1 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3 * tree.c (escaped_string::escape): Replace cast to char * by
4 const_cast<char *> (unescaped).
6 2018-06-18 Nick Clifton <nickc@redhat.com>
9 * tree.c (escaped_string): New class. Converts an unescaped
10 string into its escaped equivalent.
11 (warn_deprecated_use): Use the new class to convert the
12 deprecation message, if present.
13 (test_escaped_strings): New self test.
14 (test_c_tests): Add test_escaped_strings.
15 * doc/extend.texi (deprecated): Add a note that the
16 deprecation message is affected by the -fmessage-length
17 option, and that control characters will be escaped.
18 (#pragma GCC error): Document this pragma.
19 (#pragma GCC warning): Likewise.
20 * doc/invoke.texi (-fmessage-length): Document this option's
21 effect on the #warning and #error preprocessor directives and
22 the deprecated attribute.
24 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
26 * tree.c (decl_value_expr_lookup): Revert latest change.
27 (decl_value_expr_insert): Likewise.
29 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
31 * gimplify.c (nonlocal_vlas): Delete.
32 (nonlocal_vla_vars): Likewise.
33 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
35 (gimplify_body): Do not create and destroy nonlocal_vlas.
36 * tree-nested.c: Include diagnostic.h.
37 (use_pointer_in_frame): Tweak.
38 (lookup_field_for_decl): Add assertion and declare the transformation.
39 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
40 internal error when the reference is in a wrong context. Do not
41 create a debug decl by default.
42 (note_nonlocal_block_vlas): Delete.
43 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
44 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
45 create a debug decl by default.
46 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
47 call is in a wrong context.
48 (fixup_vla_decls): New function.
49 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
50 debug variables were created.
51 * tree.c (decl_value_expr_lookup): Add checking assertion.
52 (decl_value_expr_insert): Likewise.
54 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
57 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
58 * tree-scalar-evolution.c (interpret_expr): Likewise.
59 (expression_expensive_p): Likewise.
60 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
61 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
62 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
63 (ssa_defined_by_minus_one_stmt_p): New.
65 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
68 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
69 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
70 * dojump.c (do_jump): Likewise.
71 * expr.c (expand_expr_real_2): Check operand type's sign.
72 * fold-const.c (const_unop): Handle ABSU_EXPR.
73 (fold_abs_const): Likewise.
74 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
75 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
76 (strip_sign_op_1): Likesise.
77 * match.pd: Add new pattern to generate ABSU_EXPR.
78 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
79 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
80 * tree-eh.c (operation_could_trap_helper_p): Likewise.
81 * tree-inline.c (estimate_operator_cost): Likewise.
82 * tree-pretty-print.c (dump_generic_node): Likewise.
83 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
84 * tree.def (ABSU_EXPR): New.
86 2018-06-16 Jakub Jelinek <jakub@redhat.com>
89 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
90 documented as preserved for backward compatibility only.
91 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
93 PR rtl-optimization/86108
94 * bb-reorder.c (create_forwarder_block): Renamed to ...
95 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
96 jump from new landing pad to the second part.
97 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
100 2018-06-15 Jakub Jelinek <jakub@redhat.com>
103 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
104 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
105 Only call store_expr for halves if the mode is the same.
108 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
109 Fix up comment formatting.
111 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
113 * typed-splay-tree.h (typed_splay_tree::remove): New function.
114 (typed_splay_tree::closure,
115 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
116 (typed_splay_tree::typed_splay_tree,
117 typed_splay_tree::operator =): Declared private.
118 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
119 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
120 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
121 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
122 typed_splay_tree::splay_tree_splay,
123 typed_splay_tree::splay_tree_foreach_helper,
124 typed_splay_tree::splay_tree_insert,
125 typed_splay_tree::splay_tree_remove,
126 typed_splay_tree::splay_tree_lookup,
127 typed_splay_tree::splay_tree_predecessor,
128 typed_splay_tree::splay_tree_successor,
129 typed_splay_tree::splay_tree_min,
130 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
131 (typed_splay_tree::root, typed_splay_tree::comp,
132 typed_splay_tree::delete_key,
133 typed_splay_tree::delete_value): New data members.
134 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
135 typed_splay_tree::remove.
137 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
139 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
140 -mginv and -mno-ginv to the assembler.
141 * config/mips/mips.opt (-mcrc): New option.
143 * doc/invoke.text (-mcrc): Document.
146 2018-06-15 Nick Clifton <nickc@redhat.com>
149 * tree.c (escaped_string): New class. Converts an unescaped
150 string into its escaped equivalent.
151 (warn_deprecated_use): Use the new class to convert the
152 deprecation message, if present.
153 (test_escaped_strings): New self test.
154 (test_c_tests): Add test_escaped_strings.
155 * doc/extend.texi (deprecated): Add a note that the
156 deprecation message is affected by the -fmessage-length
157 option, and that control characters will be escaped.
158 (#pragma GCC error): Document this pragma.
159 (#pragma GCC warning): Likewise.
160 * doc/invoke.texi (-fmessage-length): Document this option's
161 effect on the #warning and #error preprocessor directives and
162 the deprecated attribute.
164 2018-06-15 Richard Biener <rguenther@suse.de>
166 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
167 here, also noting vector size used.
168 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
169 size used in MSG_OPTIMIZED_LOCATIONS dump.
170 (pass_slp_vectorize::execute): Adjust.
172 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
175 * config/arc/arc.c (arc_return_address_register): Fix
178 2018-06-15 Richard Biener <rguenther@suse.de>
181 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
182 leave useless conversion stripping to force_gimple_operand_gsi.
183 (gimplify_build2): Likewise.
184 (gimplify_build1): Likewise.
186 2018-06-15 Richard Biener <rguenther@suse.de>
189 * tree-cfg.c (move_stmt_op): unshare invariant addresses
190 before adjusting their block.
192 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
194 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
195 multilibs for *-*-rtems*.
196 * config/riscv/t-rtems: New file.
198 2018-06-14 Jakub Jelinek <jakub@redhat.com>
201 * match.pd ((A +- CST1) +- CST2): Punt if last resort
202 unsigned_type_for returns NULL.
205 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
206 subregs of multi-word pseudos unless the float mode has word size.
208 2018-06-14 Richard Biener <rguenther@suse.de>
211 * tree-vect-generic.c (build_word_mode_vector_type): Remove
212 duplicate and harmful type_hash_canon.
213 * tree.c (type_hash_canon): Assert we didn't find ourselves.
215 2018-06-14 Richard Biener <rguenther@suse.de>
218 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
221 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
223 * config/rtems.h (STDINT_LONG32): Define.
225 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
226 Prachi Godbole <prachi.godbole@imgtec.com>
228 * config/mips/mips-cpus.def: Define P6600.
229 * config/mips/mips-tables.opt: Regenerate.
230 * config/mips/mips.c (mips_ucbranch_type): New enum.
231 (mips_rtx_cost_data): Add support for P6600.
232 (mips_issue_rate): Likewise.
233 (mips_multipass_dfa_lookahead): Likewise.
234 (mips_avoid_hazard): Likewise.
235 (mips_reorg_process_insns): Likewise.
236 (mips_classify_branch_p6600): New function.
237 * config/mips/mips.h (TUNE_P6600): New define.
238 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
239 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
240 * config/mips/mips.md: Include p6600.md.
241 (processor): Add p6600.
242 * config/mips/p6600.md: New file.
243 * doc/invoke.texi: Add p6600 to supported architectures.
245 2018-06-13 Martin Sebor <msebor@redhat.com>
247 PR tree-optimization/86114
248 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
250 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
252 2018-06-13 Richard Biener <rguenther@suse.de>
254 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
255 Properly set vector type of the intermediate stmt.
256 * tree-vect-stmts.c (vectorizable_operation): The destination
257 var always has vectype_out type.
259 2018-06-13 Jeff Law <law@redhat.com>
261 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
262 integer 0 for argument to print_rtl_with_bb.
263 (rl78_reorg): Likewise.
265 2018-06-13 David Malcolm <dmalcolm@redhat.com>
267 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
268 from rtx to rtx_insn *.
269 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
271 (add_sched_insns_for_speculation): Likewise for local "target",
272 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
273 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
274 from rtx_insn ** to rtx_code_label **.
275 (reorg_emit_nops): Likewise.
276 (c6x_reorg): Likewise for local "call_labels".
277 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
279 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
280 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
281 the loops over LABEL_REFS.
282 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
284 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
285 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
286 (split_branches): Strengthen local "olabel" from rtx to
287 rtx_insn *, adding a safe_as_a cast.
288 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
290 (add_insn_after): Likewise for first two params.
291 (add_insn_before): Likewise.
292 (remove_insn): Likewise for param.
293 (emit_pattern_before_noloc): Likewise for second and third params.
294 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
295 (emit_call_insn_before_noloc): Likewise.
296 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
298 (emit_barrier_before): Likewise.
299 (emit_label_before): Strengthen "label" param from "rtx" to
300 "rtx_code_label *". Strengthen "before" param from "rtx" to
302 (emit_insn_after_1): Strengthen "after" param from "rtx" to
304 (emit_pattern_after_noloc): Likewise.
305 (emit_insn_after_noloc): Likewise.
306 (emit_jump_insn_after_noloc): Likewise.
307 (emit_call_insn_after_noloc): Likewise.
308 (emit_debug_insn_after_noloc): Likewise.
309 (emit_barrier_after): Likewise.
310 (emit_label_after): Likewise for both params.
311 (emit_pattern_after_setloc): Likewise for "after" param. Convert
312 "loc" param from "int" to "location_t".
313 (emit_insn_after_setloc): Likewise.
314 (emit_jump_insn_after_setloc): Likewise.
315 (emit_call_insn_after_setloc): Likewise.
316 (emit_debug_insn_after_setloc): Likewise.
317 (emit_pattern_before_setloc): Likewise for "before" param. Convert
318 "loc" param from "int" to "location_t".
319 (emit_pattern_before): Convert NULL_RTX to NULL.
320 (emit_insn_before_setloc): Convert "loc" param from "int" to
322 (emit_jump_insn_before_setloc): Likewise.
323 (emit_call_insn_before_setloc): Likewise.
324 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
325 rtx_insn *. Convert "loc" param from "int" to "location_t".
326 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
327 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
328 Convert 3rd param from "int" to "location_t".
329 (emit_barrier_before, emit_barrier_after, next_real_insn):
330 Strengthen param from rtx to rtx_insn *.
331 (emit_label_before): Strengthen 1st param from "rtx" to
332 "rtx_code_label *". Strengthen 2nd param from "rtx" to
334 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
335 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
336 Strengthen 2nd param from "rtx" to "rtx_insn *".
337 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
338 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
339 Likewise. Convert 3rd param from "int" to "location_t".
340 (emit_label_after): Strengthen 1st param from "rtx" to
342 (next_real_insn, remove_insn): Strengthen param from "rtx" to
344 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
345 from "rtx" to "rtx_insn *".
347 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
349 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
350 bodies streamed in with -Q.
351 * dumpfile.c (dump_files): Add lto-stream-out dump file.
352 * dumpfile.h (tree_dump_index): Add lto_stream_out.
353 * gimple-streamer-out.c: Include gimple-pretty-print.h
354 (output_bb): Dump stmts streamed.
355 * lto-section-out.c: Include print-tree.h
356 (lto_begin_section): Dump sections created.
357 (lto_output_decl_index): Dump decl encoded.
358 * lto-streamer-out.c: Include print-tree.h
359 (create_output_block): Dump output block created.
360 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
361 (output_function): Dump function output.
362 (output_constructor): Dump constructor streamed.
363 (write_global_stream): Output indexes encoded.
364 (produce_asm_for_decls): Dump streams encoded.
365 * lto-streamer.c (streamer_dump_file): New global var.
366 * lto-streamer.h (streamer_dump_file): Declare.
367 * passes.c (ipa_write_summaries): Initialize streamer dump.
368 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
371 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
374 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
375 offsets for register save directives. Emit a second batch of save
376 directives, if need be, when the function accesses prior frames.
378 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
380 * config/arc/fpu.md (fmasf4): Force operand to register.
383 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
385 * config/arc/arc-protos.h (arc_pad_return): Remove.
386 * config/arc/arc.c (machine_function): Remove force_short_suffix
388 (arc_print_operand): Adjust printing of '&'.
389 (arc_verify_short): Remove conditional printing of short suffix.
390 (arc_final_prescan_insn): Remove reference to size_reason.
391 (pad_return): New function.
392 (arc_reorg): Call pad_return.
393 (arc_pad_return): Remove.
394 (arc_init_machine_status): Remove reference to force_short_suffix.
395 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
396 (attr length): When attribute iscompact is true force to 2
397 regardless; in the case of maybe check if we want to force the
398 instruction to have 4 bytes length.
399 (nopv): Change it to generate 4 byte long nop as well.
400 (blockage): New pattern.
401 (simple_return): Remove call to arc_pad_return.
402 (p_return_i): Likewise.
404 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
406 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
408 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
410 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
413 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
415 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
416 for ARC700 and ARCv2.
418 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
421 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
422 operands[2] instead of operands[1].
425 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
427 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
428 case, check whether the outer register overlaps an unallocatable
429 register, not just whether it fits the required class.
431 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
433 * poly-int.h (can_div_trunc_p): Add new overload in which all values
435 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
436 (memrefs_conflict_p): Likewise.
437 (init_alias_analysis): Likewise.
438 * cfgexpand.c (expand_debug_expr): Likewise.
439 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
440 * cse.c (fold_rtx): Likewise.
441 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
442 * expr.c (emit_block_move_hints): Likewise.
443 (clear_storage_hints, push_block, emit_push_insn): Likewise.
444 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
445 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
446 (emit_group_store): Likewise.
447 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
448 to read the PRE/POST_MODIFY increment.
449 * calls.c (store_one_arg): Use strip_offset.
450 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
452 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
454 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
455 (simplify_binary_operation_1): Extend CONST_INT handling to
457 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
458 than a HOST_WIDE_INT.
459 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
461 (adjust_mems, add_stores): Update accodingly.
462 (vt_canonicalize_addr): Track polynomial offsets.
463 (emit_note_insn_var_location): Likewise.
464 (vt_add_function_parameter): Likewise.
465 (vt_initialize): Likewise.
467 2018-06-12 Jeff Law <law@redhat.com>
469 * config.gcc (alpha*-*-freebsd*): Remove.
470 * config/alpha/freebsd.h: Remove.
472 2018-06-12 David Malcolm <dmalcolm@redhat.com>
475 * spellcheck-tree.c (levenshtein_distance): Rename to...
476 (get_edit_distance): ...this, and update for underlying renaming.
477 * spellcheck-tree.h (levenshtein_distance): Rename to...
478 (get_edit_distance): ...this.
479 * spellcheck.c (levenshtein_distance): Rename to...
480 (get_edit_distance): ...this. Convert from Levenshtein distance
481 to Damerau-Levenshtein distance by supporting transpositions of
482 adjacent characters. Rename "v1" to "v_next" and "v0" to
484 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
485 (selftest::test_edit_distance_unit_test_oneway): ...this, and
486 update for underlying renaming.
487 (selftest::levenshtein_distance_unit_test): Rename to...
488 (selftest::test_get_edit_distance_unit): ...this, and update for
490 (selftest::test_find_closest_string): Add example from PR 69968
491 where transposition helps
492 (selftest::test_metric_conditions): Update for renaming.
493 (selftest::test_metric_conditions): Likewise.
494 (selftest::spellcheck_c_tests): Likewise.
495 * spellcheck.h (levenshtein_distance): Rename both overloads to...
496 (get_edit_distance): ...this.
497 (best_match::consider): Update for renaming.
499 2018-06-12 Martin Sebor <msebor@redhat.com>
501 PR tree-optimization/85259
502 * builtins.c (compute_objsize): Handle constant offsets.
503 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
504 true iff a warning has been issued.
505 * gimple.h (gimple_nonartificial_location): New function.
506 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
507 gimple_nonartificial_location and handle -Wno-system-headers.
508 (handle_builtin_stxncpy): Same.
510 2018-06-12 Martin Sebor <msebor@redhat.com>
513 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
515 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
517 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
518 BUILTIN_VEC_XST entries for pointer to double and long long.
520 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
523 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
525 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
528 2018-06-12 Martin Liska <mliska@suse.cz>
530 * doc/options.texi: Document IntegerRange.
532 2018-06-12 Martin Liska <mliska@suse.cz>
534 * config/i386/i386.opt: Make MPX-related options as Deprecated.
535 * opt-functions.awk: Handle Deprecated flag.
536 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
538 (read_cmdline_option): Report warning for a deprecated option.
539 * opts.h (struct cl_option): Add new field cl_deprecated.
540 (CL_ERR_DEPRECATED): New.
542 2018-06-12 Martin Liska <mliska@suse.cz>
544 * doc/options.texi: Document Deprecated option flag.
546 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
548 * config/arc/arc-arch.h (arc_extras): New enum.
549 (arc_cpu_t):Add field extra.
550 (arc_cpu_types): Consider the extras.
551 * config/arc/arc-cpus.def: Add extras info.
552 * config/arc/arc-opts.h (processor_type): Consider extra field.
553 * config/arc/arc.c (arc_override_options): Handle extra field.
555 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
557 * config/arc/arc-arch.h: Update ARC_OPTX macro.
558 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
560 * config/arc/arc.c (arc_init): Update pic warning.
561 (irq_range): Update irq range parsing warnings.
562 (arc_override_options): Update various warning messages.
563 (arc_handle_aux_attribute): Likewise.
565 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
567 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
569 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
571 * doc/sourcebuild.texi: Document usage of line number 0 in verify
572 compiler messages directives.
574 2018-06-12 Matthew Fortune <mfortune@gmail.com>
576 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
577 * config/mips/mips-tables.opt: Regenerate.
578 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
580 * doc/invoke.texi: Document -march=i6500.
582 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
584 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
585 (i6400_gpmul): Add cpu_unit.
586 (i6400_gpdiv): Likewise.
587 (i6400_msa_add_d): Update reservations.
588 (i6400_msa_int_add) Likewise.
589 (i6400_msa_short_logic3) Likewise.
590 (i6400_msa_short_logic2) Likewise.
591 (i6400_msa_short_logic) Likewise.
592 (i6400_msa_move) Likewise.
593 (i6400_msa_cmp) Likewise.
594 (i6400_msa_short_float2) Likewise.
595 (i6400_msa_div_d) Likewise.
596 (i6400_msa_long_logic1) Likewise.
597 (i6400_msa_long_logic2) Likewise.
598 (i6400_msa_mult) Likewise.
599 (i6400_msa_long_float2) Likewise.
600 (i6400_msa_long_float4) Likewise.
601 (i6400_msa_long_float5) Likewise.
602 (i6400_msa_long_float8) Likewise.
603 (i6400_fpu_fadd): Include frint type.
604 (i6400_fpu_store): New define_insn_reservation.
605 (i6400_fpu_load): Likewise.
606 (i6400_fpu_move): Likewise.
607 (i6400_fpu_fcmp): Likewise.
608 (i6400_fpu_fmadd): Likewise.
609 (i6400_int_mult): Include imul3nc type and update reservation.
610 (i6400_int_div): Include idiv3 type and update reservation.
611 (i6400_int_load): Update to check type not move_type.
612 (i6400_int_store): Likewise.
613 (i6400_int_prefetch): Set zero latency.
615 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
617 * gcc.c: Document new %@{...} sequence.
618 (LINK_COMMAND_SPEC): Use it for the -L switches.
619 (cpp_unique_options): Use it for the -I switches.
620 (at_file_argbuf): New global variable.
621 (in_at_file): Likewise.
622 (alloc_args): Create at_file_argbuf.
623 (clear_args): Truncate at_file_argbuf.
624 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
625 (open_at_file): New function.
626 (close_at_file): Likewise.
627 (create_at_file): Delete.
628 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
631 (validate_switches_from_spec): Deal with %@{...} sequence.
632 (validate_switches): Likewise.
633 (driver::finalize): Call clear_args.
635 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
637 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
639 2018-06-11 Martin Sebor <msebor@redhat.com>
641 * doc/invoke.texi (-Wall): List -Wc++17-compat.
642 (Wno-class-memaccess): Add @opindex.
643 (Wno-templates, Wno-multiple-inheritance): Same.
644 (Wno-virtual-inheritance, Wno-namespaces): Same.
645 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
646 (Wno-format-overflow, Wno-format-truncation): Same.
647 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
648 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
649 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
650 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
651 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
652 (Wno-misspelled-isr): Same.
654 2018-06-11 Martin Sebor <msebor@redhat.com>
656 * PR tree-optimization/86083
657 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
659 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
661 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
663 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
666 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
667 on the correct operand.
668 (*movdi_internal64): Ditto.
670 2018-06-11 Martin Liska <mliska@suse.cz>
672 PR tree-optimization/86089
673 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
675 2018-06-11 Julia Koval <julia.koval@intel.com>
677 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
678 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
679 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
681 2018-06-11 Olivier Hainque <hainque@adacore.com>
683 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
684 for Ada with strict dwarf2.
686 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
689 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
692 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
694 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
696 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
698 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
701 2018-06-08 Martin Liska <mliska@suse.cz>
703 * tree-cfg.h (debug_function): Fix argument type to match
706 2018-06-08 Martin Liska <mliska@suse.cz>
708 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
709 Remove usage of MPX-related (and removed) fields.
710 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
712 2018-06-08 David Malcolm <dmalcolm@redhat.com>
714 * cfg.c (debug): Use TDF_NONE rather than 0.
715 * cfghooks.c (debug): Likewise.
716 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
717 (struct dump_option_value_info): Convert to...
718 (struct kv_pair): ...this template type.
719 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
721 (optinfo_verbosity_options): Likewise.
722 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
724 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
725 than int for "optgroup_flags" param.
726 (dump_generic_expr_loc): Use dump_flags_t rather than int for
728 (dump_dec): Likewise.
729 (dump_finish): Use TDF_NONE rather than 0.
730 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
731 rather than int for "optgroup_flags" param. Use TDF_NONE rather
732 than 0. Update for change to option_ptr.
733 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
734 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
735 0. Update for changes to optinfo_verbosity_options and
737 (opt_info_switch_p): Convert optgroup_flags from int to
739 (dump_basic_block): Use dump_flags_t rather than int
740 for "dump_kind" param.
741 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
742 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
743 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
744 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
745 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
746 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
747 TDF_NONE): Convert from macros to...
748 (enum dump_flag): ...this new enum.
749 (dump_flags_t): Update to use enum.
750 (operator|, operator&, operator~, operator|=, operator&=):
751 Implement for dump_flags_t.
752 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
753 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
754 Convert from macros to...
755 (enum optgroup_flag): ...this new enum.
756 (optgroup_flags_t): New typedef.
757 (operator|, operator|=): Implement for optgroup_flags_t.
758 (struct dump_file_info): Convert field "alt_flags" to
759 dump_flags_t. Convert field "optgroup_flags" to
761 (dump_basic_block): Use dump_flags_t rather than int for param.
762 (dump_generic_expr_loc): Likewise.
763 (dump_dec): Likewise.
764 (dump_register): Convert param "optgroup_flags" to
766 (opt_info_enable_passes): Likewise.
767 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
769 * gimple-pretty-print.c (debug): Likewise.
770 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
771 (merged_store_group::apply_stores): Likewise.
772 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
773 * gimple.c (verify_gimple_pp): Likewise.
774 * graphite-poly.c (print_pbb_body): Likewise.
775 * passes.c (pass_manager::register_one_dump_file): Convert
776 local "optgroup_flags" to optgroup_flags_t.
777 * print-tree.c (print_node): Use TDF_NONE rather than 0.
779 (debug_body): Likewise.
780 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
782 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
784 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
785 (convert_mult_to_fma): Likewise.
786 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
787 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
788 * tree-vect-data-refs.c (dump_lower_bound): Convert param
789 "dump_kind" to dump_flags_t.
791 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
793 * config/rs6000/rs6000.c (min, max): Delete.
795 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
797 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
798 -mabi=spe and -mabi=no-spe.
800 2018-06-08 Martin Liska <mliska@suse.cz>
802 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
803 where we expect an existing summary.
805 2018-06-08 Martin Liska <mliska@suse.cz>
807 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
808 * ipa-inline.h (estimate_edge_growth): Likewise.
810 2018-06-08 Martin Liska <mliska@suse.cz>
812 * cgraph.c (function_version_hasher::hash): Use
813 cgraph_node::get_uid ().
814 (function_version_hasher::equal):
815 * cgraph.h (cgraph_node::get_uid): New method.
816 * ipa-inline.c (update_caller_keys): Use
817 cgraph_node::get_uid ().
818 (update_callee_keys): Likewise.
819 * ipa-utils.c (searchc): Likewise.
820 (ipa_reduced_postorder): Likewise.
821 * lto-cgraph.c (input_node): Likewise.
822 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
823 * symbol-summary.h (symtab_insertion): Likewise.
824 (symtab_removal): Likewise.
825 (symtab_duplication): Likewise.
826 * tree-pretty-print.c (dump_function_header): Likewise.
827 * tree-sra.c (convert_callers_for_node): Likewise.
829 2018-06-08 Martin Liska <mliska@suse.cz>
831 * cgraph.c (symbol_table::create_edge): Always assign a new
833 (symbol_table::free_edge): Do not recycle numbers.
834 * cgraph.h (cgraph_edge::get): New method.
835 * symbol-summary.h (symtab_removal): Use it.
836 (symtab_duplication): Likewise.
837 (call_summary::hashable_uid): Remove.
839 2018-06-08 Martin Liska <mliska@suse.cz>
841 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
842 (initialize_growth_caches): Remove.
843 (free_growth_caches): Likewise.
844 (do_estimate_edge_time): Use edge_growth_cache.
845 (do_estimate_edge_size): Likewise.
846 (do_estimate_edge_hints): Likewise.
847 * ipa-inline.c (reset_edge_caches): Likewise.
848 (recursive_inlining): Likewise.
849 (inline_small_functions): Likewise.
850 * ipa-inline.h (initialize_growth_caches): Remove.
851 (estimate_edge_size): Likewise.
852 (estimate_edge_time): Likewise.
853 (estimate_edge_hints): Likewise.
854 (reset_edge_growth_cache): Likewise.
855 * symbol-summary.h (call_summary::remove): New method.
857 2018-06-08 Martin Liska <mliska@suse.cz>
859 * ipa-cp.c (class edge_clone_summary): New summary.
860 (grow_edge_clone_vectors): Remove.
861 (ipcp_edge_duplication_hook): Remove.
862 (class edge_clone_summary_t): New call_summary class.
863 (ipcp_edge_removal_hook): Remove.
864 (edge_clone_summary_t::duplicate): New function.
865 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
866 (create_specialized_node): Likewise.
867 (ipcp_driver): Initialize edge_clone_summaries and do not
870 2018-06-08 Martin Liska <mliska@suse.cz>
872 * symbol-summary.h (get): New function.
873 (call_summary::m_initialize_when_cloning): New class member.
875 2018-06-08 Martin Liska <mliska@suse.cz>
877 * cgraph.c (cgraph_node::remove): Do not recycle uid.
878 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
879 (symbol_table::allocate_cgraph_symbol): Do not set uid.
880 * passes.c (uid_hash_t): Record removed_nodes by their uids.
881 (remove_cgraph_node_from_order): Use the removed_nodes set.
882 (do_per_function_toporder): Likwise.
883 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
884 instead of summary_uid.
885 (symtab_removal): Likewise.
886 (symtab_duplication): Likewise.
888 2018-06-08 Martin Liska <mliska@suse.cz>
890 * ipa-cp.c (ipcp_store_bits_results): Use
891 ipcp_transformation_sum.
892 (ipcp_store_vr_results): Likewise.
893 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
895 (ipcp_transformation_initialize): ... this.
896 (ipa_set_node_agg_value_chain):
897 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
898 (write_ipcp_transformation_info): Likewise.
899 (read_ipcp_transformation_info): Likewise.
900 (ipcp_update_bits): Likewise.
901 (ipcp_update_vr): Likewise.
902 (ipcp_transform_function): Likewise.
903 * ipa-prop.h: Rename ipcp_transformation_summary to
905 (class ipcp_transformation_t): New function summary.
906 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
907 (ipa_get_agg_replacements_for_node): Likewise.
909 2018-06-08 Martin Liska <mliska@suse.cz>
911 * ipa-pure-const.c (struct funct_state_d): Do it class instead
913 (class funct_state_summary_t): New function_summary class.
914 (has_function_state): Remove.
915 (get_function_state): Likewise.
916 (set_function_state): Likewise.
917 (add_new_function): Likewise.
918 (funct_state_summary_t::insert): New function.
919 (duplicate_node_data): Remove.
920 (remove_node_data): Remove.
921 (funct_state_summary_t::duplicate): New function.
922 (register_hooks): Create new funct_state_summaries.
923 (pure_const_generate_summary): Use it.
924 (pure_const_write_summary): Likewise.
925 (pure_const_read_summary): Likewise.
926 (propagate_pure_const): Likewise.
927 (propagate_nothrow): Likewise.
928 (dump_malloc_lattice): Likewise.
929 (propagate_malloc): Likewise.
930 (execute): Do not register hooks, just remove summary
932 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
935 2018-06-08 Martin Liska <mliska@suse.cz>
937 * ipa-reference.c (remove_node_data): Remove.
938 (duplicate_node_data): Likewise.
939 (class ipa_ref_var_info_summary_t): New class.
940 (class ipa_ref_opt_summary_t): Likewise.
941 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
942 (get_reference_optimization_summary): Use
943 ipa_ref_opt_sum_summaries.
944 (set_reference_vars_info): Remove.
945 (set_reference_optimization_summary): Likewise.
946 (ipa_init): Create summaries.
947 (init_function_info): Use function summary.
948 (ipa_ref_opt_summary_t::duplicate): New function.
949 (ipa_ref_opt_summary_t::remove): New function.
950 (get_read_write_all_from_node): Fix GNU coding style.
951 (propagate): Use function summary.
952 (write_node_summary_p): Fix GNU coding style.
953 (stream_out_bitmap): Likewise.
954 (ipa_reference_read_optimization_summary): Use function summary.
955 (ipa_reference_c_finalize): Do not release hooks.
957 2018-06-08 Martin Liska <mliska@suse.cz>
959 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
960 (analyze_function_body): Extract multiple calls of get_create.
961 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
962 * ipa-inline.c (recursive_inlining): Use ::get method.
963 * ipa-inline.h (estimate_edge_growth): Likewise.
965 2018-06-08 Martin Liska <mliska@suse.cz>
967 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
969 (hsa_function_summary::hsa_function_summary): Use the new enum
971 (hsa_gpu_implementation_p): Use hsa_summaries::get.
972 * hsa-gen.c (hsa_get_host_function): Likewise.
973 (get_brig_function_name): Likewise.
974 * ipa-hsa.c (process_hsa_functions): Likewise.
975 (ipa_hsa_write_summary): Likewise.
976 * symbol-summary.h (symtab_duplication): Use ::get function/
979 2018-06-08 Martin Liska <mliska@suse.cz>
981 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
983 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
984 (hsa_register_kernel): Likewise.
985 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
986 * hsa-gen.c (hsa_get_host_function): Likewise.
987 (get_brig_function_name): Likewise.
988 (generate_hsa): Likewise.
989 (pass_gen_hsail::execute): Likewise.
990 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
991 (devirtualization_time_bonus): Likewise.
992 (ipcp_propagate_stage): Likewise.
993 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
994 (edge_set_predicate): Likewise.
995 (evaluate_conditions_for_known_args): Likewise.
996 (evaluate_properties_for_edge): Likewise.
997 (ipa_fn_summary::reset): Likewise.
998 (ipa_fn_summary_t::duplicate): Likewise.
999 (dump_ipa_call_summary): Likewise.
1000 (ipa_dump_fn_summary): Likewise.
1001 (analyze_function_body): Likewise.
1002 (compute_fn_summary): Likewise.
1003 (estimate_edge_devirt_benefit): Likewise.
1004 (estimate_edge_size_and_time): Likewise.
1005 (estimate_calls_size_and_time): Likewise.
1006 (estimate_node_size_and_time): Likewise.
1007 (inline_update_callee_summaries): Likewise.
1008 (remap_edge_change_prob): Likewise.
1009 (remap_edge_summaries): Likewise.
1010 (ipa_merge_fn_summary_after_inlining): Likewise.
1011 (ipa_update_overall_fn_summary): Likewise.
1012 (read_ipa_call_summary): Likewise.
1013 (inline_read_section): Likewise.
1014 (write_ipa_call_summary): Likewise.
1015 (ipa_fn_summary_write): Likewise.
1016 (ipa_free_fn_summary): Likewise.
1017 * ipa-hsa.c (process_hsa_functions): Likewise.
1018 (ipa_hsa_write_summary): Likewise.
1019 (ipa_hsa_read_section): Likewise.
1020 * ipa-icf.c (sem_function::merge): Likewise.
1021 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1022 (do_estimate_edge_time): Likewise.
1023 (estimate_size_after_inlining): Likewise.
1024 (estimate_growth): Likewise.
1025 (growth_likely_positive): Likewise.
1026 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1027 (inline_call): Likewise.
1028 * ipa-inline.c (caller_growth_limits): Likewise.
1029 (can_inline_edge_p): Likewise.
1030 (can_inline_edge_by_limits_p): Likewise.
1031 (compute_uninlined_call_time): Likewise.
1032 (compute_inlined_call_time): Likewise.
1033 (want_inline_small_function_p): Likewise.
1034 (edge_badness): Likewise.
1035 (update_caller_keys): Likewise.
1036 (update_callee_keys): Likewise.
1037 (recursive_inlining): Likewise.
1038 (inline_small_functions): Likewise.
1039 (inline_to_all_callers_1): Likewise.
1040 (dump_overall_stats): Likewise.
1041 (early_inline_small_functions): Likewise.
1042 (early_inliner): Likewise.
1043 * ipa-inline.h (estimate_edge_growth): Likewise.
1044 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1045 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1046 * ipa-prop.h (IPA_NODE_REF): Likewise.
1047 (IPA_EDGE_REF): Likewise.
1048 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1049 (propagate_malloc): Likewise.
1050 * ipa-split.c (execute_split_functions): Likewise.
1051 * symbol-summary.h: Rename get to get_create.
1053 (get_create): Likewise.
1054 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1056 2018-06-08 Martin Liska <mliska@suse.cz>
1058 * symbol-summary.h (release): Move definition out of class
1060 (symtab_removal): Likewise.
1061 (symtab_duplication): Likewise.
1063 2018-06-08 Martin Liska <mliska@suse.cz>
1065 * symbol-summary.h (function_summary): Move constructor
1066 implementation out of class declaration.
1067 (release): Likewise.
1068 (symtab_insertion): Likewise.
1069 (symtab_removal): Likewise.
1070 (symtab_duplication): Likewise.
1073 2018-06-08 Martin Liska <mliska@suse.cz>
1075 * Makefile.in: Remove support for MPX (macros, related functions,
1076 fields in cgraph_node, ...).
1077 * builtin-types.def (BT_BND): Likewise.
1078 (BT_FN_BND_CONST_PTR): Likewise.
1079 (BT_FN_CONST_PTR_BND): Likewise.
1080 (BT_FN_VOID_PTR_BND): Likewise.
1081 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1082 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1083 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1084 (expand_builtin_mempcpy_with_bounds): Likewise.
1085 (expand_builtin_memset_with_bounds): Likewise.
1086 (expand_builtin_memset_args): Likewise.
1087 (std_expand_builtin_va_start): Likewise.
1088 (expand_builtin): Likewise.
1089 (expand_builtin_with_bounds): Likewise.
1090 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1091 (DEF_LIB_BUILTIN_CHKP): Likewise.
1092 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1093 (DEF_CHKP_BUILTIN): Likewise.
1094 (BUILT_IN_MEMCPY): Likewise.
1095 (BUILT_IN_MEMMOVE): Likewise.
1096 (BUILT_IN_MEMPCPY): Likewise.
1097 (BUILT_IN_MEMSET): Likewise.
1098 (BUILT_IN_STPCPY): Likewise.
1099 (BUILT_IN_STRCAT): Likewise.
1100 (BUILT_IN_STRCHR): Likewise.
1101 (BUILT_IN_STRCPY): Likewise.
1102 (BUILT_IN_STRLEN): Likewise.
1103 (BUILT_IN_MEMCPY_CHK): Likewise.
1104 (BUILT_IN_MEMMOVE_CHK): Likewise.
1105 (BUILT_IN_MEMPCPY_CHK): Likewise.
1106 (BUILT_IN_MEMSET_CHK): Likewise.
1107 (BUILT_IN_STPCPY_CHK): Likewise.
1108 (BUILT_IN_STRCAT_CHK): Likewise.
1109 (BUILT_IN_STRCPY_CHK): Likewise.
1110 * calls.c (store_bounds): Likewise.
1111 (emit_call_1): Likewise.
1112 (special_function_p): Likewise.
1113 (maybe_warn_nonstring_arg): Likewise.
1114 (initialize_argument_information): Likewise.
1115 (finalize_must_preallocate): Likewise.
1116 (compute_argument_addresses): Likewise.
1117 (expand_call): Likewise.
1118 * cfgexpand.c (expand_call_stmt): Likewise.
1119 (expand_return): Likewise.
1120 (expand_gimple_stmt_1): Likewise.
1121 (pass_expand::execute): Likewise.
1122 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1123 (cgraph_node::remove): Likewise.
1124 (cgraph_node::dump): Likewise.
1125 (cgraph_node::verify_node): Likewise.
1126 * cgraph.h (chkp_function_instrumented_p): Likewise.
1127 (symtab_node::get_alias_target): Likewise.
1128 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1129 (cgraph_local_p): Likewise.
1130 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1131 (cgraph_edge::rebuild_references): Likewise.
1132 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1133 (walk_polymorphic_call_targets): Likewise.
1134 (cgraph_node::expand_thunk): Likewise.
1135 (symbol_table::output_weakrefs): Likewise.
1136 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1137 (ix86_handle_option): Likewise.
1138 * config/i386/constraints.md: Likewise.
1139 * config/i386/i386-builtin-types.def (BND): Likewise.
1143 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1144 (BDESC_FIRST): Likewise.
1146 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1147 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1148 * config/i386/i386.c (enum reg_class): Likewise.
1149 (ix86_target_string): Likewise.
1150 (ix86_option_override_internal): Likewise.
1151 (ix86_conditional_register_usage): Likewise.
1152 (ix86_valid_target_attribute_inner_p): Likewise.
1153 (ix86_set_indirect_branch_type): Likewise.
1154 (ix86_set_current_function): Likewise.
1155 (ix86_function_arg_regno_p): Likewise.
1156 (init_cumulative_args): Likewise.
1157 (ix86_function_arg_advance): Likewise.
1158 (ix86_function_arg): Likewise.
1159 (ix86_pass_by_reference): Likewise.
1160 (ix86_function_value_regno_p): Likewise.
1161 (ix86_function_value_1): Likewise.
1162 (ix86_function_value_bounds): Likewise.
1163 (ix86_return_in_memory): Likewise.
1164 (ix86_setup_incoming_vararg_bounds): Likewise.
1165 (ix86_va_start): Likewise.
1166 (indirect_thunk_need_prefix): Likewise.
1167 (print_reg): Likewise.
1168 (ix86_print_operand): Likewise.
1169 (ix86_expand_call): Likewise.
1170 (ix86_output_function_return): Likewise.
1171 (reg_encoded_number): Likewise.
1172 (BDESC_VERIFYS): Likewise.
1173 (ix86_init_mpx_builtins): Likewise.
1174 (ix86_init_builtins): Likewise.
1175 (ix86_emit_cmove): Likewise.
1176 (ix86_emit_move_max): Likewise.
1177 (ix86_expand_builtin): Likewise.
1178 (ix86_builtin_mpx_function): Likewise.
1179 (ix86_get_arg_address_for_bt): Likewise.
1180 (ix86_load_bounds): Likewise.
1181 (ix86_store_bounds): Likewise.
1182 (ix86_load_returned_bounds): Likewise.
1183 (ix86_store_returned_bounds): Likewise.
1184 (ix86_class_likely_spilled_p): Likewise.
1185 (ix86_hard_regno_mode_ok): Likewise.
1186 (x86_order_regs_for_local_alloc): Likewise.
1187 (ix86_mitigate_rop): Likewise.
1188 (ix86_bnd_prefixed_insn_p): Likewise.
1189 (ix86_mpx_bound_mode): Likewise.
1190 (ix86_make_bounds_constant): Likewise.
1191 (ix86_initialize_bounds): Likewise.
1192 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1193 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1194 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1195 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1196 (TARGET_CHKP_BOUND_MODE): Likewise.
1197 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1198 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1199 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1200 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1201 * config/i386/i386.h (TARGET_MPX): Likewise.
1202 (TARGET_MPX_P): Likewise.
1203 (VALID_BND_REG_MODE): Likewise.
1204 (FIRST_BND_REG): Likewise.
1205 (LAST_BND_REG): Likewise.
1206 (enum reg_class): Likewise.
1207 (BND_REG_P): Likewise.
1208 (BND_REGNO_P): Likewise.
1209 (BNDmode): Likewise.
1210 (ADJUST_INSN_LENGTH): Likewise.
1211 * config/i386/i386.md: Likewise.
1212 * config/i386/i386.opt: Likewise.
1213 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1214 (defined): Likewise.
1215 (LINK_MPX): Likewise.
1216 (MPX_SPEC): Likewise.
1217 (LIBMPX_SPEC): Likewise.
1218 (LIBMPXWRAPPERS_SPEC): Likewise.
1219 (CHKP_SPEC): Likewise.
1220 * config/i386/predicates.md: Likewise.
1221 * dbxout.c (dbxout_type): Likewise.
1222 * doc/extend.texi: Likewise.
1223 * doc/invoke.texi: Likewise.
1224 * doc/md.texi: Likewise.
1225 * doc/tm.texi: Likewise.
1226 * doc/tm.texi.in: Likewise.
1227 * dwarf2out.c (is_base_type): Likewise.
1228 (gen_formal_types_die): Likewise.
1229 (gen_subprogram_die): Likewise.
1230 (gen_type_die_with_usage): Likewise.
1231 (gen_decl_die): Likewise.
1232 (dwarf2out_late_global_decl): Likewise.
1233 * expr.c (expand_assignment): Likewise.
1234 (emit_storent_insn): Likewise.
1235 (store_expr_with_bounds): Likewise.
1236 (store_expr): Likewise.
1237 (expand_expr_real_1): Likewise.
1238 * expr.h (store_expr_with_bounds): Likewise.
1239 * function.c (use_register_for_decl): Likewise.
1240 (struct bounds_parm_data): Likewise.
1241 (assign_parms_augmented_arg_list): Likewise.
1242 (assign_parm_find_entry_rtl): Likewise.
1243 (assign_parm_is_stack_parm): Likewise.
1244 (assign_parm_load_bounds): Likewise.
1245 (assign_bounds): Likewise.
1246 (assign_parms): Likewise.
1247 (expand_function_start): Likewise.
1248 * gcc.c (CHKP_SPEC): Likewise.
1249 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1250 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1251 (wrestrict_dom_walker::check_call): Likewise.
1252 * gimple.c (gimple_build_call_from_tree): Likewise.
1253 * gimple.h (enum gf_mask): Likewise.
1254 (gimple_call_with_bounds_p): Likewise.
1255 (gimple_call_set_with_bounds): Likewise.
1256 * gimplify.c (gimplify_init_constructor): Likewise.
1257 * ipa-cp.c (initialize_node_lattices): Likewise.
1258 (propagate_constants_across_call): Likewise.
1259 (find_more_scalar_values_for_callers_subset): Likewise.
1260 * ipa-hsa.c (process_hsa_functions): Likewise.
1261 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1262 * ipa-icf.c (sem_function::merge): Likewise.
1263 * ipa-inline.c (early_inliner): Likewise.
1264 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1265 (warn_function_cold): Likewise.
1266 (propagate_pure_const): Likewise.
1267 * ipa-ref.h (enum GTY): Likewise.
1268 * ipa-split.c (find_retbnd): Likewise.
1269 (consider_split): Likewise.
1270 (split_function): Likewise.
1271 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1272 * ipa.c (walk_polymorphic_call_targets): Likewise.
1273 (symbol_table::remove_unreachable_nodes): Likewise.
1274 (process_references): Likewise.
1275 (cgraph_build_static_cdtor_1): Likewise.
1276 * lto-cgraph.c (lto_output_node): Likewise.
1277 (output_refs): Likewise.
1278 (compute_ltrans_boundary): Likewise.
1279 (input_overwrite_node): Likewise.
1280 (input_node): Likewise.
1281 (input_cgraph_1): Likewise.
1282 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1283 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1284 (class pass_chkp_instrumentation_passes): Likewise.
1285 (make_pass_chkp_instrumentation_passes): Likewise.
1286 * passes.def: Likewise.
1287 * rtl.h (struct GTY): Likewise.
1288 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1289 * stor-layout.c (layout_type): Likewise.
1290 * symtab.c: Likewise.
1291 * target.def: Likewise.
1292 * targhooks.c (default_chkp_bound_type): Likewise.
1293 (default_chkp_bound_mode): Likewise.
1294 (default_builtin_chkp_function): Likewise.
1295 (default_chkp_function_value_bounds): Likewise.
1296 (default_chkp_make_bounds_constant): Likewise.
1297 (default_chkp_initialize_bounds): Likewise.
1298 * targhooks.h (default_chkp_bound_type): Likewise.
1299 (default_chkp_bound_mode): Likewise.
1300 (default_builtin_chkp_function): Likewise.
1301 (default_chkp_function_value_bounds): Likewise.
1302 (default_chkp_make_bounds_constant): Likewise.
1303 (default_chkp_initialize_bounds): Likewise.
1304 * toplev.c (compile_file): Likewise.
1305 (process_options): Likewise.
