.texi: Document __builtin_speculation_barrier.
14169 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14172 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14173 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14174 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14176 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14177 integral modes instead of "ss" and "sd".
14178 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14179 vectors with 32-bit and 64-bit elements.
14180 (vecdupssescalarmodesuffix): New mode attribute.
14181 (vec_dup<mode>): Use it.
14183 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14186 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14187 frame if argument is passed on stack.
14189 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14192 * ree.c (combine_reaching_defs): Optimize also
14193 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14194 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14196 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14198 PR middle-end/83189
14199 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14201 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14203 PR middle-end/83718
14204 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14205 after they are computed.
14207 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14209 PR tree-optimization/83695
14210 * gimple-loop-linterchange.cc
14211 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14212 reset cached scev information after interchange.
14213 (pass_linterchange::execute): Remove call to scev_reset_htab.
14215 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14217 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14218 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14219 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14220 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14221 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14222 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14223 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14224 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14225 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14226 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14227 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14228 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14229 (V_lane_reg): Likewise.
14230 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14232 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14233 (vfmal_lane_low<mode>_intrinsic,
14234 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14235 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14236 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14237 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14238 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14239 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14241 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14243 * config/arm/arm-cpus.in (fp16fml): New feature.
14244 (ALL_SIMD): Add fp16fml.
14245 (armv8.2-a): Add fp16fml as an option.
14246 (armv8.3-a): Likewise.
14247 (armv8.4-a): Add fp16fml as part of fp16.
14248 * config/arm/arm.h (TARGET_FP16FML): Define.
14249 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14251 * config/arm/arm-modes.def (V2HF): Define.
14252 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14253 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14254 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14255 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14256 vfmsl_low, vfmsl_high): New set of builtins.
14257 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14258 (vfml_op): New code attribute.
14259 (VFMLHALVES): New int iterator.
14260 (VFML, VFMLSEL): New mode attributes.
14261 (V_reg): Define mapping for V2HF.
14262 (V_hi, V_lo): New mode attributes.
14263 (VF_constraint): Likewise.
14264 (vfml_half, vfml_half_selector): New int attributes.
14265 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14267 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14268 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14270 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14271 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14272 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14273 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14275 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14276 Document new effective target and option set.
14278 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14280 * config/arm/arm-cpus.in (armv8_4): New feature.
14281 (ARMv8_4a): New fgroup.
14282 (armv8.4-a): New arch.
14283 * config/arm/arm-tables.opt: Regenerate.
14284 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14285 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14286 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14287 Add matching rules for -march=armv8.4-a and extensions.
14288 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14290 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14293 * config/rx/rx.md (BW): New mode attribute.
14294 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14296 2018-01-11 Richard Biener <rguenther@suse.de>
14298 PR tree-optimization/83435
14299 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14300 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14301 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14303 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14304 Alan Hayward <alan.hayward@arm.com>
14305 David Sherwood <david.sherwood@arm.com>
14307 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14309 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14310 (aarch64_print_address_internal): Use it to check for a zero offset.
14312 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14313 Alan Hayward <alan.hayward@arm.com>
14314 David Sherwood <david.sherwood@arm.com>
14316 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14317 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14318 Return a poly_int64 rather than a HOST_WIDE_INT.
14319 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14320 rather than a HOST_WIDE_INT.
14321 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14322 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14323 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14324 final_offset from HOST_WIDE_INT to poly_int64.
14325 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14326 to_constant when getting the number of units in an Advanced SIMD
14328 (aarch64_builtin_vectorized_function): Check for a constant number
14330 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14332 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14333 attribute instead of GET_MODE_NUNITS.
14334 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14335 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14336 GET_MODE_SIZE for fixed-size registers.
14337 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14338 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14339 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14340 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14341 (aarch64_print_operand, aarch64_print_address_internal)
14342 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14343 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14344 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14345 Handle polynomial GET_MODE_SIZE.
14346 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14347 wider than SImode without modification.
14348 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14349 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14350 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14351 passing and returning SVE modes.
14352 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14353 rather than GEN_INT.
14354 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14355 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14356 (aarch64_allocate_and_probe_stack_space): Likewise.
14357 (aarch64_layout_frame): Cope with polynomial offsets.
14358 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14359 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14360 polynomial offsets.
14361 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14362 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14363 poly_int64 rather than a HOST_WIDE_INT.
14364 (aarch64_get_separate_components, aarch64_process_components)
14365 (aarch64_expand_prologue, aarch64_expand_epilogue)
14366 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14367 (aarch64_anchor_offset): New function, split out from...
14368 (aarch64_legitimize_address): ...here.
14369 (aarch64_builtin_vectorization_cost): Handle polynomial
14370 TYPE_VECTOR_SUBPARTS.
14371 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14373 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14374 number of elements from the PARALLEL rather than the mode.
14375 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14376 rather than GET_MODE_BITSIZE.
14377 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14378 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14379 (aarch64_expand_vec_perm_const_1): Handle polynomial
14380 d->perm.length () and d->perm elements.
14381 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14382 Apply to_constant to d->perm elements.
14383 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14384 polynomial CONST_VECTOR_NUNITS.
14385 (aarch64_move_pointer): Take amount as a poly_int64 rather
14387 (aarch64_progress_pointer): Avoid temporary variable.
14388 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14389 the mode attribute instead of GET_MODE.
14391 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14392 Alan Hayward <alan.hayward@arm.com>
14393 David Sherwood <david.sherwood@arm.com>
14395 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14396 x exists before using it.
14397 (aarch64_add_constant_internal): Rename to...
14398 (aarch64_add_offset_1): ...this. Replace regnum with separate
14399 src and dest rtxes. Handle the case in which they're different,
14400 including when the offset is zero. Replace scratchreg with an rtx.
14401 Use 2 additions if there is no spare register into which we can
14402 move a 16-bit constant.
14403 (aarch64_add_constant): Delete.
14404 (aarch64_add_offset): Replace reg with separate src and dest
14405 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14406 Use aarch64_add_offset_1.
14407 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14408 an rtx rather than an int. Take the delta as a poly_int64
14409 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14410 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14411 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14412 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14413 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14414 and aarch64_add_sp.
14415 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14416 aarch64_add_constant.
14418 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14420 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14421 Use scalar_float_mode.
14423 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14425 * config/aarch64/aarch64-simd.md
14426 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14427 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14428 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14429 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14430 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14431 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14432 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14433 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14434 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14435 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14437 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14440 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14441 targ_options->x_arm_arch_string is non NULL.
14443 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14445 * config/aarch64/aarch64.h
14446 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14448 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14451 * expmed.c (emit_store_flag_force): Swap if const op0
14452 and change VOIDmode to mode of op0.
14454 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14456 PR rtl-optimization/83761
14457 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14458 than bytes to mode_for_size.
14460 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14462 PR middle-end/83189
14463 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14464 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14467 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14469 PR middle-end/83575
14470 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14471 when in layout mode.
14472 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14473 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14476 2018-01-10 Michael Collison <michael.collison@arm.com>
14478 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14479 * config/aarch64/aarch64-option-extension.def: Add
14480 AARCH64_OPT_EXTENSION of 'fp16fml'.
14481 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14482 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14483 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14484 * config/aarch64/constraints.md (Ui7): New constraint.
14485 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14486 (VFMLA_SEL_W): Ditto.
14489 (VFMLA16_LOW): New int iterator.
14490 (VFMLA16_HIGH): Ditto.
14491 (UNSPEC_FMLAL): New unspec.
14492 (UNSPEC_FMLSL): Ditto.
14493 (UNSPEC_FMLAL2): Ditto.
14494 (UNSPEC_FMLSL2): Ditto.
14495 (f16mac): New code attribute.
14496 * config/aarch64/aarch64-simd-builtins.def
14497 (aarch64_fmlal_lowv2sf): Ditto.
14498 (aarch64_fmlsl_lowv2sf): Ditto.
14499 (aarch64_fmlalq_lowv4sf): Ditto.
14500 (aarch64_fmlslq_lowv4sf): Ditto.
14501 (aarch64_fmlal_highv2sf): Ditto.
14502 (aarch64_fmlsl_highv2sf): Ditto.
14503 (aarch64_fmlalq_highv4sf): Ditto.
14504 (aarch64_fmlslq_highv4sf): Ditto.
14505 (aarch64_fmlal_lane_lowv2sf): Ditto.
14506 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14507 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14508 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14509 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14510 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14511 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14512 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14513 (aarch64_fmlal_lane_highv2sf): Ditto.
14514 (aarch64_fmlsl_lane_highv2sf): Ditto.
14515 (aarch64_fmlal_laneq_highv2sf): Ditto.
14516 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14517 (aarch64_fmlalq_lane_highv4sf): Ditto.
14518 (aarch64_fmlsl_lane_highv4sf): Ditto.
14519 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14520 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14521 * config/aarch64/aarch64-simd.md:
14522 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14523 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14524 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14525 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14526 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14527 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14528 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14529 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14530 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14531 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14532 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14533 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14534 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14535 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14536 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14537 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14538 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14539 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14540 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14541 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14542 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14543 (vfmlsl_low_u32): Ditto.
14544 (vfmlalq_low_u32): Ditto.
14545 (vfmlslq_low_u32): Ditto.
14546 (vfmlal_high_u32): Ditto.
14547 (vfmlsl_high_u32): Ditto.
14548 (vfmlalq_high_u32): Ditto.
14549 (vfmlslq_high_u32): Ditto.
14550 (vfmlal_lane_low_u32): Ditto.
14551 (vfmlsl_lane_low_u32): Ditto.
14552 (vfmlal_laneq_low_u32): Ditto.
14553 (vfmlsl_laneq_low_u32): Ditto.
14554 (vfmlalq_lane_low_u32): Ditto.
14555 (vfmlslq_lane_low_u32): Ditto.
14556 (vfmlalq_laneq_low_u32): Ditto.
14557 (vfmlslq_laneq_low_u32): Ditto.
14558 (vfmlal_lane_high_u32): Ditto.
14559 (vfmlsl_lane_high_u32): Ditto.
14560 (vfmlal_laneq_high_u32): Ditto.
14561 (vfmlsl_laneq_high_u32): Ditto.
14562 (vfmlalq_lane_high_u32): Ditto.
14563 (vfmlslq_lane_high_u32): Ditto.
14564 (vfmlalq_laneq_high_u32): Ditto.
14565 (vfmlslq_laneq_high_u32): Ditto.