1306 * tree-core.h (DEF_BUILTIN): Likewise.
1307 (DEF_BUILTIN_CHKP): Likewise.
1308 * tree-inline.c (declare_return_variable): Likewise.
1309 (remap_gimple_stmt): Likewise.
1310 (copy_bb): Likewise.
1311 (initialize_inlined_parameters): Likewise.
1312 (expand_call_inline): Likewise.
1313 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1314 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1315 (make_pass_ipa_chkp_produce_thunks): Likewise.
1316 (make_pass_chkp): Likewise.
1317 (make_pass_chkp_opt): Likewise.
1318 (make_pass_chkp_instrumentation_passes): Likewise.
1319 * tree-pretty-print.c (dump_generic_node): Likewise.
1320 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1321 * tree-ssa-dce.c (propagate_necessity): Likewise.
1322 (eliminate_unnecessary_stmts): Likewise.
1323 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1324 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1325 * tree-ssa-sccvn.h: Likewise.
1326 * tree-ssa-strlen.c (get_string_length): Likewise.
1327 (valid_builtin_call): Likewise.
1328 (adjust_last_stmt): Likewise.
1329 (handle_builtin_strchr): Likewise.
1330 (handle_builtin_strcpy): Likewise.
1331 (handle_builtin_stxncpy): Likewise.
1332 (handle_builtin_memcpy): Likewise.
1333 (handle_builtin_strcat): Likewise.
1334 (strlen_check_and_optimize_stmt): Likewise.
1335 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1336 * tree-streamer-in.c: Likewise.
1337 * tree-streamer.c (record_common_node): Likewise.
1338 * tree.c (tree_code_size): Likewise.
1339 (wide_int_to_tree_1): Likewise.
1340 (type_contains_placeholder_1): Likewise.
1341 (build_common_tree_nodes): Likewise.
1342 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1343 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1344 (POINTER_BOUNDS_P): Likewise.
1345 (BOUNDED_TYPE_P): Likewise.
1346 (BOUNDED_P): Likewise.
1347 (CALL_WITH_BOUNDS_P): Likewise.
1348 (pointer_bounds_type_node): Likewise.
1349 * value-prof.c (gimple_ic): Likewise.
1350 * var-tracking.c (vt_add_function_parameters): Likewise.
1351 * varasm.c (make_decl_rtl): Likewise.
1352 (assemble_start_function): Likewise.
1353 (output_constant): Likewise.
1354 (maybe_assemble_visibility): Likewise.
1355 * varpool.c (ctor_for_folding): Likewise.
1356 * chkp-builtins.def: Remove.
1357 * ipa-chkp.c: Remove.
1358 * ipa-chkp.h: Remove.
1359 * rtl-chkp.c: Remove.
1360 * rtl-chkp.h: Remove.
1361 * tree-chkp-opt.c: Remove.
1362 * tree-chkp.c: Remove.
1363 * tree-chkp.h: Remove.
1365 2018-06-07 Carl Love <cel@us.ibm.com>
1367 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1368 vextract_fp_from_shortl): Add BE support.
1370 2018-06-07 Paul Koning <ni1d@arrl.net>
1372 * compare-elim.c (try_merge_compare): Don't merge compare if
1373 address contains a side effect.
1374 (try_eliminate_compare): Likewise.
1376 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1378 * config.gcc: Support "tremont".
1379 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1380 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1382 * config/i386/i386.c (m_TREMONT): Define.
1383 (processor_target_table): Add "tremont".
1384 (PTA_TREMONT): Define.
1385 (ix86_lea_outperforms): Add TARGET_TREMONT.
1386 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1387 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1388 and M_INTEL_GOLDMONT_PLUS.
1389 (fold_builtin_cpu): Add "tremont".
1390 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1391 (ix86_option_override_internal): Add "tremont".
1392 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1393 (processor_type): Add PROCESSOR_TREMONT.
1394 * config/i386/x86-tune.def: Add m_TREMONT.
1395 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1397 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1399 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1400 symbol defined for msp430i* devices to be lower case.
1402 gcc/testsuite/gcc.target/msp430/
1403 * msp430i-device-symbol.c: New test.
1404 * msp430f-device-symbol.c: New test.
1405 * msp430.h: New test header file.
1407 2018-06-07 Richard Biener <rguenther@suse.de>
1409 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1410 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1411 Properly wrap signed arithmetic if overflow wraps.
1413 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1415 PR tree-optimization/69615
1416 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1417 of a cast from a same precision integral SSA_NAME in a bb dominated
1418 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1419 cast to utype if rhs2 has already a compatible type.
1421 2018-06-07 Richard Biener <rguenther@suse.de>
1423 PR tree-optimization/85935
1424 * graphite-scop-detection.c (find_params_in_bb): Analyze
1425 condition operands with respect to the correct loop. Assert
1426 the analysis doesn't fail.
1428 2018-06-04 Carl Love <cel@us.ibm.com>
1430 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1431 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1432 as it is slightly cheaper.
1433 (first_match_or_eos_index_<mode>):
1434 Calculate index using natural element order.
1435 (first_match_index_<mode>):
1436 Calculate index using natural element order.
1437 (first_match_or_eos_index_<mode>):
1438 Calculate index using natural order.
1439 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1440 for BE and LE modes.
1441 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1442 P9V_BUILTIN_VCLZLSBB_V16QI.
1443 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1446 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1448 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1449 indentation and line wrap for many prototypes. Add missing
1450 @smallexample directives around block of prototypes for vec_xl and
1453 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1455 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1456 track if we pass or return IEEE 128-bit floating point.
1457 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1458 C++ mangling that is compatible with GCC 8.1.
1459 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1460 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1461 floating point types.
1462 (rs6000_function_arg_advance_1): Likewise.
1463 (rs6000_mangle_type): Optionally generate mangled names that match
1464 what GCC 8.1 generated for IEEE 128-bit floating point types.
1465 (rs6000_globalize_decl_name): If we have an external function that
1466 passes or returns IEEE 128-bit types, generate a weak reference
1467 from the mangled name used in GCC 8.1 to the current mangled
1469 (rs6000_init_builtins): Make __ibm128 use the long double type if
1470 long double is IBM extended double. Make __float128 use the long
1471 double type if long double is IEEE 128-bit.
1474 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1475 macro for __ibm128 built-in functions.
1476 (PACK_IF): Add __ibm128 pack/unpack functions.
1477 (UNPACK_IF): Likewise.
1478 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1479 enable long double built-in functions if long double is IEEE
1480 128-bit floating point.
1481 (rs6000_invalid_builtin): Update long double built-in function
1483 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1484 functions, adjust the built-in function to use the long double
1485 built-in function if __ibm128 and long double are the same type.
1486 * doc/extend.texi (PowerPC builtins): Update documention for
1487 __builtin_{,un}pack_longdouble. Add documentation for
1488 __builtin_{,un}pack_ibm128.
1490 2018-06-06 Jim Wilson <jimw@sifive.com>
1492 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1493 (struct machine_function): New field interrupt_mode.
1494 (riscv_handle_type_attribute): New function. Add forward declaration.
1495 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1496 (riscv_expand_epilogue): Check interrupt_mode field.
1497 (riscv_set_current_function): Check interrupt attribute args and
1498 set interrupt_mode field.
1499 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1500 (riscv_sret, riscv_uret): New.
1501 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1502 new arguments to interrupt attribute.
1504 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1507 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1508 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1510 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1512 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1513 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1514 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1515 several redundant entries.
1517 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1519 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1520 type from "rtx" to "rtx_insn *".
1521 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1522 for local "call_insn", removing cast.
1523 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1525 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1527 PR tree-optimization/86066
1528 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1529 for BIT_INSERT_EXPR stores.
1531 2018-06-06 Richard Biener <rguenther@suse.de>
1533 PR tree-optimization/86062
1534 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1535 component refs ontop
1536 of to be offsetted base.
1538 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1540 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1541 to be static and remove check on interrupt attribute name.
1543 gcc/testsuite/gcc.target/msp430/
1544 * function-attributes-4.c: New test.
1545 * static-interrupts.c: New test.
1547 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1549 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1550 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1552 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1555 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1557 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1558 Remove second argument, change how error is called.
1559 (aarch64_layout_arg): Remove second argument from
1560 aarch64_err_no_fpadvsimd call.
1561 (aarch64_init_cumulative_args): Ditto.
1562 (aarch64_gimplify_va_arg_expr): Ditto.
1563 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1565 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1567 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1568 (*simple_return_indirect_internal<mode>): Rename from
1569 simple_return_indirect_internal. Use W mode iterator.
1570 (rstorssp): New expander.
1571 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1572 (clrssbsy): New expander.
1573 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1575 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1577 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1579 (cmse_check_pointed_object): Likewise.
1581 2018-06-05 Martin Liska <mliska@suse.cz>
1583 PR gcov-profile/47618
1584 * doc/invoke.texi: Document how -fprofile-dir format
1587 2018-06-05 Richard Biener <rguenther@suse.de>
1589 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1590 removal pretend DOM info isn't available so we do not update
1591 it and only remove edges, not dominated blocks. Actually free
1592 DOM info in case we removed something. Remove unreachable blocks.
1593 (mfb_keep_latches): Work with either DOM info or marked backedges.
1594 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1595 first. Mark backedges if DOM info isn't available.
1596 (Re-)compute DOM info after cleanup_control_flow_pre.
1598 2018-06-05 Richard Biener <rguenther@suse.de>
1600 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1601 (locus_discrim_hasher::hash): Adjust.
1602 (locus_discrim_hasher::equal): Likewise.
1603 (next_discriminator_for_locus): Work on line directly.
1604 (same_line_p): Pass in expanded locus1 as well.
1605 (assign_discriminators): Avoid redundant location expansions.
1607 2018-06-05 Richard Biener <rguenther@suse.de>
1609 PR tree-optimization/86046
1610 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1611 if required after clearing TREE_ADDRESSABLE.
1613 2018-06-05 Richard Biener <rguenther@suse.de>
1615 PR tree-optimization/86047
1616 * tree-ssa-loop.c (for_each_index): Glob handling of all
1617 decls and constants and really handle all of them.
1619 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1622 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1623 qualifier_void_pointer and qualifier_const_void_pointer.
1624 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1625 (arm_init_builtins): Handle the above.
1626 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1627 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1628 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1631 2018-06-05 Martin Liska <mliska@suse.cz>
1633 * auto-profile.c (read_autofdo_file): Do not use
1634 gcov_ctr_summary struct.
1635 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1636 * coverage.c (struct counts_entry): Likewise.
1637 (read_counts_file): Read just single summary entry.
1638 (get_coverage_counts): Use gcov_summary struct.
1639 * coverage.h (get_coverage_counts): Likewise.
1640 * gcov-dump.c (dump_working_sets): Likewise.
1641 (tag_summary): Dump just single summary.
1642 * gcov-io.c (gcov_write_summary): Write just histogram
1644 (gcov_read_summary): Read just single summary.
1645 (compute_working_sets): Use gcov_summary struct.
1646 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1647 of GCOV_COUNTERS_SUMMABLE.
1648 (GCOV_COUNTERS_SUMMABLE): Remove.
1649 (GCOV_FIRST_VALUE_COUNTER): Replace with
1650 GCOV_COUNTER_V_INTERVAL.
1651 (struct gcov_ctr_summary): Remove.
1652 (struct gcov_summary): Directly use fields of former
1654 (compute_working_sets): Use gcov_summary struct.
1655 * gcov.c (read_count_file): Do not use ctrs fields.
1656 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1658 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1660 * profile.c: Likewise.
1661 * profile.h: Likewise.
1663 2018-06-05 Martin Liska <mliska@suse.cz>
1665 PR gcov-profile/84846
1666 * gcov.c (output_lines): Print working directory only
1667 in intermediate format.
1669 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1671 * config/s390/s390-builtin-types.def: Add void function type.
1672 * config/s390/s390-builtins.def: Use the function type for the
1675 2018-06-04 Jim Wilson <jimw@sifive.com>
1677 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1679 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1680 and maybe_eh_return. Change regno to unsigned int. Use new args to
1681 handle EH_RETURN_DATA_REGNO registers properly.
1682 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1683 (riscv_expand_epilogue): Update comment. Change argument name and
1684 type. Update code to use new name and type. Pass new args to
1685 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1687 * config/riscv/riscv.md (NORMAL_RETURN): New.
1688 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1689 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1690 (eh_return): Call gen_eh_return_internal and emit barrier.
1691 (eh_return_internal): Call riscv_expand_epilogue.
1693 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1695 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1696 bit_insertion field and declare can_be_merged_into method.
1697 (merged_store_group::can_be_merged_into): New method.
1698 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1699 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1700 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1702 2018-06-04 Richard Biener <rguenther@suse.de>
1704 PR tree-optimization/85955
1705 * builtins.c (fold_builtin_sincos): Convert pointers to
1706 destination to appropriate type before dereferencing.
1708 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1710 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1712 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1714 * expr.c (expand_expr_real_1): Force the operand into memory if
1715 its TYPE_MODE is BLKmode and if there is no integer mode for
1716 the number of bits being extracted.
1718 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1722 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1723 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1725 2018-06-04 Richard Biener <rguenther@suse.de>
1727 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1728 (cleanup_tree_cfg_noloop): ... single caller. Do
1729 start_recording_case_labels later.
1731 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1733 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1734 to _IMMINTRIN_H_INCLUDED.
1735 * config/i386/pconfigintrin.h: Ditto.
1736 * config/i386/waitpkgintrin.h: Ditto.
1737 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1738 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1739 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1740 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1741 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1742 waitpkgintrin.h and cldemoteintrin.h.
1744 2018-06-04 Richard Biener <rguenther@suse.de>
1746 PR tree-optimization/86038
1747 * tracer.c (find_best_successor): Check probability for
1748 being initialized, bail out if not.
1750 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1753 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1754 of bits to ignore when comparing architectures.
1756 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1758 PR tree-optimization/69615
1759 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1760 maximum or minimum of the type, try to merge it also as if
1761 range1 is + [-, x - 1] or + [x + 1, -].
1764 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1766 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1768 PR tree-optimization/86034
1769 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1770 the unsigned bitfield type in a bit insertion sequence if it does not
1771 have a larger precision than the bitfield size.
1772 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1774 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1776 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1778 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1780 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1781 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1782 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1783 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1785 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1787 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1788 Disable -fdelete-null-pointer-checks for ELF toolchain.
1790 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1791 Kito Cheng <kito.cheng@gmail.com>
1793 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1794 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1795 (nds32*-*-*): Add glibc and uclibc conditions.
1796 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1797 (TARGET_EXCEPT_UNWIND_INFO): Define.
1798 * config/nds32/elf.h: New file.
1799 * config/nds32/linux.h: New file.
1800 * config/nds32/nds32-elf.opt: New file.
1801 * config/nds32/nds32-linux.opt: New file.
1802 * config/nds32/nds32-fp-as-gp.c
1803 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1804 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1806 (nds32_asm_file_end): Ditto.
1807 (nds32_print_operand): Ditto.
1808 (nds32_insert_attributes): Ditto.
1809 (nds32_init_libfuncs): New function.
1810 (TARGET_HAVE_TLS): Define.
1811 (TARGET_INIT_LIBFUNCS): Define.
1812 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1814 (TARGET_ELF): Apply different mcmodel setting.
1815 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1816 been migrated into elf.h and linux.h files.
1817 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1818 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1819 (mcmodel): The content has been migrated into nds32-elf.opt and
1820 nds32-linux.opt files.
1821 * config/nds32/t-elf: New file.
1822 * config/nds32/t-linux: New file.
1824 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1825 Shiva Chen <shiva0217@gmail.com>
1827 * config/nds32/constants.md (unspec_volatile_element): Add
1828 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1829 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1831 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1832 (make_pass_nds32_fp_as_gp): Declare.
1833 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1835 (nds32_asm_function_end_prologue): Remove unused asm output.
1836 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1837 (nds32_asm_file_start): Output necessary fp_as_gp information.
1838 (nds32_option_override): Adjust register usage.
1839 (nds32_expand_prologue): Consider fp_as_gp situation.
1840 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1841 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1844 (simple_return): Ditto.
1845 (omit_fp_begin): Output special directive for fp_as_gp.
1846 (omit_fp_end): Output special directive for fp_as_gp.
1847 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1848 mforbid-fp-as-gp): New options.
1850 2018-06-01 Mark Wielaard <mark@klomp.org>
1852 * dwarf2out.c (dwarf2out_finish): Remove generation of
1853 DW_AT_loclists_base.
1855 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1857 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1858 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1859 (struct merged_store_group): Add bit_insertion field.
1860 (dump_char_array): Use standard hexadecimal format.
1861 (merged_store_group::merged_store_group): Set bit_insertion to false.
1862 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1863 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1864 also print the mask in the dump file.
1865 (pass_store_merging::gate): Minor tweak.
1866 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1867 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1868 stores with INTEGER_CST stores.
1869 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1870 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1871 and use it throughout. Generate bit insertion sequences if need be.
1872 (pass_store_merging::process_store): Remove redundant condition.
1873 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1875 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1877 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1878 the 128-bit floating point types. Fix function comment.
1880 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1882 * config/aarch64/aarch64-simd.md
1883 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1885 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1888 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1890 PR tree-optimization/85989
1891 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1893 (backprop::intersect_uses): Check it when deciding whether this
1894 is a backedge reference.
1895 (backprop::process_block): Add each phi to m_visited_phis
1896 after visiting it, then clear it at the end.
1898 2018-06-01 Richard Biener <rguenther@suse.de>
1900 * tree-vectorizer.h (vect_dr_stmt): New function.
1901 (vect_get_load_cost): Adjust.
1902 (vect_get_store_cost): Likewise.
1903 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1904 Use vect_dr_stmt instead of DR_SMTT.
1905 (vect_record_base_alignments): Likewise.
1906 (vect_calculate_target_alignment): Likewise.
1907 (vect_compute_data_ref_alignment): Likewise and make static.
1908 (vect_update_misalignment_for_peel): Likewise.
1909 (vect_verify_datarefs_alignment): Likewise.
1910 (vector_alignment_reachable_p): Likewise.
1911 (vect_get_data_access_cost): Likewise. Pass down
1912 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1913 (vect_get_peeling_costs_all_drs): Likewise.
1914 (vect_peeling_hash_get_lowest_cost): Likewise.
1915 (vect_enhance_data_refs_alignment): Likewise.
1916 (vect_find_same_alignment_drs): Likewise.
1917 (vect_analyze_data_refs_alignment): Likewise.
1918 (vect_analyze_group_access_1): Likewise.
1919 (vect_analyze_group_access): Likewise.
1920 (vect_analyze_data_ref_access): Likewise.
1921 (vect_analyze_data_ref_accesses): Likewise.
1922 (vect_vfa_segment_size): Likewise.
1923 (vect_small_gap_p): Likewise.
1924 (vectorizable_with_step_bound_p): Likewise.
1925 (vect_prune_runtime_alias_test_list): Likewise.
1926 (vect_analyze_data_refs): Likewise.
1927 (vect_supportable_dr_alignment): Likewise.
1928 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1929 (vect_gen_prolog_loop_niters): Likewise.
1930 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1931 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1933 (vect_recog_mask_conversion_pattern): Likewise.
1934 (vect_try_gather_scatter_pattern): Likewise.
1935 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1936 to vect_get_store_cost.
1937 (vect_get_store_cost): Get stmt_info instead of DR.
1938 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1939 (vect_get_load_cost): Get stmt_info instead of DR.
1941 2018-06-01 Richard Biener <rguenther@suse.de>
1944 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1945 vars wrapped in MEM_REFs.
1947 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1949 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1950 Fix subreg tests so that we only return a choice between
1951 GENERAL_REGS and FP_REGS if the original classes included both.
1953 2018-06-01 Richard Biener <rguenther@suse.de>
1956 * tree-ssa-structalias.c (get_function_part_constraint):
1957 Handle NULL fi->decl.
1958 (find_func_aliases_for_call): Properly handle indirect
1959 fi from direct call.
1960 (find_func_clobbers): Likewise.
1961 (ipa_pta_execute): Likewise.
1962 (create_variable_info_for): For functions that are ifunc_resolver
1963 resolve to a varinfo that contains the result of the resolver call.
1964 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1967 2018-05-31 Michael Collison <michael.collison@arm.com>
1969 * config/aarch64/aarch64.md:
1970 (*fix_to_zero_extenddfdi2): New pattern.
1971 * gcc.target/aarch64/fix_extend1.c: New testcase.
1973 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1977 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1978 and BUILT_IN_STRNCMP_EQ.
1979 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1980 BUILT_IN_STRNCMP_EQ.
1981 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1982 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1983 (gimple_fold_builtin): Likewise.
1984 * tree-ssa-strlen.c (compute_string_length): New function.
1985 (determine_min_obsize): New function.
1986 (handle_builtin_string_cmp): New function to handle calls to
1987 string compare functions.
1988 (strlen_optimize_stmt): Add handling to builtin string compare
1990 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1991 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1992 * tree.c (build_common_builtin_nodes): Add new defines of
1993 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1995 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1998 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1999 functions with naked attribute.
2001 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2003 * config/i386/sse.md (avx_vec_concat<mode>):
2004 Substitute concat_tg_mode mode attribute with xtg_mode.
2005 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2006 (concat_tg_mode): Remove mode attribute.
2008 2018-05-31 Martin Sebor <msebor@redhat.com>
2011 * calls.c (alloc_max_size): Correct a logic error/typo.
2012 Treat excessive arguments as infinite. Warn for invalid arguments.
2013 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2015 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2018 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2019 and movx for Haswell.
2021 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2022 Cesar Philippidis <cesar@codesourcery.com>
2025 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2026 when emitting error on private/firstprivate reductions.
2027 * omp-low.c (lower_omp_target): Avoid reference-type processing
2028 on pointers for firstprivate clause.
2030 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2032 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2035 * config/aarch64/aarch64-simd.md
2036 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2037 (aarch64_ld1_x3_<mode>): Likewise
2038 (aarch64_st1x2<VALLDIF:mode>): Likewise
2039 (aarch64_st1_x2_<mode>): Likewise
2040 (aarch64_st1x3<VALLDIF:mode>): Likewise
2041 (aarch64_st1_x3_<mode>): Likewise
2042 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2043 (vld1_s8_x3): Likewise.
2044 (vld1_u16_x3): Likewise.
2045 (vld1_s16_x3): Likewise.
2046 (vld1_u32_x3): Likewise.
2047 (vld1_s32_x3): Likewise.
2048 (vld1_u64_x3): Likewise.
2049 (vld1_s64_x3): Likewise.
2050 (vld1_f16_x3): Likewise.
2051 (vld1_f32_x3): Likewise.
2052 (vld1_f64_x3): Likewise.
2053 (vld1_p8_x3): Likewise.
2054 (vld1_p16_x3): Likewise.
2055 (vld1_p64_x3): Likewise.
2056 (vld1q_u8_x3): Likewise.
2057 (vld1q_s8_x3): Likewise.
2058 (vld1q_u16_x3): Likewise.
2059 (vld1q_s16_x3): Likewise.
2060 (vld1q_u32_x3): Likewise.
2061 (vld1q_s32_x3): Likewise.
2062 (vld1q_u64_x3): Likewise.
2063 (vld1q_s64_x3): Likewise.
2064 (vld1q_f16_x3): Likewise.
2065 (vld1q_f32_x3): Likewise.
2066 (vld1q_f64_x3): Likewise.
2067 (vld1q_p8_x3): Likewise.
2068 (vld1q_p16_x3): Likewise.
2069 (vld1q_p64_x3): Likewise.
2070 (vst1_s64_x2): Likewise.
2071 (vst1_u64_x2): Likewise.
2072 (vst1_f64_x2): Likewise.
2073 (vst1_s8_x2): Likewise.
2074 (vst1_p8_x2): Likewise.
2075 (vst1_s16_x2): Likewise.
2076 (vst1_p16_x2): Likewise.
2077 (vst1_s32_x2): Likewise.
2078 (vst1_u8_x2): Likewise.
2079 (vst1_u16_x2): Likewise.
2080 (vst1_u32_x2): Likewise.
2081 (vst1_f16_x2): Likewise.
2082 (vst1_f32_x2): Likewise.
2083 (vst1_p64_x2): Likewise.
2084 (vst1q_s8_x2): Likewise.
2085 (vst1q_p8_x2): Likewise.
2086 (vst1q_s16_x2): Likewise.
2087 (vst1q_p16_x2): Likewise.
2088 (vst1q_s32_x2): Likewise.
2089 (vst1q_s64_x2): Likewise.
2090 (vst1q_u8_x2): Likewise.
2091 (vst1q_u16_x2): Likewise.
2092 (vst1q_u32_x2): Likewise.
2093 (vst1q_u64_x2): Likewise.
2094 (vst1q_f16_x2): Likewise.
2095 (vst1q_f32_x2): Likewise.
2096 (vst1q_f64_x2): Likewise.
2097 (vst1q_p64_x2): Likewise.
2098 (vst1_s64_x3): Likewise.
2099 (vst1_u64_x3): Likewise.
2100 (vst1_f64_x3): Likewise.
2101 (vst1_s8_x3): Likewise.
2102 (vst1_p8_x3): Likewise.
2103 (vst1_s16_x3): Likewise.
2104 (vst1_p16_x3): Likewise.
2105 (vst1_s32_x3): Likewise.
2106 (vst1_u8_x3): Likewise.
2107 (vst1_u16_x3): Likewise.
2108 (vst1_u32_x3): Likewise.
2109 (vst1_f16_x3): Likewise.
2110 (vst1_f32_x3): Likewise.
2111 (vst1_p64_x3): Likewise.
2112 (vst1q_s8_x3): Likewise.
2113 (vst1q_p8_x3): Likewise.
2114 (vst1q_s16_x3): Likewise.
2115 (vst1q_p16_x3): Likewise.
2116 (vst1q_s32_x3): Likewise.
2117 (vst1q_s64_x3): Likewise.
2118 (vst1q_u8_x3): Likewise.
2119 (vst1q_u16_x3): Likewise.
2120 (vst1q_u32_x3): Likewise.
2121 (vst1q_u64_x3): Likewise.
2122 (vst1q_f16_x3): Likewise.
2123 (vst1q_f32_x3): Likewise.
2124 (vst1q_f64_x3): Likewise.
2125 (vst1q_p64_x3): Likewise.
2127 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2129 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2130 user_label_prefix to name.
2132 * tree-core.h: Update comment about the format of NAME string
2133 passed to handler in attribute_spec.
2135 * config/msp430/msp430.md: Remove erroneous subreg expression from
2136 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2137 zero_extend{q,h}isi2.
2139 2018-05-30 Borislav Petkov <bp@suse.de>
2141 * doc/extend.texi: Document some architecture specific
2142 constraints and sort entries.
2144 2018-05-30 Martin Sebor <msebor@redhat.com>
2147 * builtins.c (expand_builtin_stpcpy_1): New function.
2148 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2149 only if the former succeeds.
2151 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2153 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2156 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2158 * doc/invoke.texi (-flinker-output): Document
2160 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2162 * passes.c (ipa_write_summaries): Only modify statements if body
2164 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2165 incrementally linking.
2166 (ipa_passes): Likewise.
2167 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2168 pass down resolution info.
2169 * common.opt (flag_incremental_link): Update info.
2170 * gcc.c (plugin specs): Turn flinker-output=* to
2171 -plugin-opt=-linker-output-known
2172 * toplev.c (compile_file): Also cut compilation when doing incremental
2174 * flag-types. (enum lto_partition_model): Add
2175 LTO_LINKER_OUTPUT_NOLTOREL.
2176 (invoke.texi): Add -flinker-output docs.
2177 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2178 link same way as WPA; do not stream in dead initializers.
2180 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2181 darf2out_register_external_decl): Support incremental link.
2183 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2185 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2187 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2189 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2190 it down to simple_object_copy_lto_debug_sections.
2191 (run_gcc): Determine incremental LTO link time and configure
2192 lto1 into non-wpa mode, disable renaming of debug sections.
2194 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2196 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2197 descriptions of various incorrectly documented functions.
2199 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2202 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2203 address check not strict.
2205 2018-05-30 Richard Biener <rguenther@suse.de>
2207 PR tree-optimization/85964
2208 * tracer.c (better_p): Drop initialized count check, we only
2209 call the function with initialized counts now.
2210 (find_best_successor): Do find a best edge if one
2211 has uninitialized count.
2212 (find_best_predecessor): Likewise. Do BB frequency check only
2213 if count is initialized.
2215 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2217 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2218 (aarch64_ldrstr_offset_compare): New.
2219 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2220 load/store orderings.
2221 (aarch64_gen_adjusted_ldpstp): Likewise.
2223 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2225 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2226 Check for subset of GENERAL_REGS and FP_REGS.
2227 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2230 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2232 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2233 and wi::to_poly_offset. Add the current offset and then check
2234 whether the sum fits, rather than using an unchecked addition of
2235 a checked term. Check for a shwi rather than a uhwi.
2236 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2237 (store_constructor): Use poly_int_tree_p.
2238 (expand_expr_real_1): Likewise.
2239 * function.c (assign_temp): Likewise.
2240 * fold-const.c (const_binop): Use poly_int_tree_p and
2242 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2244 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2245 to_poly_offset for MEM offsets.
2246 * ipa-icf.c (sem_variable::equals): Likewise.
2247 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2248 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2249 wi::to_poly_offset for BIT_FIELD_REF offsets.
2250 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2252 * var-tracking.c (emit_note_insn_var_location): Use
2253 tree_to_poly_uint64.
2255 2018-05-29 Jim Wilson <jimw@sifive.com>
2257 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2259 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2262 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2263 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2265 (sse4_1_round<mode>2): Use nonimmediate_operand
2266 for operand 1 predicate.
2268 2018-05-29 Martin Sebor <msebor@redhat.com>
2269 Richard Biener <rguenther@suse.de>
2272 * calls.c (get_size_range): Call determine_value_range instead
2273 of get_value_range..
2274 * tree-vrp.h (determine_value_range): Declared new function.
2275 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2277 2018-05-29 Richard Biener <rguenther@suse.de>
2279 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2280 sure to use non-pattern stmts for get_earlier_stmt arguments.
2281 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2282 called on pattern stmts.
2283 (get_later_stmt): Likewise.
2285 2018-05-29 Martin Liska <mliska@suse.cz>
2287 PR gcov-profile/85759
2288 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2291 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2293 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2295 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2299 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2300 VEC_PACK_FLOAT_EXPR): New tree codes.
2301 * tree-pretty-print.c (op_code_prio): Handle
2302 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2303 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2304 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2305 * tree-inline.c (estimate_operator_cost): Likewise.
2306 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2307 * fold-const.c (const_binop): Likewise.
2308 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2309 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2310 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2311 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2312 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2313 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2314 * expr.c (expand_expr_real_2): Likewise.
2315 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2316 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2317 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2319 * optabs.c (expand_widen_pattern_expr): For
2320 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2321 sign from result type rather than operand's type.
2322 (expand_binop_directly): For vec_packu_float_optab and
2323 vec_packs_float_optab allow result type to be different from operand's
2325 * optabs-tree.c (optab_for_tree_code): Handle
2326 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2327 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2328 * tree-vect-generic.c (expand_vector_operations_1): Handle
2329 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2330 VEC_PACK_FLOAT_EXPR.
2331 * tree-vect-stmts.c (supportable_widening_operation): Handle
2333 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2334 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2335 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2336 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2337 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2339 (vec_pack<floatprefix>_float_<mode>): New expander.
2340 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2342 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2343 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2344 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2345 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2346 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2348 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2349 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2350 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2351 VEC_PACK_FLOAT_EXPR): Document.
2353 2018-05-29 Richard Biener <rguenther@suse.de>
2355 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2357 (stmt_vec_info_vec): Make pointer.
2358 (init_stmt_vec_info_vec): Remove.
2359 (free_stmt_vec_info_vec): Likewise.
2360 (set_stmt_vec_info_vec): New function.
2361 (free_stmt_vec_infos): Likewise.
2362 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2363 (set_vinfo_for_stmt): Likewise.
2364 (get_earlier_stmt): Likewise.
2365 (get_later_stmt): Likewise.
2366 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2367 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2368 (vec_info::~vec_info): Free stmt_vec_infos.
2369 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2370 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2371 (pass_slp_vectorize::execute): Likewise.
2372 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2373 (free_stmt_vec_info_vec): Likewise.
2374 (set_stmt_vec_info_vec): New function.
2375 (free_stmt_vec_infos): Likewise.
2376 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2377 the global stmt_vec_info_vec.
2378 * tree-parloops.c (gather_scalar_reductions): Use
2379 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2382 2018-05-29 Richard Biener <rguenther@suse.de>
2384 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2386 2018-05-29 Martin Liska <mliska@suse.cz>
2387 David Malcolm <dmalcolm@redhat.com>
2389 * vec.c (test_reverse): New.
2390 (vec_c_tests): Add new test.
2391 * vec.h (vl_ptr>::reverse): New function.
2393 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2395 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2397 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2400 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2402 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2404 2018-05-28 Richard Biener <rguenther@suse.de>
2406 PR tree-optimization/85933
2407 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2408 look at stmts marked as vectorizable.
2410 2018-05-28 Richard Biener <rguenther@suse.de>
2412 PR tree-optimization/85934
2413 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2414 vector boolean check before scalar optimization.
2416 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2418 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2421 2018-05-28 Mark Wielaard <mark@klomp.org>
2423 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2424 if it is an expression containing a minus sign.
2426 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2428 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2430 2018-05-27 Paul Koning <ni1d@arrl.net>
2432 * config/pdp11/pdp11.md (truncsihi2): Remove.
2434 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2435 Chung-Ju Wu <jasonwucj@gmail.com>
2437 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2439 (unaligned_store_dw): Ditto.
2440 * config/nds32/nds32-memory-manipulation.c
2441 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2442 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2443 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2444 (emit_setmem_word_loop): Rename to ...
2445 (emit_setmem_doubleword_loop): ... this.
2446 (nds32_gen_dup_4_byte_to_word_value): New function.
2447 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2448 (nds32_expand_setmem_loop): Refine implementation.
2449 (nds32_expand_setmem_loop_v3m): Ditto.
2450 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2453 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2455 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2457 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2459 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2460 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2461 machine->attr_no_prologue_p.
2462 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2463 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2464 (nds32_expand_epilogue): Consider attr_naked_p.
2465 (nds32_expand_epilogue_v3pop): Likewise.
2466 (nds32_can_use_return_insn): Likewise.
2467 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2468 attr_no_prologue_p fields.
2469 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2471 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2474 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2476 * config/i386/sse.md
2477 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2479 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2481 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2483 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2485 (*<floatsuffix>floatv2div2sf2): Rename to ...
2486 (*float<floatunssuffix>v2div2sf2): ... this.
2487 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2488 (float<floatunssuffix>v2div2sf2_mask): ... this.
2489 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2490 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2491 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2493 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2495 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2497 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2499 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2501 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2503 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2504 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2505 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2506 gen_ufix_truncv8dfv8si2.
2507 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2508 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2509 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2510 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2511 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2512 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2513 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2514 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2516 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2520 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2522 2018-05-25 Jim Wilson <jimw@sifive.com>
2524 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2525 * config/riscv/riscv.c (struct machine_function): Add
2526 interrupt_handler_p and attribute_checked_p fields.
2527 (riscv_attribute_table): Add interrupt.
2528 (riscv_interrupt_type_p): New.
2529 (riscv_save_reg_p): Save extra regs for interrupt handler.
2530 (riscv_use_save_libcall): Return false for interrupt handler.
2531 (riscv_first_stack_step): Add forward declaration.
2532 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2533 for interrupt handler with large frame. Use it for saved reg list.
2534 (riscv_expand_prologue): Move flag_stack_usage_info support to
2535 eliminate duplication.
2536 (riscv_expand_epilogue): Generate mret for interrupt handler.
2537 (riscv_epilogue_uses): New.
2538 (riscv_can_use_return_insn): Return false for interrupt handler.
2539 (riscv_function_ok_for_sibcall): Likewise.
2540 (riscv_set_current_function): Add interrupt handler support.
2541 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2542 * config/riscv/riscv.md (UNSPECV_MRET): New.
2544 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2546 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2548 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2550 PR tree-optimization/85712
2551 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2552 this candidate has already been replaced in-situ by a copy.
2554 2018-05-25 Jason Merrill <jason@redhat.com>
2556 PR c++/80485 - inline function non-zero address.
2557 * symtab.c (nonzero_address): Check DECL_COMDAT.
2559 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2562 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2563 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2564 extension of SImode operation. Use const123_operand predicate.
2565 (*saddsi_1): Remove.
2566 (*saddl_se_1): Ditto.
2568 (*ssubl_se_1): Ditto.
2569 * config/alpha/predicates.md (const123_operand): New predicate.
2570 * config/alpha/constraints.md (P): Use IN_RANGE.
2572 2018-05-25 Richard Biener <rguenther@suse.de>
2574 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2576 (ref_maybe_used_by_stmt_p): Likewise.
2577 (stmt_may_clobber_ref_p): Likewise.
2578 (stmt_may_clobber_ref_p_1): Likewise.
2579 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2581 (ref_maybe_used_by_stmt_p): Likewise.
2582 (stmt_may_clobber_ref_p): Likewise.
2583 (stmt_may_clobber_ref_p_1): Likewise.
2584 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2585 the alias oracle to disambiguate DRs with stmts DR analysis
2587 (vect_analyze_data_refs): Do not give up on not analyzable
2588 DRs for BB vectorization. Remove code truncating the dataref
2591 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2594 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2595 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2596 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2598 2018-05-25 Richard Biener <rguenther@suse.de>
2600 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2601 function, combining stmt data ref gathering and fatal analysis
2603 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2604 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2605 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2606 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2607 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2609 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2611 PR tree-optimization/85720
2612 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2613 SCC if all partitions are builtins.
2614 (version_loop_by_alias_check): New parameter. Generate cancelable
2615 runtime alias check if all partitions are builtins.
2616 (distribute_loop): Update call to above function.
2618 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2620 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2621 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2622 (parm_default_def_partition_arg): Ditto.
2623 (set_parm_default_def_partition): Ditto.
2624 (get_parm_default_def_partitions): Ditto and make it static.
2625 (get_undefined_value_partitions): Ditto and make it static.
2626 (remove_ssa_form): Refactor call to init_var_map here.
2627 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2628 computation for loop region.
2629 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2630 (register_default_def): Delete.
2631 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2632 (parm_default_def_partition_arg): Ditto.
2633 (set_parm_default_def_partition): Ditto.
2634 (get_parm_default_def_partitions): Ditto and make it static.
2635 (get_undefined_value_partitions): Ditto and make it static.
2636 (coalesce_with_default, coalesce_with_default): Update comment.
2637 (create_coalesce_list_for_region): New func factored out from
2638 create_outofssa_var_map.
2639 (populate_coalesce_list_for_outofssa): New func factored out from
2640 create_outofssa_var_map and coalesce_ssa_name.
2641 (create_outofssa_var_map): Delete.
2642 (coalesce_ssa_name): Refactor to support live range computation.
2643 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2644 (get_parm_default_def_partitions): Delete.
2645 (get_undefined_value_partitions): Ditto.
2646 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2647 computation for loop region.
2648 (new_tree_live_info, loe_visit_block): Ditto.
2649 (live_worklist, set_var_live_on_entry): Ditto.
2650 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2651 * tree-ssa-live.h (struct _var_map): New fields.
2652 (init_var_map): Change decl.
2653 (region_contains_p): New.
2655 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2657 * tree-ssa-live.h (live_merge_and_clear): Delete.
2659 2018-05-25 Richard Biener <rguenther@suse.de>
2662 * tree-dump.c (dequeue_and_dump): Remove access to removed
2663 operand 2 of a SWITCH_EXPR.
2665 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2667 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2668 multiplication and division.
2669 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2670 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2671 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2672 (cond_udiv_optab, cond_umod_optab): New optabs.
2673 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2674 (IFN_COND_RDIV): New internal functions.
2675 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2676 TRUNC_MOD_EXPR and RDIV_EXPR.
2677 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2678 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2680 (SVE_INT_BINARY): Include mult.
2681 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2682 (optab, sve_int_op): Handle mult.
2683 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2685 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2686 for SVE_INT_BINARY_SD.
2688 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2690 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2691 (optab, sve_int_op): Handle div and udiv.
2692 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2693 for SVE_INT_BINARY_SD.
2694 (*<optab><mode>3): New insn for the same.
2696 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2698 * tree-vect-patterns.c: Include predict.h.
2699 (vect_recog_divmod_pattern): Restrict check for division support
2700 to when optimizing for size.
2702 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2704 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2705 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2706 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2707 (gimple_match_op::set_op): Likewise.
2708 (gimple_resimplify4): Declare.
2709 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2710 (expr::gen_transform): Likewise.
2711 (decision_tree::gen): Generate a simplification routine for 4 operands.
2712 * gimple-match-head.c (gimple_simplify): Add an overload for
2713 4 operands. In the top-level function, handle up to 4 call
2714 arguments and call gimple_resimplify4.
2715 (gimple_resimplify4): New function.
2716 (build_call_internal): Pass a fourth operand.
2717 (maybe_push_to_seq): Likewise.
2718 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2719 Fold VEC_COND_EXPRs of an operation and a default value into
2720 an IFN_COND_* function if possible.
2721 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2723 (SVE_COND_FP_BINARY): Include them.
2724 (optab, sve_fp_op): Handle them.
2725 (SVE_INT_BINARY_REV): New code iterator.
2726 (SVE_COND_FP_BINARY_REV): New int iterator.
2727 (commutative): New int attribute.
2728 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2730 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2732 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2733 (*cond_<optab><mode>): New patterns for reversed operands.
2735 2018-05-25 Richard Biener <rguenther@suse.de>
2737 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2738 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2739 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2740 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2741 * tree-vect-data-refs.c (everywhere): Adjust users.
2742 * tree-vect-loop.c (everywhere): Likewise.
2743 * tree-vect-slp.c (everywhere): Likewise.
2744 * tree-vect-stmts.c (everywhere): Likewise.
2745 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2747 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2749 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2751 (gcc_cv_as_section_exclude): ... this.
2752 Try Solaris as #exclude syntax.
2753 * configure: Regenerate.
2754 * config.in: Regenerate.
2755 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2757 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2758 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2760 * varasm.c (default_elf_asm_named_section): Don't check if
2761 HAVE_GAS_SECTION_EXCLUDE is defined.
2763 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2765 * doc/md.texi: Update the documentation of the cond_* optabs
2766 to mention the new final operand. Fix GET_MODE_NUNITS call.
2767 Describe the scalar case too.
2768 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2769 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2771 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2772 (get_conditional_internal_fn): Update comment.
2773 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2774 accumulator value as a final argument to conditional functions.
2775 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2776 a define_expand and add an "else" operand. Assert for now that
2777 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2778 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2779 (*cond_<optab><mode>): New patterns.
2780 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2781 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2782 (UNSPEC_COND_EOR): Delete.
2783 (optab): Remove associated mappings.
2784 (SVE_INT_BINARY): New code iterator.
2785 (sve_int_op): Remove int attribute and add "minus" to the code
2787 (SVE_COND_INT_OP): Delete.
2788 (SVE_COND_FP_OP): Rename to...
2789 (SVE_COND_FP_BINARY): ...this.
2791 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2793 * optabs.c (can_reuse_operands_p): New function.
2794 (maybe_legitimize_operands): Try to reuse the results for
2797 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2799 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2800 Add {q} suffix to insn mnemonic.
2802 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2804 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2805 (msp430_warn_func_return): New.
2807 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2809 * fold-const.c (tree_nonzero_bits): New function.
2810 * fold-const.h (tree_nonzero_bits): Likewise.
2811 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2812 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2814 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2818 * varasm.c (assemble_alias): Check ifunc_resolver only on
2821 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2824 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2825 when memory input operand is handled.
2827 2018-05-24 Luis Machado <luis.machado@linaro.org>
2829 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2831 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2833 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2835 * match.pd: Delay FMA folds until after vectorization.
2837 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2840 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2841 address check not strict.
2843 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2845 * gimple-match.h (gimple_match_op): New class.
2846 (mprts_hook): Replace parameters with a gimple_match_op *.
2847 (maybe_build_generic_op): Likewise.
2848 (gimple_simplified_result_is_gimple_val): Replace parameters with
2849 a const gimple_match_op *.
2850 (gimple_simplify): Replace code_helper * and tree * parameters with
2851 a gimple_match_op * parameter.
2852 (gimple_resimplify1): Replace code_helper *, tree and tree *
2853 parameters with a gimple_match_op * parameter.
2854 (gimple_resimplify2): Likewise.
2855 (gimple_resimplify3): Likewise.
2856 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2857 parameters with a gimple_match_op * parameter.
2858 * gimple-match-head.c (gimple_simplify): Change prototypes of
2859 auto-generated functions to take a gimple_match_op * instead of
2860 separate code_helper * and tree * parameters. Make the same
2861 change in the top-level overload and update calls to the
2862 gimple_resimplify routines. Update calls to the auto-generated
2863 functions and to maybe_push_res_to_seq in the publicly-facing
2864 operation-specific gimple_simplify overloads.
2865 (gimple_match_op::MAX_NUM_OPS): Define.
2866 (gimple_resimplify1): Replace rcode and ops with a single res_op
2867 parameter. Update call to gimple_simplify.
2868 (gimple_resimplify2): Likewise.
2869 (gimple_resimplify3): Likewise.
2870 (mprts_hook): Replace parameters with a gimple_match_op *.
2871 (maybe_build_generic_op): Likewise.
2872 (build_call_internal): Replace type, nargs and ops with
2873 a gimple_match_op *.
2874 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2875 with a single gimple_match_op *. Update calls to mprts_hook,
2876 build_call_internal and gimple_simplified_result_is_gimple_val.
2877 Factor out code that is common to the tree_code and combined_fn cases.
2878 * genmatch.c (expr::gen_transform): Replace tem_code and
2879 tem_ops with a gimple_match_op called tem_op. Update calls
2880 to the gimple_resimplify functions and maybe_push_res_to_seq.
2881 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2882 res_ops. Update call to the gimple_resimplify functions.
2883 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2884 (decision_tree::gen): Make the functions take a gimple_match_op *
2885 called res_op instead of separate res_code and res_ops parameters.
2886 Update call accordingly.
2887 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2888 and ops with a single res_op parameter. Update calls to
2889 maybe_build_generic_op and maybe_push_res_to_seq.
2890 (fold_stmt_1): Update calls to gimple_simplify and
2891 replace_stmt_with_simplification.
2892 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2893 and gimple_simplified_result_is_gimple_val.
2894 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2896 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2897 with a gimple_match_op *.
2898 (vn_nary_build_or_lookup): Likewise. Update call to
2899 vn_nary_build_or_lookup_1.
2900 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2901 gimple_match_op *. Update calls to the gimple_resimplify routines
2902 and to gimple_simplified_result_is_gimple_val.
2903 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2904 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2905 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2906 (visit_nary_op): Likewise.
2907 (visit_reference_op_load): Likewise.
2909 2018-05-23 Luis Machado <luis.machado@linaro.org>
2911 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2912 modifier for printing the step amount.
2914 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2917 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2920 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2922 * doc/sourcebuild.texi (Endianness): New subsubsection.
2924 2018-05-23 Luis Machado <luis.machado@linaro.org>
2926 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2927 <prefetch_dynamic_strides>: New const bool field.
2928 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2929 prefetch_dynamic_strides.
2930 (exynosm1_prefetch_tune): Likewise.
2931 (thunderxt88_prefetch_tune): Likewise.
2932 (thunderx_prefetch_tune): Likewise.
2933 (thunderx2t99_prefetch_tune): Likewise.
2934 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2936 (aarch64_override_options_internal): Update to set
2937 PARAM_PREFETCH_DYNAMIC_STRIDES.
2938 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2939 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2940 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2941 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2942 prefetch-dynamic-strides setting.
2944 2018-05-23 Luis Machado <luis.machado@linaro.org>
2946 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2947 <minimum_stride>: New const int field.
2948 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2949 minimum_stride field defaulting to -1.
2950 (exynosm1_prefetch_tune): Likewise.
2951 (thunderxt88_prefetch_tune): Likewise.
2952 (thunderx_prefetch_tune): Likewise.
2953 (thunderx2t99_prefetch_tune): Likewise.
2954 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2955 <default_opt_level>: Set to 3.
2956 (aarch64_override_options_internal): Update to set
2957 PARAM_PREFETCH_MINIMUM_STRIDE.
2958 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2959 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2960 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2961 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2962 stride is constant and is below the minimum stride threshold.
2964 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2966 * config/arm/arm-cpus.in (mode26): Delete.
2967 (armv4): Delete mode26 reference.
2968 * config/arm/arm.c (arm_configure_build_target): Delete use of
2971 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2973 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2975 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2976 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2978 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2979 (floatunsdidf2): Ditto.
2981 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2983 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2984 (fixuns_trunc<mode>si2_avx512f): Ditto.
2985 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2986 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2987 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2989 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2991 PR rtl-optimization/79985
2992 * df-scan.c (df_insn_refs_collect): Remove special case for
2993 global registers and asm statements.
2995 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2997 * extend.texi (Global Register Variables): Rewrite the bullet list.
2998 Note that the register is available for allocation. Note that access
2999 via inline asm must use constraints. Add note about async-signal
3000 handlers. Remove paragraph about automagic register selection.
3002 2018-05-23 Richard Biener <rguenther@suse.de>
3004 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3005 of fixed offset from memset VN.
3007 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3009 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3011 (alloc_cand_and_find_basis): Initialize first_interp field.
3012 (slsr_process_mul): Modify first_interp field.
3013 (slsr_process_add): Likewise.
3014 (slsr_process_cast): Modify first_interp field for each new
3016 (slsr_process_copy): Likewise.
3017 (dump_candidate): Dump first_interp field.
3018 (replace_mult_candidate): Process all interpretations, not just
3020 (replace_rhs_if_not_dup): Likewise.
3021 (replace_one_candidate): Likewise.
3023 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3025 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3027 (aarch64_needs_frame_chain): New function.
3028 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3030 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3033 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3034 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3035 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3036 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3037 as true for strict-align.
3038 (aarch64_can_inline_p): Perform checks even when callee has no
3039 attributes to check for strict alignment.
3040 * doc/extend.texi (AArch64 Function Attributes): Document
3042 * doc/invoke.texi: (AArch64 Options): Likewise.
3044 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3046 PR tree-optimization/85853
3047 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3048 the handling of the root of the node to...
3049 (vect_slp_analyze_node_operations_1): ...this new function,
3050 and run the whole thing with the child nodes' def types
3051 set according to their SLP node's def type.
3053 2018-05-23 Richard Biener <rguenther@suse.de>
3056 * tree-data-ref.c (create_runtime_alias_checks): Defer
3057 and ignore overflow warnings.
3059 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3061 PR tree-optimization/85822
3062 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3065 2018-05-23 Richard Biener <rguenther@suse.de>
3067 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3068 memset constants via native_interpret_expr.
3070 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3073 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3075 (cgraph_node::create_alias): Likewise.
3076 (cgraph_node::get_availability): Check ifunc_resolver instead
3077 of looking up ifunc attribute.
3078 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3079 * varasm.c (do_assemble_alias): Likewise.
3080 (assemble_alias): Likewise.
3081 (default_binds_local_p_3): Likewise.
3082 * cgraph.h (cgraph_node): Add ifunc_resolver.
3083 (cgraph_node::only_called_directly_or_aliased_p): Return false
3085 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3087 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3088 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3089 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3090 instead of looking up ifunc attribute.
3092 2018-05-22 Luis Machado <luis.machado@linaro.org>
3094 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3096 2018-05-22 Martin Sebor <msebor@redhat.com>
3099 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3100 only when expasion succeeds.
3101 (expand_builtin_strcmp): Same.
3102 (expand_builtin_strncmp): Same.
3104 2018-05-22 Martin Sebor <msebor@redhat.com>
3106 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3108 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3109 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3111 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3112 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3114 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3115 Add check for legitimate_address.
3116 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3117 (aarch64_swap_ldrstr_operands): New.
3118 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3121 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3122 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3124 * config/aarch64/aarch64.md: New patterns to generate stp
3126 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3127 single words and double words.
3128 (load_pair_sw, load_pair_dw): Likewise.
3129 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3131 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3133 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3134 for different mode ldpstp and add peephole for merged zero stores.
3136 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3138 (aarch64_gen_store_pair): Rename calls to match new patterns.
3139 (aarch64_gen_load_pair): Rename calls to match new patterns.
3140 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3141 (load_pair<DREG:mode><DREG2:mode>): ... This.
3142 (store_pair<mode>): Rename to...
3143 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3144 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3146 (V_INT_EQUIV): Handle SImode.
3147 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3150 2018-05-22 Martin Sebor <msebor@redhat.com>
3153 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3154 or ajust the presumed bound on an operation to avoid unnecessary
3157 2018-05-22 Martin Sebor <msebor@redhat.com>
3159 PR tree-optimization/85826
3160 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3161 assuming that a DECL necesarily has a constant size.
3163 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3166 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3168 2018-05-22 Richard Biener <rguenther@suse.de>
3170 PR tree-optimization/85834
3171 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3172 non-constant and non-zero memset arguments.
3174 2018-05-22 Martin Liska <mliska@suse.cz>
3177 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3179 2018-05-22 Richard Biener <rguenther@suse.de>
3181 PR tree-optimization/85863
3182 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3183 comparisons when vectype is specified.
3184 (vectorizable_condition): Do not specify vectype for
3185 vect_is_simple_cond when SLP vectorizing.
3187 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3190 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3191 define __ibm128 as long double.
3192 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3193 as a distinct type when IEEE 128-bit support is enabled.
3194 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3195 128-bit types to use the correct functions.
3196 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3197 convert between 128-bit floating point types that have different
3198 modes but the same representation, instead of using gen_lowpart to
3200 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3202 (IFKF_reg): New attributes to give the register constraints for
3204 (extend<mode>tf2_internal): New insns to mark an explicit
3205 conversion between 128-bit floating point types that have a
3206 different mode but share the same representation.
3208 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3210 PR tree-optimization/85814
3211 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3212 a null return from get_strinfo when unsharing the next
3213 strinfo in the chain.
3215 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3218 * varasm.c (weak_finish): Clean up weak_decls.
3220 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3222 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3223 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3224 UNSPEC_UADALP values.
3225 * config/aarch64/iterators.md (ABAL): New int iterator.
3228 (sur): Add mappings for the above.
3229 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3231 (aarch64_<sur>abal<mode>_4): Likewise.
3232 (aarch64_<sur>adalp<mode>_3): Likewise.
3233 (<sur>sadv16qi): New define_expand.
3235 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3237 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3238 (*movdf_internal): Ditto.
3239 (*rcpsf2_sse): Ditto.
3240 (*rsqrtsf2_sse): Ditto.
3241 (*sqrt<mode>2_sse): Ditto.
3243 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3245 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3247 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3248 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3249 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3250 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3252 * config/aarch64/arm_neon.h: Likewise.
3253 * config/aarch64/iterators.md (VQ_I): New.
3255 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3257 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3259 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3262 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3264 * config/nds32/constraints.md (S): New constraint.
3265 * config/nds32/nds32.md (call_internal): Use constraint S.
3266 (call_value_internal): Likewise.
3267 (sibcall_internal): Likewise.
3268 (sibcall_value_internal): Likewise.
3270 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3271 Chung-Ju Wu <jasonwucj@gmail.com>
3273 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3276 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3277 Chung-Ju Wu <jasonwucj@gmail.com>
3279 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3280 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3281 (nds32_rtx_costs_impl): Simplify.
3282 (nds32_address_cost_impl): Simplify.
3283 (nds32_init_rtx_costs): New function.
3284 (nds32_rtx_costs_speed_prefer): Likewise.
3285 (nds32_rtx_costs_size_prefer): Likewise.
3286 (nds32_address_cost_speed_prefer): Likewise.
3287 (nds32_address_cost_speed_fwprop): Likewise.
3288 (nds32_address_cost_size_prefer): Likewise.
3289 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3290 * config/nds32/nds32.c (nds32_option_override): Use
3291 nds32_init_rtx_costs function.
3293 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3295 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3296 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3297 (TARGET_PIPELINE_N8): Likewise.
3298 (TARGET_PIPELINE_N10): Likewise.
3299 (TARGET_PIPELINE_N13): Likewise.
3300 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3302 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3304 * config/nds32/nds32-fpu.md: Update copyright year.
3306 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3308 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3310 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3312 * config/nds32/nds32.c
3313 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3314 * config/nds32/nds32.opt (minline-asm-r15): New option.
3316 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3318 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3320 * config/nds32/nds32.md (abssi2): New pattern.
3322 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3324 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3325 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3326 Merge insn pattern from sse_cvtsi2ss<round_name> and
3327 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3328 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3329 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3330 using SWI48 mode iterator.
3331 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3332 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3333 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3334 pattern from sse_cvttss2si<round_saeonly_name>
3335 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3336 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3337 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3338 using SWI48 mode iterator.
3339 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3340 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3341 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3342 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3343 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3344 using SWI48 mode iterator.
3345 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3346 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3347 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3348 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3349 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3350 SWI48 mode iterator.
3351 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3352 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3353 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3354 pattern from sse_cvttsd2si<round_saeonly_name>
3355 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3357 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3359 * config/nds32/nds32-md-auxiliary.c
3360 (nds32_valid_smw_lwm_base_p): Refine.
3361 (nds32_output_smw_single_word): Refine.
3362 (nds32_output_smw_double_word): New.
3363 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3365 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3367 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3368 (nds32_output_stack_pop): Refine.
3369 (nds32_expand_unaligned_load): Refine.
3370 (nds32_expand_unaligned_store): Refine.
3372 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3373 Chung-Ju Wu <jasonwucj@gmail.com>
3375 * config/nds32/constants.md: Add TP_REGNUM constant.
3376 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3377 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3379 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3380 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3381 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3382 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3383 and PIC code generation.
3384 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3386 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3388 * config/nds32/nds32.md: Support TLS and PIC.
3389 * config/nds32/nds32.c: Support TLS and PIC.
3390 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3391 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3394 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3396 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3397 mode with E_ prefix.
3399 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3400 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3403 * config/nds32/nds32-md-auxiliary.c
3404 (symbolic_reference_mentioned_p): New.
3405 (nds32_legitimize_ict_address): New.
3406 (nds32_expand_ict_move): New.
3407 (nds32_indirect_call_referenced_p): New.
3408 (nds32_symbol_binds_local_p): Delete.
3409 (nds32_long_call_p): Modify.
3410 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3411 * config/nds32/nds32-protos.h
3412 (symbolic_reference_mentioned_p): Declare.
3413 (nds32_legitimize_ict_address): Declare.
3414 (nds32_expand_ict_move): Declare.
3415 (nds32_indirect_call_referenced_p): Declare.
3416 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3417 (nds32_relax_group): Use nds32_ict_const_p as condition.
3418 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3419 (nds32_asm_file_start): Output ict_model directive in asm code.
3420 (nds32_legitimate_address_p): Consider indirect call.
3421 (nds32_print_operand): Consider indirect call.
3422 (nds32_print_operand_address): Consider indirect call.
3423 (nds32_insert_attributes): Handle "indirect_call" attribute.
3424 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3425 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3426 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3427 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3428 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3429 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3430 (TARGET_ICT_MODEL_SMALL): Define.
3431 (TARGET_ICT_MODEL_LARGE): Define.
3432 * config/nds32/nds32.md (movsi): Consider ict model.
3433 (call, call_value): Consider ict model.
3434 (sibcall, sibcall_value): Consider ict model.
3435 * config/nds32/nds32.opt (mict-model): New option.
3436 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3439 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3440 Monk Chiang <sh.chiang04@gmail.com>
3441 Jim Wilson <jimw@sifive.com>
3443 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3444 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3445 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3446 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3447 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3448 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3449 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3450 compute save_libcall_adjustment properly.
3451 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3452 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3453 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3454 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3455 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3456 (ABI_SPEC): Handle mabi=ilp32e.
3457 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3458 (RVE): Add RVE mask.
3459 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3460 <-march>: Add rv32e as an example.
3462 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3465 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3466 (intra_create_variable_infos): Handle C++ constructors.
3468 2018-05-18 Martin Liska <mliska@suse.cz>
3470 * passes.def: Remove a redundant pass.
3472 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3475 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3477 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3479 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3481 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3483 (armv2, armv2a, armv3, armv3m): Delete architectures.
3484 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3485 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3486 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3488 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3489 (*mulsidi3adddi): Likewise.
3490 (mulsidi3): Likewise.
3491 (*mulsidi3_nov6): Likewise.
3492 (umulsidi3): Likewise.
3493 (umulsidi3_nov6): Likewise.
3494 (umaddsidi4): Likewise.
3495 (*umulsidi3adddi): Likewise.
3496 (smulsi3_highpart): Likewise.
3497 (*smulsi3_highpart_nov6): Likewise.
3498 (umulsi3_highpart): Likewise.
3499 (*umulsi3_highpart_nov6): Likewise.
3500 * config/arm/arm.h (arm_arch3m): Delete.
3501 * config/arm/arm.c (arm_arch3m): Delete.
3502 (arm_option_override_internal): Update armv3-related comment.
3503 (arm_configure_build_target): Delete use of isa_bit_mode32.
3504 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3505 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3506 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3508 (mulusa3): Likewise.
3509 * config/arm/arm-protos.h (arm_arch3m): Delete.
3510 * config/arm/arm-tables.opt: Regenerate.
3511 * config/arm/arm-tune.md: Likewise.
3512 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3513 deleted architectures.
3515 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3517 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3518 (armv5t, armv5te): New features.
3519 (ARMv5, ARMv5e): Delete fgroups.
3520 (ARMv5t, ARMv5te): Adjust for above changes.
3522 (armv5, armv5e): Delete arches.
3523 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3525 (*call_reg_arm): Likewise.
3526 (*call_value_reg_armv5): Likewise.
3527 (*call_value_reg_arm): Likewise.
3528 (*call_symbol): Likewise.
3529 (*call_value_symbol): Likewise.
3530 (*sibcall_insn): Likewise.
3531 (*sibcall_value_insn): Likewise.
3533 (prefetch): Likewise.
3534 (define_split and define_peephole2 dependent on arm_arch5):
3536 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3538 (TARGET_ARM_QBIT): Likewise.
3539 (TARGET_DSP_MULTIPLY): Likewise.
3540 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3541 (arm_arch5, arm_arch5e): Delete.
3542 (arm_arch5t, arm_arch5te): Declare.
3543 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3544 (arm_arch5t): Declare.
3545 (arm_option_reconfigure_globals): Update for the above.
3546 (arm_options_perform_arch_sanity_checks): Update comment, replace
3547 use of arm_arch5 with arm_arch5t.
3548 (use_return_insn): Likewise.
3549 (arm_emit_call_insn): Likewise.
3550 (output_return_instruction): Likewise.
3551 (arm_final_prescan_insn): Likewise.
3552 (arm_coproc_builtin_available): Likewise.
3553 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3554 arm_arch5e with arm_arch5t and arm_arch5te.
3555 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3556 (arm_arch5t, arm_arch5te): Declare.
3557 * config/arm/arm-tables.opt: Regenerate.
3558 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3559 * config/arm/t-multilib: Likewise.
3560 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3561 instead of arm_arch5.
3562 (*call_reg_thumb1): Likewise.
3563 (*call_value_reg_thumb1_v5): Likewise.
3564 (*call_value_reg_thumb1): Likewise.
3565 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3567 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3569 2018-05-18 Martin Liska <mliska@suse.cz>
3571 PR gcov-profile/84846
3572 * doc/gcov.texi: Document -t option of gcov tool.
3574 2018-05-18 Martin Liska <mliska@suse.cz>
3576 PR gcov-profile/84846
3577 * gcov.c (print_usage): Add new -t option.
3578 (process_args): Handle the option.
3579 (generate_results): Use stdout as output when requested by
3582 2018-05-18 Martin Liska <mliska@suse.cz>
3584 PR gcov-profile/84846
3585 * coverage.c (coverage_init): Write PWD to .gcno file.
3586 * doc/gcov.texi: Document how working directory is printed.
3587 * gcov-dump.c (dump_gcov_file): Print PWD.
3588 * gcov.c (output_intermediate_file): Likewise.
3589 (read_graph_file): Read PWD string.
3590 (output_lines): Print PWD.
3592 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3595 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3596 for retval and return false if all args to phi are zero.
3598 2018-05-18 Richard Biener <rguenther@suse.de>
3600 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3602 (evrp_dom_walker::before_dom_children): Call it.
3604 2018-05-18 Richard Biener <rguenther@suse.de>
3606 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3607 results when processing array refs with variable index.
3609 2018-05-18 Toon Moene <toon@moene.org>
3611 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3612 directly after that of -floop-interchange. Indicate that both
3613 options are enabled by default when specifying -O3.
3615 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3617 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3618 iterator. Delete separate integer-mode vec_set<mode> expander.
3619 (aarch64_simd_vec_setv2di): Delete.
3620 (vec_setv2di): Delete.
3621 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3622 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3623 the "w, r" alternative.
3625 2018-05-18 Martin Liska <mliska@suse.cz>
3627 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3628 * tree-pass.h (make_pass_lower_switch_O0): New function.
3629 * tree-switch-conversion.c (node_has_low_bound): Remove.
3630 (node_has_high_bound): Likewise.
3631 (node_is_bounded): Likewise.
3632 (class pass_lower_switch): Make it a template type and create
3634 (pass_lower_switch::execute): Add template argument.
3635 (make_pass_lower_switch): New function.
3636 (make_pass_lower_switch_O0): New function.
3637 (do_jump_if_equal): Remove.
3638 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3639 all the hard work to tree optimization passes.
3641 2018-05-18 Martin Liska <mliska@suse.cz>
3643 * dbgcnt.c (limit_low): Renamed from limit.
3644 (limit_high): New variable.
3645 (dbg_cnt_is_enabled): Check for upper limit.
3646 (dbg_cnt): Adjust dumping.
3647 (dbg_cnt_set_limit_by_index): Add new argument for high
3649 (dbg_cnt_set_limit_by_name): Likewise.
3650 (dbg_cnt_process_single_pair): Parse new format.
3651 (dbg_cnt_process_opt): Use strtok.
3652 (dbg_cnt_list_all_counters): Remove 'value' and add
3654 * doc/invoke.texi: Document changes.
3656 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3658 * doc/sourcebuild.texi (scalar_all_fma): Document.
3659 * tree.def (FMA_EXPR): Delete.
3660 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3661 * internal-fn.c (ternary_direct): New macro.
3662 (expand_ternary_optab_fn): Likewise.
3663 (direct_ternary_optab_supported_p): Likewise.
3664 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3665 * builtins.c (fold_builtin_fma): Delete.
3666 (fold_builtin_3): Don't call it.
3667 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3668 * expr.c (expand_expr_real_2): Likewise.
3669 * fold-const.c (operand_equal_p): Likewise.
3670 (fold_ternary_loc): Likewise.
3671 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3672 * gimple.c (DEFTREECODE): Likewise.
3673 * gimplify.c (gimplify_expr): Likewise.
3674 * optabs-tree.c (optab_for_tree_code): Likewise.
3675 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3676 * tree-eh.c (operation_could_trap_p): Likewise.
3677 (stmt_could_throw_1_p): Likewise.
3678 * tree-inline.c (estimate_operator_cost): Likewise.
3679 * tree-pretty-print.c (dump_generic_node): Likewise.
3680 (op_code_prio): Likewise.
3681 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3682 * tree-ssa-operands.c (get_expr_operands): Likewise.
3683 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3684 * fold-const-call.h (fold_fma): Delete.
3685 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3686 CFN_FNMA and CFN_FNMS.
3688 * genmatch.c (combined_fn): New enum.
3689 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3690 (commutative_op): New function.
3691 (commutate): Use it. Handle more than 2 operands.
3692 (dt_operand::gen_gimple_expr): Use commutative_op.
3693 (parser::parse_expr): Allow :c to be used with non-binary
3694 operators if the commutative operand is known.
3695 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3696 CFN_FMS, CFN_FNMA and CFN_FNMS.
3697 (backprop::process_assign_use): Remove FMA_EXPR handling.
3698 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3699 (gen_hsa_fma): New function.
3700 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3701 IFN_FNMA and IFN_FNMS.
3702 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3703 * gimple-fold.h (follow_all_ssa_edges): Declare.
3704 * gimple-fold.c (follow_all_ssa_edges): New function.
3705 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3706 gimple_build interface and use follow_all_ssa_edges to fold the result.
3707 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3708 instead of checking for optabs directly.
3709 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3710 rather than FMA_EXPRs.
3711 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3712 call to IFN_FMA instead of an FMA_EXPR.
3714 2018-05-17 Jim Wilson <jimw@sifive.com>
3716 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3717 sign extended promoted subreg, and the range does not have the sign bit
3718 set, then do a sign extend.
3720 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3721 test, check for sign extended subreg and/or constant operands, and
3722 do a sign extend in that case.
3724 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3726 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3727 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3729 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3730 Change logics_shift_reg to logics_shift_imm.
3731 (thunderx2t99_fp_loadpair_basic): Delete.
3732 (thunderx2t99_fp_storepair_basic): Delete.
3733 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3734 (thunderx2t99_asimd_polynomial): Delete.
3735 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3736 and neon_fp_mul_d_scalar_q.
3737 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3738 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3739 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3740 (thunderx2t99_asimd_lut): Add missing tbl types.
3741 (thunderx2t99_asimd_ext): Delete.
3742 (thunderx2t99_asimd_load1_1_mult): Delete.
3743 (thunderx2t99_asimd_load1_2_mult): Delete.
3744 (thunderx2t99_asimd_load1_ldp): New.
3745 (thunderx2t99_asimd_load1): New.
3746 (thunderx2t99_asimd_load2): Add missing *load2* types.
3747 (thunderx2t99_asimd_load3): New.
3748 (thunderx2t99_asimd_load4): New.
3749 (thunderx2t99_asimd_store1_1_mult): Delete.
3750 (thunderx2t99_asimd_store1_2_mult): Delete.
3751 (thunderx2t99_asimd_store2_mult): Delete.
3752 (thunderx2t99_asimd_store2_onelane): Delete.
3753 (thunderx2t99_asimd_store_stp): New.
3754 (thunderx2t99_asimd_store1): New.
3755 (thunderx2t99_asimd_store2): New.
3756 (thunderx2t99_asimd_store3): New.
3757 (thunderx2t99_asimd_store4): New.
3759 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3761 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3762 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3764 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3765 Segher Boessenkool <segher@kernel.crashing.org>
3768 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3771 2018-05-17 Richard Biener <rguenther@suse.de>
3773 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3774 for pruning loop and prune defs feeding only already visited PHIs.
3776 2018-05-17 Richard Biener <rguenther@suse.de>
3778 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3780 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3781 Richard Biener <rguenther@suse.de>
3783 PR tree-optimization/85793
3784 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3785 for VMAT_ELEMENTWISE.
3787 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3789 * internal-fn.h (lookup_internal_fn): Declare
3790 * internal-fn.c (lookup_internal_fn): New function.
3791 * gimple.c (gimple_build_call_from_tree): Handle calls to
3793 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3794 internal function names.
3795 * tree-pretty-print.c (dump_generic_node): Likewise.
3796 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3798 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3800 * gimple-fold.h (gimple_build): Make the function forms take
3801 combined_fn rather than built_in_function.
3802 (gimple_simplify): Likewise.
3803 * gimple-match-head.c (gimple_simplify): Likewise.
3804 * gimple-fold.c (gimple_build): Likewise.
3805 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3806 rather than gimple_build_call_internal.
3807 (get_initial_defs_for_reduction): Likewise.
3808 (vect_create_epilog_for_reduction): Likewise.
3809 (vectorizable_live_operation): Likewise.
3811 2018-05-17 Martin Liska <mliska@suse.cz>
3813 * gimple-ssa-sprintf.c (format_directive): Do not use
3814 space in between 'G_' and '('.
3816 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3819 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3820 even if the mask is not all ones.
3823 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3825 (ix86_gimple_fold_builtin): Likewise.
3828 * config/i386/i386.c: Include tree-vector-builder.h.
3829 (ix86_vector_shift_count): New function.
3830 (ix86_fold_builtin): Fold shift builtins by scalar count.
3831 (ix86_gimple_fold_builtin): Likewise.
3833 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3834 _mm512_setzero): New intrinsics.
3836 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3837 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3839 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3840 code generation for cases where splatting a value is not useful.
3841 * simplify-rtx.c (simplify_ternary_operation): Simplify
3842 vec_merge across a vec_duplicate and a paradoxical subreg forming
3843 a vector mode to a vec_concat.
3845 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3847 * config.gcc: Support "goldmont-plus".
3848 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3850 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3851 PROCESSOR_GOLDMONT_PLUS.
3852 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3853 (processor_target_table): Add "goldmont-plus".
3854 (PTA_GOLDMONT_PLUS): Define.
3855 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3856 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3857 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3858 (fold_builtin_cpu): Add "goldmont-plus".
3859 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3860 (ix86_option_override_internal): Add "goldmont-plus".
3861 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3862 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3863 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3864 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3866 2018-05-17 Richard Biener <rguenther@suse.de>
3868 PR tree-optimization/85757
3869 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3870 remove defs that only feed that PHI from further processing.
3872 2018-05-16 Jim Wilson <jimw@sifive.com>
3874 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3876 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3878 2018-05-16 Mark Wielaard <mark@klomp.org>
3880 * dwarf2out.c (count_index_strings): New function.
3881 (output_indirect_strings): Call count_index_strings and generate
3882 header for dwarf_version >= 5.
3884 2018-05-16 Mark Wielaard <mark@klomp.org>
3886 * dwarf2out.c (dwarf_FORM): New function.
3887 (set_indirect_string): Use dwarf_FORM.
3888 (reset_indirect_string): Likewise.
3889 (size_of_die): Likewise.
3890 (value_format): Likewise.
3891 (output_die): Likewise.
3892 (add_skeleton_AT_string): Likewise.
3893 (output_macinfo_op): Likewise.
3894 (index_string): Likewise.
3895 (output_index_string_offset): Likewise.
3896 (output_index_string): Likewise.
3897 (count_index_strings): Likewise.
3899 2018-05-16 Carl Love <cel@us.ibm.com>
3901 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3902 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3904 2018-05-16 Martin Jambor <mjambor@suse.cz>
3906 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3907 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3909 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3911 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3912 (fnma<mode>4): Likewise.
3913 (fms<mode>4): Likewise.
3914 (fnms<mode>4): Likewise.
3915 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3916 (aarch64_fnma<mode>4): Likewise.
3917 (aarch64_fms<mode>4): Likewise.
3918 (aarch64_fnms<mode>4): Likewise.
3919 (aarch64_fnmadd<mode>4): Likewise.
3921 2018-05-16 Jason Merrill <jason@redhat.com>
3923 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3925 2018-05-16 Richard Biener <rguenther@suse.de>
3927 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3928 (dump_stmt_cost): Declare.
3929 (add_stmt_cost): Dump cost we add.
3930 (add_stmt_costs): New function.
3931 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3933 (vect_analyze_stmt): Adjust prototype.
3934 (vectorizable_condition): Likewise.
3935 (vectorizable_live_operation): Likewise.
3936 (vectorizable_reduction): Likewise.
3937 (vectorizable_induction): Likewise.
3938 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3939 cost vector to pass to vectorizable_ and record afterwards.
3940 (vect_model_reduction_cost): Take cost vector argument and adjust.
3941 (vect_model_induction_cost): Likewise.
3942 (vectorizable_reduction): Likewise.
3943 (vectorizable_induction): Likewise.
3944 (vectorizable_live_operation): Likewise.
3945 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3946 SLP_TREE_NUMBER_OF_VEC_STMTS.
3947 (vect_analyze_slp_cost_1): Remove.
3948 (vect_analyze_slp_cost): Likewise.
3949 (vect_slp_analyze_node_operations): Take visited args and
3950 a target cost vector. Avoid processing already visited stmt sets.
3951 (vect_slp_analyze_operations): Use a local cost vector to gather
3952 costs and register those of non-discarded instances.
3953 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3954 (vect_schedule_slp_instance): Remove copying of
3955 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3957 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3958 adding cost. Record cost entry location.
3959 (vect_prologue_cost_for_slp_op): Function to compute cost of
3960 a constant or invariant generated for SLP vect in the prologue,
3961 split out from vect_analyze_slp_cost_1.
3962 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3963 (vect_model_promotion_demotion_cost): Likewise.
3964 (vect_model_store_cost): Likewise, make static.
3965 (vect_model_load_cost): Likewise.
3966 (vectorizable_bswap): Add cost vector arg and adjust.
3967 (vectorizable_call): Likewise.
3968 (vectorizable_simd_clone_call): Likewise.
3969 (vectorizable_conversion): Likewise.
3970 (vectorizable_assignment): Likewise.
3971 (vectorizable_shift): Likewise.
3972 (vectorizable_operation): Likewise.
3973 (vectorizable_store): Likewise.
3974 (vectorizable_load): Likewise.
3975 (vectorizable_condition): Likewise.
3976 (vectorizable_comparison): Likewise.
3977 (can_vectorize_live_stmts): Likewise.
3978 (vect_analyze_stmt): Likewise.
3979 (vect_transform_stmt): Adjust calls to vectorizable_*.
3980 * tree-vectorizer.c: Include gimple-pretty-print.h.
3981 (dump_stmt_cost): New function.
3983 2018-05-16 Richard Biener <rguenther@suse.de>
3985 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3986 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3987 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3988 (check_name): New callback.
3989 (dse_classify_store): Track cycles via a visited bitmap of PHI
3990 defs and simplify handling of in-loop and across loop dead stores
3991 and properly fail for loop-variant refs. Handle byte-tracking with
3992 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3995 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3997 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3998 (vect_get_mask_type_for_stmt): Likewise.
3999 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4001 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4002 to determine the statement's vector type and the vector type that
4003 should be used for calculating nunits. Deal with cases in which
4004 the type has to be deferred.
4005 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4006 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4007 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4008 (vect_determine_vf_for_stmt): New functions, split out from...
4009 (vect_determine_vectorization_factor): ...here.
4010 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4011 (vect_get_mask_type_for_stmt): New functions, split out from
4012 vect_determine_vectorization_factor.
4014 2018-05-16 Richard Biener <rguenther@suse.de>
4016 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4017 verify the [VEC_]COND_EXPR embedded comparison.
4019 2018-05-15 Martin Sebor <msebor@redhat.com>
4021 PR tree-optimization/85753
4022 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4023 RECORD_TYPE in addition to ARRAY_TYPE.
4025 2018-05-15 Martin Sebor <msebor@redhat.com>
4028 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4030 2018-05-15 Richard Biener <rguenther@suse.de>
4032 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4033 add by_clobber_p one. Change algorithm to collect all defs
4034 representing uses we need to walk and try reducing them to
4035 a single one before failing.
4036 (dse_dom_walker::dse_optimize_stmt): Adjust.
4038 2018-05-13 Mark Wielaard <mark@klomp.org>
4040 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4041 (size_of_loc_descr): Likewise.
4042 (output_loc_operands): Likewise.
4043 (output_loc_operands_raw): Likewise.
4044 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4045 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4046 (hash_loc_operands): Likewise.
4047 (compare_loc_operands): Likewise.
4049 2018-05-14 Mark Wielaard <mark@klomp.org>
4051 * dwarf2out.c (count_index_addrs): New function.
4052 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4054 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4056 PR tree-optimization/83648
4057 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4058 return value as malloc candidate.
4060 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4063 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4064 param as true in call to suggest_attribute.
4066 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4068 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4069 -mreadonly-in-sdata.
4071 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4073 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4075 (aarch64_crypto_aesd_fused): Likewise.
4077 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4079 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4080 (movsi_aarch64): Likewise.
4081 (load_pairsi): Likewise.
4082 (load_pairdi): Likewise.
4083 (store_pairsi): Likewise.
4084 (store_pairdi): Likewise.
4085 (load_pairsf): Likewise.
4086 (load_pairdf): Likewise.
4087 (store_pairsf): Likewise.
4088 (store_pairdf): Likewise.
4089 (zero_extend): Likewise.
4090 (trunc): Swap alternatives.
4091 (fcvt_target): Add '?' to prefer w over r.
4093 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4096 * config/i386/i386.md: Disallow non-commutative arithmetics in
4097 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4098 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4099 in the peephole2 before it.
4101 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4103 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4104 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4105 (ix86_handle_option): Handle -mcldemote.
4106 * config.gcc: New header.
4107 * config/i386/cldemoteintrin.h: New file.
4108 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4109 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4111 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4112 OPTION_MASK_ISA_CLDEMOTE.
4113 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4114 (ix86_valid_target_attribute_inner_p): Ditto.
4115 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4116 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4117 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4118 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4119 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4121 * config/i386/i386.opt: Add -mcldemote.
4122 * config/i386/x86intrin.h: New header.
4123 * doc/invoke.texi: Add -mcldemote.
4125 2018-05-14 Richard Biener <rguenther@suse.de>
4127 * doc/match-and-simplify.texi: Adjust :s documentation.
4129 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4131 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4132 intended memcpy size.
4133 (REORDER_45): Likewise.
4135 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4137 * sort.cc: New file.
4138 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4139 * vec.c (qsort_chk): Use gcc_qsort.
4140 * Makefile.in (OBJS-libcommon): Add sort.o.
4141 (build/sort.o): New target. Use it...
4142 (BUILD_RTL): ... here, and...
4143 (build/gencfn-macros): ... here, and...
4144 (build/genmatch): ... here.
4146 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4147 Chung-Ju Wu <jasonwucj@gmail.com>
4149 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4150 * config/nds32/nds32-graywolf.md: New file.
4151 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4152 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4154 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4155 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4156 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4157 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4158 * config/nds32/pipelines.md: Include n15 settings.
4160 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4161 Chung-Ju Wu <jasonwucj@gmail.com>
4163 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4164 * config/nds32/nds32-n13.md: New file.
4165 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4166 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4168 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4169 * config/nds32/nds32.md (pipeline_model): Add n13.
4170 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4171 * config/nds32/pipelines.md: Include n13 settings.
4173 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4174 Chung-Ju Wu <jasonwucj@gmail.com>
4176 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4177 * config/nds32/nds32-n10.md: New file.
4178 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4179 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4181 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4182 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4183 * config/nds32/nds32.md (pipeline_model): Add n10.
4184 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4185 * config/nds32/pipelines.md: Include n10 settings.
4187 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4188 Kito Cheng <kito.cheng@gmail.com>
4189 Chung-Ju Wu <jasonwucj@gmail.com>
4191 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4192 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4193 Add enum values for DSP extension instructions.
4194 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4196 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4197 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4199 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4200 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4201 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4202 * config/nds32/nds32-intrinsic.md: Likewise.
4203 * config/nds32/nds32_intrinsic.h: Likewise.
4204 * config/nds32/nds32-md-auxiliary.c: Likewise.
4205 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4206 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4207 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4208 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4209 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4210 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4211 TYPE_DMAC in switch statement.
4212 * config/nds32/nds32.c: New checking and implementation for DSP
4213 extension instructions.
4214 * config/nds32/nds32.h: Likewise.
4215 * config/nds32/nds32.md: Likewise.
4216 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4217 * config/nds32/predicates.md: Implement new predicates for DSP
4220 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4222 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4223 Reformat alternatives and attributes so it is easier to identify
4224 which constraints/attributes go with which instruction.
4225 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4226 (mov<mode>_softfloat32, FMOVE64): Likewise.
4227 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4228 (mov<mode>_softfloat64, FMOVE64): Likewise.
4230 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4232 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4234 (Basic PowerPC Built-in Functions): The new name of the
4235 subsection previously known as "PowerPC Built-in Functions".
4236 (Basic PowerPC Built-in Functions Available on all Configurations):
4238 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4239 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4240 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4241 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4243 2018-05-11 Martin Jambor <mjambor@suse.cz>
4246 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4249 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4252 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4254 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4256 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4257 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4258 (ix86_handle_option): Handle -mwaitpkg.
4259 * config.gcc: New header.
4260 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4261 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4262 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4264 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4265 OPTION_MASK_ISA_WAITPKG.
4266 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4267 (ix86_option_override_internal): Add PTA_WAITPKG.
4268 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4269 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4270 IX86_BUILTIN_TPAUSE.
4271 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4272 __builtin_ia32_umwait and __builtin_ia32_tpause.
4273 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4274 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4275 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4276 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4277 UNSPECV_TPAUSE): New.
4278 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4279 * config/i386/i386.opt: Add -mwaitpkg.
4280 * config/i386/waitpkgintrin.h: New file.
4281 * config/i386/x86intrin.h: New header.
4282 * doc/invoke.texi: Add -mwaitpkg.
4284 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4287 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4289 (cortex-m0): Use armv6s-m isa.
4290 (cortex-m0plus): Likewise.
4291 (cortex-m1): Likewise.
4292 (cortex-m0.small-multiply): Likewise.
4293 (cortex-m0plus.small-multiply): Likewise.
4294 (cortex-m1.small-multiply): Likewise.
4296 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4297 Jakub Jelinek <jakub@redhat.com>
4299 PR tree-optimization/85692
4300 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4301 source permute as well.
4303 2018-05-11 Martin Liska <mliska@suse.cz>
4306 * doc/extend.texi: Document LLVM style format for no_sanitize
4309 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4311 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4312 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4313 (mode_supports_vsx_dform_quad): Likewise.
4314 (mode_supports_vmx_dform): Move these functions to be next to the
4315 other mode_supports functions.
4316 (mode_supports_dq_form): Likewise.
4317 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4318 mode_supports_dq_form.
4319 (reg_offset_addressing_ok_p): Likewise.
4320 (offsettable_ok_by_alignment): Likewise.
4321 (rs6000_legitimate_offset_address_p): Likewise.
4322 (legitimate_lo_sum_address_p): Likewise.
4323 (rs6000_legitimize_address): Likewise.
4324 (rs6000_legitimize_reload_address): Likewise.
4325 (rs6000_secondary_reload_inner): Likewise.
4326 (rs6000_preferred_reload_class): Likewise.
4327 (rs6000_output_move_128bit): Likewise.
4329 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4331 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4332 Generate SImode target register for null target.
4333 <case IX86_BUILTIN_XGETBV>: Ditto.
4334 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4335 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4337 2018-05-10 Carl Love <cel@us.ibm.com>
4339 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4340 dcbtt and dcbtstt if operands[2] is 0.
4342 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4345 * config/i386/sse.md (usadv64qi): New expander.
4347 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4349 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4350 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4351 -maltivec=be support.
4352 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4353 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4354 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4355 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4356 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4357 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4358 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4359 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4360 altivec_vsumsws): Adjust.
4361 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4362 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4363 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4365 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4366 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4367 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4368 (altivec_lve<VI_char>x): Delete expand.
4369 (*altivec_lve<VI_char>x_internal): Rename to...
4370 (altivec_lve<VI_char>x): ... this.
4371 (altivec_lvxl_<mode>): Delete expand.
4372 (*altivec_lvxl_<mode>_internal): Rename to ...
4373 (altivec_lvxl_<mode>): ... this.
4374 (altivec_stvxl_<mode>): Delete expand.
4375 (*altivec_stvxl_<mode>_internal): Rename to ...
4376 (altivec_stvxl_<mode>): ... this.
4377 (altivec_stve<VI_char>x): Delete expand.
4378 (*altivec_stve<VI_char>x_internal): Rename to ...
4379 (altivec_stve<VI_char>x): ... this.
4380 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4381 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4382 reduc_plus_scal_<mode>): Adjust.
4383 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4385 (rs6000_cpu_cpp_builtins): Adjust.
4386 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4387 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4388 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4389 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4390 -maltivec=be support.
4391 (rs6000_split_vec_extract_var): Adjust.
4392 (rs6000_split_v4si_init): Adjust.
4393 (swap_selector_for_mode): Delete.
4394 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4395 altivec_expand_stvex_be): Delete.
4396 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4397 -maltivec=be support.
4398 (rs6000_gimple_fold_builtin): Ditto.
4399 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4401 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4402 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4403 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4404 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4405 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4406 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4407 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4408 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4409 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4410 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4411 anonymous split): Adjust.
4412 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4413 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4415 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4417 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4418 when --with-gxx-include-dir is also specified.
4419 * configure: Regenerate.
4421 2018-05-09 Jim Wilson <jimw@sifive.com>
4424 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4425 * config/riscv/t-withmultilib: New.
4426 * config/riscv/withmultilib.h: New.
4427 * doc/install.texi: Document RISC-V --with-multilib-list support.
4429 2018-05-09 Richard Biener <rguenther@suse.de>
4431 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4433 (vect_bb_vectorization_profitable_p): Adjust. Compute
4434 actual scalar cost using the cost vector and the add_stmt_cost
4437 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4439 PR rtl-optimization/85645
4440 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4441 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4443 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4445 PR rtl-optimization/85645
4446 * shrink-wrap.c (spread_components): Return a boolean saying if
4447 anything was changed.
4448 (try_shrink_wrapping_separate): Iterate spread_components until
4449 nothing changes anymore.
4451 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4453 PR rtl-optimization/85645
4454 * regrename.c (build_def_use): Also kill the chains that include the
4455 destination of a REG_CFA_REGISTER note.
4457 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4459 PR rtl-optimization/85645
4460 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4461 insn that has a REG_CFA_REGISTER note.
4463 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4465 * cfgexpand.c (expand_clobber): New function.
4466 (expand_gimple_stmt_1): Use it.
4467 * tree-vect-stmts.c (vect_clobber_variable): New function,
4469 (vectorizable_simd_clone_call): ...here.
4470 (vectorizable_store): Emit a clobber either side of an
4471 IFN_STORE_LANES sequence.
4472 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4474 2018-05-09 Tom de Vries <tom@codesourcery.com>
4477 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4478 (define_insn "trap_if_false"): Add exit after trap.
4480 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4482 PR rtl-optimization/85638
4483 * bb-reorder.c: Include common/common-target.h.
4484 (create_forwarder_block): New function extracted from...
4485 (fix_up_crossing_landing_pad): ...here. Rename into...
4486 (dw2_fix_up_crossing_landing_pad): ...this.
4487 (sjlj_fix_up_crossing_landing_pad): New function.
4488 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4489 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4490 from both partitions and exit the loop after one iteration.
4492 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4495 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4497 (Basic PowerPC Built-in Functions): The new name of the
4498 subsection previously known as "PowerPC Built-in Functions".
4499 (Basic PowerPC Built-in Functions Available on all Configurations):
4501 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4503 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4505 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4507 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4510 2018-05-08 Jim Wilson <jimw@sifive.com>
4512 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4513 (LD_EMUL_SUFFIX): New.
4514 (LINK_SPEC): Use it.
4516 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4518 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4520 (Basic PowerPC Built-in Functions): The new name of the
4521 subsection previously known as "PowerPC Built-in Functions".
4522 (Basic PowerPC Built-in Functions Available on all Configurations):
4524 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4526 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4528 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4530 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4533 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4536 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4537 after cmpelim optimization.
4539 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4541 * config.gcc: Support "goldmont".
4542 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4543 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4545 * config/i386/i386.c (m_GOLDMONT): Define.
4546 (processor_target_table): Add "goldmont".
4547 (PTA_GOLDMONT): Define.
4548 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4549 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4550 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4551 (fold_builtin_cpu): Add "goldmont".
4552 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4553 (ix86_option_override_internal): Add "goldmont".