14566 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14567 (AARCH64_FL_FOR_ARCH8_4): New.
14568 (AARCH64_ISA_F16FML): New ISA flag.
14569 (TARGET_F16FML): New feature flag for fp16fml.
14570 (doc/invoke.texi): Document new fp16fml option.
14572 2018-01-10 Michael Collison <michael.collison@arm.com>
14574 * config/aarch64/aarch64-builtins.c:
14575 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14576 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14577 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14578 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14579 (AARCH64_ISA_SHA3): New ISA flag.
14580 (TARGET_SHA3): New feature flag for sha3.
14581 * config/aarch64/iterators.md (sha512_op): New int attribute.
14582 (CRYPTO_SHA512): New int iterator.
14583 (UNSPEC_SHA512H): New unspec.
14584 (UNSPEC_SHA512H2): Ditto.
14585 (UNSPEC_SHA512SU0): Ditto.
14586 (UNSPEC_SHA512SU1): Ditto.
14587 * config/aarch64/aarch64-simd-builtins.def
14588 (aarch64_crypto_sha512hqv2di): New builtin.
14589 (aarch64_crypto_sha512h2qv2di): Ditto.
14590 (aarch64_crypto_sha512su0qv2di): Ditto.
14591 (aarch64_crypto_sha512su1qv2di): Ditto.
14592 (aarch64_eor3qv8hi): Ditto.
14593 (aarch64_rax1qv2di): Ditto.
14594 (aarch64_xarqv2di): Ditto.
14595 (aarch64_bcaxqv8hi): Ditto.
14596 * config/aarch64/aarch64-simd.md:
14597 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14598 (aarch64_crypto_sha512su0qv2di): Ditto.
14599 (aarch64_crypto_sha512su1qv2di): Ditto.
14600 (aarch64_eor3qv8hi): Ditto.
14601 (aarch64_rax1qv2di): Ditto.
14602 (aarch64_xarqv2di): Ditto.
14603 (aarch64_bcaxqv8hi): Ditto.
14604 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14605 (vsha512h2q_u64): Ditto.
14606 (vsha512su0q_u64): Ditto.
14607 (vsha512su1q_u64): Ditto.
14608 (veor3q_u16): Ditto.
14609 (vrax1q_u64): Ditto.
14610 (vxarq_u64): Ditto.
14611 (vbcaxq_u16): Ditto.
14612 * config/arm/types.md (crypto_sha512): New type attribute.
14613 (crypto_sha3): Ditto.
14614 (doc/invoke.texi): Document new sha3 option.
14616 2018-01-10 Michael Collison <michael.collison@arm.com>
14618 * config/aarch64/aarch64-builtins.c:
14619 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14620 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14621 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14622 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14623 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14624 (AARCH64_ISA_SM4): New ISA flag.
14625 (TARGET_SM4): New feature flag for sm4.
14626 * config/aarch64/aarch64-simd-builtins.def
14627 (aarch64_sm3ss1qv4si): Ditto.
14628 (aarch64_sm3tt1aq4si): Ditto.
14629 (aarch64_sm3tt1bq4si): Ditto.
14630 (aarch64_sm3tt2aq4si): Ditto.
14631 (aarch64_sm3tt2bq4si): Ditto.
14632 (aarch64_sm3partw1qv4si): Ditto.
14633 (aarch64_sm3partw2qv4si): Ditto.
14634 (aarch64_sm4eqv4si): Ditto.
14635 (aarch64_sm4ekeyqv4si): Ditto.
14636 * config/aarch64/aarch64-simd.md:
14637 (aarch64_sm3ss1qv4si): Ditto.
14638 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14639 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14640 (aarch64_sm4eqv4si): Ditto.
14641 (aarch64_sm4ekeyqv4si): Ditto.
14642 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14643 (sm3part_op): Ditto.
14644 (CRYPTO_SM3TT): Ditto.
14645 (CRYPTO_SM3PART): Ditto.
14646 (UNSPEC_SM3SS1): New unspec.
14647 (UNSPEC_SM3TT1A): Ditto.
14648 (UNSPEC_SM3TT1B): Ditto.
14649 (UNSPEC_SM3TT2A): Ditto.
14650 (UNSPEC_SM3TT2B): Ditto.
14651 (UNSPEC_SM3PARTW1): Ditto.
14652 (UNSPEC_SM3PARTW2): Ditto.
14653 (UNSPEC_SM4E): Ditto.
14654 (UNSPEC_SM4EKEY): Ditto.
14655 * config/aarch64/constraints.md (Ui2): New constraint.
14656 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14657 * config/arm/types.md (crypto_sm3): New type attribute.
14658 (crypto_sm4): Ditto.
14659 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14660 (vsm3tt1aq_u32): Ditto.
14661 (vsm3tt1bq_u32): Ditto.
14662 (vsm3tt2aq_u32): Ditto.
14663 (vsm3tt2bq_u32): Ditto.
14664 (vsm3partw1q_u32): Ditto.
14665 (vsm3partw2q_u32): Ditto.
14666 (vsm4eq_u32): Ditto.
14667 (vsm4ekeyq_u32): Ditto.
14668 (doc/invoke.texi): Document new sm4 option.
14670 2018-01-10 Michael Collison <michael.collison@arm.com>
14672 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14673 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14674 (AARCH64_FL_FOR_ARCH8_4): New.
14675 (AARCH64_FL_V8_4): New flag.
14676 (doc/invoke.texi): Document new armv8.4-a option.
14678 2018-01-10 Michael Collison <michael.collison@arm.com>
14680 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14681 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14682 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14683 * config/aarch64/aarch64-option-extension.def: Add
14684 AARCH64_OPT_EXTENSION of 'sha2'.
14685 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14686 (crypto): Disable sha2 and aes if crypto disabled.
14687 (crypto): Enable aes and sha2 if enabled.
14688 (simd): Disable sha2 and aes if simd disabled.
14689 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14691 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14692 (TARGET_SHA2): New feature flag for sha2.
14693 (TARGET_AES): New feature flag for aes.
14694 * config/aarch64/aarch64-simd.md:
14695 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14696 conditional on TARGET_AES.
14697 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14698 (aarch64_crypto_sha1hsi): Make pattern conditional
14700 (aarch64_crypto_sha1hv4si): Ditto.
14701 (aarch64_be_crypto_sha1hv4si): Ditto.
14702 (aarch64_crypto_sha1su1v4si): Ditto.
14703 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14704 (aarch64_crypto_sha1su0v4si): Ditto.
14705 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14706 (aarch64_crypto_sha256su0v4si): Ditto.
14707 (aarch64_crypto_sha256su1v4si): Ditto.
14708 (doc/invoke.texi): Document new aes and sha2 options.
14710 2018-01-10 Martin Sebor <msebor@redhat.com>
14712 PR tree-optimization/83781
14713 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14716 2018-01-11 Martin Sebor <msebor@gmail.com>
14717 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14719 PR tree-optimization/83501
14720 PR tree-optimization/81703
14722 * tree-ssa-strlen.c (get_string_cst): Rename...
14723 (get_string_len): ...to this. Handle global constants.
14724 (handle_char_store): Adjust.
14726 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14727 Jim Wilson <jimw@sifive.com>
14729 * config/riscv/riscv-protos.h (riscv_output_return): New.
14730 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14731 (riscv_attribute_table, riscv_output_return),
14732 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14733 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14734 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14736 (riscv_expand_prologue): Add early return for naked function.
14737 (riscv_expand_epilogue): Likewise.
14738 (riscv_function_ok_for_sibcall): Return false for naked function.
14739 (riscv_set_current_function): New.
14740 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14741 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14742 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14743 * doc/extend.texi (RISC-V Function Attributes): New.
14745 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14747 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14748 check for 128-bit long double before checking TCmode.
14749 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14750 128-bit long doubles before checking TFmode or TCmode.
14751 (FLOAT128_IBM_P): Likewise.
14753 2018-01-10 Martin Sebor <msebor@redhat.com>
14755 PR tree-optimization/83671
14756 * builtins.c (c_strlen): Unconditionally return zero for the empty
14758 Use -Warray-bounds for warnings.
14759 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14760 for non-constant array indices with COMPONENT_REF, arrays of
14761 arrays, and pointers to arrays.
14762 (gimple_fold_builtin_strlen): Determine and set length range for
14763 non-constant character arrays.
14765 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14767 PR middle-end/81897
14768 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14771 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14773 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14775 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14778 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14779 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14780 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14781 indexed_or_indirect_operand predicate.
14782 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14783 (*vsx_le_perm_load_v8hi): Likewise.
14784 (*vsx_le_perm_load_v16qi): Likewise.
14785 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14786 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14787 (*vsx_le_perm_store_v8hi): Likewise.
14788 (*vsx_le_perm_store_v16qi): Likewise.
14789 (eight unnamed splitters): Likewise.
14791 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14793 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14794 * config/rs6000/emmintrin.h: Likewise.
14795 * config/rs6000/mmintrin.h: Likewise.
14796 * config/rs6000/xmmintrin.h: Likewise.
14798 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14801 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14803 * tree.c (tree_nop_conversion): Return true for location wrapper
14805 (maybe_wrap_with_location): New function.
14806 (selftest::check_strip_nops): New function.
14807 (selftest::test_location_wrappers): New function.
14808 (selftest::tree_c_tests): Call it.
14809 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14810 (maybe_wrap_with_location): New decl.
14811 (EXPR_LOCATION_WRAPPER_P): New macro.
14812 (location_wrapper_p): New inline function.
14813 (tree_strip_any_location_wrapper): New inline function.
14815 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14818 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14819 stack_realign_offset for the largest alignment of stack slot
14821 (ix86_find_max_used_stack_alignment): New function.
14822 (ix86_finalize_stack_frame_flags): Use it. Set
14823 max_used_stack_alignment if we don't realign stack.
14824 * config/i386/i386.h (machine_function): Add
14825 max_used_stack_alignment.
14827 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14829 * config/arm/arm.opt (-mbranch-cost): New option.
14830 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14833 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14836 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14837 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14839 2018-01-10 Richard Biener <rguenther@suse.de>
14842 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14843 early out so it also covers the case where we have a non-NULL
14846 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14848 PR tree-optimization/83753
14849 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14850 for non-strided grouped accesses if the number of elements is 1.
14852 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14855 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14856 * i386.h (TARGET_USE_GATHER): Define.