4554 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4555 (processor_type): Add PROCESSOR_GOLDMONT.
4556 * config/i386/i386.md: Add CPU "glm".
4557 * config/i386/glm.md: New file.
4558 * config/i386/x86-tune.def: Add m_GOLDMONT.
4559 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4561 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4564 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4566 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4567 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4568 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4571 * config/i386/i386.c (ix86_fold_builtin): Handle
4572 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4575 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4576 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4578 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4581 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4582 (check_arch): Likewise.
4583 (check_fpu): Return the result rather than printing it.
4584 (end arch): Fix operator precedence.
4585 (end cpu): Likewise.
4586 (END): Print the result from check_fpu.
4588 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4589 Alan Hayward <alan.hayward@arm.com>
4590 David Sherwood <david.sherwood@arm.com>
4592 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4593 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4594 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4595 (*fcmuo<mode>_and): New patterns.
4597 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4599 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4600 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4601 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4602 (cmp_op, sve_imm_con): New code attributes.
4603 (SVE_COND_INT_CMP, imm_con): Delete.
4604 (cmp_op): Remove above unspecs from int attribute.
4605 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4607 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4608 comparison-specific unspecs.
4609 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4610 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4611 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4612 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4613 (*vec_fcm<cmp_op><mode>): Rename to...
4614 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4615 (*vec_fcmuo<mode>): Rename to...
4616 (*fcmuo<mode>): ...this and adjust likewise.
4617 (*pred_fcm<cmp_op><mode>): New pattern.
4618 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4619 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4621 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4623 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4624 (aarch64_emit_sve_predicated_cond): New function.
4625 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4626 (aarch64_emit_unspec_cond_or): Replace with...
4627 (aarch64_emit_sve_or_conds): ...this new function. Use
4628 aarch64_emit_sve_ptrue_op for the individual comparisons and
4629 aarch64_emit_binop to OR them together.
4630 (aarch64_emit_inverted_unspec_cond): Replace with...
4631 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4632 aarch64_emit_sve_ptrue_op for the comparison and
4633 aarch64_emit_unop to invert the result.
4634 (aarch64_expand_sve_vec_cmp_float): Update after the above
4635 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4637 2018-05-07 Nathan Sidwell <nathan@acm.org>
4639 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4640 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4641 (Backwards Compatibility): Likewise.
4643 2018-05-07 Luis Machado <luis.machado@linaro.org>
4647 2018-05-07 Luis Machado <luis.machado@linaro.org>
4649 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4650 <prefetch_dynamic_strides>: New const bool field.
4651 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4652 prefetch_dynamic_strides.
4653 (exynosm1_prefetch_tune): Likewise.
4654 (thunderxt88_prefetch_tune): Likewise.
4655 (thunderx_prefetch_tune): Likewise.
4656 (thunderx2t99_prefetch_tune): Likewise.
4657 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4659 (aarch64_override_options_internal): Update to set
4660 PARAM_PREFETCH_DYNAMIC_STRIDES.
4661 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4662 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4663 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4664 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4665 prefetch-dynamic-strides setting.
4667 2018-05-07 Luis Machado <luis.machado@linaro.org>
4669 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4670 <minimum_stride>: New const int field.
4671 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4672 minimum_stride field.
4673 (exynosm1_prefetch_tune): Likewise.
4674 (thunderxt88_prefetch_tune): Likewise.
4675 (thunderx_prefetch_tune): Likewise.
4676 (thunderx2t99_prefetch_tune): Likewise.
4677 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4678 (aarch64_override_options_internal): Update to set
4679 PARAM_PREFETCH_MINIMUM_STRIDE.
4680 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4681 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4682 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4683 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4684 stride is constant and is below the minimum stride threshold.
4686 2018-05-07 Luis Machado <luis.machado@linaro.org>
4688 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4691 2018-05-07 Luis Machado <luis.machado@linaro.org>
4693 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4694 <prefetch_dynamic_strides>: New const bool field.
4695 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4696 prefetch_dynamic_strides.
4697 (exynosm1_prefetch_tune): Likewise.
4698 (thunderxt88_prefetch_tune): Likewise.
4699 (thunderx_prefetch_tune): Likewise.
4700 (thunderx2t99_prefetch_tune): Likewise.
4701 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4703 (aarch64_override_options_internal): Update to set
4704 PARAM_PREFETCH_DYNAMIC_STRIDES.
4705 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4706 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4707 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4708 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4709 prefetch-dynamic-strides setting.
4711 2018-05-07 Luis Machado <luis.machado@linaro.org>
4713 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4714 <minimum_stride>: New const int field.
4715 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4716 minimum_stride field.
4717 (exynosm1_prefetch_tune): Likewise.
4718 (thunderxt88_prefetch_tune): Likewise.
4719 (thunderx_prefetch_tune): Likewise.
4720 (thunderx2t99_prefetch_tune): Likewise.
4721 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4722 (aarch64_override_options_internal): Update to set
4723 PARAM_PREFETCH_MINIMUM_STRIDE.
4724 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4725 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4726 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4727 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4728 stride is constant and is below the minimum stride threshold.
4730 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4733 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4734 the type is addressable. Don't force op into register if it has
4737 2018-05-05 Roland McGrath <mcgrathr@google.com>
4740 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4741 any section for which we don't know a specific type it should have,
4742 regardless of name. Previously this was done only for the exact
4743 names ".init_array", ".fini_array", and ".preinit_array".
4744 (default_elf_asm_named_section): Add comment about
4745 relationship with default_section_type_flags and SECTION_NOTYPE.
4746 (get_section): Don't consider it a type conflict if one side has
4747 SECTION_NOTYPE and the other doesn't, as long as neither has the
4748 SECTION_BSS et al used in the default_section_type_flags logic.
4750 2018-05-05 Tom de Vries <tom@codesourcery.com>
4753 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4754 (workaround_barsyncs): New function.
4755 (nvptx_reorg): Use workaround_barsyncs.
4756 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4757 (define_expand "nvptx_membar_cta"): New define_expand.
4758 (define_insn "*nvptx_membar_cta"): New insn.
4760 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4762 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4763 To improve optimization opportunities.
4764 * builtin-types.def: The new needed builtin types for the above.
4766 2018-05-04 Richard Biener <rguenther@suse.de>
4768 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4769 * gimple-ssa-store-merging.c
4770 (imm_store_chain_info::output_merged_store): Remove redundant create,
4771 release split_store vector contents on failure.
4772 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4773 scalar stmt vector on cache hit.
4775 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4777 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4779 * config.gcc (powerpc-xilinx-eabi*): Remove.
4780 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4782 (fusion_addis_mem_combo_load): Ditto.
4783 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4785 (rs6000_cpu_cpp_builtins): Ditto.
4786 * config/rs6000/rs6000-linux.c
4787 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4788 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4789 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4791 (rs6000_setup_reg_addr_masks): Ditto.
4792 (rs6000_init_hard_regno_mode_ok): Ditto.
4793 (rs6000_option_override_internal): Ditto.
4794 (legitimate_lo_sum_address_p): Ditto.
4795 (rs6000_legitimize_address): Ditto.
4796 (rs6000_legitimize_reload_address): Ditto.
4797 (rs6000_legitimate_address_p): Ditto.
4798 (abi_v4_pass_in_fpr): Ditto.
4799 (setup_incoming_varargs): Ditto.
4800 (rs6000_gimplify_va_arg): Ditto.
4801 (rs6000_split_multireg_move): Ditto.
4802 (rs6000_savres_strategy): Ditto.
4803 (rs6000_emit_prologue_components): Ditto.
4804 (rs6000_emit_epilogue_components): Ditto.
4805 (rs6000_emit_prologue): Ditto.
4806 (rs6000_emit_epilogue): Ditto.
4807 (rs6000_elf_file_end): Ditto.
4808 (rs6000_function_value): Ditto.
4809 (rs6000_libcall_value): Ditto.
4810 * config/rs6000/rs6000.h: Ditto.
4811 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4812 (TARGET_MINMAX): ... this. New.
4813 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4814 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4815 (*movsi_internal1_single): Delete.
4816 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4817 mfpu=, mxilinx-fpu): Delete.
4818 * config/rs6000/singlefp.h: Delete.
4819 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4820 * config/rs6000/t-rs6000: Ditto.
4821 * config/rs6000/t-xilinx: Delete.
4822 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4823 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4824 (VStype_simple): Delete.
4825 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4826 * config/rs6000/xfpu.h: Delete.
4827 * config/rs6000/xfpu.md: Delete.
4828 * config/rs6000/xilinx.h: Delete.
4829 * config/rs6000/xilinx.opt: Delete.
4830 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4831 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4833 2018-05-04 Tom de Vries <tom@codesourcery.com>
4836 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4839 2018-05-04 Richard Biener <rguenther@suse.de>
4842 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4843 (expand_complex_div_wide): Likewise.
4844 (expand_complex_operations_1): Likewise.
4845 (expand_complex_libcall): Preserve EH info of the original stmt.
4846 (tree_lower_complex): Handle removed blocks.
4847 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4848 on complex multiplication and division libcall builtins.
4850 2018-05-04 Richard Biener <rguenther@suse.de>
4853 * fold-const.c (negate_expr_p): Restrict negation of operand
4854 zero of a division to when we know that can happen without
4856 (fold_negate_expr_1): Likewise.
4858 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4861 * real.h (real_nextafter): Declare.
4862 * real.c (real_nextafter): New function.
4863 * fold-const-call.c (fold_const_nextafter): New function.
4864 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4865 CASE_CFN_NEXTTOWARD.
4866 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4867 even when arg1_mode is different from arg0_mode.
4869 2018-05-03 Nathan Sidwell <nathan@acm.org>
4871 * doc/extend.texi (Deprecated Features): Remove
4873 (Backwards Compatibility): Likewise.
4874 * doc/invoke.texi (C++ Language Options): Likewise.
4875 (C++ Dialect Options): Likewise.
4877 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4880 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4881 _mm512_mask_mullox_epi64): New intrinsics.
4883 2018-05-03 Tom de Vries <tom@codesourcery.com>
4886 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4887 dump files): Add offload-tree.
4889 2018-05-03 Richard Biener <rguenther@suse.de>
4891 PR tree-optimization/85615
4892 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4893 to loops not nested in BBs loop father to avoid creating multi-entry
4896 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4898 PR tree-optimization/70291
4899 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4900 arguments. Change return type to tree. Emit libcall as a new
4901 statement rather than replacing existing one when inplace_p is true.
4902 (expand_complex_multiplication_components): New function.
4903 (expand_complex_multiplication): Expand floating-point complex
4904 multiplication using the above.
4905 (expand_complex_division): Rename inner_type parameter to type.
4906 Update expand_complex_libcall call-site.
4907 (expand_complex_operations_1): Update expand_complex_multiplication
4908 and expand_complex_division call-sites.
4910 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4913 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4914 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4915 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4916 the highest significant bit of the shift count mask is clear. In
4917 check whether and[sq]i3 is needed verify that all significant bits
4918 of the shift count other than the highest are set.
4920 2018-05-02 Tom de Vries <tom@codesourcery.com>
4923 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4924 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4925 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4926 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4927 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4928 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4929 __builtin_goacc_parlevel_size.
4931 2018-05-02 Richard Biener <rguenther@suse.de>
4933 PR tree-optimization/85597
4934 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4935 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4938 2018-05-02 Tom de Vries <tom@codesourcery.com>
4941 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4942 dump files): Add ltrans-tree.
4944 2018-05-02 Tom de Vries <tom@codesourcery.com>
4947 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4948 dump files): Add wpa-ipa.
4950 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4952 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4953 powerpc*-*-linux*paired* target.
4954 * config/rs6000/750cl.h: Delete.
4955 * config/rs6000/paired.h: Delete.
4956 * config/rs6000/paired.md: Delete.
4957 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4959 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4960 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4961 comment. Remove paired float support.
4962 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4963 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4965 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4966 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4968 * config/rs6000/rs6000.c: Remove paired float support.
4969 (paired_expand_vector_init, paired_expand_vector_move,
4970 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4971 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4972 paired_expand_builtin, paired_expand_predicate_builtin,
4973 paired_init_builtins): Delete.
4974 * config/rs6000/rs6000.h: Remove paired float support.
4975 * config/rs6000/rs6000.md: Remove paired float support.
4976 (move_from_CR_ov_bit): Delete.
4977 * config/rs6000/rs6000.opt (mpaired): Delete.
4978 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4979 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4981 2018-05-02 Richard Biener <rguenther@suse.de>
4984 * gimplify.c (gimplify_save_expr): When in SSA form allow
4985 SAVE_EXPRs to compute to SSA vars.
4987 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4990 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4991 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4992 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4993 clobber operands[2], instead use a new pseudo. Formatting fixes.
4995 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4997 PR tree-optimization/85586
4998 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4999 exit early for statements in the same group if the accesses are
5002 2018-05-02 Tom de Vries <tom@codesourcery.com>
5005 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5008 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5010 PR tree-optimization/85143
5011 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5013 2018-05-01 Tom de Vries <tom@codesourcery.com>
5016 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5017 not found" error message.
5019 2018-05-01 Tom de Vries <tom@codesourcery.com>
5022 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5023 * vec.c (test_ordered_remove_if): New function.
5024 (vec_c_tests): Call test_ordered_remove_if.
5025 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5026 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5027 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5028 VEC_ORDERED_REMOVE_IF.
5030 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5032 PR tree-optimization/82665
5033 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5034 pointer subtraction where arguments come from a memchr call.
5036 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5038 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5039 --push-state --as-needed and --pop-state instead of --as-needed and
5040 --no-as-needed if ld supports it.
5041 * configure: Regenerated.
5044 * doc/install.texi2html: Replace _002d with - and _002a with * in
5045 generated html files using sed.
5047 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5050 * gcc-rich-location.c (blank_line_before_p): New function.
5051 (use_new_line): New function.
5052 (gcc_rich_location::add_fixit_insert_formatted): New function.
5053 * gcc-rich-location.h
5054 (gcc_rich_location::add_fixit_insert_formatted): New function.
5056 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5058 * selftest.c (assert_streq): Rename "expected" and "actual" to
5059 "val1" and "val2". Extend NULL-handling to cover both inputs
5060 symmetrically, while still requiring both to be non-NULL for a pass.
5061 * selftest.h (assert_streq): Rename "expected" and "actual" to
5063 (ASSERT_EQ): Likewise.
5064 (ASSERT_EQ_AT): Likewise.
5065 (ASSERT_KNOWN_EQ): Likewise.
5066 (ASSERT_KNOWN_EQ_AT): Likewise.
5067 (ASSERT_NE): Likewise.
5068 (ASSERT_MAYBE_NE): Likewise.
5069 (ASSERT_MAYBE_NE_AT): Likewise.
5070 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5071 the assertion to pass.
5072 (ASSERT_STREQ_AT): Likewise.
5074 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5076 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5077 interaction with -pie.
5079 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5081 * selftest.h: Fix alphabetization of per-source-file selftest
5084 2018-04-30 Jason Merrill <jason@redhat.com>
5086 PR c++/61982 - dead stores to destroyed objects.
5087 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5090 2018-04-30 Jason Merrill <jason@redhat.com>
5092 * tree.c (build_clobber): New.
5093 * tree.h: Declare it.
5094 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5096 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5098 * diagnostic-show-locus.c (layout::layout): Update for
5099 location_get_source_line returning a char_span.
5100 (struct char_span): Move to input.h.
5101 (struct correction): Update for fields in char_span becoming
5103 (struct source_line): Update for location_get_source_line
5104 returning a char_span.
5105 (layout::print_line): Likewise.
5106 * edit-context.c (edited_file::print_content): Likewise.
5107 (edited_file::print_diff_hunk): Likewise.
5108 (edited_file::print_run_of_changed_lines): Likewise.
5109 (edited_file::get_num_lines): Likewise.
5110 (edited_line::edited_line): Likewise.
5111 * final.c (asm_show_source): Likewise.
5112 * input.c (location_get_source_line): Convert return type
5113 from const char * to char_span, losing the final "line_len"
5115 (dump_location_info): Update for the above.
5116 (get_substring_ranges_for_loc): Likewise. Use a char_span
5117 when handling the literal within the line.
5118 (test_reading_source_line): Update for location_get_source_line
5119 returning a char_span.
5120 * input.h (class char_span): Move here from
5121 diagnostic-show-locus.c, converting from a struct to a class.
5122 Make data members private.
5123 (char_span::operator bool): New.
5124 (char_span::length): New.
5125 (char_span::get_buffer): New.
5126 (char_span::operator[]): New.
5127 (char_span::subspan): Make const.
5128 (char_span::xstrdup): New.
5129 (location_get_source_line): Convert return type from const char *
5130 to char_span, losing the final "line_size" param.
5132 2018-04-30 Jan Hubicka <jh@suse.cz>
5134 * lto-wrapper.c (ltrans_priorities): New static var.
5135 (cmp_priority): New.
5136 (run_gcc): Read priorities and if doing parallel build order
5137 the Makefile by them.
5139 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5141 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5143 2018-04-30 Richard Biener <rguenther@suse.de>
5145 * tree-cfg.c (verify_address): Remove base argument, add
5146 flag whether to check TREE_ADDRESSABLE and do that.
5147 (verify_expr): Remove.
5148 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5149 (verify_gimple_assign_single): Likewise.
5150 (verify_gimple_switch): Likewise.
5151 (verify_expr_location_1): Dereference tp once. Add (disabled)
5152 piece from verify_expr.
5153 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5155 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5157 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5159 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5161 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5162 (small_data_pattern): Likewise.
5163 (arc_rewrite_small_data): Likewise.
5164 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5165 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5166 (get_symbol_alignment): New function.
5167 (legitimate_small_data_address_p): Likewise.
5168 (legitimate_scaled_address): Update, call
5169 legitimate_small_data_address_p.
5170 (output_sdata): New static variable.
5171 (arc_print_operand): Update how we handle small data operands.
5172 (arc_print_operand_address): Likewise.
5173 (arc_legitimate_address_p): Update, use
5174 legitimate_small_data_address_p.
5175 (arc_rewrite_small_data_p): Remove.
5176 (arc_rewrite_small_data_1): Likewise.
5177 (arc_rewrite_small_data): Likewise.
5178 (small_data_pattern): Likewise.
5179 (compact_sda_memory_operand): Update to use
5180 legitimate_small_data_address_p and get_symbol_alignment.
5181 (prepare_move_operands): Don't rewite sdata pattern.
5182 (prepare_extend_operands): Remove.
5183 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5185 (zero_extendqisi2): Likewise.
5186 (zero_extendhisi2): Likewise.
5187 (extendqihi2): Likewise.
5188 (extendqisi2): Likewise.
5189 (extendhisi2): Likewise.
5193 * config/arc/constraints.md (Usd): Change it to memory constraint.
5195 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5197 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5198 as source of std instructions.
5199 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5200 allow 6-bit constants as source for store instructions.
5201 (movdi_insn): Update instruction pattern to allow 6-bit constants
5202 as source for store instructions.
5204 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5206 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5208 2018-04-30 Nathan Sidwell <nathan@acm.org>
5209 Sandra Loosemore <sandra@codesourcery.com>
5211 * dumpfile.c (dump_open): Allow '-' for stdout.
5212 * doc/invoke.texi (Developer Options): Document dump filename
5213 determination early. Document stdin/stdout selection.
5215 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5217 Microblaze Target: PIC data text relative
5219 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5220 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5222 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5223 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5224 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5225 New addressing mode for data-text relative position indepenedent code.
5226 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5227 'ADDRESS_SYMBOLIC_TXT_REL'.
5228 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5229 (microblaze_legitimate_pic_operand): Exclude function calls from
5230 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5231 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5233 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5234 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5235 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5236 for 'address + offset'.
5237 (microblaze_expand_prologue): Add new function prologue call for
5239 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5240 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5241 table in case of TARGET_PIC_DATA_TEXT_REL.
5242 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5243 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5244 Add new macros 'UNSPEC_TEXT',
5245 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5246 + exclude function calls from 'UNSPEC_PLT' in case of data text
5248 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5249 new target hook for generating address diff vector tables in case of
5251 * doc/tm.texi : Regenerate.
5252 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5253 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5254 of addr diff vector generation.
5255 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5256 target hook definition.
5257 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5258 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5259 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5260 option for data text relative.
5262 2018-04-30 Richard Biener <rguenther@suse.de>
5264 * tree-chrec.h (evolution_function_is_constant_p): Remove
5266 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5268 2018-04-30 Richard Biener <rguenther@suse.de>
5271 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5273 2018-04-30 Richard Biener <rguenther@suse.de>
5275 PR tree-optimization/28364
5276 PR tree-optimization/85275
5277 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5278 copying first exit test.
5280 2018-04-28 Mark Wielaard <mark@klomp.org>
5282 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5284 (dwarf_AT): Handle DW_AT_addr_base.
5285 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5287 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5290 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5291 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5292 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5293 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5295 2018-04-28 Richard Biener <rguenther@suse.de>
5297 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5298 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5299 to reflect use. Only add interesting stmts.
5301 2018-04-27 Martin Jambor <mjambor@suse.cz>
5304 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5305 the jump function allows for passing through aggregate values.
5307 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5309 * input.h (in_system_header_at): Convert from macro to inline
5311 (from_macro_expansion_at): Likewise.
5312 (from_macro_definition_at): Likewise.
5314 2018-04-27 Jeff Law <law@redhat.com>
5316 * config.gcc: Mark tile* targets as deprecated/obsolete.
5318 2018-04-27 Richard Biener <rguenther@suse.de>
5320 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5323 2018-04-27 Richard Biener <rguenther@suse.de>
5325 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5327 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5329 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5330 with Yd constraint. Set "preferred_for_speed" attribute from
5331 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5333 (*movdi_internal): Ditto.
5334 (movti_interunit splitters): Remove
5335 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5336 (movdi_interunit splitters): Ditto.
5337 * config/i386/constraints.md (Ye): Remove.
5338 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5340 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5343 * config/aarch64/constraints.md (Usg): Limit to 31.
5346 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5348 PR tree-optimization/85529
5349 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5350 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5351 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5352 zero extension or masking of the MSB bit.
5353 (optimize_range_tests): Add FIRST_BB argument, pass it through
5354 to optimize_range_tests_var_bound.
5355 (maybe_optimize_range_tests, reassociate_bb): Adjust
5356 optimize_range_tests callers.
5358 2018-04-26 Richard Biener <rguenther@suse.de>
5359 Jakub Jelinek <jakub@redhat.com>
5361 * cgraph.h (symbol_table): Just declare debug method here.
5362 * symtab.c (symbol_table::debug): Define.
5364 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5366 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5368 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5370 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5371 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5372 (*movdi_internal): Substitute Yi and Yj constraint with x
5373 and Ym and Yn constraint with y constraint. Update "isa"
5374 attribute and set "preferred_for_speed" attribute from
5375 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5376 (*movsi_internal): Ditto.
5377 (*movdf_internal): Ditto.
5378 (*movsf_internal): Ditto.
5379 (*zero_extendsidi2): Ditto.
5380 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5381 (sse2_loadld): Ditto.
5382 (*vec_extract<ssevecmodelower>_0): Ditto.
5383 (*vec_extractv4si_0_zext_sse4): Ditto.
5384 (vec_concatv2di): Ditto.
5385 (*vec_dup<mode>): Ditto.
5386 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5387 * config/i386/constraints.md (Yi): Remove.
5392 2018-04-26 Nathan Sidwell <nathan@acm.org>
5394 * dumpfile.c (dump_open): New.
5395 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5396 (dump_finish): Detect stdio/stderr by value not name.
5398 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5400 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5402 2018-04-26 Tom de Vries <tom@codesourcery.com>
5405 * config/nvptx/nvptx.c (verify_neutering_jumps)
5406 (verify_neutering_labels): New function
5407 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5409 2018-04-26 Tom de Vries <tom@codesourcery.com>
5412 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5413 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5416 2018-04-26 Richard Biener <rguenther@suse.de>
5417 Tom de Vries <tom@codesourcery.com>
5420 * lto-streamer-out.c (output_function): Fixup loops if required to match
5421 discovery done in the reader.
5423 2018-04-26 Richard Biener <rguenther@suse.de>
5425 PR tree-optimization/85116
5426 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5427 have a loop exit from the single latch predecessor. Remove
5428 case of header with just condition.
5429 (ch_base::copy_headers): Exclude infinite loops from any
5431 (pass_ch::execute): Record exits.
5433 2018-04-26 Richard Biener <rguenther@suse.de>
5435 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5436 prologue cost vector and pass it to vect_get_load_cost.
5437 (vect_get_peeling_costs_all_drs): Likewise.
5438 (vect_peeling_hash_get_lowest_cost): Likewise.
5439 (vect_enhance_data_refs_alignment): Likewise.
5441 2018-04-26 Richard Biener <rguenther@suse.de>
5444 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5445 checking of integer<->pointer conversions.
5446 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5447 sign-/zero-extending pointer types.
5448 (expand_omp_for_static_chunk): Likewise.
5450 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5451 Jean Lee <xiaoyur347@gmail.com>
5453 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5454 (TARGET_ASAN_SHADOW_OFFSET): Define.
5455 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5456 true for -fsanitize=address.
5458 2018-04-25 Mark Wielaard <mark@klomp.org>
5460 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5463 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5465 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5466 than "alu", remove explicit "memory" and "imm_disp" attributes.
5467 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5470 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5471 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5472 gen_lowpart_no_emit.
5474 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5477 * config/i386/i386.c (ix86_expand_builtin): Change memory
5478 operand to XI, extend p0 to Pmode.
5479 * config/i386/i386.md: Change unspec volatile and operand
5480 1 mode to XI, change operand 0 mode to P.
5482 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5484 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5485 GET_MODE_MASK before any checking.
5486 (nds32_can_use_bset_p): Likewise.
5487 (nds32_can_use_btgl_p): Likewise.
5489 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5491 * config/nds32/nds32-doubleword.md: New define_split pattern for
5492 illegal register number.
5494 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5496 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5498 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5500 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5502 2018-04-25 Richard Biener <rguenther@suse.de>
5504 * lto-streamer.h (LTO_major_version): Bump to 8.
5506 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5508 * BASE-VER: Set to 9.0.0.
5510 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5512 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5513 in __abskf2 and __powikf2.
5515 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5518 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5519 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5520 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5521 Use the above on operand 2. Reindent.
5522 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5524 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5527 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5529 * config/i386/i386.opt (mcet): Removed.
5530 * doc/install.texi: Remove -mcet documentation.
5531 * doc/invoke.texi: Likewise.
5533 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5536 * doc/install.texi: Remove -mcet from bootstrap-cet.
5538 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5541 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5542 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5546 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5547 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5548 containing a CONST_VECTOR.
5550 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5552 * doc/install.texi: Update newlib dependency for nvptx.
5554 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5557 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5558 instead of INTVAL when shifting x left.
5560 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5562 PR tree-optimization/85478
5563 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5564 vect_grouped_store_supported for single element vectors.
5566 2018-04-24 Richard Biener <rguenther@suse.de>
5569 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5570 load cost increase to the case of non-constant step.
5572 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5575 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5576 destination if any_malformed_asm.
5578 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5581 * expr.c (store_field): In the bitfield case, if the value comes from
5582 a function call and is returned in registers by means of a PARALLEL,
5583 do not change the mode of the temporary unless BLKmode and VOIDmode.
5585 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5587 PR rtl-optimization/85423
5588 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5589 dependencies to debug insns when the previous insn is non-debug.
5591 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5593 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5594 enums into a single definition.
5595 (fls): Fix predicates and printing.
5598 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5600 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5601 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5602 and short u6 immediate.
5603 (check_if_valid_sleep_operand): Remove.
5604 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5606 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5608 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5609 flag_always_save_lp condition.
5610 * config/nds32/nds32.opt (malways-save-lp): New option.
5612 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5614 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5615 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5616 * config/nds32/nds32.h
5617 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5618 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5620 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5622 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5623 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5625 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5626 Chung-Ju Wu <jasonwucj@gmail.com>
5628 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5630 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5631 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5633 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5635 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5637 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5639 * config/nds32/nds32-protos.h (nds32_data_alignment,
5640 nds32_local_alignment): Declare.
5641 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5642 nds32_local_alignment): New functions.
5643 (TARGET_CONSTANT_ALIGNMENT): Define.
5644 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5646 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5648 * config/nds32/nds32.c
5649 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5650 (TARGET_MODES_TIEABLE_P): Likewise.
5652 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5654 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5657 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5658 Chung-Ju Wu <jasonwucj@gmail.com>
5660 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5661 for unaligned access.
5662 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5664 * config/nds32/nds32-intrinsic.md: Likewise.
5665 * config/nds32/nds32_intrinsic.h: Likewise.
5666 * config/nds32/nds32.h (nds32_builtins): Likewise.
5667 * config/nds32/nds32.opt (munaligned-access): New option.
5668 * config/nds32/nds32.c (nds32_asm_file_start): Display
5669 flag_unaligned_access status.
5671 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5673 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5674 -mno-relax is present.
5675 * config/riscv/linux.h (LINK_SPEC): Ditto.
5677 2018-04-20 Martin Sebor <msebor@redhat.com>
5680 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5682 (gimple_fold_builtin_stxcpy_chk): Same.
5683 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5685 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5688 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5689 __powikf2 when long double is IEEE 128-bit.
5691 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5693 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5694 step to make sure stack always aligned.
5696 2018-04-20 Carl Love <cel@us.ibm.com>
5699 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5700 size check for arg0.
5702 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5703 Tom de Vries <tom@codesourcery.com>
5706 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5707 Emit insns for calls too.
5708 (nvptx_find_par): Always look for worker-level predecessor insn.
5709 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5711 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5712 (nvptx_process_pars): Propagate frames for calls.
5714 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5717 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5719 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5720 (ix86_handle_option): Don't handle OPT_mibt.
5721 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5723 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5725 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5726 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5727 (ix86_target_macros): Define __CET__ with flag_cf_protection
5728 for -fcf-protection.
5729 * config/i386/i386.c (isa2_opts): Remove -mibt.
5730 * config/i386/i386.h (TARGET_IBT): Removed.
5731 (TARGET_IBT_P): Likewise.
5732 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5733 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5734 * config/i386/i386.opt (mcet): Update help message.
5737 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5738 -mcet as an alias for -mshstk.
5740 2018-04-20 Richard Biener <rguenther@suse.de>
5743 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5744 complexity by forcing a single use of the multiply operand.
5746 2018-04-20 Martin Jambor <mjambor@suse.cz>
5749 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5750 recursion dependency to only apply to non-clones.
5752 2018-04-20 Martin Jambor <mjambor@suse.cz>
5755 * ipa-cp.c (create_specialized_node): Check that clones of
5756 self-recursive edges exist during IPA-CP.
5758 2018-04-19 Toon Moene <toon@moene.org>
5760 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5763 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5765 PR tree-optimization/85467
5766 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5767 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5768 VECTOR_CST element to type.
5770 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5773 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5774 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5775 (builtin_longjmp): Likewise.
5776 (save_stack_nonlocal): New pattern.
5777 (restore_stack_nonlocal): Likewise.
5779 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5782 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5783 Replace ASM_OUTPUT_LABEL with fprintf.
5785 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5788 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5789 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5790 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5791 define __IBT__ and __SHSTK__ for -fcf-protection.
5792 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5794 (ix86_trampoline_init): Likewise.
5795 (x86_output_mi_thunk): Likewise.
5796 (ix86_notrack_prefixed_insn_p): Likewise.
5797 (ix86_option_override_internal): Don't disallow -fcf-protection.
5798 * config/i386/i386.md (rdssp<mode>): Also enable for
5800 (incssp<mode>): Likewise.
5801 (nop_endbr): Likewise.
5802 * config/i386/i386.opt (mcet): Change help message to built-in
5806 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5807 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5808 enable CET built-in functions.
5810 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5812 * common/config/i386/i386-common.c
5813 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5814 OPTION_MASK_ISA_MOVDIRI_UNSET,
5815 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5816 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5817 * config.gcc (movdirintrin.h): New header.
5818 * config/i386/cpuid.h (bit_MOVDIRI,
5819 bit_MOVDIR64B): New bits.
5820 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5822 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5823 (VOID, PVOID, PCVOID)): New function types.
5824 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5825 __builtin_ia32_directstoreu_u64,
5826 __builtin_ia32_movdir64b): New builtins.
5827 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5828 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5830 (ix86_valid_target_attribute_inner_p): Ditto.
5831 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5832 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5833 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5834 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5835 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5836 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5837 (movdiri<mode>, movdir64b_<mode>): New.
5838 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5839 * config/i386/immintrin.h: Include movdirintrin.h.
5840 * config/i386/movdirintrin.h: New file.
5841 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5843 2018-04-19 Richard Biener <rguenther@suse.de>
5846 * cfg.c (clear_bb_flags): When loop state says we have
5847 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5849 2018-04-19 Richard Biener <rguenther@suse.de>
5851 PR tree-optimization/84737
5852 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5853 copying restrict info.
5854 (vect_setup_realignment): Use it.
5855 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5856 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5857 the first DR to all generated stores.
5858 (vectorizable_load): Likewise for loads.
5860 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5862 PR tree-optimization/85446
5863 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5864 the integral and pointer types to have the same precision.
5866 * doc/install.texi: Document --disable-cet being the default and
5869 2018-04-18 Martin Liska <mliska@suse.cz>
5871 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5874 2018-04-18 Martin Liska <mliska@suse.cz>
5877 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5880 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5881 arguments if they are comparable.
5883 2018-04-18 Martin Liska <mliska@suse.cz>
5886 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5889 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5892 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5895 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5896 ENDBR after calling __morestack.
5898 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5901 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5902 by using gcc_base_ver to generate a gcc_driver_version, and use
5903 it when generating GCC_DRIVER_NAME.
5904 * configure: Regenerate.
5906 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5909 * config.gcc: Obsolete powerpc*-*-*spe*.
5911 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5914 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5915 (stabstr_D): Change type of unum from unsigned int to
5916 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5919 2018-04-17 Jim Wilson <jimw@sifive.com>
5922 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5923 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5924 Set arg_pointer_offset after using pretend_args_size.
5926 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5928 PR rtl-optimization/85431
5929 * dse.c (record_store): Ignore zero width stores.
5932 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5933 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5934 __builtin_stack_restore rather than after it.
5935 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5936 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5937 argument instead of virtual_dynamic_stack_rtx.
5939 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5941 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5943 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5944 Add note to error message to explain internal mapping of overloaded
5945 built-in function name to non-overloaded built-in function name.
5946 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5949 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5952 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5953 where the inputs overlap with the output.
5955 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5957 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5958 (=v, v) alternative and explicit "memory" attribute.
5959 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5960 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5962 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5963 "sselog1" type instead of "sselog".
5964 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5965 "sselog". Remove explicit "memory" attribute.
5966 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5967 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5969 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5970 "sselog1" type instead of "sselog". Remove explicit "memory"
5972 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5973 use "sselog1" type instead of "sselog". Remove explicit "memory"
5975 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5976 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5978 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5979 "sselog1" type instead of "sselog". Remove explicit "memory"
5981 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5982 use "sselog1" type instead of "sselog". Remove explicit "memory"
5986 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5989 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5992 2018-04-17 Martin Jambor <mjambor@suse.cz>
5995 * ipa-cp.c (create_specialized_node): Call
5996 expand_all_artificial_thunks if necessary.
5998 2018-04-17 Martin Liska <mliska@suse.cz>
6001 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6002 in message, remote space in between '_G' and '('.
6004 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6007 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6008 avx512f_vmcmp<mode>3<round_saeonly_name>,
6009 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6010 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6011 avx512f_rndscale<mode><round_saeonly_name>,
6012 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6013 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6014 Use %<iptr>2 instead of %2 for -masm=intel.
6015 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6016 avx512f_vcvttss2usi<round_saeonly_name>,
6017 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6019 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6020 avx512f_vcvttsd2usi<round_saeonly_name>,
6021 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6022 Use %q1 instead of %1 for -masm=intel.
6023 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6024 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6025 of %3 for -masm=intel.
6026 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6028 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6030 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6032 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6033 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6035 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6037 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6038 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6039 %g1 and one with %0 and %1.
6040 (avx512er_vmrcp28<mode><round_saeonly_name>,
6041 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6043 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6044 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6045 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6046 of %0 and %{%4%} for -masm=intel.
6047 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6048 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6049 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6050 order of %0 and %{%5%}%{z%} for -masm=intel.
6052 2018-04-17 Jan Hubicka <jh@suse.cz>
6055 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6057 2018-04-17 Martin Liska <mliska@suse.cz>
6060 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6061 for target_clone error message. Make default implementation
6062 clone to be a local declaration.
6063 (separate_attrs): Add new argument and check for an empty
6065 (expand_target_clones): Handle it.
6066 (ipa_target_clone): Make redirection just for target_clones
6069 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6070 Tom de Vries <tom@codesourcery.com>
6073 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6074 tiled basic blocks without omp continue statements.
6076 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6079 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6080 vec_extract expression as having side effects to make sure it gets
6083 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6086 * config/i386/i386.c (get_builtin_code_for_version): Check
6089 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6092 * gcc/config.gcc: Support "skylake".
6093 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6095 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6096 (processor_target_table): Add "skylake".
6097 (ix86_option_override_internal): Add "skylake".
6098 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6099 PROCESSOR_CANNONLAKE.
6100 (get_builtin_code_for_version): Fix priority for
6101 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6102 PROCESSOR_SKYLAKE-AVX512.
6103 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6104 (processor_type): Add PROCESSOR_SKYLAKE.
6106 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6107 Jason Merrill <jason@redhat.com>
6110 * convert.c (convert_to_integer_1): Use direct recursion for
6111 enumeral types and types with a precision less than the number
6112 of bits in their mode.
6114 2018-04-16 Julia Koval <julia.koval@intel.com>
6117 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6118 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6120 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6123 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6124 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6125 and -mno-direct-move.
6127 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6130 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6131 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6132 Ensure negative shifts result in {0}.
6134 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6136 PR rtl-optimization/79916
6137 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6138 regs (if any) to define how to gnerate SD moves when LRA is in
6141 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6143 PR rtl-optimization/85393
6144 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6145 * except.c (expand_dw2_landing_pad_for_region): Make static.
6146 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6147 a label and unconditional jump to old_bb, rather than
6148 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6151 PR rtl-optimization/85376
6152 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6153 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6154 instead of a specific value.
6156 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6157 Bin Cheng <bin.cheng@arm.com>
6159 PR tree-optimization/82965
6160 PR tree-optimization/83991
6161 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6162 by_profile_only parameter.
6163 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6164 information if the loop was predicted to iterate too many times.
6165 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6167 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6170 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6173 2018-04-13 Martin Liska <mliska@suse.cz>
6174 Jakub Jelinek <jakub@redhat.com>
6177 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6178 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6179 * builtins.c (expand_builtin_memory_copy_args): Use
6180 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6181 handle dest_addr == pc_rtx.
6183 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6186 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6187 asked to not generate direct moves.
6188 (fix_trunc<mode>si2_stfiwx): Similar.
6189 (fix_trunc<mode>si2_internal): Similar.
6191 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6194 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6195 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6196 lookup if dest in some wider mode is known to be const0_rtx and
6197 if so, record permanent equivalence for it to be ZERO_EXTEND of
6198 the narrower mode destination.
6200 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6202 * lto-streamer-out.c (output_function): Revert 259346.
6203 * omp-expand.c (expand_oacc_for): Likewise.
6205 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6207 PR rtl-optimization/85354
6208 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6209 * sel-sched.c (sel_global_init): ... here.
6211 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6214 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6215 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6216 mode for PE-COFF targets.
6217 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6218 (i386_pe_asm_lto_end): Likewise.
6219 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6220 (TARGET_ASM_LTO_END): Likewise.
6221 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6222 (i386_pe_asm_lto_start): New function.
6223 (i386_pe_asm_lto_end): Likewise.
6225 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6226 Richard Biener <rguenther@suse.de>
6229 * lto-streamer-out.c (output_function): Fix CFG loop state before
6231 * omp-expand.c (expand_oacc_for): Handle calls to internal
6232 functions like regular functions.
6234 2018-04-12 Richard Biener <rguenther@suse.de>
6237 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6238 for the early LTO debug to properly generate references to it
6239 during DIE emission. Do not re-use that for the skeleton for
6241 (dwarf2out_early_finish): Likewise.
6243 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6246 * config/i386/sse.md
6247 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6248 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6249 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6250 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6251 and output is a reg, avoid creating invalid lowpart subreg, but
6252 instead split into a 512-bit move. Don't split if not AVX512VL,
6253 input is xmm16+ reg and output is a mem.
6254 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6255 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6256 xmm16+ reg and output is a mem.
6258 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6260 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6261 also for flag_dwarf2_cfi_asm.
6263 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6265 PR rtl-optimization/85342
6266 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6267 a bool scalar var inside of the loop instead. Don't try to update
6268 recog_data.operand after failed apply_change_group.
6270 2018-04-12 Tom de Vries <tom@codesourcery.com>
6273 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6274 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6275 array with flexible array member as array without given dimension.
6276 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6277 argument for undefined param to true.
6279 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6282 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6283 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6284 from PowerPC section.
6285 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6286 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6287 help text that is too long.
6288 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6289 help text that is too long.
6290 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6291 help text that is too long.
6293 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6295 * config/alpha/alpha.md (stack_probe_internal): Rename
6296 from "probe_stack". Update all callers.
6298 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6300 PR rtl-optimization/84566
6301 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6302 sched_macro_fuse_insns.
6304 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6307 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6308 (compute_block_dependences): ... from here.
6310 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6312 PR tree-optimization/85331
6313 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6314 from int to HOST_WIDE_INT.
6316 2018-04-11 Martin Jambor <mjambor@suse.cz>
6319 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6320 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6321 not the same as the source val.
6322 (cgraph_edge_brings_value_p): New parameter.
6323 (gather_edges_for_value): Pass destination value to
6324 cgraph_edge_brings_value_p.
6325 (perhaps_add_new_callers): Likewise.
6326 (get_info_about_necessary_edges): Likewise and exclude values brought
6327 only by self-recursive edges.
6328 (create_specialized_node): Redirect only clones of self-calling edges.
6329 (+self_recursive_pass_through_p): New function.
6330 (find_more_scalar_values_for_callers_subset): Use it.
6331 (find_aggregate_values_for_callers_subset): Likewise.
6332 (known_aggs_to_agg_replacement_list): Removed.
6333 (decide_whether_version_node): Re-calculate known constants for all
6334 remaining context clones.
6336 2018-04-11 Richard Biener <rguenther@suse.de>
6339 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6340 from early DWARF output.
6341 (dwarf2out_early_finish): Output line info unconditionally into
6342 early DWARF and add reference to it.
6344 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6347 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6348 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6349 other than V2DFmode using iptr mode attribute.
6350 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6352 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6354 PR rtl-optimization/84659
6355 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6357 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6360 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6362 (output_loc_list): Pass address of a dummy size variable even in the
6363 locview handling loop.
6364 (index_location_lists): Add comment on why skip_loc_list_entry can't
6367 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6370 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6373 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6376 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6377 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6378 and -mstring-compare-inline-limit.
6380 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6383 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6384 for stack clash protection in a register whenever we need it to be in
6387 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6389 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6390 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6392 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6395 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6397 (mlong-double-): Ditto.
6398 * config/rs6000/sysv4.opt (msdata=): Ditto.
6399 (mtls-size=): Ditto.
6401 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6403 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6404 erroneous entries for
6405 "vector int vec_ldl (int, long int *)", and
6406 "vector unsigned int vec_ldl (int, unsigned long int *)".
6407 Add comments and entries for
6408 "vector bool char vec_ldl (int, bool char *)",
6409 "vector bool short vec_ldl (int, bool short *)",
6410 "vector bool int vec_ldl (int, bool int *)",
6411 "vector bool long long vec_ldl (int, bool long long *)",
6412 "vector pixel vec_ldl (int, pixel *)",
6413 "vector long long vec_ldl (int, long long *)",
6414 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6415 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6416 type tree bool_long_long_type_node and correct definition of
6417 bool_V2DI_type_node to make reference to this new type tree.
6418 (rs6000_mangle_type): Replace erroneous reference to
6419 bool_long_type_node with bool_long_long_type_node.
6420 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6421 comments to emphasize sign distinctions for char and int types and
6422 replace RS6000_BTI_bool_long constant with
6423 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6424 use of RS6000_BTI_pixel.
6425 (bool_long_type_node): Remove this macro definition.
6426 (bool_long_long_type_node): New macro definition
6428 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6430 PR rtl-optimization/85300
6431 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6432 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6433 simplify_unary_operation fails.
6435 2018-04-10 Martin Liska <mliska@suse.cz>
6437 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6438 cgraph_edge and ipa_ref.
6440 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6444 * config/i386/sse.md
6445 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6446 computation of the VEC_MERGE selector from mask.
6447 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6448 Fix decoding of the VEC_MERGE selector into mask.
6450 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6452 PR tree-optimization/85286
6453 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6455 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6457 * final.c (final_1): Set insn_last_address as well as
6458 insn_current_address.
6460 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6463 * explow.c (emit_stack_probe): Call validize_mem on memory location
6464 before passing it to gen_probe_stack. Create address operand and
6465 legitimize it for the probe_stack_address case.
6467 2018-04-09 Jan Hubicka <jh@suse.cz>
6470 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6471 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6472 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6474 (free_lang_data): Rebuild type inheritance graph.
6476 2018-04-09 Martin Sebor <msebor@redhat.com>
6478 * invoke.texi (-finline-small-functions): Mention other optimization
6480 (-findirect-inlining, -fpartial-inlining): Same.
6481 (-finline-functions-called-once): Same.
6482 (-freorder-blocks-and-partition): Same.
6484 2018-04-09 Jan Hubicka <jh@suse.cz>
6487 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6488 jumps; choose last target that matches the criteria (i.e.
6489 no partition changes for non-crossing jumps).
6490 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6491 support for redirecting crossing jumps to non-crossing.
6493 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6495 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6496 also for naked functions.
6498 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6500 * config/arc/arc.md (add_shift): New pattern.
6501 (add_shift2): Likewise.
6502 (sub_shift): Likewise.
6503 (sub_shift_cmp0_noout): Likewise.
6504 (compare_si_ashiftsi): Likewise.
6505 (xbfu_cmp0_noout): New combine pattern.
6506 (xbfu_cmp0"): Likewise.
6507 (movsi_set_cc_insn): Place the predicable variant first.
6508 (commutative_binary_cmp0_noout): Remove clobber.
6509 (commutative_binary_cmp0): New pattern.
6510 (noncommutative_binary_cmp0): Likewise.
6511 (noncommutative_binary_cmp0_noout): Likewise.
6512 (noncommutative_binary_comparison_result_used): Removed.
6513 (rsub_cmp0): New pattern.
6514 (rsub_cmp0_noout): Likewise.
6515 (extzvsi): Changed, keep only meaningful variants.
6516 (SQH, SEZ): New iterators.
6517 (SQH_postfix): New mode attribute.
6518 (SEZ_prefix): New code attribute.
6519 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6520 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6521 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6523 (noncommutative_operator): Check the availability of barrel
6526 2018-04-09 Richard Biener <rguenther@suse.de>
6528 PR tree-optimization/85284
6529 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6530 Only use the niter constraining form of simple_iv when the exit
6533 2018-04-09 Tom de Vries <tom@codesourcery.com>
6536 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6537 (define_expand "*memory_barrier"): New define_expand.
6538 (define_insn "memory_barrier"): New insn.
6540 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6542 PR rtl-optimization/80463
6543 PR rtl-optimization/83972
6544 PR rtl-optimization/83480
6546 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6547 correct producer for the insn.
6548 (tidy_control_flow): Fixup seqnos in case of debug insns.
6550 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6552 PR rtl-optimization/83913
6554 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6555 different sched-times when merging exprs.
6557 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6559 PR rtl-optimization/83962
6561 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6562 tidy_fallthru_edge and tidy_control_flow.
6564 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6566 PR rtl-optimization/83530
6568 * sel-sched.c (force_next_insn): New global variable.
6569 (remove_insn_for_debug): When force_next_insn is true, also leave only
6570 next insn in the ready list.
6571 (sel_sched_region): When the region wasn't scheduled, make another pass
6572 over it with force_next_insn set to 1.
6574 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6576 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6578 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6579 for interrupt control.
6580 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6581 functions for interrupt control.
6582 * config/nds32/nds32-intrinsic.md: Likewise.
6583 * config/nds32/nds32_intrinsic.h: Likewise.
6584 * config/nds32/nds32.h (nds32_builtins): Likewise.
6586 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6588 * config/nds32/nds32.c (nds32_init_machine_status,
6589 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6590 strict_aligned_p field.
6591 (nds32_expand_to_rtl_hook): New function.
6592 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6593 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6595 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6596 Chung-Ju Wu <jasonwucj@gmail.com>
6598 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6599 * config/nds32/nds32-n7.md: New file.
6600 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6601 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6603 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6604 * config/nds32/nds32.md (pipeline_model): Add n7.
6605 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6606 * config/nds32/pipelines.md: Include n7 settings.
6608 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6609 Chung-Ju Wu <jasonwucj@gmail.com>
6611 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6612 * config/nds32/nds32-e8.md: New file.
6613 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6614 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6616 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6617 * config/nds32/nds32.md (pipeline_model): Add e8.
6618 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6619 * config/nds32/pipelines.md: Include e8 settings.
6621 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6622 Chung-Ju Wu <jasonwucj@gmail.com>
6624 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6625 * config/nds32/nds32-n8.md: New file.
6626 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6627 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6629 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6630 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6631 * config/nds32/nds32.md (pipeline_model): Add n8.
6632 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6633 * config/nds32/pipelines.md: Include n8 settings.
6635 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6636 Chung-Ju Wu <jasonwucj@gmail.com>
6638 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6639 * config/nds32/nds32-n9-2r1w.md: New file.
6640 * config/nds32/nds32-n9-3r2w.md: New file.
6641 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6642 nds32_register_ports): New or modify for cpu n9.
6643 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6645 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6646 * config/nds32/nds32-utils.c: New file.
6647 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6648 TARGET_MUL_SLOW): Define.
6649 * config/nds32/nds32.md (pipeline_model): New attribute.
6650 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6651 New options that support cpu n9.
6652 * config/nds32/pipelines.md: Include n9 settings.
6653 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6655 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6657 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6658 information if necessary.
6659 (output_cond_branch_compare_zero): Likewise.
6660 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6661 (nds32_target_alignment): Refine for alignment.
6662 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6663 (FUNCTION_BOUNDARY): Modify.
6664 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6666 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6668 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6670 * config/nds32/constants.md (unspec_volatile_element): Add values for
6671 TLB operation and data prefetch.
6672 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6673 functions for TLB operation and data prefetch.
6674 * config/nds32/nds32-intrinsic.md: Likewise.
6675 * config/nds32/nds32_intrinsic.h: Likewise.
6676 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6677 (nds32_print_operand): Likewise.
6678 * config/nds32/nds32.h (nds32_builtins): Likewise.
6680 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6681 Andrew Pinski <pinsika@gcc.gnu.org>
6684 * match.pd: Use constant_boolean_node of correct type instead of
6685 boolean_true_node or boolean_false_node for simplifying
6686 pointer comparisons to zero.
6688 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6690 PR tree-optimization/80021
6691 * tree.c (verify_type_variant): Make error call in verify_variant_match
6692 translatable and remove final full stop.
6694 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6696 * config/nds32/constants.md (unspec_volatile_element): Add
6697 UNSPEC_VOLATILE_EH_RETURN.
6698 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6699 nds32_output_stack_pop): Support dwarf exception handling process.
6700 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6701 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6702 exception handling process.
6703 (nds32_compute_stack_frame): Likewise.
6704 (nds32_return_addr_rtx): Likewise.
6705 (nds32_initial_elimination_offset): Likewise.
6706 (nds32_expand_prologue): Likewise.
6707 (nds32_expand_epilogue): Likewise.
6708 (nds32_dynamic_chain_address): New function.
6709 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6711 (DYNAMIC_CHAIN_ADDRESS): Define.
6712 (EH_RETURN_DATA_REGNO): Define.
6713 (EH_RETURN_STACKADJ_RTX): Define.
6714 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6715 patterns for dwarf exception handling.
6717 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6719 * config/nds32/nds32.h: Clean up obsolete macros.
6721 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6723 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6724 Add enum values for particular instructions.
6725 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6726 particular intrinsic functions.
6727 * config/nds32/nds32-intrinsic.md: Likewise.
6728 * config/nds32/nds32_intrinsic.h: Likewise.
6729 * config/nds32/nds32.h (nds32_builtins): Likewise.
6730 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6731 (btst, ave): New patterns for particular instructions.
6733 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6735 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6736 Add enum values for atomic load/store and memory sync.
6737 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6739 * config/nds32/nds32-intrinsic.md: Likewise.
6740 * config/nds32/nds32_intrinsic.h: Likewise.
6741 * config/nds32/nds32.h (nds32_builtins): Likewise.
6743 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6745 PR tree-optimization/85257
6746 * fold-const.c (native_encode_vector): If not all elts could fit
6747 and off is -1, return 0 rather than offset.
6748 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6749 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6750 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6751 adjust buffer in native_interpret_expr call.
6753 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6755 * config/nds32/constants.md (unspec_volatile_element): Add cache
6756 control enum values.
6757 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6758 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6759 * config/nds32/nds32.c (nds32_cctl_names): New.
6760 (nds32_print_operand): Handle cache control register names.
6761 * config/nds32/nds32.h (nds32_builtins): New enum values.
6762 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6764 * config/nds32/nds32.md (type): Add mmu.
6765 * config/nds32/pipelines.md (simple_insn): Add mmu.
6767 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6769 * config/nds32/nds32.md (type): Remove call.
6770 * config/nds32/pipelines.md (simple_insn): Likewise.
6772 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6774 * config/nds32/constants.md (unspec_volatile_element): Add
6775 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6776 UNSPEC_VOLATILE_FMFCFG.
6777 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6778 description for fmfcfg and fmfcsr.
6779 (bdesc_1arg): Add fmtcsr.
6780 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6781 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6782 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6783 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6784 unspec_fmfcfg): New patterns.
6785 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6786 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6787 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6788 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6789 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6790 __nds32__fmfcfg): Define.
6792 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6794 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6795 intrinsic register names.
6796 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6797 intrinsic register enum values and macros.
6799 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6801 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6802 for load/store addressing form.
6803 (nds32_print_operand_address): Likewise.
6805 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6808 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6809 based on LABEL_REF. Remove useless assertion.
6810 (pic_address_needs_scratch): Fix formatting.
6811 (sparc_legitimize_pic_address): Minor tweaks.
6812 (sparc_delegitimize_address): Adjust assertion accordingly.
6813 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6814 into symbolic_operand.
6815 (movsi_high_pic_label_ref): Likewise.
6816 (movsi_lo_sum_pic_label_ref): Likewise.
6817 (movdi_pic_label_ref): Likewise.
6818 (movdi_high_pic_label_ref): Likewise.
6819 (movdi_lo_sum_pic_label_ref): Likewise.
6821 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6823 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6824 custom LIB_SPEC setup.
6826 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6827 Kito Cheng <kito.cheng@gmail.com>
6829 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6830 * config/riscv/freebsd.h: New.
6832 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6834 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6835 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6838 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6839 Kito Cheng <kito.cheng@gmail.com>
6841 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6842 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6843 * config/nds32/nds32-protos.h (nds32_output_call,
6844 nds32_output_return): Declare.
6845 * config/nds32/nds32.md: Refine all the call and return patterns.
6847 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6850 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6851 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6853 PR rtl-optimization/84872
6854 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6855 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6858 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6860 * expr.c (copy_blkmode_to_reg): Revert 254862.
6861 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6863 2018-04-06 Richard Biener <rguenther@suse.de>
6866 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6867 after seeing a component reference with an adjacent field. Treat
6868 refs to arrays at struct end of external decls similar to
6869 refs to unconstrained commons.
6871 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6874 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6875 look through SAVE_EXPRs with non-side-effects argument. Adjust
6877 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6880 2018-04-06 Richard Biener <rguenther@suse.de>
6883 * alias.c (find_base_term): New wrapper around find_base_term
6884 unwinding CSELIB_VAL_PTR changes.
6885 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6888 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6890 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6892 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6893 constant definitions.
6894 ("nop"): lr 0,0 -> nopr r0
6895 ("nop_lr0", "nop_lr1"): New insn definitions.
6897 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6899 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6900 NDS32_V3PUSH_AVAILABLE_P macro.
6902 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6903 Chung-Ju Wu <jasonwucj@gmail.com>
6905 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6906 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6907 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6908 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6909 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6910 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6911 * config/nds32/constraints.md: New constraints and checking for hard
6912 float configuration.
6913 * config/nds32/iterators.md: New mode iterator and attribute for hard
6914 float configuration.
6915 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6917 * config/nds32/nds32-fpu.md: New file.
6918 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6919 deal with hard float code generation.
6920 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6922 (abi_type, float_reg_number): New enum type.
6923 * config/nds32/nds32-predicates.c: New predicates for hard float.
6924 * config/nds32/nds32-protos.h: Declare functions for hard float.
6925 * config/nds32/nds32.c: Implementation for hard float configuration.
6926 * config/nds32/nds32.h: Definitions for hard float configuration.
6927 * config/nds32/nds32.md: Include hard float machine description and
6928 modify patterns for hard float configuration.
6929 * config/nds32/nds32.opt: New options for hard float configuration.
6930 * config/nds32/predicates.md: New predicates for hard float
6933 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6935 * common/config/nds32/nds32-common.c
6936 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6938 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6941 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6942 CONSTRUCTOR_ELT (ctor, ...)->value.
6944 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6947 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6949 2018-04-05 Tom de Vries <tom@codesourcery.com>
6952 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6955 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6956 Kito Cheng <kito.cheng@gmail.com>
6958 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6959 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6960 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6961 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6963 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6964 Kito Cheng <kito.cheng@gmail.com>
6966 * config/nds32/constraints.md (Ufe): New memory constraint.
6967 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6968 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6969 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6971 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6972 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6974 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6976 * config/nds32/nds32.md: Use optimize_size in the condition for
6977 alu-shift instructions.
6979 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6981 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6983 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6985 * config/nds32/nds32.md (negsi2): Refine pattern.
6987 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6988 Chung-Ju Wu <jasonwucj@gmail.com>
6990 * config/nds32/iterators.md (shift_rotate): New code iterator.
6991 (shift): New code attribute.
6992 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6993 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6994 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6995 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6996 bit-wise operations.
6997 (andsi3, *andsi3): Ditto.
6998 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6999 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7000 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7001 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7002 nds32_ior_operand, nds32_xor_operand): New predicates.
7004 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7006 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7007 (addsi3, subsi3): ... this.
7009 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7011 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7013 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7015 * config/nds32/nds32.md: Adjust indention.
7017 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7019 * config/nds32/nds32.md (feature): New attribute.
7021 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7023 * config/nds32/nds32.md (subtype): New attribute.
7025 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7028 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7029 expansion to perform a bitwise AND of the argument followed by a
7030 boolean negation of the result.
7032 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7034 PR rtl-optimization/84878
7035 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7036 the basic block. Assert the use reference is not artificial and that
7037 it has an associated insn.
7039 2018-04-04 Michael Matz <matz@suse.de>
7041 * builtins.c (compute_objsize): Pass correct operand
7042 to array_at_struct_end_p.
7044 2018-04-04 Richard Biener <rguenther@suse.de>
7047 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7048 from contexts for DINFO_LEVEL_TERSE and below.
7050 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7052 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7053 resiter_operand condition.
7054 * config/nds32/nds32.md (*move<mode>): Ditto.
7056 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7057 Monk Chiang <sh.chiang04@gmail.com>
7059 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7061 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7063 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7065 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7066 Kito Cheng <kito.cheng@gmail.com>
7068 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7069 nds32_cond_code_str, output_cond_branch,
7070 output_cond_branch_compare_zero, nds32_expand_cbranch,
7071 nds32_expand_cstore, nds32_expand_movcc,
7072 nds32_output_cbranchsi4_equality_zero,
7073 nds32_output_cbranchsi4_equality_reg,
7074 nds32_output_cbranchsi4_equality_reg_or_const_int,
7075 nds32_output_cbranchsi4_greater_less_zero: New functions.
7076 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7077 nds32_expand_cstore, nds32_expand_movcc,
7078 nds32_output_cbranchsi4_equality_zero,
7079 nds32_output_cbranchsi4_equality_reg,
7080 nds32_output_cbranchsi4_equality_reg_or_const_int,
7081 nds32_output_cbranchsi4_greater_less_zero): Declare.
7082 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7083 nds32_rimm11s_operand): New predicates.
7084 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7085 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7088 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7090 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7091 * config/nds32/nds32.md: Ditto.
7092 * config/nds32/pipelines.md: Ditto.
7094 2018-04-04 Richard Biener <rguenther@suse.de>
7096 PR tree-optimization/85168
7097 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7098 propagating abnormals.
7100 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7102 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7104 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7105 Kito Cheng <kito.cheng@gmail.com>
7107 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7108 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7109 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7110 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7111 * config/nds32/nds32.md (sibcall_internal): New.
7112 (sibcall_register): Remove.
7113 (sibcall_immediate): Remove.
7114 (sibcall_value_internal): New.
7115 (sibcall_value_register): Remove.
7116 (sibcall_value_immediate): Remove.
7117 * config/nds32/predicates.md (nds32_general_register_operand): New.
7118 (nds32_call_address_operand): New.
7120 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7122 PR rtl-optimization/85167
7123 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7124 bb_defs if *split_p, instead preinitialize it to NULL.
7126 PR tree-optimization/85156
7127 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7128 evaluating the argument multiple times.
7130 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7132 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7134 (_mm_cvtpd_ps): Likewise.
7135 (_mm_cvttpd_epi32): Likewise.
7136 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7137 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7138 vector, pixel, and bool following altivec.h include.
7140 2018-04-03 Martin Sebor <msebor@redhat.com>
7142 * doc/extend.texi (Common Function Attributes): Clarify.
7143 (const attribute): Likewise.
7144 (pure attribute): Likewise.
7146 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7149 * config/i386/i386.c (ix86_expand_vector_set): Use
7150 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7152 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7154 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7155 instructions when changing rounding bits to preserve precision bits
7156 in the x87 control word.
7158 2018-04-03 Martin Liska <mliska@suse.cz>
7160 PR tree-optimization/82491
7161 * rtl.h (strip_offset_and_add): Replace += suboffset with
7162 poly_uint64 () + suboffset.
7164 2018-03-29 Martin Liska <mliska@suse.cz>
7165 Martin Jambor <mjambor@suse.cz>
7168 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7169 param_type is not an integral or pointer type.
7171 2018-04-03 Richard Biener <rguenther@suse.de>
7173 * sese.h (recompute_all_dominators): Remove.
7175 2018-04-02 Martin Sebor <msebor@redhat.com>
7177 * doc/invoke.texi (-Wrestrict): Fix typos.
7179 2018-04-02 Jim Wilson <jimw@sifive.com>
7181 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7182 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7183 (<optab>di3, <optab>si3_extend): Likewise.
7184 (<optab>si3_mask, <optab>si3_mask_1): New.
7185 (<optab>di3_mask, <optab>di3_mask_1): New.
7186 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7187 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7188 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7190 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7192 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7195 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7197 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7198 (nds32_canonicalize_comparison): New function.
7200 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7201 Kito Cheng <kito.cheng@gmail.com>
7202 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7204 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7205 * config/nds32/constants.md (unspec_volatile_element): Add
7206 UNSPEC_VOLATILE_RELAX_GROUP.
7207 * config/nds32/nds32-relax-opt.c: New file.
7208 * config/nds32/nds32-predicates.c
7209 (nds32_symbol_load_store_p): New function.
7210 * config/nds32/nds32-protos.h
7211 (nds32_symbol_load_store_p): Declare function.
7212 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7213 * config/nds32/nds32.c
7214 (nds32_register_pass): New function to register pass.
7215 (nds32_register_passes): New function to register passes.
7216 * config/nds32/nds32.md (relax_group): New pattern.
7217 * config/nds32/nds32.opt (mrelax-hint): New option.
7218 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7220 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7222 * config/nds32/t-nds32: Modify files dependency.
7224 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7226 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7227 (PROFILE_HOOK): Define its implementation.
7229 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7231 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7232 type and 32-bit size.
7234 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7237 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7238 (V_128_256): New mode iterator.
7239 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7240 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7241 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7243 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7244 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7246 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7249 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7250 NaN inputs correctly.
7252 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7255 * config/rs6000/vsx.md (??r): New mode attribute.
7256 (*vsx_mov<mode>_64bit): Use it.
7257 (*vsx_mov<mode>_32bit): Likewise.
7259 2018-03-30 Martin Sebor <msebor@redhat.com>
7261 PR tree-optimization/84818
7262 * builtins.c (check_access): Use warning_n.
7264 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7267 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7269 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7272 2018-03-30 Julia Koval <julia.koval@intel.com>
7275 * x86-tune.def (movx, partial_reg_dependency): Enable for
7278 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7281 * lra-constraints.c (process_alt_operands): Move setting
7282 this_alternative_matches below.
7284 2018-03-29 Martin Liska <mliska@suse.cz>
7287 * doc/invoke.texi: Document how LTO works with debug info.
7288 Describe auto-load support of binutils. Mention 'x86-64'
7289 as valid option value of -march option.
7291 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7293 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7296 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7297 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7298 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7301 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7304 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7305 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7306 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7307 for RS6000_BTM_POWERPC64.
7308 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7309 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7310 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7315 2018-03-28 Carl Love <cel@us.ibm.com>
7318 2017-09-27 Carl Love <cel@us.ibm.com>
7320 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7321 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7322 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7325 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7327 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7328 instead of __vector bool.
7329 (_mm_max_pu8): Likewise.
7330 (_mm_min_pi16): Likewise.
7332 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7335 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7336 (DIVWEUO): Likewise.
7338 (DIVDEUO): Likewise.
7339 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7340 DIVWEUO and DIVDEUO.
7341 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7342 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7343 (div_extend): Likewise.
7344 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7346 (__builtin_divweuo): Likewise.
7347 (__builtin_divdeo): Likewise.
7348 (__builtin_divdeuo): Likewise.
7350 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7353 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7354 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7356 PR tree-optimization/82004
7357 * gimple-match-head.c (optimize_pow_to_exp): New function.
7358 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7359 Don't fold to exp if optimize_pow_to_exp is false.
7361 2018-03-28 Martin Liska <mliska@suse.cz>
7364 * calls.c (initialize_argument_information): Fix trailing space.
7365 * common.opt: Fix typo and provide better explanation for
7366 -fsanitize-coverage option.
7367 * config/i386/i386.opt: Fix typo.
7369 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7370 Martin Liska <mliska@suse.cz>
7373 * gimplify.c (asan_poison_variable): Don't do the check for
7374 gimplify_omp_ctxp here.
7375 (gimplify_decl_expr): Do it here.
7376 (gimplify_target_expr): Likewise.
7378 2018-03-28 Martin Liska <mliska@suse.cz>
7381 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7382 chkp_type_bounds_count if MPX is not enabled.
7384 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7386 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7388 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7391 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7392 function to create the function decl for complex long double
7393 multiply and divide for -mabi=ieeelongdouble.
7394 (init_float128_ieee): Call it.
7396 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7399 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7400 -fcf-protection=branch -mibt.
7401 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7403 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7406 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7407 arm_word_relocations.
7409 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7412 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7413 extern array declarations.
7415 2018-03-27 Richard Biener <rguenther@suse.de>
7418 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7419 explicit single_use checks.
7421 2018-03-27 Richard Biener <rguenther@suse.de>
7423 PR tree-optimization/85082
7424 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7427 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7429 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7430 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7431 Turn on fasynchronous-unwind-tables and funwind-tables.
7433 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7436 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7437 (*bmi_blsr_<mode>_ccz): Ditto.
7439 2018-03-26 Tom de Vries <tom@codesourcery.com>
7441 PR tree-optimization/85063
7442 * omp-general.c (offloading_function_p): New function. Factor out
7444 * omp-offload.c (pass_omp_target_link::gate): ... here.
7445 * omp-general.h (offloading_function_p): Declare.
7446 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7447 with attribute omp declare target for offloading functions.
7449 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7451 PR tree-optimization/84005
7452 * tree-data-ref.h (get_base_for_alignment): Declare.
7453 * tree-data-ref.c (get_base_for_alignment_1): New function.
7454 (get_base_for_alignment): Likewise.
7455 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7456 get_base_for_alignment to find a suitable base object, instead
7457 of always using drb->base_address.
7459 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7462 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7463 known size by default.
7465 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7468 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7469 and non BLKmode operands.
7471 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7474 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7475 Clean up attributes.
7477 2018-03-23 Richard Biener <rguenther@suse.de>
7480 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7481 we are going to emit early debug for LTO.
7483 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7486 * function.c (match_asm_constraints_1): Don't optimize if input
7487 doesn't satisfy general_operand predicate for output's mode.
7490 * alias.c (write_dependence_p): Don't require for x_canonicalized
7491 non-VOIDmode if x has VOIDmode.
7494 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7495 just don't try to optimize it rather than assert it never happens.
7497 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7499 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7500 macro expansions for definition of ST_INTERNAL_<mode> and
7501 LD_INTERNAL_<mode> builtins.
7502 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7504 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7506 (altivec_expand_st_builtin): Likewise.
7507 (altivec_expand_builtin): Remove calls to deleted functions.
7508 (rs6000_address_for_altivec): Delete this function.
7509 * config/rs6000/vector.md: Remove expands for
7510 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7512 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7515 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7516 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7517 re-computing once computed.
7518 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7519 (arm_init_machine_status): Initialize
7520 machine->static_chain_stack_bytes.
7522 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7525 * doc/extend.texi: Add four new prototypes for vec_ld.
7526 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7527 definitions for more logical presentation.
7528 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7529 entries for V1TI variants of __builtin_altivec_ld builtin.
7530 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7531 handling of V1TI variant of LVX icode pattern.
7532 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7533 (rs6000_gimple_fold_builtin): Likewise.
7534 (altivec_init_builtins): Add code to define
7535 __builtin_altivec_lvx_v1ti function.
7537 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7540 * function.c (match_asm_constraints_1): Don't do the optimization
7541 if input isn't a REG, SUBREG, MEM or constant.
7543 2018-03-22 Tom de Vries <tom@codesourcery.com>
7545 PR tree-optimization/84956
7546 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7547 bb_has_abnormal_pred.
7549 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7552 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7553 DECL_INITIAL (decl) to decl at the end.
7554 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7557 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7559 * doc/extend.texi (__builtin_tgmath): Document when complex
7560 integer types are treated as _Complex _Float64.
7562 2018-03-21 Tom de Vries <tom@codesourcery.com>
7564 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7566 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7568 PR tree-optimization/84960
7569 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7570 if it is ENTRY block, move them into single succ of ENTRY in that case.
7572 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7574 PR tree-optimization/84811
7575 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7576 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7577 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7578 (known_subrange_p): Update accordingly. Cast each value involved
7579 in the size comparison, rather than casting the result of the
7582 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7584 PR tree-optimization/84982
7585 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7586 by flipping the least significant bit rather than all bits from
7587 bitpos to bitpos + bitsize - 1.
7589 2018-03-21 Nathan Sidwell <nathan@acm.org>
7591 * doc/extend.texi (Deprecated Features): Remove mention of
7592 long-deleted deprecations.
7594 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7597 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7598 * configure: Regenerate.
7600 2018-03-21 Tom de Vries <tom@codesourcery.com>
7602 PR tree-optimization/83126
7603 * tree-parloops.c (num_phis): New function.
7604 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7606 2018-03-21 Nathan Sidwell <nathan@acm.org>
7608 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7609 mention anon-struct/union members and trailing attributes.
7611 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7613 PR tree-optimization/84969
7614 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7615 builtin memset partitions if they set different rhs values.
7617 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7619 PR rtl-optimization/84989
7620 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7621 VEC_DUPLICATE with scalar result mode.
7623 2018-03-21 Martin Liska <mliska@suse.cz>
7626 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7627 not intended return statement.
7629 2018-03-21 Martin Liska <mliska@suse.cz>
7632 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7633 (chkp_find_bound_slots_1): Limit number of iterations.
7635 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7638 * Minor grammar fixes for x86 options.
7640 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7643 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7644 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7646 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7649 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7650 (altivec_lvx_<mode>_1op): Likewise.
7651 (altivec_stvx_<mode>_2op): Likewise.
7652 (altivec_stvx_<mode>_1op): Likewise.
7653 (altivec_lvx_<VM2:mode>): New define_expand.
7654 (altivec_stvx_<VM2:mode>): Likewise.
7655 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7656 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7657 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7658 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7659 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7660 (rs6000_gen_lvx): Likewise.
7661 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7662 (altivec_expand_stv_builtin): Likewise.
7663 (altivec_expand_builtin): Likewise.
7664 * config/rs6000/vector.md: Likewise.
7666 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7669 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7672 2018-03-20 Richard Biener <rguenther@suse.de>
7675 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7676 sign-conversions as zero, fall back to standard scalar_stmt
7679 2018-03-20 Martin Liska <mliska@suse.cz>
7682 * predict.c (rebuild_frequencies): Handle case when we have
7683 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7685 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7688 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7689 flag_section_anchors.
7690 * varasm.c (use_blocks_for_decl_p): Remove hack for
7691 dw2_force_const_mem.
7694 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7696 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7697 be created, use lowpart_subreg of operands[0] rather than operands[0]
7699 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7700 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7701 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7702 and n constraint instead of aarch64_shift_imm_di and Usd.
7703 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7704 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7706 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7709 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7710 to favor GPR over NEON registers.
7711 (<shift>di3_neon): Likewise.
7713 2018-03-20 Tom de Vries <tom@codesourcery.com>
7716 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7717 (nvptx_process_pars): Emit bar.sync asap and alap.
7719 2018-03-20 Tom de Vries <tom@codesourcery.com>
7722 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7723 seen_label if seen_label is already set.
7725 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7728 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7729 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7730 Use 1U instead of 1. Formatting fixes.
7733 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7734 instead of TREE_TYPE (s1) for the return value.
7736 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7738 PR tree-optimization/84946
7739 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7740 bitsize + bitsize in poly_uint64 rather than poly_int64.
7743 * dwarf2asm.c: Include fold-const.c.
7744 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7745 of decl rather than decl itself.
7747 PR rtl-optimization/84643
7748 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7750 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7753 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7754 calling assemble_variable.
7756 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7759 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7760 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7762 2018-03-19 Jim Wilson <jimw@sifive.com>
7765 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7766 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7767 (riscv_first_stack_step): Likewise.
7768 (riscv_option_override): Use STACK_BOUNDARY instead of
7770 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7772 (BIGGEST_ALIGNMENT): Set to 128.
7773 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7774 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7777 2018-03-19 Richard Biener <rguenther@suse.de>
7779 PR tree-optimization/84933
7780 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7781 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7783 2018-03-19 Richard Biener <rguenther@suse.de>
7785 PR tree-optimization/84859
7786 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7787 (cond_if_else_store_replacement): Perform sinking operation on
7788 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7789 Generalize what a BB with a single eligible store is.
7791 2018-03-19 Richard Biener <rguenther@suse.de>
7793 PR tree-optimization/84929
7794 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7795 chrec_is_positive against non-chrec arg.
7797 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7800 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7802 2018-03-18 Martin Liska <mliska@suse.cz>
7804 PR rtl-optimization/84635
7805 * regrename.c (build_def_use): Use matches_mode only when
7808 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7810 PR tree-optimization/84913
7811 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7812 vectorize chains of COND_EXPRs.
7814 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7816 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7818 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7820 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7822 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7824 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7826 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7827 Kito Cheng <kito.cheng@gmail.com>
7829 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7830 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7831 (nds32_adjust_reg_alloc_order): New function.
7832 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7834 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7836 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7837 nds32_print_operand, nds32_print_operand_address): Use
7838 HOST_WIDE_INT_PRINT_DEC instead.
7840 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7842 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7844 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7847 * config/i386/i386.c (initial_ix86_tune_features,
7848 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7850 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7851 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7852 rather than 1u << ix86_tune. Formatting fix.
7853 (ix86_option_override_internal): Change ix86_arch_mask from
7854 unsigned int to unsigned HOST_WIDE_INT, initialize to
7855 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7856 (ix86_function_specific_restore): Likewise.
7858 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7861 * postreload.c (reload_combine_recognize_pattern): Perform
7862 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7863 truncate_int_for_mode the result for the destination's mode.
7866 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7867 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7869 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7872 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7875 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7876 regno_allocno_class_array and sorted_pseudos.
7877 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7878 insns where regno is used.
7880 2018-03-16 Martin Liska <mliska@suse.cz>
7883 * multiple_target.c (create_dispatcher_calls): Redirect
7884 reference in the symbol table.
7886 2018-03-16 Martin Liska <mliska@suse.cz>
7889 * multiple_target.c (create_dispatcher_calls): Redirect also
7892 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7896 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7897 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7898 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7900 2018-03-16 Julia Koval <julia.koval@intel.com>
7902 * doc/invoke.texi (Skylake Server): Add CLWB.
7903 Cannonlake): Remove CLWB.
7905 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7907 PR tree-optimization/84841
7908 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7910 (FLOAT_ONE_CONST_TYPE): Define.
7911 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7912 (sort_by_operand_rank): Put entries with higher constant_type last
7913 rather than first to match comments.
7915 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7917 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7920 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7923 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7924 check and use address of target always.
7926 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7929 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7930 (indirect_thunk_bnd_needed): Likewise.
7931 (indirect_thunks_used): Likewise.
7932 (indirect_thunks_bnd_used): Likewise.
7933 (indirect_return_needed): New.
7934 (indirect_return_bnd_needed): Likewise.
7935 (output_indirect_thunk_function): Add a bool argument for
7937 (output_indirect_thunk_function): Don't generate alias for
7938 function return thunk.
7939 (ix86_code_end): Call output_indirect_thunk_function to generate
7940 function return thunks.
7941 (ix86_output_function_return): Set indirect_return_bnd_needed
7942 and indirect_return_needed instead of indirect_thunk_bnd_needed
7943 and indirect_thunk_needed.
7945 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7947 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7948 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7949 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7951 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7952 Paul Hua <paul.hua.gm@gmail.com>
7955 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7957 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7959 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7960 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7963 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7966 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7967 instead of GET_MODE_SIZE when comparing Units.
7969 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7972 * varasm.c (hash_section): Return an unchangeble hash value
7973 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7974 Return !aarch64_can_use_per_function_literal_pools_p ().
7976 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7979 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7980 rather than address of cmode as last argument to prepare_cmp_insn.
7982 2018-03-15 Julia Koval <julia.koval@intel.com>
7984 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7985 F_AVX512VNNI, F_AVX512BITALG): New.
7987 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7990 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7991 insn for floating-point loads and stores.
7993 2018-03-14 Carl Love <cel@us.ibm.com>
7995 * config/rs6000/rs6000-c.c: Add macro definitions for
7996 ALTIVEC_BUILTIN_VEC_PERMXOR.
7997 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7998 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7999 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8000 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8002 * config/doc/extend.texi: Add prototypes for vec_permxor.
8004 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8007 * diagnostic-show-locus.c (class layout_point): Convert m_line
8008 from int to linenum_type.
8009 (line_span::comparator): Use linenum "compare" function when
8010 comparing line numbers.
8011 (test_line_span): New function.
8012 (layout_range::contains_point): Convert param "row" from int to
8014 (layout_range::intersects_line_p): Likewise.
8015 (layout::will_show_line_p): Likewise.
8016 (layout::print_source_line): Likewise.
8017 (layout::should_print_annotation_line_p): Likewise.
8018 (layout::print_annotation_line): Likewise.
8019 (layout::print_leading_fixits): Likewise.
8020 (layout::annotation_line_showed_range_p): Likewise.
8021 (struct line_corrections): Likewise for field m_row.
8022 (line_corrections::line_corrections): Likewise for param "row".
8023 (layout::print_trailing_fixits): Likewise.
8024 (layout::get_state_at_point): Likewise.
8025 (layout::get_x_bound_for_row): Likewise.
8026 (layout::print_line): Likewise.
8027 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8028 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8029 * input.c (selftest::test_linenum_comparisons): New function.
8030 (selftest::input_c_tests): Call it.
8031 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8032 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8033 * selftest.h (ASSERT_GT): New macro.
8034 (ASSERT_GT_AT): New macro.
8035 (ASSERT_LT): New macro.
8036 (ASSERT_LT_AT): New macro.
8038 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8040 PR rtl-optimization/84780
8041 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8043 2018-03-14 Martin Liska <mliska@suse.cz>
8045 * tree.c (record_node_allocation_statistics): Use
8046 get_stats_node_kind.
8047 (get_stats_node_kind): New function extracted from
8048 record_node_allocation_statistics.
8049 (free_node): Use get_stats_node_kind.
8051 2018-03-14 Richard Biener <rguenther@suse.de>
8053 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8054 that the value-set of ANTIC_IN doesn't grow.
8057 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8059 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8060 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8062 2018-03-14 Julia Koval <julia.koval@intel.com>
8064 * config.gcc (icelake-client, icelake-server): New.
8066 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8067 (initial_ix86_arch_features): Ditto.
8068 (PTA_SKYLAKE): Add SGX.
8069 (PTA_ICELAKE): Remove.
8070 (PTA_ICELAKE_CLIENT): New.
8071 (PTA_ICELAKE_SERVER): New.
8072 (ix86_option_override_internal): Split up icelake on icelake client and
8074 (get_builtin_code_for_version): Ditto.
8075 (fold_builtin_cpu): Ditto.
8076 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8077 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8078 * config/i386/i386.h (processor_type): Ditto.
8079 * doc/invoke.texi: Ditto.
8081 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8084 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8085 INTEGER_CST offset, add it together with bitpos / 8 and
8086 sign extend based on POINTER_SIZE.
8090 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8093 * config/i386/constraints.md (Yc): New register constraint.
8094 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8095 Use Yc constraint for alternative 2 of operand 0. Remove
8096 preferred_for_speed attribute.
8098 2018-03-14 Richard Biener <rguenther@suse.de>
8100 PR tree-optimization/84830
8101 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8102 with the old one to avoid oscillations.
8104 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8107 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8109 (assign_by_spills): Return a flag of reload assignment failure.
8110 Do not process the reload assignment failures. Do not spill other
8111 reload pseudos if they has the same reg class. Update n if
8113 (lra_assign): Add a return arg. Set up from the result of
8114 assign_by_spills call.
8115 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8116 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8117 usage_insns if it is not NULL.
8118 (spill_hard_reg_in_range): New function.
8119 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8120 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8121 function prototypes.
8122 (lra_assign): Change prototype.
8123 * lra.c (lra): Add code to deal with fails by splitting hard reg
8126 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8128 * config/riscv/riscv.opt (mrelax): New option.
8129 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8130 "norelax" when riscv_mrelax is disabled.
8131 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8133 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8136 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8137 reassociation for int modes.
8139 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8141 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8142 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8144 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8145 * config/aarch64/aarch64-sve.md
8146 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8147 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8148 (*extend<mode><Vwide>2): Rename to...
8149 (aarch64_sve_extend<mode><Vwide>2): ...this.
8150 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8151 renaming the old pattern to...
8152 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8154 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8155 define_expand, renaming the old pattern to...
8156 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8157 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8158 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8159 account when deciding which SVE instruction the optab should use.
8160 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8162 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8164 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8165 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8166 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8167 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8168 (tlsdesc_small_<mode>): Turn a define_expand and use
8169 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8170 (tlsdesc_small_advsimd_<mode>): ...this.
8171 (tlsdesc_small_sve_<mode>): New pattern.
8173 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8175 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8176 (UNSPEC_UMUL_HIGHPART): New constants.
8177 (MUL_HIGHPART): New int iteraor.
8178 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8179 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8181 (*<su>mul<mode>3_highpart): New define_insn.
8183 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8186 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8189 2018-03-13 Martin Liska <mliska@suse.cz>
8192 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8194 (sem_item_optimizer::~sem_item_optimizer): Release it.
8195 (sem_item_optimizer::merge_classes): Register variable aliases.
8196 (sem_item_optimizer::fixup_pt_set): New function.
8197 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8198 * ipa-icf.h: Declare new variables and functions.
8200 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8203 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8204 integer_pow2p@2 and test integer_pow2p in condition.
8205 (A < 0 ? C : 0): Similarly for @1.
8208 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8209 characters starting at p contain '\0' character, don't look beyond
8213 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8214 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8217 * reg-stack.c (change_stack): Change update_end var from int to
8218 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8219 also call set_block_for_insn on the newly added insns and rescan.
8222 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8223 on the last operand.
8226 * tree.c (stabilize_reference_1): Return save_expr (e) for
8227 STATEMENT_LIST even if it doesn't have side-effects.
8229 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8231 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8233 2018-03-12 Renlin Li <renlin.li@arm.com>
8235 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8236 aarch64_output_scalar_simd_mov_immediate.
8238 2018-03-12 Martin Sebor <msebor@redhat.com>
8240 PR tree-optimization/83456
8241 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8242 for perfectly overlapping calls to memcpy.
8243 (gimple_fold_builtin_memory_chk): Same.
8244 (gimple_fold_builtin_strcpy): Handle no-warning.
8245 (gimple_fold_builtin_stxcpy_chk): Same.
8246 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8248 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8250 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8251 parameter. Use it for SFmode.
8252 (rs6000_function_arg_advance_1): Adjust.
8253 (rs6000_function_arg): Adjust.
8254 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8256 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8258 PR rtl-optimization/84169
8259 PR rtl-optimization/84780
8260 * combine.c (can_combine_p): Check for a 2-insn combination whether
8261 the destination register is used between the two insns, too.
8263 2018-03-12 Richard Biener <rguenther@suse.de>
8265 PR tree-optimization/84803
8266 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8267 for refs DR analysis didn't process.
8269 2018-03-12 Richard Biener <rguenther@suse.de>
8271 PR tree-optimization/84777
8272 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8273 force-vectorize loops ignore whether we are optimizing for size.
8275 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8277 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8278 (TARGET_MD_ASM_ADJUST): Define.
8280 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8281 Kito Cheng <kito.cheng@gmail.com>
8282 Chung-Ju Wu <jasonwucj@gmail.com>
8284 * config/nds32/nds32.c (nds32_compute_stack_frame,
8285 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8286 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8287 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8288 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8289 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8290 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8291 * config/nds32/nds32.md (prologue, epilogue): Use macro
8292 NDS32_V3PUSH_AVAILABLE_P to do checking.
8294 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8297 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8298 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8299 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8300 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8301 addition of most attributes on !orig_type_die or the attribute not
8302 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8304 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8305 Chung-Ju Wu <jasonwucj@gmail.com>
8307 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8309 * config/nds32/nds32.opt (mvh): New option.
8311 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8312 Chung-Ju Wu <jasonwucj@gmail.com>
8314 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8316 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8317 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8320 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8321 Chung-Ju Wu <jasonwucj@gmail.com>
8323 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8325 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8326 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8328 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8329 Kito Cheng <kito.cheng@gmail.com>
8330 Chung-Ju Wu <jasonwucj@gmail.com>
8332 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8333 UNSPEC_FFMISM and UNSPEC_FLMISM.
8334 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8335 for ffb, ffmism and flmism.
8336 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8337 (unspec_ffmism): Ditto.
8338 (unspec_flmism): Ditto.
8339 (nds32_expand_builtin_impl): Check if string extension is available.
8340 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8341 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8343 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8346 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8349 * lra-assigns.c (assign_by_spills): Return a flag of reload
8350 assignment failure. Do not process the reload assignment
8351 failures. Do not spill other reload pseudos if they has the same
8353 (lra_assign): Add a return arg. Set up from the result of
8354 assign_by_spills call.
8355 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8356 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8357 usage_insns if it is not NULL.
8358 (spill_hard_reg_in_range): New function.
8359 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8360 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8361 function prototypes.
8362 (lra_assign): Change prototype.
8363 * lra.c (lra): Add code to deal with fails by splitting hard reg
8366 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8369 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8371 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8374 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8375 (dw_val_node): Add val_symbolic_view.
8376 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8377 (symview_upper_bound): New.
8378 (new_line_info_table): Initialize symviews_since_reset.
8379 (dwarf2out_source_line): Count symviews_since_reset and set
8380 symview_upper_bound.
8381 (dw_val_equal_p): Handle symview.
8382 (add_AT_symview): New.
8383 (print_dw_val): Handle symview.
8384 (attr_checksum, attr_checksum_ordered): Likewise.
8385 (same_dw_val_p, size_of_die): Likewise.
8386 (value_format, output_die): Likewise.
8387 (add_high_low_attributes): Use add_AT_symview for entry_view.
8388 (dwarf2out_finish): Reset symview_upper_bound, clear
8391 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8394 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8395 Add strict argument and use it.
8396 (rs6000_split_multireg_move): Update for new strict argument.
8397 (mem_operand_gpr): Disallow all non-offsettable addresses.
8398 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8400 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8403 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8404 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8405 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8408 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8409 decl, use remap_type if we want to use the type.
8411 2018-03-09 Martin Sebor <msebor@redhat.com>
8413 PR tree-optimization/84526
8414 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8416 (builtin_access::generic_overlap): Be prepared to handle non-array
8419 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8421 PR rtl-optimization/84682
8422 * lra-constraints.c (process_address_1): Check is_address flag
8423 for address constraints.
8424 (process_alt_operands): Likewise.
8425 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8426 preprocess_constraints.
8427 * recog.h (preprocess_constraints): Add oploc parameter.
8429 * recog.c (preprocess_constraints): Test address_operand for
8430 CT_ADDRESS constraints.
8432 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8435 * lra-assigns.c (assign_by_spills): Return a flag of reload
8436 assignment failure. Do not process the reload assignment
8437 failures. Do not spill other reload pseudos if they has the same
8439 (lra_assign): Add a return arg. Set up from the result of
8440 assign_by_spills call.
8441 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8442 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8443 usage_insns if it is not NULL.
8444 (spill_hard_reg_in_range): New function.
8445 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8446 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8447 function prototypes.
8448 (lra_assign): Change prototype.
8449 * lra.c (lra): Add code to deal with fails by splitting hard reg
8452 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8455 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8456 Accept complain bool parameter. Only emit errors if it is true.
8457 (arm_parse_cpu_option_name): Likewise.
8458 (arm_target_thumb_only): Adjust callers of the above.
8459 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8460 prototype to take a default true bool parameter.
8461 (arm_parse_arch_option_name): Likewise.
8463 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8464 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8468 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8469 * configure: Regenerate.
8470 * configure.ac ("linker --version-script option"): New.
8471 ("linker soname option"): New.
8473 2018-03-09 Richard Biener <rguenther@suse.de>
8475 PR tree-optimization/84775
8476 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8477 immediate uses of predicate stmts and mark them modified.
8480 PR tree-optimization/84178
8481 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8483 (version_loop_for_if_conversion): Delay update_ssa call.
8484 (tree_if_conversion): Delay update_ssa until after predicate
8487 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8490 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8491 when the function accesses prior frames.
8493 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8496 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8497 gen_llsym, otherwise call maybe_gen_llsym.
8500 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8501 has ',' character inside of it.
8503 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8506 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8507 as clobbering CC_REGNUM.