14857 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14859 2018-01-10 Martin Liska <mliska@suse.cz>
14862 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14863 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14865 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14866 CLEANUP_NO_PARTITIONING is not set.
14868 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14870 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14871 for vectors, as a partial revert of r254296.
14872 * rtl.h (const_vec_p): Delete.
14873 (const_vec_duplicate_p): Don't test for vector CONSTs.
14874 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14875 * expmed.c (make_tree): Likewise.
14878 * common.md (E, F): Use CONSTANT_P instead of checking for
14880 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14881 checking for CONST_VECTOR.
14883 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14885 PR middle-end/83575
14886 * predict.c (force_edge_cold): Handle in more sane way edges
14887 with no prediction.
14889 2018-01-09 Carl Love <cel@us.ibm.com>
14891 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14893 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14894 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14895 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14896 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14897 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14898 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14899 * config/rs6000/rs6000-protos.h: Add extern defition for
14900 rs6000_generate_float2_double_code.
14901 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14903 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14904 (float2_v2df): Add define_expand.
14906 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14909 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14910 op_mode in the force_to_mode call.
14912 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14914 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14915 instead of checking each element individually.
14916 (aarch64_evpc_uzp): Likewise.
14917 (aarch64_evpc_zip): Likewise.
14918 (aarch64_evpc_ext): Likewise.
14919 (aarch64_evpc_rev): Likewise.
14920 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14921 instead of checking each element individually. Return true without
14923 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14924 whether all selected elements come from the same input, instead of
14925 checking each element individually. Remove calls to gen_rtx_REG,
14926 start_sequence and end_sequence and instead assert that no rtl is
14929 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14931 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14932 order of HIGH and CONST checks.
14934 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14936 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14937 if the destination isn't an SSA_NAME.
14939 2018-01-09 Richard Biener <rguenther@suse.de>
14941 PR tree-optimization/83668
14942 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14944 (canonicalize_loop_form): ... here, renamed from ...
14945 (canonicalize_loop_closed_ssa_form): ... this and amended to
14946 swap successor edges for loop exit blocks to make us use
14947 the RPO order we need for initial schedule generation.
14949 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14951 PR tree-optimization/64811
14952 * match.pd: When optimizing comparisons with Inf, avoid
14953 introducing or losing exceptions from comparisons with NaN.
14955 2018-01-09 Martin Liska <mliska@suse.cz>
14958 * asan.c (shadow_mem_size): Add gcc_assert.
14960 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14962 Don't save registers in main().
14965 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14966 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14967 * config/avr/avr.c (avr_set_current_function): Don't error if
14968 naked, OS_task or OS_main are specified at the same time.
14969 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14971 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14973 * common/config/avr/avr-common.c (avr_option_optimization_table):
14974 Switch on -mmain-is-OS_task for optimizing compilations.
14976 2018-01-09 Richard Biener <rguenther@suse.de>
14978 PR tree-optimization/83572
14979 * graphite.c: Include cfganal.h.
14980 (graphite_transform_loops): Connect infinite loops to exit
14981 and remove fake edges at the end.
14983 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14985 * ipa-inline.c (edge_badness): Revert accidental checkin.
14987 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14990 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14991 symbols; not inline clones.
14993 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14996 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14997 hard registers. Formatting fixes.
14999 PR preprocessor/83722
15000 * gcc.c (try_generate_repro): Pass
15001 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15002 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15005 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15006 Kito Cheng <kito.cheng@gmail.com>
15008 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15009 (riscv_leaf_function_p): Delete.
15010 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15012 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15014 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15016 (do_ifelse): New function.
15017 (do_isel): New function.
15018 (do_sub3): New function.
15019 (do_add3): New function.
15020 (do_load_mask_compare): New function.
15021 (do_overlap_load_compare): New function.
15022 (expand_compare_loop): New function.
15023 (expand_block_compare): Call expand_compare_loop() when appropriate.
15024 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15025 option description.
15026 (-mblock-compare-inline-loop-limit): New option.
15028 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15031 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15032 Reverse order of second and third operands in first alternative.
15033 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15034 of first and second elements in UNSPEC_VPERMR vector.
15035 (altivec_expand_vec_perm_le): Likewise.
15037 2018-01-08 Jeff Law <law@redhat.com>
15039 PR rtl-optimizatin/81308
15040 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15041 (process_switch): If group_case_labels makes a change, then set
15043 (pass_convert_switch::execute): If a switch is converted, then
15044 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15046 PR rtl-optimization/81308
15047 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15050 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15052 PR target/83663 - Revert r255946
15053 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15054 generation for cases where splatting a value is not useful.
15055 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15056 across a vec_duplicate and a paradoxical subreg forming a vector
15057 mode to a vec_concat.
15059 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15061 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15062 -march=armv8.3-a variants.
15063 * config/arm/t-multilib: Likewise.
15064 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15066 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15068 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15070 (cceq_ior_compare_complement): Give it a name so I can use it, and
15071 change boolean_or_operator predicate to boolean_operator so it can
15072 be used to generate a crand.
15073 (eqne): New code iterator.
15074 (bd/bd_neg): New code_attrs.
15075 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15076 a single define_insn.
15077 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15078 decrement (bdnzt/bdnzf/bdzt/bdzf).
15079 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15080 with the new names of the branch decrement patterns, and added the
15081 names of the branch decrement conditional patterns.
15083 2018-01-08 Richard Biener <rguenther@suse.de>
15085 PR tree-optimization/83563
15086 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15089 2018-01-08 Richard Biener <rguenther@suse.de>
15091 PR middle-end/83713
15092 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15094 2018-01-08 Richard Biener <rguenther@suse.de>
15096 PR tree-optimization/83685
15097 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15098 references to abnormals.
15100 2018-01-08 Richard Biener <rguenther@suse.de>
15103 * dwarf2out.c (output_indirect_strings): Handle empty
15104 skeleton_debug_str_hash.
15105 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15107 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15109 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15110 (emit_store_direct): Likewise.
15111 (arc_trampoline_adjust_address): Likewise.
15112 (arc_asm_trampoline_template): New function.
15113 (arc_initialize_trampoline): Use asm_trampoline_template.
15114 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15115 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15116 * config/arc/arc.md (flush_icache): Delete pattern.
15118 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15120 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15121 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15124 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15127 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15128 by not USED_FOR_TARGET.
15129 (make_pass_resolve_sw_modes): Likewise.
15131 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15133 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15136 2018-01-08 Richard Biener <rguenther@suse.de>
15138 PR middle-end/83580
15139 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15141 2018-01-08 Richard Biener <rguenther@suse.de>
15143 PR middle-end/83517
15144 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15146 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15148 PR middle-end/81897
15149 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15150 basic blocks with a small number of successors.
15151 (convert_control_dep_chain_into_preds): Improve handling of
15153 (dump_predicates): Split apart into...
15154 (dump_pred_chain): ...here...
15155 (dump_pred_info): ...and here.
15156 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15157 (can_chain_union_be_invalidated_p): Improve check for invalidation
15159 (uninit_uses_cannot_happen): Avoid unnecessary if
15160 convert_control_dep_chain_into_preds yielded nothing.
15162 2018-01-06 Martin Sebor <msebor@redhat.com>
15164 PR tree-optimization/83640
15165 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15166 subtracting negative offset from size.
15167 (builtin_access::overlap): Adjust offset bounds of the access to fall
15168 within the size of the object if possible.
15170 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15172 PR rtl-optimization/83699
15173 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15174 extract_bit_field_as_subreg to cases in which the extracted
15175 value is also a vector.
15177 * lra-constraints.c (process_alt_operands): Test for the equivalence
15178 substitutions when detecting a possible reload cycle.
15180 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15183 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15184 by default if flag_selective_schedling{,2}. Formatting fixes.
15186 PR rtl-optimization/83682
15187 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15188 if it has non-VECTOR_MODE element mode.
15189 (vec_duplicate_p): Likewise.
15191 PR middle-end/83694
15192 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15193 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15195 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15198 * config/i386/i386-builtin.def
15199 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15200 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15201 Require also OPTION_MASK_ISA_AVX512F in addition to
15202 OPTION_MASK_ISA_GFNI.
15203 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15204 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15205 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15206 to OPTION_MASK_ISA_GFNI.
15207 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15208 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15209 OPTION_MASK_ISA_AVX512BW.
15210 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15211 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15212 addition to OPTION_MASK_ISA_GFNI.
15213 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15214 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15215 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15216 to OPTION_MASK_ISA_GFNI.
15217 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15218 a requirement for all ISAs rather than any of them with a few
15220 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15222 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15223 bitmasks to be enabled with 3 exceptions, instead of requiring any
15224 enabled ISA with lots of exceptions.
15225 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15226 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15227 Change avx512bw in isa attribute to avx512f.
15228 * config/i386/sgxintrin.h: Add license boilerplate.
15229 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15230 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15231 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15232 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15234 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15235 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15236 temporarily sse2 rather than sse if not enabled already.
15239 * config/i386/sse.md (VI248_VLBW): Rename to ...
15240 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15241 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15242 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15243 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15244 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15245 mode iterator instead of VI248_VLBW.
15247 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15249 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15250 (record_modified): Skip clobbers; add debug output.
15251 (param_change_prob): Use sreal frequencies.
15253 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15255 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15256 punt for user-aligned variables.
15258 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15260 * tree-chrec.c (chrec_contains_symbols): Return true for
15263 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15266 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15267 of (x|y) == x for BICS pattern.
15269 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15271 PR tree-optimization/83605
15272 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15273 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15276 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15278 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15279 * config/epiphany/rtems.h: New file.
15281 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15282 Uros Bizjak <ubizjak@gmail.com>
15285 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15286 QIreg_operand instead of register_operand predicate.
15287 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15288 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15289 comments instead of -fmitigate[-_]rop.
15291 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15294 * cgraphunit.c (symbol_table::compile): Switch to text_section
15295 before calling assembly_start debug hook.
15296 * run-rtl-passes.c (run_rtl_passes): Likewise.
15299 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15301 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15302 range_int_cst_p rather than !symbolic_range_p before calling
15303 extract_range_from_multiplicative_op_1.
15305 2018-01-04 Jeff Law <law@redhat.com>
15307 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15308 redundant test in assertion.
15310 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15312 * doc/rtl.texi: Document machine_mode wrapper classes.
15314 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15316 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15317 using tree_to_uhwi.