8509 2018-03-08 Richard Biener <rguenther@suse.de>
8512 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8513 (follow_copies_to_constant): Do not follow SSA names registered
8516 2018-03-08 Richard Biener <rguenther@suse.de>
8518 PR tree-optimization/84178
8519 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8521 (version_loop_for_if_conversion): Delay update_ssa call.
8522 (tree_if_conversion): Delay update_ssa until after predicate
8525 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8527 PR tree-optimization/84178
8528 * tree-if-conv.c (release_bb_predicate): Remove the
8529 the assertion that the stmts have NULL use_ops.
8530 Discard the statements, asserting that they haven't
8531 yet been added to a BB.
8533 2018-03-08 Richard Biener <rguenther@suse.de>
8535 PR tree-optimization/84746
8536 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8537 (phi_translate): Pass in destination ANTIC_OUT set.
8538 (phi_translate_1): Likewise. For a simplified result lookup
8539 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8540 (phi_translate_set): Adjust.
8541 (do_pre_regular_insertion): Likewise.
8542 (do_pre_partial_partial_insertion): Likewise.
8544 2018-03-08 Martin Liska <mliska@suse.cz>
8546 PR gcov-profile/84735
8547 * doc/gcov.texi: Document usage of profile files.
8548 * gcov-io.h: Document changes in the format.
8550 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8554 * dwarf2out.c (struct dw_line_info_table): Update comments for
8556 (FORCE_RESET_NEXT_VIEW): New.
8557 (FORCE_RESETTING_VIEW_P): New.
8558 (RESETTING_VIEW_P): Check for -1 too.
8559 (ZERO_VIEW_P): Likewise.
8560 (new_line_info_table): Force-reset next view.
8561 (dwarf2out_begin_function): Likewise.
8562 (dwarf2out_source_line): Simplify zero_view_p initialization.
8563 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8564 view directly. Omit view when omitting .loc at line 0.
8566 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8568 PR tree-optimization/84740
8569 * tree-switch-conversion.c (process_switch): Call build_constructors
8570 only if info.phi_count is non-zero.
8572 PR tree-optimization/84739
8573 * tree-tailcall.c (find_tail_calls): Check call arguments against
8574 DECL_ARGUMENTS (current_function_decl) rather than
8575 DECL_ARGUMENTS (func) when checking for tail recursion.
8577 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8579 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8580 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8581 Volker Reichelt's entry and add entries for people that perform
8582 GCC fuzzy testing and report numerous bugs.
8584 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8587 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8588 readonly data in sdata, if that is disabled.
8589 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8590 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8591 -mreadonly-in-sdata option.
8593 2018-03-07 Martin Sebor <msebor@redhat.com>
8595 PR tree-optimization/84468
8596 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8597 basic block when looking for nul assignment.
8599 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8602 * except.h (output_function_exception_table): Adjust prototype.
8603 * except.c (output_function_exception_table): Remove FNNAME parameter
8604 and add SECTION parameter. Ouput one part of the table at a time.
8605 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8606 the first part of the exception table and emit unwind directives.
8607 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8608 (i386_pe_seh_cold_init): Likewise.
8609 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8610 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8611 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8612 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8613 * config/i386/winnt.c: Include except.h.
8614 (struct seh_frame_state): Add reg_offset, after_prologue and
8615 in_cold_section fields.
8616 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8617 (i386_pe_seh_cold_init): New function.
8618 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8619 to seh->in_cold_section.
8620 (seh_emit_push): Record the offset of the push.
8621 (seh_emit_save): Record the offet of the save.
8622 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8623 Test seh->after_prologue to disregard the epilogue.
8624 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8625 (i386_pe_end_cold_function): New function.
8627 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8630 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8631 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8634 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8636 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8637 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8640 * multiple_target.c: Include tree-inline.h and intl.h.
8641 (expand_target_clones): Diagnose and fail if node->definition and
8642 !tree_versionable_function_p (node->decl).
8644 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8646 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8648 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8649 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8650 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8652 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8655 * combine.c (try_combine): Use reg_or_subregno instead of handling
8656 just paradoxical SUBREGs and REGs.
8658 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8660 * config/arc/arc.c (arc_finalize_pic): Remove function.
8661 (arc_must_save_register): We use single base PIC register, remove
8662 checks to save/restore the PIC register.
8663 (arc_expand_prologue): Likewise.
8664 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8666 (arc_verify_short): Likewise.
8667 (arc_attr_type): Likewise.
8668 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8669 (walk_stores): Likewise.
8670 (arc_address_cost): Make it static.
8671 (arc_verify_short): Likewise.
8672 (branch_dest): Likewise.
8673 (arc_attr_type): Likewise.
8674 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8675 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8676 (arc_final_prescan_insn): Remove inserting the nops due to
8677 hardware hazards. It is done in reorg step.
8678 (insn_length_variant_t): Remove.
8679 (insn_length_parameters_t): Likewise.
8680 (arc_insn_length_parameters): Likewise.
8681 (arc_get_insn_variants): Likewise.
8682 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8684 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8687 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8690 PR tree-optimization/84687
8691 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8693 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8695 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8697 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8698 Rename to ppc_speculation_barrier.
8699 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8700 __builtin_ppc_speculation_barrier.
8702 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8705 * combine.c (combine_simplify_rtx): Don't try to simplify if
8706 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8709 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8711 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8712 to 32 bytes when compiling for POWER9.
8714 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8717 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8718 regparm >= 3 with no arg reg available also for calls with
8719 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8722 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8724 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8726 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8729 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8731 2018-03-05 Richard Biener <rguenther@suse.de>
8733 PR tree-optimization/84486
8734 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8735 When inserting a __builtin_assume_aligned call set the LHS
8736 SSA name alignment info accordingly.
8738 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8740 PR tree-optimization/84114
8741 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8742 Avoid reassociation of FLOAT_MODE addition.
8744 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8746 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8747 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8748 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8749 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8750 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8751 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8752 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8754 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8755 __builtin_ia32_wbinvd): New builtins.
8756 (SPECIAL_ARGS2): New.
8757 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8758 (SPECIAL_ARGS2): New.
8759 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8760 (ix86_valid_target_attribute_inner_p): Ditto.
8761 (ix86_init_mmx_sse_builtins): Add special_args2.
8762 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8763 TARGET_WBNOINVD_P): New.
8764 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8765 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8766 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8767 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8768 * config/i386/pconfigintrin.h: New file.
8769 * config/i386/wbnoinvdintrin.h: Ditto.
8770 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8772 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8774 2018-03-05 Richard Biener <rguenther@suse.de>
8776 PR tree-optimization/84670
8777 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8779 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8780 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8781 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8782 doesn't grow if all successors (recursively) were visited at least
8785 2018-03-05 Richard Biener <rguenther@suse.de>
8787 PR tree-optimization/84650
8788 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8789 if executed in the loop pipeline.
8791 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8793 * doc/configfiles.texi (Configuration Files): Move info about
8794 conditionalizing $target-protos.h to...
8795 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8796 differs from $target-protos.h.
8798 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8799 Chung-Ju Wu <jasonwucj@gmail.com>
8801 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8802 * config/nds32/nds32-multiple.md (setmemsi): Define.
8803 * config/nds32/nds32-memory-manipulation.c
8804 (nds32_gen_dup_4_byte_to_word_value): New.
8805 (emit_setmem_word_loop): New.
8806 (emit_setmem_byte_loop): New.
8807 (nds32_expand_setmem_loop): New.
8808 (nds32_expand_setmem_loop_v3m): New.
8809 (nds32_expand_setmem_unroll): New.
8810 (nds32_expand_setmem): New.
8812 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8813 Chung-Ju Wu <jasonwucj@gmail.com>
8815 * config/nds32/nds32-memory-manipulation.c
8816 (nds32_emit_load_store): New.
8817 (nds32_emit_post_inc_load_store): New.
8818 (nds32_emit_mem_move): New.
8819 (nds32_emit_mem_move_block): New.
8820 (nds32_expand_movmemsi_loop_unknown_size): New.
8821 (nds32_expand_movmemsi_loop_known_size): New.
8822 (nds32_expand_movmemsi_loop): New.
8823 (nds32_expand_movmemsi_unroll): New.
8824 (nds32_expand_movmemqi): Rename ...
8825 (nds32_expand_movmemsi): ... to this.
8826 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8827 (movmemsi): ... to this.
8828 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8829 (nds32_expand_movmemsi): ... to this.
8831 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8832 Monk Chiang <sh.chiang04@gmail.com>
8833 Chung-Ju Wu <jasonwucj@gmail.com>
8835 * config/nds32/nds32-protos.h
8836 (nds32_expand_load_multiple): New arguments.
8837 (nds32_expand_store_multiple): Ditto.
8838 (nds32_valid_multiple_load_store): Rename ...
8839 (nds32_valid_multiple_load_store_p): ... to this.
8840 * config/nds32/nds32-memory-manipulation.c
8841 (nds32_expand_load_multiple): Refine implementation.
8842 (nds32_expand_store_multiple): Ditto.
8843 * config/nds32/nds32-multiple.md
8844 (load_multiple): Update nds32_expand_load_multiple interface.
8845 (store_multiple): Update nds32_expand_store_multiple interface.
8846 * config/nds32/nds32-predicates.c
8847 (nds32_valid_multiple_load_store): Rename ...
8848 (nds32_valid_multiple_load_store_p): ... to this and refine
8850 * config/nds32/predicates.md
8851 (nds32_load_multiple_and_update_address_operation): New predicate.
8852 (nds32_store_multiple_and_update_address_operation): New predicate.
8854 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8855 Chung-Ju Wu <jasonwucj@gmail.com>
8857 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8858 (combo): New attribute.
8859 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8861 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8863 * config/nds32/nds32.opt: Change -mcmodel= default value.
8865 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8866 Monk Chiang <sh.chiang04@gmail.com>
8867 Chung-Ju Wu <jasonwucj@gmail.com>
8869 * config/nds32/constants.md (unspec_element): New enum.
8870 * config/nds32/constraints.md (Umw): New constraint.
8871 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8872 * config/nds32/nds32-intrinsic.md: Likewise.
8873 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8874 (nds32_valid_smw_lwm_base_p): New.
8875 (nds32_output_smw_single_word): New.
8876 (nds32_output_lmw_single_word): New.
8877 (nds32_expand_unaligned_load): New.
8878 (nds32_expand_unaligned_store): New.
8879 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8880 (nds32_output_smw_single_word): Declare.
8881 (nds32_output_lmw_single_word): Declare.
8882 (nds32_expand_unaligned_load): Declare.
8883 (nds32_expand_unaligned_store): Declare.
8884 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8885 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8886 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8887 NDS32_BUILTIN_UASTORE_DW.
8888 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8891 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8892 Kito Cheng <kito.cheng@gmail.com>
8893 Chung-Ju Wu <jasonwucj@gmail.com>
8895 * config/nds32/nds32-intrinsic.c
8896 (nds32_expand_builtin_null_ftype_reg): Delete.
8897 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8898 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8899 (nds32_read_argument): New.
8900 (nds32_legitimize_target): Ditto.
8901 (nds32_legitimize_argument): Ditto.
8902 (nds32_check_constant_argument): Ditto.
8903 (nds32_expand_unop_builtin): Ditto.
8904 (nds32_expand_unopimm_builtin): Ditto.
8905 (nds32_expand_binop_builtin): Ditto.
8906 (nds32_builtin_decl_impl): Ditto.
8907 (builtin_description): Ditto.
8908 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8909 (nds32_init_builtins_impl): Ditto.
8910 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8911 (nds32_builtin_decl): New.
8912 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8913 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8915 2018-03-02 Jeff Law <law@redhat.com>
8917 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8918 (redundant_insn, fill_simple_delay_slots): Likewise.
8919 (fill_slots_from_thread): Likewise.
8920 * resource.c (mark_referenced_resources): Likewise.
8921 (mark_set_resources, find_dead_or_set_registers): Likewise.
8923 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8925 * substring-locations.h (format_warning_va): Formatting fix for
8927 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8929 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8930 * substring-locations.c: Include intl.h.
8931 (format_warning_va): Turned into small wrapper around
8932 format_warning_n_va, renamed to ...
8933 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8934 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8936 (format_warning_at_substring_n): New function.
8937 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8938 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8939 format_warning_at_substring with just a shorter name instead of
8940 const function pointer.
8941 (fmtwarn_n): New function.
8942 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8943 appropriate, get rid of all the fmtstr temporaries, move conditionals
8944 with G_() wrapped string literals directly into fmtwarn arguments,
8945 cast dir.len to (int), formatting fixes.
8947 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8949 * doc/invoke.texi: Remove "Cilk Plus" references.
8951 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8952 Richard Biener <rguenther@suse.de>
8955 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8956 for error or warning attributes if CALL_FROM_THUNK_P is set.
8959 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8962 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8963 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8966 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8967 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8969 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8970 instead of -1U in last predictors element's probability member.
8972 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8975 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8976 arguments if they are comparable.
8978 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8980 PR tree-optimization/84634
8981 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8982 masks and masked_loop_p with a single loop_masks, making sure it's
8983 null for bb vectorization.
8985 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8987 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8988 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8989 loop->force_vectorize to check whether there is no alias.
8991 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8994 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8996 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8998 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8999 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9000 instead of a loop around prev_real_insn.
9001 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9005 * config/i386/i386.c (ix86_print_operand): Use conditional
9006 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9009 2018-03-02 Richard Biener <rguenther@suse.de>
9011 PR tree-optimization/84427
9012 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9013 (bitmap_set_subtract_values): Rewrite to handle multiple
9016 (prune_clobbered_mems): Likewise.
9017 (phi_translate): Take edge instead of pred/phiblock.
9018 (phi_translate_1): Likewise.
9019 (phi_translate_set): Likewise. Insert all translated
9020 exprs for a value into the set, keeping possibly multiple
9021 expressions per value.
9022 (compute_antic_aux): Adjust for phi_translate changes.
9023 When intersecting union the expressions and prune those
9024 not in the final value set, keeping possibly multiple
9025 expressions per value. Do not use value-insertion
9026 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9027 all expressions. Add verification that the value-sets
9028 only shrink during iteration.
9029 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9030 (do_pre_regular_insertion): Likewise.
9031 (do_pre_partial_partial_insertion): Likewise.
9033 2018-03-02 Richard Biener <rguenther@suse.de>
9036 * config/darwin.c (saved_debug_info_level): New static global.
9037 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9038 (darwin_asm_lto_end): Restore debug info generation settings.
9040 2018-03-01 Martin Liska <mliska@suse.cz>
9043 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9046 2018-03-01 Richard Biener <rguenther@suse.de>
9049 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9050 type annotation with LTO when debug was disabled at compile-time.
9052 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9054 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9056 (mips_final_postscan_insn): Likewise.
9058 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9060 PR rtl-optimization/84528
9061 * alias.c (init_alias_target): Add commentary.
9062 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9063 a unique base value if the frame pointer is not eliminated
9064 to the stack pointer.
9066 2018-03-01 Tom de Vries <tom@codesourcery.com>
9068 PR rtl-optimization/83327
9069 * lra-int.h (hard_regs_spilled_into): Declare.
9070 * lra.c (hard_regs_spilled_into): Define.
9071 (init_reg_info): Init hard_regs_spilled_into.
9072 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9073 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9074 (process_bb_lives): Handle hard_regs_spilled_into.
9075 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9076 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9078 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9080 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9081 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9082 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9083 * config/rs6000/aix72.h: New file.
9085 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9087 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9088 instead of warning_at with conditional singular and plural messages
9092 * stor-layout.c (update_alignment_for_field): For
9093 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9094 && !DECL_PACKED (field), do the alignment update, just use
9095 only desired_align instead of MAX (type_align, desired_align)
9097 (place_field): Don't do known_align < desired_align handling
9098 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9099 is non-NULL, instead do it after rli->prev_field handling and
9100 only if not within a bitfield word. For DECL_PACKED (field)
9101 use type_align of BITS_PER_UNIT.
9103 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9105 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9106 superfluous parentheses and trailing spaces.
9108 2018-02-28 Richard Biener <rguenther@suse.de>
9110 PR tree-optimization/84584
9111 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9112 SCoPs with fake exit edge.
9114 2018-02-28 Martin Liska <mliska@suse.cz>
9117 * timevar.c (timer::print): Fix format to properly print 100%
9120 2018-02-28 Richard Biener <rguenther@suse.de>
9123 * genmatch.c (capture_info::walk_match): Do not mark
9124 captured expressions without operands as expr_p given
9125 they act more like predicates and should be subject to
9126 "lost tail" side-effect preserving.
9128 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9130 PR rtl-optimization/81611
9131 * auto-inc-dec.c (attempt_change): Move dead note from
9132 mem_insn if it's the next use of regno
9133 (find_address): Take address use of reg holding
9134 non-incremented value. Add parm to limit search to the named
9136 (merge_in_block): Attempt to use a mem insn that is the next
9137 use of the original regno.
9139 2018-02-27 Martin Sebor <msebor@redhat.com>
9142 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9143 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9145 2018-02-27 Martin Sebor <msebor@redhat.com>
9147 PR translation/84207
9148 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9149 n argument to unsigned HOST_WIDE_INT.
9150 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9151 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9152 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9153 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9155 2018-02-27 Richard Biener <rguenther@suse.de>
9157 PR tree-optimization/84512
9158 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9159 Do not use the estimate returned from record_stmt_cost for
9160 the scalar iteration cost but sum properly using add_stmt_cost.
9162 2018-02-27 Richard Biener <rguenther@suse.de>
9164 PR tree-optimization/84466
9165 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9166 Adjust last change to less strictly validate use operands.
9168 2018-02-27 Martin Liska <mliska@suse.cz>
9170 PR gcov-profile/84548
9171 * gcov.c (process_file): Allow partial overlap and consider it
9172 also as group functions.
9173 (output_lines): Properly calculate range of lines for a group.
9175 2018-02-27 Martin Liska <mliska@suse.cz>
9177 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9178 'ggc' suffixes. Change first column width.
9179 (timer::print): Fix formatting of the column.
9181 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9183 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9184 preserve inline entry blocks for the sake of debug inline
9185 entry point markers alone.
9186 (remove_unused_locals): Suggest in comments a better place to
9187 force the preservation of inline entry blocks that are
9188 otherwise unused, but do not preserve them.
9190 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9192 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9194 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9197 * config/i386/constraints.md (Bs): Replace
9198 ix86_indirect_branch_register with
9199 TARGET_INDIRECT_BRANCH_REGISTER.
9201 * config/i386/i386.md (indirect_jump): Likewise.
9202 (tablejump): Likewise.
9203 (*sibcall_memory): Likewise.
9204 (*sibcall_value_memory): Likewise.
9205 Peepholes of indirect call and jump via memory: Likewise.
9206 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9207 (*sibcall_value_GOT_32): Likewise.
9208 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9209 (GOT_memory_operand): Likewise.
9210 (call_insn_operand): Likewise.
9211 (sibcall_insn_operand): Likewise.
9212 (GOT32_symbol_operand): Likewise.
9213 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9215 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9217 PR rtl-optimization/83496
9218 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9219 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9220 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9221 redundant insn, if any.
9222 (relax_delay_slots): Likewise.
9223 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9225 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9227 PR tree-optimization/83965
9228 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9229 that grouped statements are part of a reduction chain. Return
9230 true if the statement is not marked as a reduction itself but
9232 (vect_recog_dot_prod_pattern): Don't check whether the statement
9233 is part of a group here.
9234 (vect_recog_sad_pattern): Likewise.
9235 (vect_recog_widen_sum_pattern): Likewise.
9237 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9240 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9242 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9245 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9247 (ix86_output_indirect_function_return): New prototype.
9248 (ix86_split_simple_return_pop_internal): Likewise.
9249 * config/i386/i386.c (indirect_return_via_cx): New.
9250 (indirect_return_via_cx_bnd): Likewise.
9251 (indirect_thunk_name): Handle return va CX_REG.
9252 (output_indirect_thunk_function): Create alias for
9253 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9254 (ix86_output_indirect_jmp): Remove the bool argument.
9255 (ix86_output_indirect_function_return): New function.
9256 (ix86_split_simple_return_pop_internal): Likewise.
9257 * config/i386/i386.md (*indirect_jump): Don't pass false
9258 to ix86_output_indirect_jmp.
9259 (*tablejump_1): Likewise.
9260 (simple_return_pop_internal): Change it to define_insn_and_split.
9261 Call ix86_split_simple_return_pop_internal to split it for
9263 (simple_return_indirect_internal): Call
9264 ix86_output_indirect_function_return instead of
9265 ix86_output_indirect_jmp.
9267 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9270 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9271 memset and value initialization afterwards.
9273 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9275 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9277 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9280 * common/config/aarch64/aarch64-common.c
9281 (aarch_option_optimization_table[]): Switch
9282 off fomit-frame-pointer
9284 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9285 Chung-Ju Wu <jasonwucj@gmail.com>
9287 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9289 (store_multiple): Ditto.
9291 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9293 * config.gcc: Add --with-cpu support for nds32 target.
9294 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9295 * config/nds32/nds32.opt: Add -mcpu= option.
9297 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9299 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9300 isel=yes): Warn for these deprecated options.
9302 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9304 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9305 ISA_2_5_MASKS_EMBEDDED.
9307 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9309 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9310 p->max as pointers rather than using iterative_hash_expr.
9312 2018-02-23 Carl Love <cel@us.ibm.com>
9314 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9315 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9317 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9318 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9319 P8V_BUILTIN_VEC_VUNSIGNED2.
9321 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9324 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9325 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9327 * lra-constraints.c (curr_insn_transform): Set up
9328 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9330 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9331 * lra-eliminations.c (spill_pseudos): Ditto.
9332 (process_insn_for_elimination): Ditto.
9333 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9334 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9335 LRA_NON_CLOBBERED_ALT.
9337 2018-02-22 Martin Sebor <msebor@redhat.com>
9339 PR tree-optimization/84480
9340 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9341 to maybe_diag_stxncpy_trunc. Call it.
9342 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9343 from gimple_fold_builtin_strcpy. Print inlining stack.
9344 (handle_builtin_stxncpy): Print inlining stack.
9345 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9347 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9350 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9351 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9352 and -fcheck-pointer-bounds are used together.
9353 (indirect_thunk_prefix): New enum.
9354 (indirect_thunk_need_prefix): New function.
9355 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9356 "_nt" instead of "_bnd" for NOTRACK prefix.
9357 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9358 (output_indirect_thunk_function): Likewise.
9360 (ix86_code_end): Update output_indirect_thunk_function calls.
9361 (ix86_output_indirect_branch_via_reg): Replace
9362 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9363 (ix86_output_indirect_branch_via_push): Likewise.
9364 (ix86_output_function_return): Likewise.
9365 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9366 incompatible with -fcf-protection=branch and
9367 -fcheck-pointer-bounds.
9369 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9372 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9373 Change gcc_assert call to output_operand_lossage.
9375 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9377 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9379 2018-02-22 DJ Delorie <dj@redhat.com>
9380 Sebastian Perta <sebastian.perta@renesas.com>
9381 Oleg Endo <olegendo@gcc.gnu.org>
9383 * config/rx/rx.c (rx_rtx_costs): New function.
9384 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9386 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9388 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9390 2018-02-22 Martin Liska <mliska@suse.cz>
9393 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9394 Add "native" as a possible value.
9396 2018-02-22 Martin Liska <mliska@suse.cz>
9399 * config/i386/i386.c (ix86_option_override_internal):
9400 Add "native" as a possible value for -march and -mtune.
9402 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9405 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9406 to all type variants.
9408 PR tree-optimization/84503
9409 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9410 width as info->bitpos + info->bitsize - start.
9411 (merged_store_group::merge_overlapping): Simplify width computation.
9412 (check_no_overlap): New function.
9413 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9414 start + width and last_order of the group, fail if check_no_overlap
9416 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9417 to group if check_no_overlap fails.
9419 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9421 * config/rs6000/altivec.md: Delete contraint arguments to
9422 define_expand, define_split, and define_peephole2, and in
9423 define_insn_and_split if always unused.
9424 * config/rs6000/darwin.md: Ditto.
9425 * config/rs6000/dfp.md: Ditto.
9426 * config/rs6000/rs6000.md: Ditto.
9427 * config/rs6000/sync.md: Ditto.
9428 * config/rs6000/vector.md: Ditto.
9429 * config/rs6000/vsx.md: Ditto.
9431 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9433 * config/rs6000/altivec.md: Write output control strings as braced
9434 blocks instead of double-quoted strings.
9435 * config/rs6000/darwin.md: Ditto.
9436 * config/rs6000/rs6000.md: Ditto.
9437 * config/rs6000/vector.md: Ditto.
9438 * config/rs6000/vsx.md: Ditto.
9440 2018-02-21 Jason Merrill <jason@redhat.com>
9442 PR c++/84314 - ICE with templates and fastcall attribute.
9443 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9445 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9447 * ipa-cp.c (determine_versionability): Fix comment typos.
9449 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9452 * ipa-cp.c (determine_versionability): Do not version functions caling
9455 2018-02-21 Martin Liska <mliska@suse.cz>
9458 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9459 Add "native" as a possible value.
9460 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9461 the macro when native cpu detection is available.
9463 2018-02-21 Martin Liska <mliska@suse.cz>
9466 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9467 Add "native" as a possible value.
9468 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9469 when native cpu detection is available.
9471 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9472 Martin Sebor <msebor@redhat.com>
9474 PR tree-optimization/84478
9475 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9477 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9478 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9479 support which is conservatively correct, for 2 only stay conservative
9480 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9481 argument to the 2 argument get_range_strlen, adjust 6 arg
9482 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9484 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9485 (gimple_fold_builtin_strlen): Pass true as last argument to
9488 2018-02-20 Martin Sebor <msebor@redhat.com>
9491 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9492 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9493 (builtin_memref::builtin_memref): Factor out parts into
9494 set_base_and_offset and call it.
9496 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9499 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9500 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9501 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9502 search at the associated MODE_INT.
9504 2018-02-20 Jeff Law <law@redhat.com>
9507 PR tree-optimization/81592
9509 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9510 for range data rather than using global data.
9511 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9512 range data rather than using global data.
9513 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9514 pass it to children as needed.
9515 (struct directive::fmtresult): Similarly.
9516 (struct directive::set_width): Similarly.
9517 (struct directive::set_precision): Similarly.
9518 (format_integer, format_directive, parse_directive): Similarly.
9519 (format_none): Accept unnamed vr_values parameter.
9520 (format_percent, format_floating, format_character): Similarly.
9521 (format_string, format_plain): Similarly.
9522 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9523 the EVRP range analyzer for range data rather than using global data.
9524 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9525 gimple-ssa-evrp-analyze.h
9526 (class sprintf_dom_walker): Add after_dom_children member function.
9527 Add evrp_range_analyzer member.
9528 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9529 range analyzer as needed.
9530 (sprintf_dom_walker::after_dom_children): New member function.
9531 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9533 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9534 (evrp_range_analyzer::pop_to_marker): Likewise.
9536 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9538 PR tree-optimization/84419
9539 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9540 with the required type if its current type is compatible but
9543 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9546 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9547 after vectorization.
9549 2018-02-20 Martin Liska <mliska@suse.cz>
9552 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9553 possible values if we don't have a hint.
9555 2018-02-20 Martin Liska <mliska@suse.cz>
9559 * final.c (shorten_branches): Build align_tab array with one
9561 * opts.c (finish_options): Add alignment option limit check.
9562 (MAX_CODE_ALIGN): Likewise.
9563 (MAX_CODE_ALIGN_VALUE): Likewise.
9564 * doc/invoke.texi: Document maximum allowed option value for
9565 all -falign-* options.
9567 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9570 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9571 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9572 * var-tracking.c (emit_note_insn_var_location): Remove all references
9573 to NOTE_INSN_CALL_ARG_LOCATION.
9574 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9575 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9576 Use copy_rtx_if_shared.
9577 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9578 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9579 (dwarf2out_var_location): Remove handling of
9580 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9582 * final.c (final_scan_insn): Remove all references to
9583 NOTE_INSN_CALL_ARG_LOCATION.
9584 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9585 before dumping final insns.
9586 * except.c (emit_note_eh_region_end): Remove all references to
9587 NOTE_INSN_CALL_ARG_LOCATION.
9588 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9589 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9590 * config/arc/arc.c (hwloop_optimize): Likewise.
9591 * config/arm/arm.c (create_fix_barrier): Likewise.
9592 * config/s390/s390.c (s390_chunkify_start): Likewise.
9593 * config/sh/sh.c (find_barrier): Likewise.
9594 * config/i386/i386.c (rest_of_insert_endbranch,
9595 ix86_seh_fixup_eh_fallthru): Likewise.
9596 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9597 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9598 * config/frv/frv.c (frv_function_prologue): Likewise.
9599 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9601 (note_outside_basic_block_p): Remove all references to
9602 NOTE_INSN_CALL_ARG_LOCATION.
9603 * gengtype.c (adjust_field_rtx_def): Likewise.
9604 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9606 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9607 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9610 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9613 PR tree-optimization/84452
9614 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9615 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9618 2018-02-19 Martin Liska <mliska@suse.cz>
9621 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9623 2018-02-19 Martin Liska <mliska@suse.cz>
9624 Richard Sandiford <richard.sandiford@linaro.org>
9626 PR tree-optimization/82491
9627 * gimple-fold.c (get_base_constructor): Make earlier bail out
9630 2018-02-19 Carl Love <cel@us.ibm.com>
9632 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9633 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9635 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9636 P8V_BUILTIN_VEC_NEG.
9638 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9640 * config/rl78/rl78.md (movdf): New define expand.
9642 2018-02-19 Martin Liska <mliska@suse.cz>
9645 * doc/invoke.texi: Fix typo.
9646 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9648 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9650 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9651 handle rs6000_single_float and rs6000_double_float specially for
9654 2018-02-16 Jeff Law <law@redhat.com>
9656 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9659 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9662 * ipa-inline.c (inline_small_functions): Fix a typo.
9664 2018-02-16 Nathan Sidwell <nathan@acm.org>
9666 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9668 2018-02-16 Carl Love <cel@us.ibm.com>
9670 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9671 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9672 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9673 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9674 expansion to P8V_BUILTIN_VEC_FLOAT2.
9676 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9678 PR rtl-optimization/70023
9679 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9680 src_regno into account.
9682 2018-02-16 Carl Love <cel@us.ibm.com>
9684 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9685 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9686 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9687 * config/rs6000/rs6000.c: Remove case statements for
9688 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9689 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9690 and P9V_BUILTIN_VEC_VINSERT4B.
9691 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9692 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9693 * config/rs6000/vsx.md:
9694 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9697 2018-02-16 Carl Love <cel@us.ibm.com>
9699 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9701 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9703 * config/rs6000/rs6000-c.c: Add the definitions for
9704 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9705 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9706 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9707 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9708 definition for insert4b and define insn *insert3b_internal.
9709 * doc/extend.texi: Add documentation for vec_extract4b.
9711 2018-02-16 Nathan Sidwell <nathan@acm.org>
9713 * doc/extend.texi (Backwards Compatibility): Mention friend
9714 injection. Note for-scope is deprecated.
9715 * doc/invoke.texi (-ffriend-injection): Deprecate.
9717 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9719 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9720 that moved to I2, also allow destinations that are a paradoxical
9721 subreg (instead of a normal reg).
9723 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9726 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9729 2018-02-16 Richard Biener <rguenther@suse.de>
9731 PR tree-optimization/84037
9732 PR tree-optimization/84016
9734 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9735 Adjust vec_construct for the fact we need additional higher latency
9736 128bit inserts for AVX256 and AVX512 vector builds.
9737 (ix86_add_stmt_cost): Scale vector construction cost for
9740 2018-02-16 Richard Biener <rguenther@suse.de>
9742 PR tree-optimization/84417
9743 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9744 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9745 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9747 2018-02-16 Martin Liska <mliska@suse.cz>
9750 * internal-fn.def (ASAN_CHECK): Set proper flags.
9751 (ASAN_MARK): Likewise.
9753 2018-02-16 Julia Koval <julia.koval@intel.com>
9755 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9756 from PTA_CANNONLAKE.
9758 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9761 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9762 Use ++iter rather than iter++ for std::list iterators.
9763 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9764 defer deleting them until all nodes in the forest are processed. Do
9765 free even leaf nodes. Change to_process into auto_vec.
9768 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9769 * vec.h (vec_default_construct): Use memset instead of placement new
9770 if BROKEN_VALUE_INITIALIZATION is defined.
9771 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9772 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9775 PR rtl-optimization/83723
9776 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9777 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9778 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9780 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9782 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9784 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9786 PR rtl-optimization/81443
9787 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9788 from inner REGs to paradoxical SUBREGs.
9790 2018-02-16 Richard Biener <rguenther@suse.de>
9792 PR tree-optimization/84399
9793 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9794 For operands we can analyze at their definition make sure we can
9795 analyze them at each use as well.
9797 2018-02-16 Richard Biener <rguenther@suse.de>
9799 PR tree-optimization/84190
9800 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9801 volatile accesses if the decl isn't volatile.
9803 2018-02-15 Jason Merrill <jason@redhat.com>
9805 PR c++/84314 - ICE with templates and fastcall attribute.
9806 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9807 TYPE_CANONICAL on an existing type.
9809 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9811 PR tree-optimization/84383
9812 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9813 dstoff nor call operand_equal_p if dstbase is NULL.
9815 PR tree-optimization/84334
9816 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9817 also a CONSTANT_CLASS_P, punt.
9819 2018-02-14 Jim Wilson <jimw@sifive.com>
9821 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9822 first SMALL_OPERAND check. New local min_second_step. Move assert
9823 to where locals are set. Add TARGET_RVC support.
9824 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9826 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9828 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9830 2018-02-14 Martin Sebor <msebor@redhat.com>
9832 PR tree-optimization/83698
9833 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9834 arrays constrain the offset range to their bounds.
9835 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9836 (builtin_access::overlap): Avoid setting the size of overlap if it's
9838 (maybe_diag_overlap): Also consider arrays when deciding what values
9839 of offsets to include in diagnostics.
9841 2018-02-14 Martin Sebor <msebor@redhat.com>
9844 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9845 that correspond to the kind of a declaration.
9847 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9850 * config/pa/pa.md: Load address of PIC label using the linkage table
9851 if the label is nonlocal.
9853 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9855 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9856 warning message if user requests -maltivec=be.
9857 * doc/invoke.texi: Document deprecation of -maltivec=be.
9859 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9862 * config/rs6000/rs6000-c.c: Update definitions for
9863 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9864 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9866 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9869 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9870 add _get_ssp intrinsics. Remove argument from
9871 __builtin_ia32_rdssp[d|q].
9872 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9873 * config/i386/i386-builtin.def: Remove argument from
9874 __builtin_ia32_rdssp[d|q].
9875 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9876 ix86_expand_special_args_builtin for _rdssp[d|q].
9877 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9878 Clear register before usage.
9879 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9880 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9882 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9884 PR tree-optimization/84357
9885 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9886 operand 1 of an ARRAY_REF too.
9888 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9891 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9892 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9894 (set_of_reg): New struct.
9895 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9896 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9897 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9899 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9900 Split into bitclr, bitset, bitinvert patterns if appropriate.
9901 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9902 use rx_fuse_in_memory_bitop.
9903 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9904 to named insn, correct maximum insn length.
9906 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9909 * machmode.def: Define a complex mode for PARTIAL_INT.
9910 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9912 * doc/rtl.texi: Document CSPImode.
9913 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9915 (msp430_hard_regno_nregs_with_padding): Likewise.
9917 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9920 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9922 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9924 PR rtl-optimization/84169
9925 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9926 we generated a parallel as new i3 and we split that to new i2 and i3
9927 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9928 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9929 those to i2, not i1. Partially rewrite this scan code.
9931 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9934 * stor-layout.c (place_field): For variable length fields, adjust
9935 offset_align afterwards not just based on the field's alignment,
9936 but also on the size.
9939 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9940 of exps and logs in the use_exp2 case.
9942 2018-02-13 Jeff Law <law@redhat.com>
9944 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9947 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9950 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9954 * common.opt (gas-loc-support, gas-locview-support): New.
9955 (ginline-points, ginternal-reset-location-views): New.
9956 * doc/invoke.texi: Document them. Use @itemx where intended.
9957 (gvariable-location-views): Adjust.
9958 * target.def (reset_location_view): New.
9959 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9960 (TARGET_RESET_LOCATION_VIEW): New.
9961 * doc/tm.texi: Rebuilt.
9962 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9963 (dwarf2out_default_as_locview_support): New.
9964 (output_asm_line_debug_info): Use option variables.
9965 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9966 (output_loc_list): Likewise.
9967 (add_high_low_attributes): Check option variables.
9968 Don't output entry view attribute in strict mode.
9969 (gen_inlined_subroutine_die): Check option variables.
9970 (dwarf2out_inline_entry): Likewise.
9971 (init_sections_and_labels): Likewise.
9972 (dwarf2out_early_finish): Likewise.
9973 (maybe_reset_location_view): New, from...
9974 (dwarf2out_var_location): ... here. Call it.
9975 * debug.h (dwarf2out_default_as_loc_support): Declare.
9976 (dwarf2out_default_as_locview_support): Declare.
9977 * hooks.c (hook_int_rtx_insn_0): New.
9978 * hooks.h (hook_int_rtx_insn_0): Declare.
9979 * toplev.c (process_options): Take -gas-loc-support and
9980 -gas-locview-support from dwarf2out. Enable
9981 -gvariable-location-views by default only with locview
9982 assembler support. Enable -ginternal-reset-location-views by
9983 default only if the target defines the corresponding hook.
9984 Enable -ginline-points by default if location views are
9985 enabled; force it disabled if statement frontiers are
9987 * tree-inline.c (expand_call_inline): Check option variables.
9988 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9990 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9992 PR tree-optimization/84321
9993 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9994 handling. Also check whether the anti-range contains any values
9995 that satisfy the mask; switch to a VR_RANGE if not.
9997 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10000 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10002 2018-02-13 Martin Jambor <mjambor@suse.cz>
10005 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10006 of call statements, also set location of a load to a temporary.
10008 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10010 * config/rl78/rl78.c (add_vector_labels): New function.
10011 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10012 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10013 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10014 which checks that no arguments are passed.
10015 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10016 * doc/extend.texi: Documentation for the new attribute.
10018 2018-02-13 Andreas Schwab <schwab@suse.de>
10020 * config/riscv/linux.h (CPP_SPEC): Define.
10022 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10025 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10026 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10027 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10028 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10029 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10030 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10031 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10032 temporarily for AES and PCLMUL builtins.
10034 PR tree-optimization/84339
10035 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10036 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10039 PR middle-end/84309
10040 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10041 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10042 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10044 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10046 * omp-simd-clone.h: New file.
10047 * omp-simd-clone.c: Include omp-simd-clone.h.
10048 (expand_simd_clones): No longer static.
10049 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10050 cgraph.h and omp-simd-clone.h.
10051 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10052 (vect_recog_widen_shift_pattern): Formatting fix.
10053 (vect_pattern_recog_1): Don't check optab for calls.
10056 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10057 operands[2] into a REG before using gen_lowpart on it.
10059 2018-02-12 Jeff Law <law@redhat.com>
10062 * config/sh/sh.c (find_barrier): Consider a sibling call
10065 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10066 successfully back substituting a reg.
10068 2018-02-12 Richard Biener <rguenther@suse.de>
10070 PR tree-optimization/84037
10071 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10072 parameter, move visited init to caller.
10073 (vect_slp_analyze_operations): Separate cost from validity
10074 check, initialize visited once for all instances.
10075 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10077 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10079 (vect_model_promotion_demotion_cost): Likewise.
10080 (vectorizable_bswap): Guard cost modeling with !slp_node
10081 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10083 (vectorizable_call): Likewise.
10084 (vectorizable_conversion): Likewise.
10085 (vectorizable_assignment): Likewise.
10086 (vectorizable_shift): Likewise.
10087 (vectorizable_operation): Likewise.
10088 (vectorizable_store): Likewise.
10089 (vectorizable_load): Likewise.
10090 (vectorizable_condition): Likewise.
10091 (vectorizable_comparison): Likewise.
10093 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10096 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10097 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10098 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10100 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10102 PR middle-end/83665
10103 * params.def (inline-min-speedup): Increase from 8 to 15.
10104 (max-inline-insns-auto): Decrease from 40 to 30.
10105 * ipa-split.c (consider_split): Add some buffer for function to
10106 be considered inlining candidate.
10107 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10110 2018-02-12 Richard Biener <rguenther@suse.de>
10112 PR tree-optimization/84037
10113 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10114 matched stmts if we cannot swap the non-matched ones.
10116 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10118 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10119 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10120 _mm_maskz_scalef_round_ss): New intrinsics.
10121 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10122 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10123 __builtin_ia32_scalefss_round): Remove.
10124 (__builtin_ia32_scalefsd_mask_round,
10125 __builtin_ia32_scalefss_mask_round): New intrinsics.
10126 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10127 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10128 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10129 "<round_constraint>")): Changed to ...
10130 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10131 "<round_scalar_constraint>")): ... this.
10132 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10133 %0, %1, %2<round_op3>}"): Changed to ...
10134 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10135 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10136 %2<round_scalar_mask_op3>}"): ... this.
10137 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10139 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10141 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10142 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10143 (_mm_maskz_sqrt_round_ss): New intrinsics.
10144 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10145 (__builtin_ia32_sqrtsd_mask_round)
10146 (__builtin_ia32_sqrtss_mask_round): New builtins.
10147 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10148 (__builtin_ia32_sqrtss_round): Remove.
10149 (__builtin_ia32_sqrtsd_mask_round)
10150 (__builtin_ia32_sqrtss_mask_round): New builtins.
10151 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10152 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10153 ((match_operand:VF_128 1 "vector_operand"
10154 "xBm,<round_constraint>")): Changed to ...
10155 ((match_operand:VF_128 1 "vector_operand"
10156 "xBm,<round_scalar_constraint>")): ... this.
10157 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10158 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10159 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10160 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10161 %<iptr>1<round_scalar_mask_op3>}): ... this.
10162 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10163 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10165 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10168 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10169 Cast vec_cmpeq result to correct type.
10170 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10171 Cast vec_cmpgt result to correct type.
10173 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10175 * final.c (final_scan_insn_1): Renamed from...
10176 (final_scan_insn): ... this. New wrapper, to recover
10177 seen from the outermost call in recursive ones.
10178 * config/sparc/sparc.c (output_return): Drop seen from call.
10179 (output_sibcall): Likewise.
10180 * config/visium/visium.c (output_branch): Likewise.
10182 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10184 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10187 2018-02-10 Alan Modra <amodra@gmail.com>
10190 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10191 Specify LR as an input.
10193 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10196 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10197 remove_member_access_dummy_vars): New functions.
10198 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10199 lower_omp_1, execute_lower_omp): Use them.
10201 PR rtl-optimization/84308
10202 * shrink-wrap.c (spread_components): Release todo vector.
10204 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10206 PR rtl-optimization/57193
10207 * ira-color.c (struct allocno_color_data): Add member
10208 conflict_allocno_hard_prefs.
10209 (update_conflict_allocno_hard_prefs): New.
10210 (bucket_allocno_compare_func): Add a preference based on
10211 conflict_allocno_hard_prefs.
10212 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10213 (color_allocnos): Remove a dead code. Initiate
10214 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10216 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10219 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10220 constraint from =wa to wa. Avoid a subreg on the output operand,
10221 instead use a pseudo and subreg it in a move.
10222 (p9_xxbrd_<mode>): Changed to ...
10223 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10224 (p9_xxbrd_v2df): New expander.
10225 (p9_xxbrw_<mode>): Changed to ...
10226 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10227 (p9_xxbrw_v4sf): New expander.
10229 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10231 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10233 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10236 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10237 multiply in 32-bit mode.
10238 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10239 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10242 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10244 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10245 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10246 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10247 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10249 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10252 * dwarf2out.c (is_trivial_indirect_ref): New function.
10253 (dwarf2out_late_global_decl): Do not generate a location
10254 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10255 and that are not defined in the current unit.
10257 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10259 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10260 instead of a libcall for UNORDERED.
10262 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10265 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10266 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10268 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10271 * config/s390/s390.c (s390_set_current_function): Invoke
10272 s390_indirect_branch_settings also if fndecl didn't change.
10274 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10276 * config/rs6000/rs6000.md (blockage): Set length to zero.
10278 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10280 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10282 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10285 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10286 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10290 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10291 PARALLEL incoming that failed vt_get_decl_and_offset check.
10293 PR middle-end/84237
10294 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10295 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10297 (get_variable_section): For decls in named .bss* sections pass true as
10298 second argument to bss_initializer_p.
10300 2018-02-09 Marek Polacek <polacek@redhat.com>
10301 Jakub Jelinek <jakub@redhat.com>
10304 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10305 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10306 Sync some changes from cxx_fold_indirect_ref.
10308 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10310 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10312 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10313 (BLOCK_INLINE_ENTRY_LABEL): New.
10314 (dwarf2out_var_location): Disregard inline entry markers.
10315 (inline_entry_data): New struct.
10316 (inline_entry_data_hasher): New hashtable type.
10317 (inline_entry_data_hasher::hash): New.
10318 (inline_entry_data_hasher::equal): New.
10319 (inline_entry_data_table): New variable.
10320 (add_high_low_attributes): Add DW_AT_entry_pc and
10321 DW_AT_GNU_entry_view attributes if a pending entry is found
10322 in inline_entry_data_table. Add old entry_pc attribute only
10323 if debug nonbinding markers are disabled.
10324 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10325 markers are enabled.
10326 (block_within_block_p, dwarf2out_inline_entry): New.
10327 (dwarf2out_finish): Check that no entries remained in
10328 inline_entry_data_table.
10329 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10330 (final_scan_insn, notice_source_line): Likewise.
10331 (rest_of_clean_state): Skip inline entry markers.
10332 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10334 * gimple.c (gimple_build_debug_inline_entry): New.