15319 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15321 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15322 the VEC_PERM_EXPR fold to fail.
15324 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15327 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15328 to switched_sections.
15330 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15333 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15334 test for d.testing.
15336 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15339 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15340 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15342 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15345 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15346 is BLKmode and bitpos not zero or mode change is needed.
15348 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15351 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15354 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15357 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15358 instead of MULT rtx. Update all corresponding splitters.
15359 (*saddl_se): Ditto.
15360 (*ssub<modesuffix>): Ditto.
15361 (*ssubl_se): Ditto.
15362 (*cmp_sadd_di): Update split patterns.
15363 (*cmp_sadd_si): Ditto.
15364 (*cmp_sadd_sidi): Ditto.
15365 (*cmp_ssub_di): Ditto.
15366 (*cmp_ssub_si): Ditto.
15367 (*cmp_ssub_sidi): Ditto.
15368 * config/alpha/predicates.md (const23_operand): New predicate.
15369 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15370 Look for ASHIFT, not MULT inner operand.
15371 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15373 2018-01-04 Martin Liska <mliska@suse.cz>
15375 PR gcov-profile/83669
15376 * gcov.c (output_intermediate_file): Add version to intermediate
15378 * doc/gcov.texi: Document new field 'version' in intermediate
15379 file format. Fix location of '-k' option of gcov command.
15381 2018-01-04 Martin Liska <mliska@suse.cz>
15384 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15386 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15388 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15390 2018-01-03 Martin Sebor <msebor@redhat.com>
15392 PR tree-optimization/83655
15393 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15394 checking calls with invalid arguments.
15396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15398 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15399 (vectorizable_mask_load_store): Delete.
15400 (vectorizable_call): Return false for masked loads and stores.
15401 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15402 instead of gimple_assign_rhs1.
15403 (vectorizable_load): Handle IFN_MASK_LOAD.
15404 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15408 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15410 (vectorizable_mask_load_store): ...here.
15411 (vectorizable_load): ...and here.
15413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15415 * tree-vect-stmts.c (vect_build_all_ones_mask)
15416 (vect_build_zero_merge_argument): New functions, split out from...
15417 (vectorizable_load): ...here.
15419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15421 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15423 (vectorizable_mask_load_store): ...here.
15424 (vectorizable_store): ...and here.
15426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15428 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15430 (vectorizable_mask_load_store): ...here.
15432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15434 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15435 (vect_model_store_cost): Take a vec_load_store_type instead of a
15437 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15438 (vect_model_store_cost): Take a vec_load_store_type instead of a
15440 (vectorizable_mask_load_store): Update accordingly.
15441 (vectorizable_store): Likewise.
15442 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15446 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15447 IFN_MASK_LOAD calls here rather than...
15448 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15451 Alan Hayward <alan.hayward@arm.com>
15452 David Sherwood <david.sherwood@arm.com>
15454 * expmed.c (extract_bit_field_1): For vector extracts,
15455 fall back to extract_bit_field_as_subreg if vec_extract
15458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15459 Alan Hayward <alan.hayward@arm.com>
15460 David Sherwood <david.sherwood@arm.com>
15462 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15463 they are variable or constant sized.
15464 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15465 slots for constant-sized data.
15467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15468 Alan Hayward <alan.hayward@arm.com>
15469 David Sherwood <david.sherwood@arm.com>
15471 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15472 handling COND_EXPRs with boolean comparisons, try to find a better
15473 basis for the mask type than the boolean itself.
15475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15477 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15478 is calculated and how it can be overridden.
15479 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15480 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15482 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15486 Alan Hayward <alan.hayward@arm.com>
15487 David Sherwood <david.sherwood@arm.com>
15489 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15490 Remove the mode argument.
15491 (aarch64_simd_valid_immediate): Remove the mode and inverse
15493 * config/aarch64/iterators.md (bitsize): New iterator.
15494 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15495 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15496 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15497 aarch64_simd_valid_immediate.
15498 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15499 (aarch64_reg_or_bic_imm): Likewise.
15500 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15501 with an insn_type enum and msl with a modifier_type enum.
15502 Replace element_width with a scalar_mode. Change the shift
15503 to unsigned int. Add constructors for scalar_float_mode and
15504 scalar_int_mode elements.
15505 (aarch64_vect_float_const_representable_p): Delete.
15506 (aarch64_can_const_movi_rtx_p)
15507 (aarch64_simd_scalar_immediate_valid_for_move)
15508 (aarch64_simd_make_constant): Update call to
15509 aarch64_simd_valid_immediate.
15510 (aarch64_advsimd_valid_immediate_hs): New function.
15511 (aarch64_advsimd_valid_immediate): Likewise.
15512 (aarch64_simd_valid_immediate): Remove mode and inverse
15513 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15514 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15515 and aarch64_float_const_representable_p on the result.
15516 (aarch64_output_simd_mov_immediate): Remove mode argument.
15517 Update call to aarch64_simd_valid_immediate and use of
15518 simd_immediate_info.
15519 (aarch64_output_scalar_simd_mov_immediate): Update call
15522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15523 Alan Hayward <alan.hayward@arm.com>
15524 David Sherwood <david.sherwood@arm.com>
15526 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15527 (mode_nunits): Likewise CONST_MODE_NUNITS.
15528 * machmode.def (ADJUST_NUNITS): Document.
15529 * genmodes.c (mode_data::need_nunits_adj): New field.
15530 (blank_mode): Update accordingly.
15531 (adj_nunits): New variable.
15532 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15534 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15535 listed in adj_nunits.
15536 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15537 listed in adj_nunits. Don't emit case statements for such modes.
15538 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15539 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15540 nothing if adj_nunits is nonnull.
15541 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15542 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15543 (emit_mode_fbit): Update use of print_maybe_const_decl.
15544 (emit_move_size): Likewise. Treat the array as non-const
15546 (emit_mode_adjustments): Handle adj_nunits.
15548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15550 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15551 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15552 (VECTOR_MODES): Use it.
15553 (make_vector_modes): Take the prefix as an argument.
15555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15556 Alan Hayward <alan.hayward@arm.com>
15557 David Sherwood <david.sherwood@arm.com>
15559 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15560 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15561 for MODE_VECTOR_BOOL.
15562 * machmode.def (VECTOR_BOOL_MODE): Document.
15563 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15564 (make_vector_bool_mode): New function.
15565 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15567 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15568 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15570 * stor-layout.c (int_mode_for_mode): Likewise.
15571 * tree.c (build_vector_type_for_mode): Likewise.
15572 * varasm.c (output_constant_pool_2): Likewise.
15573 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15574 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15575 for MODE_VECTOR_BOOL.
15576 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15577 of mode class checks.
15578 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15579 instead of a list of mode class checks.
15580 (expand_vector_scalar_condition): Likewise.
15581 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15584 Alan Hayward <alan.hayward@arm.com>
15585 David Sherwood <david.sherwood@arm.com>
15587 * machmode.h (mode_size): Change from unsigned short to
15589 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15590 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15591 or if measurement_type is not polynomial.
15592 (fixed_size_mode::includes_p): Check for constant-sized modes.
15593 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15594 return a poly_uint16 rather than an unsigned short.
15595 (emit_mode_size): Change the type of mode_size from unsigned short
15596 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15597 (emit_mode_adjustments): Cope with polynomial vector sizes.
15598 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15600 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15602 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15603 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15604 * caller-save.c (setup_save_areas): Likewise.
15605 (replace_reg_with_saved_mem): Likewise.
15606 * calls.c (emit_library_call_value_1): Likewise.
15607 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15608 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15609 (gen_lowpart_for_combine): Likewise.
15610 * convert.c (convert_to_integer_1): Likewise.
15611 * cse.c (equiv_constant, cse_insn): Likewise.
15612 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15613 (cselib_subst_to_values): Likewise.
15614 * dce.c (word_dce_process_block): Likewise.
15615 * df-problems.c (df_word_lr_mark_ref): Likewise.
15616 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15617 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15618 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15619 (rtl_for_decl_location): Likewise.
15620 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15621 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15622 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15623 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15624 (expand_expr_real_1): Likewise.
15625 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15626 (pad_below): Likewise.
15627 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15628 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15629 * ira.c (get_subreg_tracking_sizes): Likewise.
15630 * ira-build.c (ira_create_allocno_objects): Likewise.
15631 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15632 (ira_sort_regnos_for_alter_reg): Likewise.
15633 * ira-costs.c (record_operand_costs): Likewise.
15634 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15635 (resolve_simple_move): Likewise.
15636 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15637 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15638 (lra_constraints): Likewise.
15639 (CONST_POOL_OK_P): Reject variable-sized modes.
15640 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15641 (add_pseudo_to_slot, lra_spill): Likewise.
15642 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15643 * optabs-query.c (get_best_extraction_insn): Likewise.
15644 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15645 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15646 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15647 * recog.c (offsettable_address_addr_space_p): Likewise.
15648 * regcprop.c (maybe_mode_change): Likewise.
15649 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15650 * regrename.c (build_def_use): Likewise.
15651 * regstat.c (dump_reg_info): Likewise.
15652 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15653 (find_reloads, find_reloads_subreg_address): Likewise.
15654 * reload1.c (eliminate_regs_1): Likewise.
15655 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15656 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15657 (simplify_binary_operation_1, simplify_subreg): Likewise.
15658 * targhooks.c (default_function_arg_padding): Likewise.
15659 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15660 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15661 (verify_gimple_assign_ternary): Likewise.
15662 * tree-inline.c (estimate_move_cost): Likewise.
15663 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15664 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15665 (get_address_cost_ainc): Likewise.
15666 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15667 (vect_supportable_dr_alignment): Likewise.
15668 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15669 (vectorizable_reduction): Likewise.
15670 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15671 (vectorizable_operation, vectorizable_load): Likewise.
15672 * tree.c (build_same_sized_truth_vector_type): Likewise.
15673 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15674 * var-tracking.c (emit_note_insn_var_location): Likewise.
15675 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15676 (ADDR_VEC_ALIGN): Likewise.
15678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15679 Alan Hayward <alan.hayward@arm.com>
15680 David Sherwood <david.sherwood@arm.com>
15682 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15684 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15685 or if measurement_type is polynomial.
15686 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15687 * combine.c (make_extraction): Likewise.
15688 * dse.c (find_shift_sequence): Likewise.
15689 * dwarf2out.c (mem_loc_descriptor): Likewise.
15690 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15691 (extract_bit_field, extract_low_bits): Likewise.
15692 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15693 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15694 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15695 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15696 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15697 * reload.c (find_reloads): Likewise.
15698 * reload1.c (alter_reg): Likewise.
15699 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15700 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15701 * tree-if-conv.c (predicate_mem_writes): Likewise.
15702 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15703 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15704 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15705 * valtrack.c (dead_debug_insert_temp): Likewise.
15706 * varasm.c (mergeable_constant_section): Likewise.
15707 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15710 Alan Hayward <alan.hayward@arm.com>
15711 David Sherwood <david.sherwood@arm.com>
15713 * expr.c (expand_assignment): Cope with polynomial mode sizes
15714 when assigning to a CONCAT.
15716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15717 Alan Hayward <alan.hayward@arm.com>
15718 David Sherwood <david.sherwood@arm.com>
15720 * machmode.h (mode_precision): Change from unsigned short to
15722 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15724 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15725 or if measurement_type is not polynomial.
15726 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15727 in which the mode is already known to be a scalar_int_mode.
15728 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15729 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15731 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15732 for GET_MODE_PRECISION.
15733 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15734 for GET_MODE_PRECISION.
15735 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15737 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15738 (expand_field_assignment, make_extraction): Likewise.
15739 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15740 (get_last_value): Likewise.
15741 * convert.c (convert_to_integer_1): Likewise.
15742 * cse.c (cse_insn): Likewise.
15743 * expr.c (expand_expr_real_1): Likewise.
15744 * lra-constraints.c (simplify_operand_subreg): Likewise.
15745 * optabs-query.c (can_atomic_load_p): Likewise.
15746 * optabs.c (expand_atomic_load): Likewise.
15747 (expand_atomic_store): Likewise.
15748 * ree.c (combine_reaching_defs): Likewise.
15749 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15750 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15751 * tree.h (type_has_mode_precision_p): Likewise.
15752 * ubsan.c (instrument_si_overflow): Likewise.
15754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15755 Alan Hayward <alan.hayward@arm.com>
15756 David Sherwood <david.sherwood@arm.com>
15758 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15759 polynomial numbers of units.
15760 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15761 (valid_vector_subparts_p): New function.
15762 (build_vector_type): Remove temporary shim and take the number
15763 of units as a poly_uint64 rather than an int.
15764 (build_opaque_vector_type): Take the number of units as a
15765 poly_uint64 rather than an int.
15766 * tree.c (build_vector_from_ctor): Handle polynomial
15767 TYPE_VECTOR_SUBPARTS.
15768 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15769 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15770 (build_vector_from_val): If the number of units is variable,
15771 use build_vec_duplicate_cst for constant operands and
15772 VEC_DUPLICATE_EXPR otherwise.
15773 (make_vector_type): Remove temporary is_constant ().
15774 (build_vector_type, build_opaque_vector_type): Take the number of
15775 units as a poly_uint64 rather than an int.
15776 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15778 * cfgexpand.c (expand_debug_expr): Likewise.
15779 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15780 (store_constructor, expand_expr_real_1): Likewise.
15781 (const_scalar_mask_from_tree): Likewise.
15782 * fold-const-call.c (fold_const_reduction): Likewise.
15783 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15784 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15785 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15786 (fold_relational_const): Likewise.
15787 (native_interpret_vector): Likewise. Change the size from an
15788 int to an unsigned int.
15789 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15790 TYPE_VECTOR_SUBPARTS.
15791 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15792 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15793 duplicating a non-constant operand into a variable-length vector.
15794 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15795 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15796 * ipa-icf.c (sem_variable::equals): Likewise.
15797 * match.pd: Likewise.
15798 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15799 * print-tree.c (print_node): Likewise.
15800 * stor-layout.c (layout_type): Likewise.
15801 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15802 * tree-cfg.c (verify_gimple_comparison): Likewise.
15803 (verify_gimple_assign_binary): Likewise.
15804 (verify_gimple_assign_ternary): Likewise.
15805 (verify_gimple_assign_single): Likewise.
15806 * tree-pretty-print.c (dump_generic_node): Likewise.
15807 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15808 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15809 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15810 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15811 (vect_shift_permute_load_chain): Likewise.
15812 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15813 (expand_vector_condition, optimize_vector_constructor): Likewise.
15814 (lower_vec_perm, get_compute_type): Likewise.
15815 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15816 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15817 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15818 (vect_recog_mask_conversion_pattern): Likewise.
15819 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15820 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15821 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15822 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15823 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15824 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15825 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15826 (supportable_widening_operation): Likewise.
15827 (supportable_narrowing_operation): Likewise.
15828 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15830 * varasm.c (output_constant): Likewise.
15832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15833 Alan Hayward <alan.hayward@arm.com>
15834 David Sherwood <david.sherwood@arm.com>
15836 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15837 so that both the length == 3 and length != 3 cases set up their
15838 own permute vectors. Add comments explaining why we know the
15839 number of elements is constant.
15840 (vect_permute_load_chain): Likewise.
15842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15843 Alan Hayward <alan.hayward@arm.com>
15844 David Sherwood <david.sherwood@arm.com>
15846 * machmode.h (mode_nunits): Change from unsigned char to
15848 (ONLY_FIXED_SIZE_MODES): New macro.
15849 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15850 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15851 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15853 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15854 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15855 or if measurement_type is not polynomial.
15856 * genmodes.c (ZERO_COEFFS): New macro.
15857 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15859 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15860 Use ZERO_COEFFS when emitting initializers.
15861 * data-streamer.h (bp_pack_poly_value): New function.
15862 (bp_unpack_poly_value): Likewise.
15863 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15864 for GET_MODE_NUNITS.
15865 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15866 for GET_MODE_NUNITS.
15867 * tree.c (make_vector_type): Remove temporary shim and make
15868 the real function take the number of units as a poly_uint64
15869 rather than an int.
15870 (build_vector_type_for_mode): Handle polynomial nunits.
15871 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15872 * emit-rtl.c (const_vec_series_p_1): Likewise.
15873 (gen_rtx_CONST_VECTOR): Likewise.
15874 * fold-const.c (test_vec_duplicate_folding): Likewise.
15875 * genrecog.c (validate_pattern): Likewise.
15876 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15877 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15878 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15879 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15880 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15881 * rtlanal.c (subreg_get_info): Likewise.
15882 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15883 (vect_grouped_load_supported): Likewise.
15884 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15885 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15886 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15887 (simplify_const_unary_operation, simplify_binary_operation_1)
15888 (simplify_const_binary_operation, simplify_ternary_operation)
15889 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15890 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15891 instead of CONST_VECTOR_NUNITS.
15892 * varasm.c (output_constant_pool_2): Likewise.
15893 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15894 explicit-encoded elements in the XVEC for variable-length vectors.
15896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15898 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15901 Alan Hayward <alan.hayward@arm.com>
15902 David Sherwood <david.sherwood@arm.com>
15904 * coretypes.h (fixed_size_mode): Declare.
15905 (fixed_size_mode_pod): New typedef.
15906 * builtins.h (target_builtins::x_apply_args_mode)
15907 (target_builtins::x_apply_result_mode): Change type to
15908 fixed_size_mode_pod.
15909 * builtins.c (apply_args_size, apply_result_size, result_vector)
15910 (expand_builtin_apply_args_1, expand_builtin_apply)
15911 (expand_builtin_return): Update accordingly.
15913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15915 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15916 * cselib.c (cselib_hash_rtx): Likewise.
15917 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15918 CONST_VECTOR encoding.
15920 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15921 Jeff Law <law@redhat.com>
15924 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15925 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15926 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15927 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15930 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15931 explicitly probe *sp in a noreturn function if there were any callee
15932 register saves or frame pointer is needed.
15934 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15937 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15938 BLKmode for ternary, binary or unary expressions.
15941 * var-tracking.c (delete_vta_debug_insn): New inline function.
15942 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15943 insns from get_insns () to NULL instead of each bb separately.
15944 Use delete_vta_debug_insn. No longer static.
15945 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15946 delete_vta_debug_insns callers.
15947 * rtl.h (delete_vta_debug_insns): Declare.
15948 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15949 instead of variable_tracking_main.
15951 2018-01-03 Martin Sebor <msebor@redhat.com>
15953 PR tree-optimization/83603
15954 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15955 arguments past the endof the argument list in functions declared
15956 without a prototype.
15957 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15958 Avoid checking when arguments are null.
15960 2018-01-03 Martin Sebor <msebor@redhat.com>
15963 * doc/extend.texi (attribute const): Fix a typo.
15964 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15965 issuing -Wsuggest-attribute for void functions.
15967 2018-01-03 Martin Sebor <msebor@redhat.com>
15969 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15970 offset_int::from instead of wide_int::to_shwi.
15971 (maybe_diag_overlap): Remove assertion.
15972 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15973 * gimple-ssa-sprintf.c (format_directive): Same.
15974 (parse_directive): Same.
15975 (sprintf_dom_walker::compute_format_length): Same.
15976 (try_substitute_return_value): Same.
15978 2018-01-03 Jeff Law <law@redhat.com>
15980 PR middle-end/83654
15981 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15982 non-constant residual for zero at runtime and avoid probing in
15983 that case. Reorganize code for trailing problem to mirror handling
15986 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15988 PR tree-optimization/83501
15989 * tree-ssa-strlen.c (get_string_cst): New.
15990 (handle_char_store): Call get_string_cst.
15992 2018-01-03 Martin Liska <mliska@suse.cz>
15994 PR tree-optimization/83593
15995 * tree-ssa-strlen.c: Include tree-cfg.h.
15996 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15997 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15998 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16000 (strlen_dom_walker::before_dom_children): Call
16001 gimple_purge_dead_eh_edges. Dump tranformation with details
16003 (strlen_dom_walker::before_dom_children): Update call by adding
16004 new argument cleanup_eh.
16005 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16007 2018-01-03 Martin Liska <mliska@suse.cz>
16010 * cif-code.def (VARIADIC_THUNK): New enum value.
16011 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16014 2018-01-03 Jan Beulich <jbeulich@suse.com>
16016 * sse.md (mov<mode>_internal): Tighten condition for when to use
16017 vmovdqu<ssescalarsize> for TI and OI modes.
16019 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16021 Update copyright years.
16023 2018-01-03 Martin Liska <mliska@suse.cz>
16026 * ipa-visibility.c (function_and_variable_visibility): Skip
16027 functions with noipa attribure.