10335 * gimple.h (enum gimple_debug_subcode): Add
10336 GIMPLE_DEBUG_INLINE_ENTRY.
10337 (gimple_build_debug_inline_entry): Declare.
10338 (gimple_debug_inline_entry_p): New.
10339 (gimple_debug_nonbind_marker_p): Adjust.
10340 * insn-notes.def (INLINE_ENTRY): New.
10341 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10342 inline entry marker notes.
10343 (print_insn): Likewise.
10344 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10345 (INSN_DEBUG_MARKER_KIND): Likewise.
10346 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10347 * tree-inline.c (expand_call_inline): Build and insert
10348 debug_inline_entry stmt.
10349 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10350 inline entry blocks early, if nonbind markers are enabled.
10351 (dump_scope_block): Dump fragment info.
10352 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10353 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10354 (gimple_build_debug_inline_entry): New.
10355 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10356 Enable/disable inline entry points too.
10357 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10358 (DEBUG_INSN): Describe inline entry markers.
10360 * common.opt (gvariable-location-views): New.
10361 (gvariable-location-views=incompat5): New.
10362 * config.in: Rebuilt.
10363 * configure: Rebuilt.
10364 * configure.ac: Test assembler for view support.
10365 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10366 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10367 * dwarf2out.c (var_loc_view): New typedef.
10368 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10369 (dwarf2out_locviews_in_attribute): New.
10370 (dwarf2out_locviews_in_loclist): New.
10371 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10372 (enum dw_line_info_opcode): Add LI_adv_address.
10373 (struct dw_line_info_table): Add view.
10374 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10375 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10376 (zero_view_p): New variable.
10377 (ZERO_VIEW_P): New macro.
10378 (output_asm_line_debug_info): New.
10379 (struct var_loc_node): Add view.
10380 (add_AT_view_list, AT_loc_list): New.
10381 (add_var_loc_to_decl): Add view param. Test it against last.
10382 (new_loc_list): Add view params. Record them.
10383 (AT_loc_list_ptr): Handle loc and view lists.
10384 (view_list_to_loc_list_val_node): New.
10385 (print_dw_val): Handle dw_val_class_view_list.
10386 (size_of_die): Likewise.
10387 (value_format): Likewise.
10388 (loc_list_has_views): New.
10389 (gen_llsym): Set vl_symbol too.
10390 (maybe_gen_llsym, skip_loc_list_entry): New.
10391 (dwarf2out_maybe_output_loclist_view_pair): New.
10392 (output_loc_list): Output view list or entries too.
10393 (output_view_list_offset): New.
10394 (output_die): Handle dw_val_class_view_list.
10395 (output_dwarf_version): New.
10396 (output_compilation_unit_header): Use it.
10397 (output_skeleton_debug_sections): Likewise.
10398 (output_rnglists, output_line_info): Likewise.
10399 (output_pubnames, output_aranges): Update version comments.
10400 (output_one_line_info_table): Output view numbers in asm comments.
10401 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10402 Call maybe_gen_llsym.
10403 (loc_list_from_tree_1): Adjust.
10404 (add_AT_location_description): Create view list attribute if
10405 needed, check it's absent otherwise.
10406 (convert_cfa_to_fb_loc_list): Adjust.
10407 (maybe_emit_file): Call output_asm_line_debug_info for test.
10408 (dwarf2out_var_location): Reset views as needed. Precompute
10409 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10410 attribute. Set view.
10411 (new_line_info_table): Reset next view.
10412 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10413 (dwarf2out_source_line): Likewise. Output view resets and labels to
10414 the assembler, or select appropriate line info opcodes.
10415 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10416 (optimize_string_length): Catch it. Adjust.
10417 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10418 dw_val_class_view_list, and remove it if no longer needed.
10419 (hash_loc_list): Hash view numbers.
10420 (loc_list_hasher::equal): Compare them.
10421 (optimize_location_lists): Check whether a view list symbol is
10422 needed, and whether the locview attribute is present, and
10423 whether they match. Remove the locview attribute if no longer
10425 (index_location_lists): Call skip_loc_list_entry for test.
10426 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10427 Use output_dwarf_version.
10428 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10429 (struct dw_val_node): Add val_view_list.
10430 * final.c (SEEN_NEXT_VIEW): New.
10431 (set_next_view_needed): New.
10432 (clear_next_view_needed): New.
10433 (maybe_output_next_view): New.
10434 (final_start_function): Rename to...
10435 (final_start_function_1): ... this. Take pointer to FIRST,
10436 add SEEN parameter. Emit param bindings in the initial view.
10437 (final_start_function): Reintroduce SEEN-less interface.
10438 (final): Rename to...
10439 (final_1): ... this. Take SEEN parameter. Output final pending
10440 next view at the end.
10441 (final): Reintroduce seen-less interface.
10442 (final_scan_insn): Output pending next view before switching
10443 sections or ending a block. Mark the next view as needed when
10444 outputting variable locations. Notify debug backend of section
10445 changes, and of location view changes.
10446 (rest_of_handle_final): Adjust.
10447 * toplev.c (process_options): Autodetect value for debug variable
10448 location views option. Warn on incompat5 without -gdwarf-5.
10449 * doc/invoke.texi (gvariable-location-views): New.
10450 (gvariable-location-views=incompat5): New.
10451 (gno-variable-location-views): New.
10453 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10455 PR tree-optimization/84136
10456 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10457 that the result of find_edge is non-NULL.
10459 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10462 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10463 storing integer register in SImode. Fix cost of 256 and 512
10464 byte aligned SSE register store.
10466 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10468 * config/i386/i386.c (ix86_multiplication_cost): Fix
10469 multiplication cost for TARGET_AVX512DQ.
10471 2018-02-08 Marek Polacek <polacek@redhat.com>
10473 PR tree-optimization/84238
10474 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10477 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10479 PR tree-optimization/84265
10480 * tree-vect-stmts.c (vectorizable_store): Don't treat
10481 VMAT_CONTIGUOUS accesses as grouped.
10482 (vectorizable_load): Likewise.
10484 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10486 PR tree-optimization/81635
10487 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10488 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10489 (test_round_for_mask): New functions.
10490 (wide_int_cc_tests): Call test_round_for_mask.
10491 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10492 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10493 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10494 range returned by get_range_info.
10496 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10499 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10500 * symtab.c: Include builtins.h
10501 (symtab_node::output_to_lto_symbol_table_p): Move here
10502 from lto-streamer-out.c:output_symbol_p.
10503 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10504 (output_symbol_p): Move all logic to symtab.c
10505 (produce_symtab): Update.
10507 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10509 * config/s390/s390-opts.h (enum indirect_branch): Define.
10510 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10511 (s390_indirect_branch_via_thunk)
10512 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10513 (enum s390_indirect_branch_type): Define.
10514 * config/s390/s390.c (struct s390_frame_layout, struct
10515 machine_function): Remove.
10516 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10517 (indirect_branch_table_label_no, indirect_branch_table_name):
10519 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10520 (enum s390_indirect_branch_option): Define.
10521 (s390_return_addr_from_memory): New function.
10522 (s390_handle_string_attribute): New function.
10523 (s390_attribute_table): Add new attribute handler.
10524 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10525 (s390_indirect_branch_via_thunk): New function.
10526 (s390_indirect_branch_via_inline_thunk): New function.
10527 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10528 sibling call optimization for non z10 compiles.
10529 (s390_emit_call): Force indirect branch target to be a single
10530 register. Add r1 clobber for non-z10 compiles.
10531 (s390_emit_epilogue): Emit return jump via return_use expander.
10532 (s390_reorg): Handle JUMP_INSNs as execute targets.
10533 (s390_option_override_internal): Perform validity checks for the
10534 new command line options.
10535 (s390_indirect_branch_attrvalue): New function.
10536 (s390_indirect_branch_settings): New function.
10537 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10538 (s390_output_indirect_thunk_function): New function.
10539 (s390_code_end): Implement target hook.
10540 (s390_case_values_threshold): Implement target hook.
10541 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10543 * config/s390/s390.h (struct s390_frame_layout)
10544 (struct machine_function): Move here from s390.c.
10545 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10546 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10547 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10548 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10549 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10550 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10551 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10552 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10553 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10554 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10555 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10556 (mnemonic attribute): Add values which aren't recognized
10558 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10559 pattern for branch conversion. Fix mnemonic attribute.
10560 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10561 indirect branch via thunk if requested.
10562 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10563 ("*indirect_jump"): Disable for branch conversion using out of
10565 ("indirect_jump_via_thunk<mode>_z10")
10566 ("indirect_jump_via_thunk<mode>")
10567 ("indirect_jump_via_inlinethunk<mode>_z10")
10568 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10569 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10570 ("casesi_jump_via_inlinethunk<mode>_z10")
10571 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10572 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10573 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10574 ("*indirect2_jump"): Disable for branch conversion.
10575 ("casesi_jump"): Turn into expander and expand patterns for branch
10577 ("return_use"): New expander.
10578 ("*return"): Emit return via thunk and rename it to ...
10579 ("*return<mode>"): ... this one.
10580 * config/s390/s390.opt: Add new options and and enum for the
10583 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10585 * lra-constraints.c (match_reload): Unconditionally use
10586 gen_lowpart_SUBREG, rather than selecting between that
10587 and equivalent gen_rtx_SUBREG code.
10589 2018-02-08 Richard Biener <rguenther@suse.de>
10591 PR tree-optimization/84233
10592 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10593 changed flag instead of boguously re-using phi_inserted.
10595 2018-02-08 Martin Jambor <mjambor@suse.cz>
10597 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10598 static local variables.
10600 2018-02-08 Richard Biener <rguenther@suse.de>
10602 PR tree-optimization/84278
10603 * tree-vect-stmts.c (vectorizable_store): When looking for
10604 smaller vector types to perform grouped strided loads/stores
10605 make sure the mode is supported by the target.
10606 (vectorizable_load): Likewise.
10608 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10610 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10611 Increase LDP/STP opportunities by adding adjacent callee-saves.
10613 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10615 PR rtl-optimization/84068
10616 PR rtl-optimization/83459
10617 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10619 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10621 PR tree-optimization/84224
10622 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10623 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10624 non-zero arguments.
10626 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10629 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10630 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10631 count, remove one USE.
10633 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10635 * doc/install.texi (Configuration): Document the
10636 --with-long-double-format={ibm,ieee} PowerPC configuration
10640 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10641 Convert from define_expand to be define_insn_and_split. Rework
10642 float/double/_Float128 conversions to QI/HI/SImode to work with
10643 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10644 conversions to QI/HImode types did a store and then a load to
10645 truncate the value. For conversions to VSX registers, don't split
10646 the insn, instead emit the code directly. Use the code iterator
10647 any_fix to combine signed and unsigned conversions.
10648 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10649 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10650 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10651 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10652 (fix_<mode>di2_hw): Likewise.
10653 (fixuns_<mode>di2_hw): Likewise.
10654 (fix_<mode>si2_hw): Likewise.
10655 (fixuns_<mode>si2_hw): Likewise.
10656 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10657 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10658 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10659 fix<uns>_trunc<SFDF:mode>si2_p8.
10660 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10661 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10662 (fix<uns>_<mode>_mem): Likewise.
10663 (fctiw<u>z_<mode>_mem): Likewise.
10664 (fix<uns>_<mode>_mem): Likewise.
10665 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10666 the register allocator from doing a direct move to the GPRs to do
10667 a store, and instead use the ISA 3.0 store byte/half-word from
10668 vector register instruction. For IEEE 128-bit floating point,
10669 also optimize stores of 32-bit ints.
10670 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10672 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10674 * genextract.c (push_pathstr_operand): New function to support
10676 (walk_rtx): Call push_pathstr_operand.
10677 (print_path): Support [a-zA-Z].
10679 2018-02-07 Richard Biener <rguenther@suse.de>
10681 PR tree-optimization/84037
10682 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10683 (cse_and_gimplify_to_preheader): Declare.
10684 (vect_get_place_in_interleaving_chain): Likewise.
10685 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10687 (_loop_vec_info::~_loop_vec_info): Delete it.
10688 (cse_and_gimplify_to_preheader): New function.
10689 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10690 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10691 (vectorizable_load): Likewise. For grouped stores always base
10692 the IV on the first element.
10693 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10694 condition before gimplifying.
10696 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10698 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10699 *DIV_EXPR and *MOD_EXPR.
10701 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10704 * config/i386/i386.c (ix86_option_override_internal): Mask out
10705 the CF_SET bit when checking -fcf-protection.
10707 2018-02-07 Tom de Vries <tom@codesourcery.com>
10710 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10713 2018-02-07 Richard Biener <rguenther@suse.de>
10715 PR tree-optimization/84204
10716 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10719 PR tree-optimization/84205
10720 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10721 special-case isl_ast_op_zdiv_r.
10723 PR tree-optimization/84223
10724 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10725 Only add conditions from within the region.
10726 (gather_bbs::after_dom_children): Adjust.
10728 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10731 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10732 * config/avr/avr.md: Only post-reload split REG-REG moves if
10733 either register is GENERAL_REG_P.
10735 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10737 PR tree-optimization/84235
10738 * tree-ssa-scopedtables.c
10739 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10740 if the subtraction is performed in floating point type where NaNs are
10741 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10742 build 1. Formatting fix.
10744 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10747 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10748 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10749 and skip it regardless of bb boundaries. Use CALL_P macro,
10750 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10751 unnecessarily, formatting fix.
10753 2018-02-06 Michael Collison <michael.collison@arm.com>
10755 * config/arm/thumb2.md:
10756 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10757 (*thumb_mov_notscc): Ditto.
10759 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10762 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10763 unsigned_fix, not "s".
10765 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10767 * configure.ac (gcc_fn_eh_frame_ro): New function.
10768 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10769 correct .eh_frame permissions.
10770 * configure: Regenerate.
10772 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10774 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10775 irrelevant options.
10777 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10779 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10780 Display warning message for -mno-speculate-indirect-jumps.
10782 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10784 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10786 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10788 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10790 PR tree-optimization/84225
10791 * tree-eh.c (find_trapping_overflow): Only call
10792 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10794 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10797 * config/i386/i386.c: Reimplement the check of possible options
10798 -mibt/-mshstk conbination. Change error messages.
10799 * doc/invoke.texi: Fix a typo: remove extra '='.
10801 2018-02-06 Marek Polacek <polacek@redhat.com>
10803 PR tree-optimization/84228
10804 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10806 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10809 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10810 emitted arch directives.
10811 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10812 __ARM_FEATURE_COPROC before changing architectures.
10814 2018-02-06 Richard Biener <rguenther@suse.de>
10816 * config/i386/i386.c (print_reg): Fix typo.
10817 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10819 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10821 * configure: Regenerate.
10823 2018-02-05 Martin Sebor <msebor@redhat.com>
10825 PR tree-optimization/83369
10826 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10829 2018-02-05 Martin Liska <mliska@suse.cz>
10831 * doc/invoke.texi: Cherry-pick upstream r323995.
10833 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10835 * ira.c (ira_init_register_move_cost): Adjust comment.
10837 2018-02-05 Martin Liska <mliska@suse.cz>
10839 PR gcov-profile/84137
10840 * doc/gcov.texi: Fix typo in documentation.
10842 2018-02-05 Martin Liska <mliska@suse.cz>
10844 PR gcov-profile/83879
10845 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10846 using dlopen functionality.
10848 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10850 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10851 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10852 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10853 _mm_maskz_range_round_ss): New intrinsics.
10854 (__builtin_ia32_rangesd128_round)
10855 (__builtin_ia32_rangess128_round): Remove.
10856 (__builtin_ia32_rangesd128_mask_round,
10857 __builtin_ia32_rangess128_mask_round): New builtins.
10858 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10859 __builtin_ia32_rangess128_round): Remove.
10860 (__builtin_ia32_rangesd128_mask_round,
10861 __builtin_ia32_rangess128_mask_round): New builtins.
10862 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10863 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10864 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10865 "<round_saeonly_constraint>")): Changed to ...
10866 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10867 "<round_saeonly_scalar_constraint>")): ... this.
10868 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10869 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10870 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10871 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10872 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10874 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10876 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10878 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10879 Remove all values except native, 8540 and 8548.
10881 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10883 * config/i386/i386.c (ix86_output_function_return): Pass
10884 INVALID_REGNUM, instead of -1, as invalid register number to
10885 indirect_thunk_name and output_indirect_thunk.
10887 2018-02-02 Julia Koval <julia.koval@intel.com>
10889 * config.gcc: Add -march=icelake.
10890 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10891 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10892 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10893 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10894 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10895 (processor_target_table): Add icelake.
10896 (ix86_option_override_internal): Handle new PTAs.
10897 (get_builtin_code_for_version): Handle icelake.
10898 (M_INTEL_COREI7_ICELAKE): New.
10899 (fold_builtin_cpu): Handle icelake.
10900 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10901 * doc/invoke.texi: Add -march=icelake.
10903 2018-02-02 Julia Koval <julia.koval@intel.com>
10905 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10906 to wide_int_bitmask.
10907 * wide-int-bitmask.h: New.
10909 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10912 * config/i386/i386.md: Replace Pmode with word_mode in
10913 builtin_setjmp_setup and builtin_longjmp to support x32.
10915 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10919 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10921 (rs6000_supported_cpu_names): New static variable.
10922 (linux_cpu_translation_table): Likewise.
10923 (elf_platform) <cpu>: Define new static variable and use it.
10924 Translate kernel AT_PLATFORM name to canonical name if needed.
10925 Error if platform name is unknown.
10927 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10930 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10932 2018-02-01 Jeff Law <law@redhat.com>
10935 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10936 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10937 the scratch if RELEASE_VIA_POP is false.
10938 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10939 If we have to save a temporary register, decrement SIZE appropriately.
10940 Pass new arguments to release_scratch_register_on_entry.
10941 (ix86_adjust_stack_and_probe): Likewise.
10942 (ix86_emit_probe_stack_range): Pass new arguments to
10943 release_scratch_register_on_entry.
10945 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10947 PR rtl-optimization/84157
10948 * combine.c (change_zero_ext): Use REG_P predicate in
10949 front of HARD_REGISTER_P predicate.
10951 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10953 * config/avr/avr.c (avr_option_override): Move disabling of
10954 -fdelete-null-pointer-checks to...
10955 * common/config/avr/avr-common.c (avr_option_optimization_table):
10958 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10960 PR tree-optimization/81635
10961 * tree-data-ref.c (split_constant_offset_1): For types that
10962 wrap on overflow, try to use range info to prove that wrapping
10965 2018-02-01 Renlin Li <renlin.li@arm.com>
10968 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10969 TAILCALL_ADDR_REGS.
10970 (aarch64_register_move_cost): Likewise.
10971 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10972 TAILCALL_ADDR_REGS.
10973 (REG_CLASS_NAMES): Likewise.
10974 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10975 TAILCALL_ADDR_REGS. Remove IP registers.
10976 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10978 2018-02-01 Richard Biener <rguenther@suse.de>
10980 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10981 for specifying RPO order and allow NULL for that.
10982 * domwalk.c (dom_walker::dom_walker): Likewise.
10983 (dom_walker::walk): Handle NULL RPO order.
10984 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10986 (rewrite_update_dom_walker): Likewise.
10987 (mark_def_dom_walker): Likewise.
10989 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10991 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10992 (aarch64_maybe_expand_sve_subreg_move): Declare.
10993 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10994 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10996 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10997 that are semantically a reverse operation.
10998 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10999 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11000 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11002 (aarch64_can_change_mode_class): For big-endian, forbid changes
11003 between two SVE modes if they have different element sizes.
11005 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11007 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11008 the TImode handling for big-endian targets.
11010 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11012 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11013 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11015 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11016 Remove BSWAP handing for big-endian targets and use the form of
11017 LD1RQ appropariate for the mode.
11019 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11021 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11022 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11023 duplicated element.
11025 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11028 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11029 check for operands that need to go through aarch64_sve_reload_be.
11031 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11033 PR tree-optimization/81661
11034 PR tree-optimization/84117
11035 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11036 * tree-eh.c: Include gimplify.h.
11037 (find_trapping_overflow, replace_trapping_overflow,
11038 rewrite_to_non_trapping_overflow): New functions.
11039 * tree-vect-loop.c: Include tree-eh.h.
11040 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11041 * tree-data-ref.c: Include tree-eh.h.
11042 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11044 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11046 PR rtl-optimization/84123
11047 * combine.c (change_zero_ext): Check if hard register satisfies
11048 can_change_dest_mode before calling gen_lowpart_SUBREG.
11050 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11053 * ira.c (ira_init_register_move_cost): Remove assert.
11055 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11057 PR rtl-optimization/84071
11058 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11059 * doc/tm.texi: Regenerate.
11061 2018-01-31 Richard Biener <rguenther@suse.de>
11063 PR tree-optimization/84132
11064 * tree-data-ref.c (analyze_miv_subscript): Properly
11065 check whether evolution_function_is_affine_multivariate_p
11066 before calling gcd_of_steps_may_divide_p.
11068 2018-01-31 Julia Koval <julia.koval@intel.com>
11071 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11072 * config/i386/i386.md (rdpid_rex64) New.
11073 (rdpid): Make 32bit only.
11075 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11078 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11079 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11081 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11084 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11086 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11088 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11090 PR rtl-optimization/84071
11091 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11092 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11094 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11096 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11097 (arc_attribute_table): Add 'aux' attribute.
11098 (arc_in_small_data_p): Consider aux like variables.
11099 (arc_is_aux_reg_p): New function.
11100 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11101 (arc_get_aux_arg): New function.
11102 (prepare_move_operands): Handle aux-register access.
11103 (arc_handle_aux_attribute): New function.
11104 * doc/extend.texi (ARC Variable attributes): Add subsection.
11106 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11108 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11109 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11110 (arc_attribute_table): Add 'uncached' attribute.
11111 (arc_print_operand): Print '.di' flag for uncached memory
11113 (arc_in_small_data_p): Do not consider for small data the uncached
11115 (arc_is_uncached_mem_p): New function.
11116 * config/arc/predicates.md (compact_store_memory_operand): Check
11117 for uncached memory accesses.
11118 (nonvol_nonimm_operand): Likewise.
11119 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11121 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11124 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11125 falign-loops=): Add Optimization flag.
11127 2018-01-30 Jeff Law <law@redhat.com>
11130 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11131 INT_REGISTERS_SAVED. Check it prior to calling
11132 get_scratch_register_on_entry.
11133 (ix86_adjust_stack_and_probe): Similarly.
11134 (ix86_emit_probe_stack_range): Similarly.
11135 (ix86_expand_prologue): Corresponding changes.
11137 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11140 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11141 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11143 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11146 * lra-constraints.c (curr_insn_transform): Process AND in the
11149 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11151 PR rtl-optimization/83986
11152 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11153 dependence against last_pending_memory_flush in addition to
11154 pending_jump_insns.
11156 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11158 PR tree-optimization/81611
11159 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11162 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11165 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11168 2018-01-30 Richard Biener <rguenther@suse.de>
11169 Jakub Jelinek <jakub@redhat.com>
11171 PR tree-optimization/84111
11172 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11173 inner loops added during recursion, as they don't have up-to-date
11176 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11179 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11180 (can_inline_edge_by_limits_p): ... here.
11181 (can_early_inline_edge_p, check_callers,
11182 update_caller_keys, update_callee_keys, recursive_inlining,
11183 add_new_edges_to_heap, speculation_useful_p,
11184 inline_small_functions,
11185 inline_small_functions, flatten_function,
11186 inline_to_all_callers_1): Update.
11188 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11190 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11193 2018-01-30 Richard Biener <rguenther@suse.de>
11195 PR tree-optimization/83008
11196 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11197 invariant and constant vector uses in stmts when they need
11198 more than one stmt.
11200 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11203 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11204 * configure: Regenerate.
11206 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11208 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11210 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11211 Use gen_rtx_REG rather than gen_lowpart.
11213 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11215 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11216 rather than 0 when creating partial subregs.
11218 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11220 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11223 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11226 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11227 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11228 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11229 flags. This restores the settings used before the 2017-07-24.
11230 Turning off pre increment/decrement/modify allows IVOPTS to
11231 optimize DF/SF loops where the index is an int.
11233 2018-01-29 Richard Biener <rguenther@suse.de>
11234 Kelvin Nilsen <kelvin@gcc.gnu.org>
11237 * tree-vect-stmts.c (vectorizable_call): Don't call
11238 targetm.vectorize_builtin_md_vectorized_function if callee is
11241 2018-01-22 Carl Love <cel@us.ibm.com>
11243 * doc/extend.tex: Fix typo in second arg in
11244 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11246 2018-01-29 Richard Biener <rguenther@suse.de>
11248 PR tree-optimization/84086
11249 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11250 (flush_ssaname_freelist): When SSA names were released reset
11251 the SCEV hash table.
11253 2018-01-29 Richard Biener <rguenther@suse.de>
11255 PR tree-optimization/84057
11256 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11257 removed paths when removing edges.
11259 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11261 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11262 -mfunction-return=@var{choice}.
11264 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11266 PR diagnostic/84034
11267 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11268 Handle CR like TAB.
11269 (layout::print_source_line): Likewise.
11270 (test_get_line_width_without_trailing_whitespace): Add test cases.
11272 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11274 PR middle-end/84040
11275 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11278 2018-01-26 Jim Wilson <jimw@sifive.com>
11280 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11282 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11285 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11287 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11288 and CMP + SUB-immediate -> SUBS.
11290 2018-01-26 Martin Sebor <msebor@redhat.com>
11292 PR tree-optimization/83896
11293 * tree-ssa-strlen.c (get_string_len): Rename...
11294 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11295 Avoid assuming length is constant.
11296 (handle_char_store): Use HOST_WIDE_INT for string length.
11298 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11301 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11302 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11304 2018-01-26 Richard Biener <rguenther@suse.de>
11306 PR rtl-optimization/84003
11307 * dse.c (record_store): Only record redundant stores when
11308 the earlier store aliases at least all accesses the later one does.
11310 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11312 PR rtl-optimization/83985
11313 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11314 REG_CFA_RESTORE insns.
11315 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11316 REG_CFA_RESTORE insns here.
11319 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11320 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11322 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11324 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11325 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11326 (arc_init): Likewise.
11327 (arc_override_options): Likewise.
11328 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11330 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11332 * config/arc/arc.h (TARGET_DBNZ): Define.
11333 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11334 properly set the tune attribute.
11335 (dbnz): Use TARGET_DBNZ guard.
11336 * config/arc/arc.opt (mtune): Add core3 option.
11338 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11340 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11341 recognize new pic like addresses.
11342 (arc_delegitimize_address): Clean up.
11344 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11346 * config/arc/arc-arches.def: Option mrf16 valid for all
11348 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11349 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11350 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11351 * config/arc/arc-tables.opt: Regenerate.
11352 * config/arc/arc.c (arc_conditional_register_usage): Handle
11353 reduced register file case.
11354 (arc_file_start): Set must have build attributes.
11355 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11356 mrf16 option value.
11357 * config/arc/arc.opt (mrf16): Add new option.
11358 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11359 * config/arc/genmultilib.awk: Handle new mrf16 option.
11360 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11361 * config/arc/t-multilib: Regenerate.
11362 * doc/invoke.texi (ARC Options): Document mrf16 option.
11364 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11366 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11367 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11368 (arc_attribute_table): Add 'secure_call' attribute.
11369 (arc_print_operand): Print secure call operand.
11370 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11372 (arc_is_secure_call_p): New function. * config/arc/arc.md
11373 (call_i): Add support for sjli instruction.
11374 (call_value_i): Likewise.
11375 * config/arc/constraints.md (Csc): New constraint.
11377 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11378 John Eric Martin <John.Martin@emmicro-us.com>
11380 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11381 * config/arc/arc.c (_arc_jli_section): New struct.
11382 (arc_jli_section): New type.
11383 (rc_jli_sections): New static variable.
11384 (arc_handle_jli_attribute): New function.
11385 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11386 (arc_file_end): New function.
11387 (TARGET_ASM_FILE_END): Define.
11388 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11389 (arc_add_jli_section): New function.
11390 (jli_call_scan): Likewise.
11391 (arc_reorg): Call jli_call_scan.
11392 (arc_output_addsi): Remove 'S' from printing asm operand.
11393 (arc_is_jli_call_p): New function.
11394 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11396 (movhi_insn): Likewise.
11397 (movsi_insn): Likewise.
11398 (movsi_set_cc_insn): Likewise.
11399 (loadqi_update): Likewise.
11400 (load_zeroextendqisi_update): Likewise.
11401 (load_signextendqisi_update): Likewise.
11402 (loadhi_update): Likewise.
11403 (load_zeroextendhisi_update): Likewise.
11404 (load_signextendhisi_update): Likewise.
11405 (loadsi_update): Likewise.
11406 (loadsf_update): Likewise.
11407 (movsicc_insn): Likewise.
11408 (bset_insn): Likewise.
11409 (bxor_insn): Likewise.
11410 (bclr_insn): Likewise.
11411 (bmsk_insn): Likewise.
11412 (bicsi3_insn): Likewise.
11413 (cmpsi_cc_c_insn): Likewise.
11414 (movsi_ne): Likewise.
11415 (movsi_cond_exec): Likewise.
11416 (clrsbsi2): Likewise.
11417 (norm_f): Likewise.
11427 (call_i): Remove 'S' asm letter, add jli instruction.
11428 (call_value_i): Likewise.
11429 * config/arc/arc.op (mjli-always): New option.
11430 * config/arc/constraints.md (Cji): New constraint.
11431 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11433 (subsf3_fpx): Likewise.
11434 (mulsf3_fpx): Likewise.
11435 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11437 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11438 function attrbutes.
11439 * doc/invoke.texi (ARC): Document mjli-always option.
11441 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11443 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11444 avoid addition with 0 and use incw and decw where possible.
11446 2018-01-26 Richard Biener <rguenther@suse.de>
11448 PR tree-optimization/81082
11449 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11450 association if it requires casting to unsigned.
11451 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11452 from fold_plusminus_mult_expr to catch important cases late when
11453 range info is available.
11455 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11457 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11458 * configure.ac (hidden_linkonce): New test.
11459 * configure: Regenerate.
11460 * config.in: Regenerate.
11462 2018-01-26 Julia Koval <julia.koval@intel.com>
11464 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11465 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11466 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11467 _mm_mask_bitshuffle_epi64_mask): Fix type.
11468 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11469 USI_FTYPE_V4DI_V4DI_USI): Remove.
11470 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11471 __builtin_ia32_vpshufbitqmb256_mask,
11472 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11473 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11474 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11476 2018-01-26 Alan Modra <amodra@gmail.com>
11479 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11480 UNSPEC_VBPERMQ. Sort other unspecs.
11482 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11484 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11486 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11488 PR middle-end/83055
11489 * predict.c (drop_profile): Do not push/pop cfun; update also
11491 (handle_missing_profiles): Fix logic looking for zero profiles.
11493 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11495 PR middle-end/83977
11496 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11497 on functions with #pragma omp declare simd or functions with simd
11499 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11500 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11501 Remove trailing \n from warning_at calls.
11503 2018-01-25 Tom de Vries <tom@codesourcery.com>
11506 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11507 for neutered workers.
11509 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11512 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11513 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11515 2018-01-24 Jeff Law <law@redhat.com>
11518 * i386.c (get_probe_interval): Move to earlier point.
11519 (ix86_compute_frame_layout): If -fstack-clash-protection and
11520 the frame is larger than the probe interval, then use pushes
11521 to save registers rather than reg->mem moves.
11522 (ix86_expand_prologue): Remove conditional for int_registers_saved
11525 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11528 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11529 min/max for never referenced object.
11531 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11533 PR middle-end/83977
11534 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11536 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11537 attributes from DECL_ATTRIBUTES (decl) without affecting
11538 DECL_ATTRIBUTES (current_function_decl).
11539 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11540 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11542 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11544 PR tree-optimization/83979
11545 * fold-const.c (fold_comparison): Use constant_boolean_node
11546 instead of boolean_{true,false}_node.
11548 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11550 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11553 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11555 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11556 Simplify the clause that sets the length attribute.
11557 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11558 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11559 clause that sets the length attribute.
11560 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11562 2018-01-24 Tom de Vries <tom@codesourcery.com>
11565 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11566 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11567 Add strict parameter.
11568 (prevent_branch_around_nothing): Insert dummy insn between branch to
11569 label and label with no ptx insn inbetween.
11570 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11572 2018-01-24 Tom de Vries <tom@codesourcery.com>
11575 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11576 for neutered threads in warp.
11577 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11579 2018-01-24 Richard Biener <rguenther@suse.de>
11581 PR tree-optimization/83176
11582 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11585 2018-01-24 Richard Biener <rguenther@suse.de>
11587 PR tree-optimization/82819
11588 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11589 code generating pluses that are no-ops in the target precision.
11591 2018-01-24 Richard Biener <rguenther@suse.de>
11593 PR middle-end/84000
11594 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11596 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11598 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11599 to merge probabilities.
11600 * predict.c (probably_never_executed): Also mark as cold functions
11601 with global 0 profile and guessed local profile.
11602 * profile-count.c (profile_probability::combine_with_count): New
11604 * profile-count.h (profile_probability::operator*,
11605 profile_probability::operator*=, profile_probability::operator/,
11606 profile_probability::operator/=): Reduce precision to adjusted
11607 and set value to guessed on contradictory divisions.
11608 (profile_probability::combine_with_freq): Remove.
11609 (profile_probability::combine_wiht_count): Declare.
11610 (profile_count::force_nonzero):: Set to adjusted.
11611 (profile_count::probability_in):: Set quality to adjusted.
11612 * tree-ssa-tail-merge.c (replace_block_by): Use
11613 combine_with_count.
11615 2018-01-23 Andrew Waterman <andrew@sifive.com>
11616 Jim Wilson <jimw@sifive.com>
11618 * config/riscv/riscv.c (riscv_stack_boundary): New.
11619 (riscv_option_override): Set riscv_stack_boundary. Handle
11620 riscv_preferred_stack_boundary_arg.
11621 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11622 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11623 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11624 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11625 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11626 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11628 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11631 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11632 of struct ix86_frame.
11633 (ix86_expand_epilogue): Likewise. Add a local variable for
11634 the reg_save_offset field in struct ix86_frame.
11636 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11638 PR tree-optimization/82604
11639 * tree-loop-distribution.c (enum partition_kind): New enum item
11640 PKIND_PARTIAL_MEMSET.
11641 (partition_builtin_p): Support above new enum item.
11642 (generate_code_for_partition): Ditto.
11643 (compute_access_range): Differentiate cases that equality can be
11644 proven at all loops, the innermost loops or no loops.
11645 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11646 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11647 (finalize_partitions, distribute_loop): Don't fuse partition of
11648 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11649 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11650 parloop is enabled.
11652 2018-01-23 Martin Liska <mliska@suse.cz>
11654 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11655 order to ignore the predictor.
11656 (PRED_POLYMORPHIC_CALL): Likewise.
11657 (PRED_RECURSIVE_CALL): Likewise.
11659 2018-01-23 Martin Liska <mliska@suse.cz>
11661 * tree-profile.c (tree_profiling): Print function header to
11662 aware reader which function we are working on.
11663 * value-prof.c (gimple_find_values_to_profile): Do not print
11664 not interesting value histograms.
11666 2018-01-23 Martin Liska <mliska@suse.cz>
11668 * profile-count.h (enum profile_quality): Add
11669 profile_uninitialized as the first value. Do not number values
11670 as they are zero based.
11671 (profile_count::verify): Update sanity check.
11672 (profile_probability::verify): Likewise.
11674 2018-01-23 Nathan Sidwell <nathan@acm.org>
11676 * doc/invoke.texi (ffor-scope): Deprecate.
11678 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11680 PR tree-optimization/83510
11681 * domwalk.c (set_all_edges_as_executable): New function.
11682 (dom_walker::dom_walker): Convert bool param
11683 "skip_unreachable_blocks" to enum reachability. Move setup of
11684 edge flags to set_all_edges_as_executable and only do it when
11685 reachability is REACHABLE_BLOCKS.
11686 * domwalk.h (enum dom_walker::reachability): New enum.
11687 (dom_walker::dom_walker): Convert bool param
11688 "skip_unreachable_blocks" to enum reachability.
11689 (set_all_edges_as_executable): New decl.
11690 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11691 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11693 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11694 but converting true to REACHABLE_BLOCKS.
11695 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11697 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11698 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11699 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11701 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11702 if check_all_array_refs will be called.
11704 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11706 * tree.c (selftest::test_location_wrappers): Add more test
11709 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11711 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11712 (selftest::test_bit_in_range): Likewise.
11714 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11717 * doc/sourcebuild.texi (vect_float): Say that the selector
11718 only describes the situation when -funsafe-math-optimizations is on.
11719 (vect_float_strict): Document.
11721 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11723 PR tree-optimization/83965
11724 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11725 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11726 instead of checking only for a reduction.
11727 (vect_recog_widen_sum_pattern): Likewise.
11729 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11731 * predict.c (probably_never_executed): Only use precise profile info.
11732 (compute_function_frequency): Skip after inlining hack since we now
11733 have quality checking.
11735 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11737 * profile-count.h (profile_probability::very_unlikely,
11738 profile_probability::unlikely, profile_probability::even): Set
11739 precision to guessed.
11741 2018-01-23 Richard Biener <rguenther@suse.de>
11743 PR tree-optimization/83963
11744 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11745 Properly terminate dominator walk when crossing the exit edge not
11746 when visiting its source block.
11748 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11751 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11752 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11754 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11756 PR tree-optimization/83957
11757 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11758 semicolon after for body surrounded by braces.
11760 PR tree-optimization/83081
11761 * profile-count.h (profile_probability::split): New method.
11762 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11763 Use profile_probability::split.
11764 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11765 when splitting a single conditional jump into 2.
11767 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11769 PR tree-optimization/69452
11770 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11775 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11776 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11777 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11779 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11781 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11783 * config/rl78/rl78.md (movdi): New define_expand.
11784 * config/rl78/rl78.c (rl78_split_movdi): New function.
11786 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11789 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11791 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11792 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11793 128-bit to produce an UNSPEC move to get the double word with the
11794 signbit and then a shift directly to do signbit.
11795 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11796 implementation with a new version that just does either a direct
11797 move or a regular move. Move memory interface to separate insns.
11798 Move insns so they are next to the expander.
11799 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11800 with signbit move. Split big and little endian case.
11801 (signbit<mode>2_dm_mem_le): Likewise.
11802 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11803 (signbit<mode>2_dm2): Likewise.
11805 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11807 * config/rl78/rl78.md (anddi3): New define_expand.
11809 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11811 * config/rl78/rl78.md (umindi3): New define_expand.
11813 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11815 * config/rl78/rl78.md (smindi3): New define_expand.
11817 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11819 * config/rl78/rl78.md (smaxdi3): New define_expand.
11821 2018-01-22 Carl Love <cel@us.ibm.com>
11823 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11824 LVX_V1TI): Add macro expansion.
11825 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11826 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11827 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11828 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11829 Change check to determine if the instruction is a byte reversing
11830 entry. Fix typo in comment.
11831 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11832 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11833 Add def_builtin calls for new builtins.
11834 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11835 Add define_insn expansion.
11837 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11839 * config/rl78/rl78.md (umaxdi3): New define_expand.
11841 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11843 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11844 for non-QImode registers.
11846 2018-01-22 Richard Biener <rguenther@suse.de>
11848 PR tree-optimization/83963
11849 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11850 including the loop exit block.
11851 (scop_detection::merge_sese): Likewise.
11852 (scop_detection::add_scop): Do it here instead.
11854 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11856 * doc/sourcebuild.texi (arm_softfloat): Document.
11858 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11861 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11862 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11863 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11865 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11866 David Edelsohn <dje.gcc@gmail.com>
11869 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11870 Change "crset eq" to "crset 2".
11871 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11872 (*call_indirect_aix<mode>_nospec): Likewise.
11873 (*call_value_indirect_aix<mode>_nospec): Likewise.
11874 (*call_indirect_elfv2<mode>_nospec): Likewise.
11875 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11876 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11877 change assembly output from . to $.
11878 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11879 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11880 (*tablejump<mode>_internal1_nospec): Likewise.
11882 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11885 * config/sh/sh_optimize_sett_clrt.cc:
11886 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11888 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11890 PR tree-optimization/83940
11891 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11892 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11893 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11894 use it to pass back the definition type.
11895 (vect_check_store_rhs): Likewise rhs_dt_out.
11896 (vect_build_gather_load_calls): Add a mask_dt argument and use
11897 it instead of a call to vect_is_simple_use.
11898 (vectorizable_store): Update calls to vect_check_load_store_mask
11899 and vect_check_store_rhs. Use the dt returned by the latter instead
11900 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11901 instead of calls to vect_is_simple_use. Pass the scalar rather
11902 than the vector operand to vect_is_simple_use when handling
11903 second and subsequent copies of an rhs value.
11904 (vectorizable_load): Update calls to vect_check_load_store_mask
11905 and vect_build_gather_load_calls. Use the cached mask_dt and
11906 gs_info.offset_dt instead of calls to vect_is_simple_use.
11908 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11910 PR middle-end/83945
11911 * tree-emutls.c: Include gimplify.h.
11912 (lower_emutls_2): New function.
11913 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11914 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11915 it before further processing.
11918 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11919 UINTVAL (trueop1) instead of INTVAL (op1).
11921 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11925 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11926 INCOMING_FRAME_SP_OFFSET if not defined.
11927 (scan_trace): Add ENTRY argument. If true and
11928 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11929 emit a note to adjust the CFA offset.
11930 (create_cfi_notes): Adjust scan_trace callers.
11931 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11932 INCOMING_FRAME_SP_OFFSET in the CIE.
11933 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11934 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11936 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11937 * doc/tm.texi: Regenerated.
11939 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11941 PR rtl-optimization/83147
11942 * lra-constraints.c (remove_inheritance_pseudos): Use
11943 lra_substitute_pseudo_within_insn.
11945 2018-01-19 Tom de Vries <tom@codesourcery.com>
11946 Cesar Philippidis <cesar@codesourcery.com>
11949 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11951 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11954 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11955 spaces for function labels.
11957 2018-01-19 Martin Liska <mliska@suse.cz>
11959 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11960 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11961 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11962 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11963 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11964 (PRED_CONST_RETURN): Change from 69 to 65.
11965 (PRED_NULL_RETURN): Change from 91 to 71.
11966 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11967 (PRED_LOOP_GUARD): Change from 66 to 73.
11969 2018-01-19 Martin Liska <mliska@suse.cz>
11971 * predict.c (predict_insn_def): Add new assert.
11972 (struct branch_predictor): Change type to signed integer.
11973 (test_prediction_value_range): Amend test to cover
11974 PROB_UNINITIALIZED.
11975 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11976 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11977 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11978 (PRED_LOOP_IV_COMPARE): Likewise.
11979 * predict.h (PROB_UNINITIALIZED): Define new constant.
11981 2018-01-19 Martin Liska <mliska@suse.cz>
11983 * predict.c (dump_prediction): Add new format for
11984 analyze_brprob.py script which is enabled with -details
11986 * profile-count.h (precise_p): New function.
11988 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11990 PR tree-optimization/83922
11991 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11992 there are no statements that need masking.
11993 (vect_active_double_reduction_p): New function.
11994 (vect_analyze_loop_operations): Use it when handling phis that
11995 are not in the loop header.
11997 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11999 PR tree-optimization/83914
12000 * tree-vect-loop.c (vectorizable_induction): Don't convert
12001 init_expr or apply the peeling adjustment for inductions
12002 that are nested within the vectorized loop.
12004 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12006 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12009 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12013 * function.h (gimplify_parameters): Add gimple_seq * argument.
12014 * function.c: Include gimple.h and options.h.
12015 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12016 for the added local temporaries if needed.
12017 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12018 if there are any parameter cleanups, wrap whole body into a
12019 try/finally with the cleanups.
12021 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12024 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12025 Use GET_MODE_CLASS for scalar floating point.
12027 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12031 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12032 Fix call of call_cgraph_insertion_hooks.
12034 2018-01-18 Martin Sebor <msebor@redhat.com>
12036 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12038 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12041 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12044 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12047 * common.opt: (-ffile-prefix-map): New option.
12048 * opts.c (common_handle_option): Defer it.
12049 * opts-global.c (handle_common_deferred_options): Handle it.
12050 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12051 * file-prefix-map.h: New file.
12052 (remap_debug_filename, add_debug_prefix_map): ...here.
12053 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12054 * final.c (debug_prefix_map, add_debug_prefix_map
12055 remap_debug_filename): Move to...
12056 * file-prefix-map.c: New file.
12057 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12058 generalize, get rid of alloca(), use strrchr() instead of strchr().
12059 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12060 Implement in terms of add_prefix_map().
12061 (remap_macro_filename, remap_debug_filename): Implement in term of
12063 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12064 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12065 * dbxout.c: Include file-prefix-map.h.
12066 * varasm.c: Likewise.
12067 * vmsdbgout.c: Likewise.
12068 * xcoffout.c: Likewise.
12069 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12070 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12071 * doc/invoke.texi (-ffile-prefix-map): Document.
12072 (-fdebug-prefix-map): Update description.
12074 2018-01-18 Martin Liska <mliska@suse.cz>
12076 * config/i386/i386.c (indirect_thunk_name): Document that also
12078 (output_indirect_thunk): Document why both instructions
12079 (pause and lfence) are generated.
12081 2018-01-18 Richard Biener <rguenther@suse.de>
12083 PR tree-optimization/83887
12084 * graphite-scop-detection.c
12085 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12086 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12087 (scop_detection::merge_sese): Re-implement with a flood-fill
12088 algorithm that properly finds a SESE region if it exists.
12090 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12093 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12094 pointer_diff optimizations use view_convert instead of convert.
12096 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12098 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12099 Generate different code for -mno-speculate-indirect-jumps.
12100 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12101 (*call_indirect_aix<mode>): Disable for
12102 -mno-speculate-indirect-jumps.