16029 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16031 * gcc.c (process_command): Update copyright notice dates.
16032 * gcov-dump.c (print_version): Ditto.
16033 * gcov.c (print_version): Ditto.
16034 * gcov-tool.c (print_version): Ditto.
16035 * gengtype.c (create_file): Ditto.
16036 * doc/cpp.texi: Bump @copying's copyright year.
16037 * doc/cppinternals.texi: Ditto.
16038 * doc/gcc.texi: Ditto.
16039 * doc/gccint.texi: Ditto.
16040 * doc/gcov.texi: Ditto.
16041 * doc/install.texi: Ditto.
16042 * doc/invoke.texi: Ditto.
16044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16046 * vector-builder.h (vector_builder::m_full_nelts): Change from
16047 unsigned int to poly_uint64.
16048 (vector_builder::full_nelts): Update prototype accordingly.
16049 (vector_builder::new_vector): Likewise.
16050 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16051 (vector_builder::operator ==): Likewise.
16052 (vector_builder::finalize): Likewise.
16053 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16054 Take the number of elements as a poly_uint64 rather than an
16056 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16057 from unsigned int to poly_uint64.
16058 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16059 (vec_perm_indices::new_vector): Likewise.
16060 (vec_perm_indices::length): Likewise.
16061 (vec_perm_indices::nelts_per_input): Likewise.
16062 (vec_perm_indices::input_nelts): Likewise.
16063 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16064 number of elements per input as a poly_uint64 rather than an
16065 unsigned int. Use the original encoding for variable-length
16066 vectors, rather than clamping each individual element.
16067 For the second and subsequent elements in each pattern,
16068 clamp the step and base before clamping their sum.
16069 (vec_perm_indices::series_p): Handle polynomial element counts.
16070 (vec_perm_indices::all_in_range_p): Likewise.
16071 (vec_perm_indices_to_tree): Likewise.
16072 (vec_perm_indices_to_rtx): Likewise.
16073 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16074 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16075 (tree_vector_builder::new_binary_operation): Handle polynomial
16076 element counts. Return false if we need to know the number
16077 of elements at compile time.
16078 * fold-const.c (fold_vec_perm): Punt if the number of elements
16079 isn't known at compile time.
16081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16083 * vec-perm-indices.h (vec_perm_builder): Change element type
16084 from HOST_WIDE_INT to poly_int64.
16085 (vec_perm_indices::element_type): Update accordingly.
16086 (vec_perm_indices::clamp): Handle polynomial element_types.
16087 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16088 (vec_perm_indices::all_in_range_p): Likewise.
16089 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16091 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16092 polynomial vec_perm_indices element types.
16093 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16094 * fold-const.c (fold_vec_perm): Likewise.
16095 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16096 * tree-vect-generic.c (lower_vec_perm): Likewise.
16097 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16098 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16099 element type to HOST_WIDE_INT.
16101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16102 Alan Hayward <alan.hayward@arm.com>
16103 David Sherwood <david.sherwood@arm.com>
16105 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16106 rather than an int. Use plus_constant.
16107 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16108 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16111 Alan Hayward <alan.hayward@arm.com>
16112 David Sherwood <david.sherwood@arm.com>
16114 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16115 a HOST_WIDE_INT to a poly_int64.
16117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16118 Alan Hayward <alan.hayward@arm.com>
16119 David Sherwood <david.sherwood@arm.com>
16121 * calls.c (load_register_parameters): Cope with polynomial
16122 mode sizes. Require a constant size for BLKmode parameters
16123 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16124 forces a parameter to be padded at the lsb end in order to
16125 fill a complete number of words, require the parameter size
16126 to be ordered wrt UNITS_PER_WORD.
16128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16129 Alan Hayward <alan.hayward@arm.com>
16130 David Sherwood <david.sherwood@arm.com>
16132 * reload1.c (spill_stack_slot_width): Change element type
16133 from unsigned int to poly_uint64_pod.
16134 (alter_reg): Treat mode sizes as polynomial.
16136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16137 Alan Hayward <alan.hayward@arm.com>
16138 David Sherwood <david.sherwood@arm.com>
16140 * reload.c (complex_word_subreg_p): New function.
16141 (reload_inner_reg_of_subreg, push_reload): Use it.
16143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16144 Alan Hayward <alan.hayward@arm.com>
16145 David Sherwood <david.sherwood@arm.com>
16147 * lra-constraints.c (process_alt_operands): Reject matched
16148 operands whose sizes aren't ordered.
16149 (match_reload): Refer to this check here.
16151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16152 Alan Hayward <alan.hayward@arm.com>
16153 David Sherwood <david.sherwood@arm.com>
16155 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16156 that the mode size is in the set {1, 2, 4, 8, 16}.
16158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16159 Alan Hayward <alan.hayward@arm.com>
16160 David Sherwood <david.sherwood@arm.com>
16162 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16163 Use plus_constant instead of gen_rtx_PLUS.
16165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16166 Alan Hayward <alan.hayward@arm.com>
16167 David Sherwood <david.sherwood@arm.com>
16169 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16170 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16171 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16172 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16173 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16174 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16175 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16176 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16177 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16178 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16180 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16181 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16182 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16183 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16184 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16185 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16186 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16187 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16188 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16189 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16191 * expr.c (emit_move_resolve_push): Treat the input and result
16192 of PUSH_ROUNDING as a poly_int64.
16193 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16194 (emit_push_insn): Likewise.
16195 * lra-eliminations.c (mark_not_eliminable): Likewise.
16196 * recog.c (push_operand): Likewise.
16197 * reload1.c (elimination_effects): Likewise.
16198 * rtlanal.c (nonzero_bits1): Likewise.
16199 * calls.c (store_one_arg): Likewise. Require the padding to be
16200 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16207 Use plus_constant instead of gen_rtx_PLUS.
16209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16210 Alan Hayward <alan.hayward@arm.com>
16211 David Sherwood <david.sherwood@arm.com>
16213 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16214 rather than an int.
16216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16217 Alan Hayward <alan.hayward@arm.com>
16218 David Sherwood <david.sherwood@arm.com>
16220 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16221 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16222 via stack temporaries. Treat the mode size as polynomial too.
16224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16225 Alan Hayward <alan.hayward@arm.com>
16226 David Sherwood <david.sherwood@arm.com>
16228 * expr.c (expand_expr_real_2): When handling conversions involving
16229 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16230 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16231 as a poly_uint64 too.
16233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16234 Alan Hayward <alan.hayward@arm.com>
16235 David Sherwood <david.sherwood@arm.com>
16237 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16240 Alan Hayward <alan.hayward@arm.com>
16241 David Sherwood <david.sherwood@arm.com>
16243 * combine.c (can_change_dest_mode): Handle polynomial
16244 REGMODE_NATURAL_SIZE.
16245 * expmed.c (store_bit_field_1): Likewise.
16246 * expr.c (store_constructor): Likewise.
16247 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16248 and polynomial REGMODE_NATURAL_SIZE.
16249 (gen_lowpart_common): Likewise.
16250 * reginfo.c (record_subregs_of_mode): Likewise.
16251 * rtlanal.c (read_modify_subreg_p): Likewise.
16253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16254 Alan Hayward <alan.hayward@arm.com>
16255 David Sherwood <david.sherwood@arm.com>
16257 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16258 numbers of elements.
16260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16261 Alan Hayward <alan.hayward@arm.com>
16262 David Sherwood <david.sherwood@arm.com>
16264 * match.pd: Cope with polynomial numbers of vector elements.
16266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16267 Alan Hayward <alan.hayward@arm.com>
16268 David Sherwood <david.sherwood@arm.com>
16270 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16271 in a POINTER_PLUS_EXPR.
16273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16274 Alan Hayward <alan.hayward@arm.com>
16275 David Sherwood <david.sherwood@arm.com>
16277 * omp-simd-clone.c (simd_clone_subparts): New function.
16278 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16279 (ipa_simd_modify_function_body): Likewise.
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-vect-generic.c (nunits_for_known_piecewise_op): New function.
16286 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16287 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16288 (expand_vector_condition, vector_element): Likewise.
16289 (subparts_gt): New function.
16290 (get_compute_type): Use subparts_gt.
16291 (count_type_subparts): Delete.
16292 (expand_vector_operations_1): Use subparts_gt instead of
16293 count_type_subparts.
16295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16296 Alan Hayward <alan.hayward@arm.com>
16297 David Sherwood <david.sherwood@arm.com>
16299 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16300 (vect_compile_time_alias): ...this new function. Do the calculation
16301 on poly_ints rather than trees.
16302 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16310 (vect_schedule_slp_instance): Likewise.
16312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16313 Alan Hayward <alan.hayward@arm.com>
16314 David Sherwood <david.sherwood@arm.com>
16316 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16317 constant and extern definitions for variable-length vectors.
16318 (vect_get_constant_vectors): Note that the number of units
16319 is known to be constant.
16321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16322 Alan Hayward <alan.hayward@arm.com>
16323 David Sherwood <david.sherwood@arm.com>
16325 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16326 of units as polynomial. Choose between WIDE and NARROW based
16329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16330 Alan Hayward <alan.hayward@arm.com>
16331 David Sherwood <david.sherwood@arm.com>
16333 * tree-vect-stmts.c (simd_clone_subparts): New function.
16334 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16337 Alan Hayward <alan.hayward@arm.com>
16338 David Sherwood <david.sherwood@arm.com>
16340 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16341 vectors as polynomial. Use build_index_vector for
16342 IFN_GOMP_SIMD_LANE.
16344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16345 Alan Hayward <alan.hayward@arm.com>
16346 David Sherwood <david.sherwood@arm.com>
16348 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16349 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16350 for variable-length vectors.
16351 (vectorizable_mask_load_store): Treat the number of units as
16352 polynomial, asserting that it is constant if the condition has
16353 already been enforced.
16354 (vectorizable_store, vectorizable_load): Likewise.
16356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16357 Alan Hayward <alan.hayward@arm.com>
16358 David Sherwood <david.sherwood@arm.com>
16360 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16361 of units as polynomial. Punt if we can't tell at compile time
16362 which vector contains the final result.
16364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16365 Alan Hayward <alan.hayward@arm.com>
16366 David Sherwood <david.sherwood@arm.com>
16368 * tree-vect-loop.c (vectorizable_induction): Treat the number
16369 of units as polynomial. Punt on SLP inductions. Use an integer
16370 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16371 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
16379 * tree.c (build_index_vector): New function.
16380 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16381 of units as polynomial, forcibly converting it to a constant if
16382 vectorizable_reduction has already enforced the condition.
16383 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16384 to create a {1,2,3,...} vector.
16385 (vectorizable_reduction): Treat the number of units as polynomial.
16386 Choose vectype_in based on the largest scalar element size rather
16387 than the smallest number of units. Enforce the restrictions
16390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16391 Alan Hayward <alan.hayward@arm.com>
16392 David Sherwood <david.sherwood@arm.com>
16394 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16395 number of units as polynomial.
16397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16398 Alan Hayward <alan.hayward@arm.com>
16399 David Sherwood <david.sherwood@arm.com>
16401 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16402 * target.def (autovectorize_vector_sizes): Return the vector sizes
16403 by pointer, using vector_sizes rather than a bitmask.
16404 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16405 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16406 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16408 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16409 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16410 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16411 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16412 * omp-general.c (omp_max_vf): Likewise.
16413 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16414 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16415 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16416 * tree-vect-slp.c (vect_slp_bb): Likewise.
16417 * doc/tm.texi: Regenerate.
16418 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16420 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16421 the vector size as a poly_uint64 rather than an unsigned int.
16422 (current_vector_size): Change from an unsigned int to a poly_uint64.
16423 (get_vectype_for_scalar_type): Update accordingly.
16424 * tree.h (build_truth_vector_type): Take the size and number of
16425 units as a poly_uint64 rather than an unsigned int.
16426 (build_vector_type): Add a temporary overload that takes
16427 the number of units as a poly_uint64 rather than an unsigned int.
16428 * tree.c (make_vector_type): Likewise.
16429 (build_truth_vector_type): Take the number of units as a poly_uint64
16430 rather than an unsigned int.
16432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16433 Alan Hayward <alan.hayward@arm.com>
16434 David Sherwood <david.sherwood@arm.com>
16436 * target.def (get_mask_mode): Take the number of units and length
16437 as poly_uint64s rather than unsigned ints.
16438 * targhooks.h (default_get_mask_mode): Update accordingly.
16439 * targhooks.c (default_get_mask_mode): Likewise.
16440 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16441 * doc/tm.texi: Regenerate.
16443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16444 Alan Hayward <alan.hayward@arm.com>
16445 David Sherwood <david.sherwood@arm.com>
16447 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16448 * omp-general.c (omp_max_vf): Likewise.
16449 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16450 (expand_omp_simd): Handle polynomial safelen.
16451 * omp-low.c (omplow_simd_context): Add a default constructor.
16452 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16453 (lower_rec_simd_input_clauses): Update accordingly.
16454 (lower_rec_input_clauses): Likewise.
16456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16457 Alan Hayward <alan.hayward@arm.com>
16458 David Sherwood <david.sherwood@arm.com>
16460 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16461 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16462 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16463 (vect_analyze_slp_cost): Likewise.
16464 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16465 (vect_model_load_cost): Likewise.
16467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16468 Alan Hayward <alan.hayward@arm.com>
16469 David Sherwood <david.sherwood@arm.com>
16471 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16472 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16473 from an unsigned int * to a poly_uint64_pod *.
16474 (calculate_unrolling_factor): New function.
16475 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16478 Alan Hayward <alan.hayward@arm.com>
16479 David Sherwood <david.sherwood@arm.com>
16481 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16482 from an unsigned int to a poly_uint64.
16483 (_loop_vec_info::slp_unrolling_factor): Likewise.
16484 (_loop_vec_info::vectorization_factor): Change from an int
16486 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16487 (vect_get_num_vectors): New function.
16488 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16489 (vect_get_num_copies): Use vect_get_num_vectors.
16490 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16491 to an unsigned int *.
16492 (vect_analyze_data_refs): Change min_vf from an int * to a
16494 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16495 than an unsigned HOST_WIDE_INT.
16496 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16497 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16498 to an unsigned int *.
16499 (vect_analyze_data_ref_dependences): Likewise.
16500 (vect_compute_data_ref_alignment): Handle polynomial vf.
16501 (vect_enhance_data_refs_alignment): Likewise.
16502 (vect_prune_runtime_alias_test_list): Likewise.
16503 (vect_shift_permute_load_chain): Likewise.
16504 (vect_supportable_dr_alignment): Likewise.
16505 (dependence_distance_ge_vf): Take the vectorization factor as a
16506 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16507 (vect_analyze_data_refs): Change min_vf from an int * to a
16509 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16510 vfm1 as a poly_uint64 rather than an int. Make the same change
16511 for the returned bound_scalar.
16512 (vect_gen_vector_loop_niters): Handle polynomial vf.
16513 (vect_do_peeling): Likewise. Update call to
16514 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16515 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16517 * tree-vect-loop.c (vect_determine_vectorization_factor)
16518 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16519 (vect_get_known_peeling_cost): Likewise.
16520 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16521 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16522 (vect_transform_loop): Likewise. Use the lowest possible VF when
16523 updating the upper bounds of the loop.
16524 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16525 rather than an int.
16526 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16527 polynomial unroll factors.
16528 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16529 (vect_make_slp_decision): Likewise.
16530 (vect_supported_load_permutation_p): Likewise, and polynomial
16532 (vect_analyze_slp_cost): Handle polynomial vf.
16533 (vect_slp_analyze_node_operations): Likewise.
16534 (vect_slp_analyze_bb_1): Likewise.
16535 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16536 than an unsigned HOST_WIDE_INT.
16537 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16538 (vectorizable_load): Handle polynomial vf.
16539 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16541 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16544 Alan Hayward <alan.hayward@arm.com>
16545 David Sherwood <david.sherwood@arm.com>
16547 * match.pd: Handle bit operations involving three constants
16548 and try to fold one pair.
16550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16552 * tree-vect-loop-manip.c: Include gimple-fold.h.
16553 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16554 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16555 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16556 Add a path that uses a step of VF instead of 1, but disable it
16558 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16559 and niters_no_overflow parameters. Update calls to
16560 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16561 Create a new SSA name if the latter choses to use a ste other
16562 than zero, and return it via niters_vector_mult_vf_var.
16563 * tree-vect-loop.c (vect_transform_loop): Update calls to
16564 vect_do_peeling, vect_gen_vector_loop_niters and
16565 slpeel_make_loop_iterate_ntimes.
16566 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16567 (vect_gen_vector_loop_niters): Update declarations after above changes.
16569 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16571 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16572 128-bit round to integer instructions.
16573 (ceil<mode>2): Likewise.
16574 (btrunc<mode>2): Likewise.
16575 (round<mode>2): Likewise.
16577 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16579 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16580 unaligned VSX load/store on P8/P9.
16581 (expand_block_clear): Allow the use of unaligned VSX
16582 load/store on P8/P9.
16584 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16586 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16588 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16589 swap associated with both a load and a store.
16591 2018-01-02 Andrew Waterman <andrew@sifive.com>
16593 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16594 * config/riscv/riscv.md (clear_cache): Use it.
16596 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16598 * web.c: Remove out-of-date comment.
16600 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16602 * expr.c (fixup_args_size_notes): Check that any existing
16603 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16604 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16605 (emit_single_push_insn): ...here.
16607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16609 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16610 (const_vector_encoded_nelts): New function.
16611 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16612 (const_vector_int_elt, const_vector_elt): Declare.
16613 * emit-rtl.c (const_vector_int_elt_1): New function.
16614 (const_vector_elt): Likewise.
16615 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16616 of CONST_VECTOR_ELT.
16618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16620 * expr.c: Include rtx-vector-builder.h.
16621 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16622 directly on the tree encoding.
16623 (const_vector_from_tree): Likewise.
16624 * optabs.c: Include rtx-vector-builder.h.
16625 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16626 sequence of "u" values.
16627 * vec-perm-indices.c: Include rtx-vector-builder.h.
16628 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16629 directly on the vec_perm_indices encoding.
16631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16633 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16634 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16635 * rtx-vector-builder.h: New file.
16636 * rtx-vector-builder.c: Likewise.
16637 * rtl.h (rtx_def::u2): Add a const_vector field.
16638 (CONST_VECTOR_NPATTERNS): New macro.
16639 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16640 (CONST_VECTOR_DUPLICATE_P): Likewise.
16641 (CONST_VECTOR_STEPPED_P): Likewise.
16642 (CONST_VECTOR_ENCODED_ELT): Likewise.
16643 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16644 (unwrap_const_vec_duplicate): Likewise.
16645 (const_vec_series_p): Check for a non-duplicated vector encoding.
16646 Say that the function only returns true for integer vectors.
16647 * emit-rtl.c: Include rtx-vector-builder.h.
16648 (gen_const_vec_duplicate_1): Delete.
16649 (gen_const_vector): Call gen_const_vec_duplicate instead of
16650 gen_const_vec_duplicate_1.
16651 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16652 (gen_const_vec_duplicate): Use rtx_vector_builder.
16653 (gen_const_vec_series): Likewise.
16654 (gen_rtx_CONST_VECTOR): Likewise.
16655 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16656 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16657 Build a new vector rather than modifying a CONST_VECTOR in-place.
16658 (handle_special_swappables): Update call accordingly.
16659 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16660 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16661 Build a new vector rather than modifying a CONST_VECTOR in-place.
16662 (handle_special_swappables): Update call accordingly.
16664 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16666 * simplify-rtx.c (simplify_const_binary_operation): Use
16667 CONST_VECTOR_ELT instead of XVECEXP.
16669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16671 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16672 the selector elements to be different from the data elements
16673 if the selector is a VECTOR_CST.
16674 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16675 ssizetype for the selector.
16677 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16679 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16680 before testing each element individually.
16681 * tree-vect-generic.c (lower_vec_perm): Likewise.
16683 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16685 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16686 * selftest-run-tests.c (selftest::run_tests): Call it.
16687 * vector-builder.h (vector_builder::operator ==): New function.
16688 (vector_builder::operator !=): Likewise.
16689 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16690 (vec_perm_indices::all_from_input_p): New function.
16691 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16692 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16693 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16694 instead of reading the VECTOR_CST directly. Detect whether both
16695 vector inputs are the same before constructing the vec_perm_indices,
16696 and update the number of inputs argument accordingly. Use the
16697 utility functions added above. Only construct sel2 if we need to.
16699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16701 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16702 the broadcast of the low byte.
16703 (expand_mult_highpart): Use an explicit encoding for the permutes.
16704 * optabs-query.c (can_mult_highpart_p): Likewise.
16705 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16706 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16707 (vectorizable_bswap): Likewise.
16708 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16709 explicit encoding for the power-of-2 permutes.
16710 (vect_permute_store_chain): Likewise.
16711 (vect_grouped_load_supported): Likewise.
16712 (vect_permute_load_chain): Likewise.
16714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16716 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16717 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16718 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16719 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16720 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16721 (vect_gen_perm_mask_any): Likewise.
16723 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16725 * int-vector-builder.h: New file.
16726 * vec-perm-indices.h: Include int-vector-builder.h.
16727 (vec_perm_indices): Redefine as an int_vector_builder.
16728 (auto_vec_perm_indices): Delete.
16729 (vec_perm_builder): Redefine as a stand-alone class.
16730 (vec_perm_indices::vec_perm_indices): New function.
16731 (vec_perm_indices::clamp): Likewise.
16732 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16733 (vec_perm_indices::new_vector): New function.
16734 (vec_perm_indices::new_expanded_vector): Update for new
16735 vec_perm_indices class.
16736 (vec_perm_indices::rotate_inputs): New function.
16737 (vec_perm_indices::all_in_range_p): Operate directly on the
16738 encoded form, without computing elided elements.
16739 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16740 encoding. Update for new vec_perm_indices class.
16741 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16742 the given vec_perm_builder.
16743 (expand_vec_perm_var): Update vec_perm_builder constructor.
16744 (expand_mult_highpart): Use vec_perm_builder instead of
16745 auto_vec_perm_indices.
16746 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16747 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16748 or double series encoding as appropriate.
16749 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16750 vec_perm_indices instead of auto_vec_perm_indices.
16751 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16752 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16753 (vect_permute_store_chain): Likewise.
16754 (vect_grouped_load_supported): Likewise.
16755 (vect_permute_load_chain): Likewise.
16756 (vect_shift_permute_load_chain): Likewise.
16757 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16758 (vect_transform_slp_perm_load): Likewise.
16759 (vect_schedule_slp_instance): Likewise.
16760 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16761 (vectorizable_mask_load_store): Likewise.
16762 (vectorizable_bswap): Likewise.
16763 (vectorizable_store): Likewise.
16764 (vectorizable_load): Likewise.
16765 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16766 vec_perm_indices instead of auto_vec_perm_indices. Use
16767 tree_to_vec_perm_builder to read the vector from a tree.
16768 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16769 vec_perm_builder instead of a vec_perm_indices.
16770 (have_whole_vector_shift): Use vec_perm_builder and
16771 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16772 truncation to calc_vec_perm_mask_for_shift.
16773 (vect_create_epilog_for_reduction): Likewise.
16774 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16775 from auto_vec_perm_indices to vec_perm_indices.
16776 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16777 instead of changing individual elements.
16778 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16779 the vector in d.perm.
16780 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16781 from auto_vec_perm_indices to vec_perm_indices.
16782 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16783 instead of changing individual elements.
16784 (arm_vectorize_vec_perm_const): Use new_vector to install
16785 the vector in d.perm.
16786 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16787 Update vec_perm_builder constructor.
16788 (rs6000_expand_interleave): Likewise.
16789 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16790 (rs6000_expand_interleave): Likewise.
16792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16794 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16795 to qimode could truncate the indices.
16796 * optabs.c (expand_vec_perm_var): Likewise.
16798 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16800 * Makefile.in (OBJS): Add vec-perm-indices.o.
16801 * vec-perm-indices.h: New file.
16802 * vec-perm-indices.c: Likewise.
16803 * target.h (vec_perm_indices): Replace with a forward class
16805 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16806 * optabs.h: Include vec-perm-indices.h.
16807 (expand_vec_perm): Delete.
16808 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16809 (expand_vec_perm_const): Declare.
16810 * target.def (vec_perm_const_ok): Replace with...
16811 (vec_perm_const): ...this new hook.
16812 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16813 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16814 * doc/tm.texi: Regenerate.
16815 * optabs.def (vec_perm_const): Delete.
16816 * doc/md.texi (vec_perm_const): Likewise.
16817 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16818 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16819 expand_vec_perm for constant permutation vectors. Assert that
16820 the mode of variable permutation vectors is the integer equivalent
16821 of the mode that is being permuted.
16822 * optabs-query.h (selector_fits_mode_p): Declare.
16823 * optabs-query.c: Include vec-perm-indices.h.
16824 (selector_fits_mode_p): New function.
16825 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16826 is defined, instead of checking whether the vec_perm_const_optab
16827 exists. Use targetm.vectorize.vec_perm_const instead of
16828 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16829 fit in the vector mode before using a variable permute.
16830 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16831 vec_perm_indices instead of an rtx.
16832 (expand_vec_perm): Replace with...
16833 (expand_vec_perm_const): ...this new function. Take the selector
16834 as a vec_perm_indices rather than an rtx. Also take the mode of
16835 the selector. Update call to shift_amt_for_vec_perm_mask.
16836 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16837 Use vec_perm_indices::new_expanded_vector to expand the original
16838 selector into bytes. Check whether the indices fit in the vector
16839 mode before using a variable permute.
16840 (expand_vec_perm_var): Make global.
16841 (expand_mult_highpart): Use expand_vec_perm_const.
16842 * fold-const.c: Includes vec-perm-indices.h.
16843 * tree-ssa-forwprop.c: Likewise.
16844 * tree-vect-data-refs.c: Likewise.
16845 * tree-vect-generic.c: Likewise.
16846 * tree-vect-loop.c: Likewise.
16847 * tree-vect-slp.c: Likewise.
16848 * tree-vect-stmts.c: Likewise.
16849 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16851 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16852 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16853 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16854 (aarch64_vectorize_vec_perm_const): ...this new function.
16855 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16856 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16857 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16858 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16859 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16860 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16861 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16863 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16864 check for NEON modes.
16865 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16866 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16867 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16868 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16870 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16871 the old VEC_PERM_CONST conditions.
16872 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16873 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16874 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16875 (ia64_vectorize_vec_perm_const_ok): Merge into...
16876 (ia64_vectorize_vec_perm_const): ...this new function.
16877 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16878 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16879 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16880 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16881 * config/mips/mips.c (mips_expand_vec_perm_const)
16882 (mips_vectorize_vec_perm_const_ok): Merge into...
16883 (mips_vectorize_vec_perm_const): ...this new function.
16884 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16885 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16886 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16887 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16888 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16889 (rs6000_expand_vec_perm_const): Delete.
16890 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16892 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16893 (altivec_expand_vec_perm_const_le): Take each operand individually.
16894 Operate on constant selectors rather than rtxes.
16895 (altivec_expand_vec_perm_const): Likewise. Update call to
16896 altivec_expand_vec_perm_const_le.
16897 (rs6000_expand_vec_perm_const): Delete.
16898 (rs6000_vectorize_vec_perm_const_ok): Delete.
16899 (rs6000_vectorize_vec_perm_const): New function.
16900 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16901 an element count and rtx array.
16902 (rs6000_expand_extract_even): Update call accordingly.
16903 (rs6000_expand_interleave): Likewise.
16904 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16905 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16906 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16907 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16908 (rs6000_expand_vec_perm_const): Delete.
16909 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16910 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16911 (altivec_expand_vec_perm_const_le): Take each operand individually.
16912 Operate on constant selectors rather than rtxes.
16913 (altivec_expand_vec_perm_const): Likewise. Update call to
16914 altivec_expand_vec_perm_const_le.
16915 (rs6000_expand_vec_perm_const): Delete.
16916 (rs6000_vectorize_vec_perm_const_ok): Delete.
16917 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16918 reference to the SPE evmerge intructions.
16919 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16920 an element count and rtx array.
16921 (rs6000_expand_extract_even): Update call accordingly.
16922 (rs6000_expand_interleave): Likewise.
16923 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16924 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16926 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16930 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16931 vector mode and that that mode matches the mode of the data
16933 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16934 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16935 directly using expand_vec_perm_1 when forcing selectors into
16937 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16941 * optabs-query.h (can_vec_perm_p): Delete.
16942 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16943 * optabs-query.c (can_vec_perm_p): Split into...
16944 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16945 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16946 particular selector is valid.
16947 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16948 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16949 (vect_grouped_load_supported): Likewise.
16950 (vect_shift_permute_load_chain): Likewise.
16951 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16952 (vect_transform_slp_perm_load): Likewise.
16953 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16954 (vectorizable_bswap): Likewise.
16955 (vect_gen_perm_mask_checked): Likewise.
16956 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16957 implementations of variable permutation vectors into account
16958 when deciding which selector to use.
16959 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16960 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16961 with a false third argument.
16962 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16963 to test whether the constant selector is valid and can_vec_perm_var_p
16964 to test whether a variable selector is valid.
16966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16968 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16969 * optabs-query.c (can_vec_perm_p): Likewise.
16970 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16971 instead of vec_perm_indices.
16972 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16973 (vect_gen_perm_mask_checked): Likewise,
16974 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16975 (vect_gen_perm_mask_checked): Likewise,
16977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16979 * optabs-query.h (qimode_for_vec_perm): Declare.
16980 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16981 (qimode_for_vec_perm): ...this new function.
16982 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16984 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16986 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16987 does not have a conditional at the top.
16989 2018-01-02 Richard Biener <rguenther@suse.de>
16991 * ipa-inline.c (big_speedup_p): Fix expression.
16993 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16996 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16999 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17003 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17004 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17005 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17006 cond_taken_branch_cost 3->4.
17008 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17010 PR tree-optimization/83581
17011 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17012 TODO_cleanup_cfg if any changes have been made.
17014 PR middle-end/83608
17015 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17016 convert_modes if target mode has the right side, but different mode
17019 PR middle-end/83609
17020 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17021 last argument when extracting from CONCAT. If either from_real or
17022 from_imag is NULL, use expansion through memory. If result is not
17023 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17024 the parts directly to inner mode, if even that fails, use expansion
17027 PR middle-end/83623
17028 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17029 check for bswap in mode rather than HImode and use that in expand_unop
17032 Copyright (C) 2018 Free Software Foundation, Inc.
17034 Copying and distribution of this file, with or without modification,
17035 are permitted in any medium without royalty provided the copyright
17036 notice and this notice are preserved.