12103 (*call_indirect_aix<mode>_nospec): New define_insn.
12104 (*call_value_indirect_aix<mode>): Disable for
12105 -mno-speculate-indirect-jumps.
12106 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12107 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12108 -mno-speculate-indirect-jumps.
12109 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12111 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12113 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12114 long double type, set the flags for noting the default long double
12115 type, even if we don't pass or return a long double type.
12117 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12120 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12123 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12125 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12126 support for merge[hl].
12127 (fold_mergehl_helper): New helper function.
12128 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12129 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12130 (altivec_vmrglw_direct): Add xxmrglw insn.
12132 2018-01-17 Andrew Waterman <andrew@sifive.com>
12134 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12135 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12137 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12140 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12141 call the lto_location_cache before reading the
12142 DECL_SOURCE_LOCATION of the types.
12144 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12145 Richard Sandiford <richard.sandiford@linaro.org>
12147 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12148 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12149 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12150 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12151 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12153 * config/aarch64/constraints.md (aarch64_movti_operand):
12155 * config/aarch64/predicates.md (Uti): Add new constraint.
12157 2018-01-17 Carl Love <cel@us.ibm.com>
12159 * config/rs6000/vsx.md (define_expand xl_len_r,
12160 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12161 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12163 (define_expand, define_insn): Move the shift left from the
12164 define_insn to the define_expand for lxvl and stxvl instructions.
12165 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12166 and XL_LEN_R definitions to PURE.
12168 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12170 * config/i386/i386.c (indirect_thunk_name): Declare regno
12171 as unsigned int. Compare regno with INVALID_REGNUM.
12172 (output_indirect_thunk): Ditto.
12173 (output_indirect_thunk_function): Ditto.
12174 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12175 in the call to output_indirect_thunk_function.
12177 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12179 PR middle-end/83884
12180 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12181 rather than the size of inner_type to determine the stack slot size
12182 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12184 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12187 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12190 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12192 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12193 endian Linux systems to optionally enable multilibs for selecting
12194 the long double type if the user configured an explicit type.
12195 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12196 have no long double multilibs if not defined.
12197 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12198 warn if the user used -mabi={ieee,ibm}longdouble and we built
12199 multilibs for long double.
12200 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12201 appropriate multilib option.
12202 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12204 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12205 for building long double multilibs.
12206 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12208 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12210 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12213 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12215 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12218 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12221 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12224 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12226 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12227 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12229 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12231 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12232 different rtl trees depending on TARGET_64BIT.
12233 (rs6000_gen_lvx): Likewise.
12235 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12237 * config/visium/visium.md (nop): Tweak comment.
12238 (hazard_nop): Likewise.
12240 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12242 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12243 -mspeculate-indirect-jumps.
12244 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12245 for -mno-speculate-indirect-jumps.
12246 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12247 (*call_value_indirect_elfv2<mode>): Disable for
12248 -mno-speculate-indirect-jumps.
12249 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12250 (indirect_jump): Emit different RTL for
12251 -mno-speculate-indirect-jumps.
12252 (*indirect_jump<mode>): Disable for
12253 -mno-speculate-indirect-jumps.
12254 (*indirect_jump<mode>_nospec): New define_insn.
12255 (tablejump): Emit different RTL for
12256 -mno-speculate-indirect-jumps.
12257 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12258 (tablejumpsi_nospec): New define_expand.
12259 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12260 (tablejumpdi_nospec): New define_expand.
12261 (*tablejump<mode>_internal1): Disable for
12262 -mno-speculate-indirect-jumps.
12263 (*tablejump<mode>_internal1_nospec): New define_insn.
12264 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12267 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12269 * caller-save.c (insert_save): Drop unnecessary parameter. All
12272 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12273 Richard Biener <rguenth@suse.de>
12276 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12277 return early, inline manually is_gimple_sizepos. Make sure if we
12278 call gimplify_expr we don't end up with a gimple constant.
12279 * tree.c (variably_modified_type_p): Don't return true for
12280 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12281 * gimplify.h (is_gimple_sizepos): Remove.
12283 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12285 PR tree-optimization/83857
12286 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12287 vectorizable_live_operation for pure SLP statements.
12288 (vectorizable_live_operation): Handle PHIs.
12290 2018-01-16 Richard Biener <rguenther@suse.de>
12292 PR tree-optimization/83867
12293 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12294 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12296 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12299 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12300 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12301 If off is not INTEGER_CST, issue a may not be aligned warning
12302 rather than isn't aligned. Use isn%'t rather than isn't.
12303 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12305 <case MULT_EXPR>: Improve the case when bottom and one of the
12306 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12307 operand, in that case check if the other operand is multiple of
12308 bottom divided by the INTEGER_CST operand.
12310 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12313 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12314 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12315 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12316 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12317 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12318 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12319 (pa_function_arg_size): New function.
12321 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12323 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12324 in a separate statement.
12326 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12328 PR tree-optimization/83847
12329 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12330 group gathers and scatters.
12332 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12334 PR rtl-optimization/86620
12335 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12337 PR rtl-optimization/83213
12338 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12339 to last if both are JUMP_INSNs.
12341 PR tree-optimization/83843
12342 * gimple-ssa-store-merging.c
12343 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12344 store_immediate_info for bswap/nop orig_stores.
12346 2018-01-15 Andrew Waterman <andrew@sifive.com>
12348 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12350 <UDIV>: Increase cost if !TARGET_DIV.
12352 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12354 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12355 (define_attr "cr_logical_3op"): New.
12356 (cceq_ior_compare): Adjust.
12357 (cceq_ior_compare_complement): Adjust.
12358 (*cceq_rev_compare): Adjust.
12359 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12360 (is_cracked_insn): Adjust.
12361 (insn_must_be_first_in_group): Adjust.
12362 * config/rs6000/40x.md: Adjust.
12363 * config/rs6000/440.md: Adjust.
12364 * config/rs6000/476.md: Adjust.
12365 * config/rs6000/601.md: Adjust.
12366 * config/rs6000/603.md: Adjust.
12367 * config/rs6000/6xx.md: Adjust.
12368 * config/rs6000/7450.md: Adjust.
12369 * config/rs6000/7xx.md: Adjust.
12370 * config/rs6000/8540.md: Adjust.
12371 * config/rs6000/cell.md: Adjust.
12372 * config/rs6000/e300c2c3.md: Adjust.
12373 * config/rs6000/e500mc.md: Adjust.
12374 * config/rs6000/e500mc64.md: Adjust.
12375 * config/rs6000/e5500.md: Adjust.
12376 * config/rs6000/e6500.md: Adjust.
12377 * config/rs6000/mpc.md: Adjust.
12378 * config/rs6000/power4.md: Adjust.
12379 * config/rs6000/power5.md: Adjust.
12380 * config/rs6000/power6.md: Adjust.
12381 * config/rs6000/power7.md: Adjust.
12382 * config/rs6000/power8.md: Adjust.
12383 * config/rs6000/power9.md: Adjust.
12384 * config/rs6000/rs64.md: Adjust.
12385 * config/rs6000/titan.md: Adjust.
12387 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12389 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12390 ix86_indirect_branch_register logic.
12392 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12394 * config/i386/constraints.md (Bs): Update
12395 ix86_indirect_branch_register check. Don't check
12396 ix86_indirect_branch_register with GOT_memory_operand.
12398 * config/i386/predicates.md (GOT_memory_operand): Don't check
12399 ix86_indirect_branch_register here.
12400 (GOT32_symbol_operand): Likewise.
12402 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12404 * config/i386/predicates.md (constant_call_address_operand):
12405 Rewrite ix86_indirect_branch_register logic.
12406 (sibcall_insn_operand): Likewise.
12408 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12410 * config/i386/constraints.md (Bs): Replace
12411 ix86_indirect_branch_thunk_register with
12412 ix86_indirect_branch_register.
12414 * config/i386/i386.md (indirect_jump): Likewise.
12415 (tablejump): Likewise.
12416 (*sibcall_memory): Likewise.
12417 (*sibcall_value_memory): Likewise.
12418 Peepholes of indirect call and jump via memory: Likewise.
12419 * config/i386/i386.opt: Likewise.
12420 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12421 (GOT_memory_operand): Likewise.
12422 (call_insn_operand): Likewise.
12423 (sibcall_insn_operand): Likewise.
12424 (GOT32_symbol_operand): Likewise.
12426 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12428 PR middle-end/83837
12429 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12430 type rather than type addr's type points to.
12431 (expand_omp_atomic_mutex): Likewise.
12432 (expand_omp_atomic): Likewise.
12434 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12437 * config/i386/i386.c (output_indirect_thunk_function): Use
12438 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12439 for __x86_return_thunk.
12441 2018-01-15 Richard Biener <rguenther@suse.de>
12443 PR middle-end/83850
12444 * expmed.c (extract_bit_field_1): Fix typo.
12446 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12449 * config/arm/iterators.md (VF): New mode iterator.
12450 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12451 Remove integer-related logic from pattern.
12452 (neon_vabd<mode>_3): Likewise.
12454 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12456 PR middle-end/82694
12457 * common.opt (fstrict-overflow): No longer an alias.
12458 (fwrapv-pointer): New option.
12459 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12460 also for pointer types based on flag_wrapv_pointer.
12461 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12462 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12463 opts->x_flag_wrapv got set.
12464 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12465 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12466 POINTER_TYPE_OVERFLOW_UNDEFINED.
12467 * match.pd: Likewise in address comparison pattern.
12468 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12470 2018-01-15 Richard Biener <rguenther@suse.de>
12473 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12474 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12475 Reset type names to their identifier if their TYPE_DECL doesn't
12476 have linkage (and thus is used for ODR and devirt).
12477 (save_debug_info_for_decl): Remove.
12478 (save_debug_info_for_type): Likewise.
12479 (add_tree_to_fld_list): Adjust.
12480 * tree-pretty-print.c (dump_generic_node): Make dumping of
12481 type names more robust.
12483 2018-01-15 Richard Biener <rguenther@suse.de>
12485 * BASE-VER: Bump to 8.0.1.
12487 2018-01-14 Martin Sebor <msebor@redhat.com>
12490 * builtins.c (check_access): Avoid warning when the no-warning bit
12493 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12495 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12496 * ira-color (allocno_hard_regs_compare): Likewise.
12498 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12501 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12502 Use .pushsection/.popsection.
12504 2018-01-14 Martin Sebor <msebor@redhat.com>
12507 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12509 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12511 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12512 entry from extra_headers.
12513 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12514 extra_headers, make the list bitwise identical to the i?86-*-* one.
12516 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12518 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12519 -mcmodel=large with -mindirect-branch=thunk,
12520 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12521 -mfunction-return=thunk-extern.
12522 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12523 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12524 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12526 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12528 * config/i386/i386.c (print_reg): Print the name of the full
12529 integer register without '%'.
12530 (ix86_print_operand): Handle 'V'.
12531 * doc/extend.texi: Document 'V' modifier.
12533 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12535 * config/i386/constraints.md (Bs): Disallow memory operand for
12536 -mindirect-branch-register.
12538 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12539 (GOT_memory_operand): Likewise.
12540 (call_insn_operand): Likewise.
12541 (sibcall_insn_operand): Likewise.
12542 (GOT32_symbol_operand): Likewise.
12543 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12544 for -mindirect-branch-register.
12545 (tablejump): Likewise.
12546 (*sibcall_memory): Likewise.
12547 (*sibcall_value_memory): Likewise.
12548 Disallow peepholes of indirect call and jump via memory for
12549 -mindirect-branch-register.
12550 (*call_pop): Replace m with Bw.
12551 (*call_value_pop): Likewise.
12552 (*sibcall_pop_memory): Replace m with Bs.
12553 * config/i386/i386.opt (mindirect-branch-register): New option.
12554 * doc/invoke.texi: Document -mindirect-branch-register option.
12556 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12558 * config/i386/i386-protos.h (ix86_output_function_return): New.
12559 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12560 set function_return_type.
12561 (indirect_thunk_name): Add ret_p to indicate thunk for function
12563 (output_indirect_thunk_function): Pass false to
12564 indirect_thunk_name.
12565 (ix86_output_indirect_branch_via_reg): Likewise.
12566 (ix86_output_indirect_branch_via_push): Likewise.
12567 (output_indirect_thunk_function): Create alias for function
12568 return thunk if regno < 0.
12569 (ix86_output_function_return): New function.
12570 (ix86_handle_fndecl_attribute): Handle function_return.
12571 (ix86_attribute_table): Add function_return.
12572 * config/i386/i386.h (machine_function): Add
12573 function_return_type.
12574 * config/i386/i386.md (simple_return_internal): Use
12575 ix86_output_function_return.
12576 (simple_return_internal_long): Likewise.
12577 * config/i386/i386.opt (mfunction-return=): New option.
12578 (indirect_branch): Mention -mfunction-return=.
12579 * doc/extend.texi: Document function_return function attribute.
12580 * doc/invoke.texi: Document -mfunction-return= option.
12582 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12584 * config/i386/i386-opts.h (indirect_branch): New.
12585 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12586 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12587 with local indirect jump when converting indirect call and jump.
12588 (ix86_set_indirect_branch_type): New.
12589 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12590 (indirectlabelno): New.
12591 (indirect_thunk_needed): Likewise.
12592 (indirect_thunk_bnd_needed): Likewise.
12593 (indirect_thunks_used): Likewise.
12594 (indirect_thunks_bnd_used): Likewise.
12595 (INDIRECT_LABEL): Likewise.
12596 (indirect_thunk_name): Likewise.
12597 (output_indirect_thunk): Likewise.
12598 (output_indirect_thunk_function): Likewise.
12599 (ix86_output_indirect_branch_via_reg): Likewise.
12600 (ix86_output_indirect_branch_via_push): Likewise.
12601 (ix86_output_indirect_branch): Likewise.
12602 (ix86_output_indirect_jmp): Likewise.
12603 (ix86_code_end): Call output_indirect_thunk_function if needed.
12604 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12606 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12607 (ix86_attribute_table): Add indirect_branch.
12608 * config/i386/i386.h (machine_function): Add indirect_branch_type
12609 and has_local_indirect_jump.
12610 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12612 (tablejump): Likewise.
12613 (*indirect_jump): Use ix86_output_indirect_jmp.
12614 (*tablejump_1): Likewise.
12615 (simple_return_indirect_internal): Likewise.
12616 * config/i386/i386.opt (mindirect-branch=): New option.
12617 (indirect_branch): New.
12620 (thunk-inline): Likewise.
12621 (thunk-extern): Likewise.
12622 * doc/extend.texi: Document indirect_branch function attribute.
12623 * doc/invoke.texi: Document -mindirect-branch= option.
12625 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12628 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12630 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12632 * ipa-inline.c (want_inline_small_function_p): Return false if
12633 inlining has already failed with CIF_FINAL_ERROR.
12634 (update_caller_keys): Call want_inline_small_function_p before
12636 (update_callee_keys): Likewise.
12638 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12640 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12642 (rs6000_quadword_masked_address_p): Likewise.
12643 (quad_aligned_load_p): Likewise.
12644 (quad_aligned_store_p): Likewise.
12645 (const_load_sequence_p): Add comment to describe the outer-most loop.
12646 (mimic_memory_attributes_and_flags): New function.
12647 (rs6000_gen_stvx): Likewise.
12648 (replace_swapped_aligned_store): Likewise.
12649 (rs6000_gen_lvx): Likewise.
12650 (replace_swapped_aligned_load): Likewise.
12651 (replace_swapped_load_constant): Capitalize argument name in
12652 comment describing this function.
12653 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12654 and stores that access quad-word aligned addresses and replace
12655 with stvx or lvx instructions when appropriate.
12656 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12657 New function prototype.
12658 (rs6000_quadword_masked_address_p): Likewise.
12659 (rs6000_gen_lvx): Likewise.
12660 (rs6000_gen_stvx): Likewise.
12661 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12662 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12663 when memory address is aligned.
12664 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12665 this split to select lvx instruction when memory address is aligned.
12666 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12667 instruction when memory address is aligned.
12668 (*vsx_le_perm_load_v16qi): Likewise.
12669 (four unnamed splitters): Modify to select the stvx instruction
12670 when memory is aligned.
12672 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12674 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12675 which appears in the queue multiple times.
12677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12678 Alan Hayward <alan.hayward@arm.com>
12679 David Sherwood <david.sherwood@arm.com>
12681 * tree-vectorizer.h (vec_lower_bound): New structure.
12682 (_loop_vec_info): Add check_nonzero and lower_bounds.
12683 (LOOP_VINFO_CHECK_NONZERO): New macro.
12684 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12685 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12686 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12687 fields. Make seg_len the distance travelled, not including the
12689 (dr_direction_indicator): Declare.
12690 (dr_zero_step_indicator): Likewise.
12691 (dr_known_forward_stride_p): Likewise.
12692 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12694 (runtime_alias_check_p): Allow runtime alias checks with
12696 (operator ==): Compare access_size and align.
12697 (prune_runtime_alias_test_list): Rework for new distinction between
12698 the access_size and seg_len.
12699 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12701 (get_segment_min_max): New function.
12702 (create_intersect_range_checks): Use it.
12703 (dr_step_indicator): New function.
12704 (dr_direction_indicator): Likewise.
12705 (dr_zero_step_indicator): Likewise.
12706 (dr_known_forward_stride_p): Likewise.
12707 * tree-loop-distribution.c (data_ref_segment_size): Return
12708 DR_STEP * (niters - 1).
12709 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12711 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12712 (vect_preserves_scalar_order_p): New function, split out from...
12713 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12714 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12715 (vect_vfa_access_size): New function.
12716 (vect_vfa_align): Likewise.
12717 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12718 (dump_lower_bound): New function.
12719 (vect_check_lower_bound): Likewise.
12720 (vect_small_gap_p): Likewise.
12721 (vectorizable_with_step_bound_p): Likewise.
12722 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12723 depencies if the vectorization factor is 1. Convert the checks
12724 for nonzero steps into checks on the bounds of DR_STEP. Try using
12725 a bunds check for variable steps if the minimum required step is
12726 relatively small. Update calls to the dr_with_seg_len
12727 constructor and to vect_compile_time_alias.
12728 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12730 (vect_loop_versioning): Call it.
12731 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12733 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * doc/sourcebuild.texi (vect_scatter_store): Document.
12740 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12742 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12744 * genopinit.c (main): Add supports_vec_scatter_store and
12745 supports_vec_scatter_store_cached to target_optabs.
12746 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12747 IFN_MASK_SCATTER_STORE.
12748 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12750 * internal-fn.h (internal_store_fn_p): Declare.
12751 (internal_fn_stored_value_index): Likewise.
12752 * internal-fn.c (scatter_store_direct): New macro.
12753 (expand_scatter_store_optab_fn): New function.
12754 (direct_scatter_store_optab_supported_p): New macro.
12755 (internal_store_fn_p): New function.
12756 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12757 IFN_MASK_SCATTER_STORE.
12758 (internal_fn_mask_index): Likewise.
12759 (internal_fn_stored_value_index): New function.
12760 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12761 for scatter stores.
12762 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12763 * optabs-query.c (supports_vec_scatter_store_p): New function.
12764 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12765 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12766 true for scatter stores.
12767 (vect_gather_scatter_fn_p): Handle scatter stores too.
12768 (vect_check_gather_scatter): Consider using scatter stores if
12769 supports_vec_scatter_store_p.
12770 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12771 scatter stores too.
12772 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12773 internal_fn_stored_value_index.
12774 (check_load_store_masking): Handle scatter stores too.
12775 (vect_get_store_rhs): Make public.
12776 (vectorizable_call): Use internal_store_fn_p.
12777 (vectorizable_store): Handle scatter store internal functions.
12778 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12779 when deciding whether the end of the group has been reached.
12780 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12781 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12782 (mask_scatter_store<mode>): New insns.
12784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12785 Alan Hayward <alan.hayward@arm.com>
12786 David Sherwood <david.sherwood@arm.com>
12788 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12789 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12790 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12792 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12793 Use vect_truncate_gather_scatter_offset if we can't treat the
12794 operation as a normal gather load or scatter store.
12795 (get_group_load_store_type): Take the gather_scatter_info
12796 as argument. Try using a gather load or scatter store for
12797 single-element groups.
12798 (get_load_store_type): Update calls to get_group_load_store_type
12799 and vect_use_strided_gather_scatters_p.
12801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12802 Alan Hayward <alan.hayward@arm.com>
12803 David Sherwood <david.sherwood@arm.com>
12805 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12806 optional tree argument.
12807 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12809 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12810 but continue to use the current value as a fallback.
12811 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12812 to compare the updates.
12813 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12814 (get_load_store_type): Use it when handling a strided access.
12815 (vect_get_strided_load_store_ops): New function.
12816 (vect_get_data_ptr_increment): Likewise.
12817 (vectorizable_load): Handle strided gather loads. Always pass
12818 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12821 Alan Hayward <alan.hayward@arm.com>
12822 David Sherwood <david.sherwood@arm.com>
12824 * doc/md.texi (gather_load@var{m}): Document.
12825 (mask_gather_load@var{m}): Likewise.
12826 * genopinit.c (main): Add supports_vec_gather_load and
12827 supports_vec_gather_load_cached to target_optabs.
12828 * optabs-tree.c (init_tree_optimization_optabs): Use
12829 ggc_cleared_alloc to allocate target_optabs.
12830 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12831 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12833 * internal-fn.h (internal_load_fn_p): Declare.
12834 (internal_gather_scatter_fn_p): Likewise.
12835 (internal_fn_mask_index): Likewise.
12836 (internal_gather_scatter_fn_supported_p): Likewise.
12837 * internal-fn.c (gather_load_direct): New macro.
12838 (expand_gather_load_optab_fn): New function.
12839 (direct_gather_load_optab_supported_p): New macro.
12840 (direct_internal_fn_optab): New function.
12841 (internal_load_fn_p): Likewise.
12842 (internal_gather_scatter_fn_p): Likewise.
12843 (internal_fn_mask_index): Likewise.
12844 (internal_gather_scatter_fn_supported_p): Likewise.
12845 * optabs-query.c (supports_at_least_one_mode_p): New function.
12846 (supports_vec_gather_load_p): Likewise.
12847 * optabs-query.h (supports_vec_gather_load_p): Declare.
12848 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12849 and memory_type field.
12850 (NUM_PATTERNS): Bump to 15.
12851 * tree-vect-data-refs.c: Include internal-fn.h.
12852 (vect_gather_scatter_fn_p): New function.
12853 (vect_describe_gather_scatter_call): Likewise.
12854 (vect_check_gather_scatter): Try using internal functions for
12855 gather loads. Recognize existing calls to a gather load function.
12856 (vect_analyze_data_refs): Consider using gather loads if
12857 supports_vec_gather_load_p.
12858 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12859 (vect_get_gather_scatter_offset_type): Likewise.
12860 (vect_convert_mask_for_vectype): Likewise.
12861 (vect_add_conversion_to_patterm): Likewise.
12862 (vect_try_gather_scatter_pattern): Likewise.
12863 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12864 (vect_vect_recog_func_ptrs): Add it.
12865 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12866 internal_fn_mask_index and internal_gather_scatter_fn_p.
12867 (check_load_store_masking): Take the gather_scatter_info as an
12868 argument and handle gather loads.
12869 (vect_get_gather_scatter_ops): New function.
12870 (vectorizable_call): Check internal_load_fn_p.
12871 (vectorizable_load): Likewise. Handle gather load internal
12873 (vectorizable_store): Update call to check_load_store_masking.
12874 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12875 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12876 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12877 (aarch64_gather_scale_operand_d): New predicates.
12878 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12879 (mask_gather_load<mode>): New insns.
12881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12882 Alan Hayward <alan.hayward@arm.com>
12883 David Sherwood <david.sherwood@arm.com>
12885 * optabs.def (fold_left_plus_optab): New optab.
12886 * doc/md.texi (fold_left_plus_@var{m}): Document.
12887 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12888 * internal-fn.c (fold_left_direct): Define.
12889 (expand_fold_left_optab_fn): Likewise.
12890 (direct_fold_left_optab_supported_p): Likewise.
12891 * fold-const-call.c (fold_const_fold_left): New function.
12892 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12893 * tree-parloops.c (valid_reduction_p): New function.
12894 (gather_scalar_reductions): Use it.
12895 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12896 (vect_finish_replace_stmt): Declare.
12897 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12898 (needs_fold_left_reduction_p): New function, split out from...
12899 (vect_is_simple_reduction): ...here. Accept reductions that
12900 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12901 (vect_force_simple_reduction): Also store the reduction type in
12902 the assignment's STMT_VINFO_REDUC_TYPE.
12903 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12904 (merge_with_identity): New function.
12905 (vect_expand_fold_left): Likewise.
12906 (vectorize_fold_left_reduction): Likewise.
12907 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12908 scalar phi in place for it. Check for target support and reject
12909 cases that would reassociate the operation. Defer the transform
12910 phase to vectorize_fold_left_reduction.
12911 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12912 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12913 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12917 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12918 call to ifc_temp_var.
12920 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12921 Alan Hayward <alan.hayward@arm.com>
12922 David Sherwood <david.sherwood@arm.com>
12924 * target.def (legitimize_address_displacement): Take the original
12925 offset as a poly_int.
12926 * targhooks.h (default_legitimize_address_displacement): Update
12928 * targhooks.c (default_legitimize_address_displacement): Likewise.
12929 * doc/tm.texi: Regenerate.
12930 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12931 as an argument, moving assert of ad->disp == ad->disp_term to...
12932 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12933 Try calling targetm.legitimize_address_displacement before expanding
12934 the address rather than afterwards, and adjust for the new interface.
12935 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12936 Match the new hook interface. Handle SVE addresses.
12937 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12938 new hook interface.
12940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12942 * Makefile.in (OBJS): Add early-remat.o.
12943 * target.def (select_early_remat_modes): New hook.
12944 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12945 * doc/tm.texi: Regenerate.
12946 * targhooks.h (default_select_early_remat_modes): Declare.
12947 * targhooks.c (default_select_early_remat_modes): New function.
12948 * timevar.def (TV_EARLY_REMAT): New timevar.
12949 * passes.def (pass_early_remat): New pass.
12950 * tree-pass.h (make_pass_early_remat): Declare.
12951 * early-remat.c: New file.
12952 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12954 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12957 Alan Hayward <alan.hayward@arm.com>
12958 David Sherwood <david.sherwood@arm.com>
12960 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12961 vfm1 with a bound_epilog parameter.
12962 (vect_do_peeling): Update calls accordingly, and move the prologue
12963 call earlier in the function. Treat the base bound_epilog as 0 for
12964 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12965 this base when peeling for gaps.
12966 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12967 with fully-masked loops.
12968 (vect_estimate_min_profitable_iters): Handle the single peeled
12969 iteration in that case.
12971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12972 Alan Hayward <alan.hayward@arm.com>
12973 David Sherwood <david.sherwood@arm.com>
12975 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12976 single-element interleaving even if the size is not a power of 2.
12977 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12978 accesses for single-element interleaving if the group size is
12981 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12982 Alan Hayward <alan.hayward@arm.com>
12983 David Sherwood <david.sherwood@arm.com>
12985 * doc/md.texi (fold_extract_last_@var{m}): Document.
12986 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12987 * optabs.def (fold_extract_last_optab): New optab.
12988 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12989 * internal-fn.c (fold_extract_direct): New macro.
12990 (expand_fold_extract_optab_fn): Likewise.
12991 (direct_fold_extract_optab_supported_p): Likewise.
12992 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12993 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12994 EXTRACT_LAST_REDUCTION.
12995 (get_initial_def_for_reduction): Do not create an initial vector
12996 for EXTRACT_LAST_REDUCTION reductions.
12997 (vectorizable_reduction): Leave the scalar phi in place for
12998 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12999 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13000 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13001 transform phase to vectorizable_condition.
13002 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13004 (vect_finish_stmt_generation): ...here.
13005 (vect_finish_replace_stmt): New function.
13006 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13007 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13009 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13012 Alan Hayward <alan.hayward@arm.com>
13013 David Sherwood <david.sherwood@arm.com>
13015 * doc/md.texi (extract_last_@var{m}): Document.
13016 * optabs.def (extract_last_optab): New optab.
13017 * internal-fn.def (EXTRACT_LAST): New internal function.
13018 * internal-fn.c (cond_unary_direct): New macro.
13019 (expand_cond_unary_optab_fn): Likewise.
13020 (direct_cond_unary_optab_supported_p): Likewise.
13021 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13022 loops using EXTRACT_LAST.
13023 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13024 (extract_last_<mode>): ...this optab.
13025 (vec_extract<mode><Vel>): Update accordingly.
13027 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13028 Alan Hayward <alan.hayward@arm.com>
13029 David Sherwood <david.sherwood@arm.com>
13031 * target.def (empty_mask_is_expensive): New hook.
13032 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13033 * doc/tm.texi: Regenerate.
13034 * targhooks.h (default_empty_mask_is_expensive): Declare.
13035 * targhooks.c (default_empty_mask_is_expensive): New function.
13036 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13037 if the target says that empty masks are expensive.
13038 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13040 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13043 Alan Hayward <alan.hayward@arm.com>
13044 David Sherwood <david.sherwood@arm.com>
13046 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13047 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13048 (vect_use_loop_mask_for_alignment_p): New function.
13049 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13050 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13051 niters_skip argument. Make sure that the first niters_skip elements
13052 of the first iteration are inactive.
13053 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13054 Update call to vect_set_loop_masks_directly.
13055 (get_misalign_in_elems): New function, split out from...
13056 (vect_gen_prolog_loop_niters): ...here.
13057 (vect_update_init_of_dr): Take a code argument that specifies whether
13058 the adjustment should be added or subtracted.
13059 (vect_update_init_of_drs): Likewise.
13060 (vect_prepare_for_masked_peels): New function.
13061 (vect_do_peeling): Skip prologue peeling if we're using a mask
13062 instead. Update call to vect_update_inits_of_drs.
13063 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13065 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13066 alignment. Do not include the number of peeled iterations in
13067 the minimum threshold in that case.
13068 (vectorizable_induction): Adjust the start value down by
13069 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13070 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13071 Take the number of skipped iterations into account when calculating
13073 * tree-vect-stmts.c (vect_gen_while_not): New function.
13075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13076 Alan Hayward <alan.hayward@arm.com>
13077 David Sherwood <david.sherwood@arm.com>
13079 * doc/sourcebuild.texi (vect_fully_masked): Document.
13080 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13081 default value to 0.
13082 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13084 (vect_analyze_loop_2): ...here. Don't check the vectorization
13085 factor against the number of loop iterations if the loop is
13088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13089 Alan Hayward <alan.hayward@arm.com>
13090 David Sherwood <david.sherwood@arm.com>
13092 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13093 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13094 (dump_groups): Update accordingly.
13095 (iv_use::mem_type): New member variable.
13096 (address_p): New function.
13097 (record_use): Add a mem_type argument and initialize the new
13099 (record_group_use): Add a mem_type argument. Use address_p.
13100 Remove obsolete null checks of base_object. Update call to record_use.
13101 (find_interesting_uses_op): Update call to record_group_use.
13102 (find_interesting_uses_cond): Likewise.
13103 (find_interesting_uses_address): Likewise.
13104 (get_mem_type_for_internal_fn): New function.
13105 (find_address_like_use): Likewise.
13106 (find_interesting_uses_stmt): Try find_address_like_use before
13107 calling find_interesting_uses_op.
13108 (addr_offset_valid_p): Use the iv mem_type field as the type
13109 of the addressed memory.
13110 (add_autoinc_candidates): Likewise.
13111 (get_address_cost): Likewise.
13112 (split_small_address_groups_p): Use address_p.
13113 (split_address_groups): Likewise.
13114 (add_iv_candidate_for_use): Likewise.
13115 (autoinc_possible_for_pair): Likewise.
13116 (rewrite_groups): Likewise.
13117 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13118 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13119 (get_alias_ptr_type_for_ptr_address): New function.
13120 (rewrite_use_address): Rewrite address uses in calls that were
13121 identified by find_address_like_use.
13123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13124 Alan Hayward <alan.hayward@arm.com>
13125 David Sherwood <david.sherwood@arm.com>
13127 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13129 * gimple-expr.h (is_gimple_addressable: Likewise.
13130 * gimple-expr.c (is_gimple_address): Likewise.
13131 * internal-fn.c (expand_call_mem_ref): New function.
13132 (expand_mask_load_optab_fn): Use it.
13133 (expand_mask_store_optab_fn): Likewise.
13135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13136 Alan Hayward <alan.hayward@arm.com>
13137 David Sherwood <david.sherwood@arm.com>
13139 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13140 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13141 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13142 (cond_umax@var{mode}): Document.
13143 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13144 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13145 (cond_umin_optab, cond_umax_optab): New optabs.
13146 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13147 (COND_IOR, COND_XOR): New internal functions.
13148 * internal-fn.h (get_conditional_internal_fn): Declare.
13149 * internal-fn.c (cond_binary_direct): New macro.
13150 (expand_cond_binary_optab_fn): Likewise.
13151 (direct_cond_binary_optab_supported_p): Likewise.
13152 (get_conditional_internal_fn): New function.
13153 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13154 Cope with reduction statements that are vectorized as calls rather
13156 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13157 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13158 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13159 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13160 (UNSPEC_COND_EOR): New unspecs.
13161 (optab): Add mappings for them.
13162 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13163 (sve_int_op, sve_fp_op): New int attributes.
13165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13166 Alan Hayward <alan.hayward@arm.com>
13167 David Sherwood <david.sherwood@arm.com>
13169 * optabs.def (while_ult_optab): New optab.
13170 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13171 * internal-fn.def (WHILE_ULT): New internal function.
13172 * internal-fn.h (direct_internal_fn_supported_p): New override
13173 that takes two types as argument.
13174 * internal-fn.c (while_direct): New macro.
13175 (expand_while_optab_fn): New function.
13176 (convert_optab_supported_p): Likewise.
13177 (direct_while_optab_supported_p): New macro.
13178 * wide-int.h (wi::udiv_ceil): New function.
13179 * tree-vectorizer.h (rgroup_masks): New structure.
13180 (vec_loop_masks): New typedef.
13181 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13182 and fully_masked_p.
13183 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13184 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13185 (vect_max_vf): New function.
13186 (slpeel_make_loop_iterate_ntimes): Delete.
13187 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13188 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13189 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13190 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13191 internal-fn.h, stor-layout.h and optabs-query.h.
13192 (vect_set_loop_mask): New function.
13193 (add_preheader_seq): Likewise.
13194 (add_header_seq): Likewise.
13195 (interleave_supported_p): Likewise.
13196 (vect_maybe_permute_loop_masks): Likewise.
13197 (vect_set_loop_masks_directly): Likewise.
13198 (vect_set_loop_condition_masked): Likewise.
13199 (vect_set_loop_condition_unmasked): New function, split out from
13200 slpeel_make_loop_iterate_ntimes.
13201 (slpeel_make_loop_iterate_ntimes): Rename to..
13202 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13203 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13204 (vect_do_peeling): Update call accordingly.
13205 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13207 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13208 mask_compare_type, can_fully_mask_p and fully_masked_p.
13209 (release_vec_loop_masks): New function.
13210 (_loop_vec_info): Use it to free the loop masks.
13211 (can_produce_all_loop_masks_p): New function.
13212 (vect_get_max_nscalars_per_iter): Likewise.
13213 (vect_verify_full_masking): Likewise.
13214 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13215 retries, and free the mask rgroups before retrying. Check loop-wide
13216 reasons for disallowing fully-masked loops. Make the final decision
13217 about whether use a fully-masked loop or not.
13218 (vect_estimate_min_profitable_iters): Do not assume that peeling
13219 for the number of iterations will be needed for fully-masked loops.
13220 (vectorizable_reduction): Disable fully-masked loops.
13221 (vectorizable_live_operation): Likewise.
13222 (vect_halve_mask_nunits): New function.
13223 (vect_double_mask_nunits): Likewise.
13224 (vect_record_loop_mask): Likewise.
13225 (vect_get_loop_mask): Likewise.
13226 (vect_transform_loop): Handle the case in which the final loop
13227 iteration might handle a partial vector. Call vect_set_loop_condition
13228 instead of slpeel_make_loop_iterate_ntimes.
13229 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13230 (check_load_store_masking): New function.
13231 (prepare_load_store_mask): Likewise.
13232 (vectorizable_store): Handle fully-masked loops.
13233 (vectorizable_load): Likewise.
13234 (supportable_widening_operation): Use vect_halve_mask_nunits for
13236 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13237 (vect_gen_while): New function.
13238 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13239 (aarch64_uqdec<mode>): New insn.
13241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13242 Alan Hayward <alan.hayward@arm.com>
13243 David Sherwood <david.sherwood@arm.com>
13245 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13246 (reduc_xor_scal_optab): New optabs.
13247 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13248 (reduc_xor_scal_@var{m}): Document.
13249 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13250 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13251 internal functions.
13252 * fold-const-call.c (fold_const_call): Handle them.
13253 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13254 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13255 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13256 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13257 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13258 (UNSPEC_XORV): New unspecs.
13259 (optab): Add entries for them.
13260 (BITWISEV): New int iterator.
13261 (bit_reduc_op): New int attributes.
13263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13264 Alan Hayward <alan.hayward@arm.com>
13265 David Sherwood <david.sherwood@arm.com>
13267 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13268 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13269 * optabs.def (vec_shl_insert_optab): New optab.
13270 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13271 (duplicate_and_interleave): Likewise.
13272 * tree-vect-loop.c: Include internal-fn.h.
13273 (neutral_op_for_slp_reduction): New function, split out from
13274 get_initial_defs_for_reduction.
13275 (get_initial_def_for_reduction): Handle option 2 for variable-length
13276 vectors by loading the neutral value into a vector and then shifting
13277 the initial value into element 0.
13278 (get_initial_defs_for_reduction): Replace the code argument with
13279 the neutral value calculated by neutral_op_for_slp_reduction.
13280 Use gimple_build_vector for constant-length vectors.
13281 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13282 but the first group_size elements have a neutral value.
13283 Use duplicate_and_interleave otherwise.
13284 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13285 Update call to get_initial_defs_for_reduction. Handle SLP
13286 reductions for variable-length vectors by creating one vector
13287 result for each scalar result, with the elements associated
13288 with other scalar results stubbed out with the neutral value.
13289 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13290 Require IFN_VEC_SHL_INSERT for double reductions on
13291 variable-length vectors, or SLP reductions that have
13292 a neutral value. Require can_duplicate_and_interleave_p
13293 support for variable-length unchained SLP reductions if there
13294 is no neutral value, such as for MIN/MAX reductions. Also require
13295 the number of vector elements to be a multiple of the number of
13296 SLP statements when doing variable-length unchained SLP reductions.
13297 Update call to vect_create_epilog_for_reduction.
13298 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13299 and remove initial values.
13300 (duplicate_and_interleave): Make public.
13301 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13302 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13305 Alan Hayward <alan.hayward@arm.com>
13306 David Sherwood <david.sherwood@arm.com>
13308 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13309 (can_duplicate_and_interleave_p): New function.
13310 (vect_get_and_check_slp_defs): Take the vector of statements
13311 rather than just the current one. Remove excess parentheses.
13312 Restriction rejectinon of vect_constant_def and vect_external_def
13313 for variable-length vectors to boolean types, or types for which
13314 can_duplicate_and_interleave_p is false.
13315 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13316 (duplicate_and_interleave): New function.
13317 (vect_get_constant_vectors): Use gimple_build_vector for
13318 constant-length vectors and suitable variable-length constant
13319 vectors. Use duplicate_and_interleave for other variable-length
13320 vectors. Don't defer the update when inserting new statements.
13322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13323 Alan Hayward <alan.hayward@arm.com>
13324 David Sherwood <david.sherwood@arm.com>
13326 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13327 min_profitable_iters doesn't go negative.
13329 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13330 Alan Hayward <alan.hayward@arm.com>
13331 David Sherwood <david.sherwood@arm.com>
13333 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13334 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13335 * optabs.def (vec_mask_load_lanes_optab): New optab.
13336 (vec_mask_store_lanes_optab): Likewise.
13337 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13338 (MASK_STORE_LANES): Likewise.
13339 * internal-fn.c (mask_load_lanes_direct): New macro.
13340 (mask_store_lanes_direct): Likewise.
13341 (expand_mask_load_optab_fn): Handle masked operations.
13342 (expand_mask_load_lanes_optab_fn): New macro.
13343 (expand_mask_store_optab_fn): Handle masked operations.
13344 (expand_mask_store_lanes_optab_fn): New macro.
13345 (direct_mask_load_lanes_optab_supported_p): Likewise.
13346 (direct_mask_store_lanes_optab_supported_p): Likewise.
13347 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13349 (vect_load_lanes_supported): Likewise.
13350 * tree-vect-data-refs.c (strip_conversion): New function.
13351 (can_group_stmts_p): Likewise.
13352 (vect_analyze_data_ref_accesses): Use it instead of checking
13353 for a pair of assignments.
13354 (vect_store_lanes_supported): Take a masked_p parameter.
13355 (vect_load_lanes_supported): Likewise.
13356 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13357 vect_store_lanes_supported and vect_load_lanes_supported.
13358 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13359 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13360 parameter. Don't allow gaps for masked accesses.
13361 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13362 and vect_load_lanes_supported.
13363 (get_load_store_type): Take a masked_p parameter and update
13364 call to get_group_load_store_type.
13365 (vectorizable_store): Update call to get_load_store_type.
13366 Handle IFN_MASK_STORE_LANES.
13367 (vectorizable_load): Update call to get_load_store_type.
13368 Handle IFN_MASK_LOAD_LANES.
13370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13371 Alan Hayward <alan.hayward@arm.com>
13372 David Sherwood <david.sherwood@arm.com>
13374 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13376 * config/aarch64/aarch64-protos.h
13377 (aarch64_sve_struct_memory_operand_p): Declare.
13378 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13379 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13380 (VPRED, vpred): Handle SVE structure modes.
13381 * config/aarch64/constraints.md (Utx): New constraint.
13382 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13383 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13384 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13385 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13386 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13387 structure modes. Split into pieces after RA.
13388 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13389 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13391 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13392 SVE structure modes.
13393 (aarch64_classify_address): Likewise.
13394 (sizetochar): Move earlier in file.
13395 (aarch64_print_operand): Handle SVE register lists.
13396 (aarch64_array_mode): New function.
13397 (aarch64_sve_struct_memory_operand_p): Likewise.
13398 (TARGET_ARRAY_MODE): Redefine.
13400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13401 Alan Hayward <alan.hayward@arm.com>
13402 David Sherwood <david.sherwood@arm.com>
13404 * target.def (array_mode): New target hook.
13405 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13406 * doc/tm.texi: Regenerate.
13407 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13408 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13409 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13410 targetm.array_mode.
13411 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13414 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13415 Alan Hayward <alan.hayward@arm.com>
13416 David Sherwood <david.sherwood@arm.com>
13418 * fold-const.c (fold_binary_loc): Check the argument types
13419 rather than the result type when testing for a vector operation.
13421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13423 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13424 * doc/tm.texi: Regenerate.
13426 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13427 Alan Hayward <alan.hayward@arm.com>
13428 David Sherwood <david.sherwood@arm.com>
13430 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13431 (sve): Document new AArch64 extension.
13432 * doc/md.texi (w): Extend the description of the AArch64
13433 constraint to include SVE vectors.
13434 (Upl, Upa): Document new AArch64 predicate constraints.
13435 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13437 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13438 (msve-vector-bits=): New option.
13439 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13440 SVE when these are disabled.
13441 (sve): New extension.
13442 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13443 modes. Adjust their number of units based on aarch64_sve_vg.
13444 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13445 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13446 aarch64_addr_query_type.
13447 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13448 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13449 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13450 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13451 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13452 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13453 (aarch64_simd_imm_zero_p): Delete.
13454 (aarch64_check_zero_based_sve_index_immediate): Declare.
13455 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13456 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13457 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13458 (aarch64_sve_float_mul_immediate_p): Likewise.
13459 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13460 rather than an rtx.
13461 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13462 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13463 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13464 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13465 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13466 (aarch64_regmode_natural_size): Likewise.
13467 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13468 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13470 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13471 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13472 for VG and the SVE predicate registers.
13473 (V_ALIASES): Add a "z"-prefixed alias.
13474 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13475 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13476 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13477 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13478 (REG_CLASS_NAMES): Add entries for them.
13479 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13480 and the predicate registers.
13481 (aarch64_sve_vg): Declare.
13482 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13483 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13484 (REGMODE_NATURAL_SIZE): Define.
13485 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13487 * config/aarch64/aarch64.c: Include cfgrtl.h.
13488 (simd_immediate_info): Add a constructor for series vectors,
13489 and an associated step field.
13490 (aarch64_sve_vg): New variable.
13491 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13492 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13493 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13494 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13495 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13496 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13497 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13498 (aarch64_get_mask_mode): New functions.
13499 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13500 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13501 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13502 predicate modes and predicate registers. Explicitly restrict
13503 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13504 to store a vector mode if it is recognized by
13505 aarch64_classify_vector_mode.
13506 (aarch64_regmode_natural_size): New function.
13507 (aarch64_hard_regno_caller_save_mode): Return the original mode
13509 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13510 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13511 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13512 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13514 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13515 does not overlap dest if the function is frame-related. Handle
13517 (aarch64_split_add_offset): New function.
13518 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13519 them aarch64_add_offset.
13520 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13521 and update call to aarch64_sub_sp.
13522 (aarch64_add_cfa_expression): New function.
13523 (aarch64_expand_prologue): Pass extra temporary registers to the
13524 functions above. Handle the case in which we need to emit new
13525 DW_CFA_expressions for registers that were originally saved
13526 relative to the stack pointer, but now have to be expressed
13527 relative to the frame pointer.
13528 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13530 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13531 IP0 and IP1 values for SVE frames.
13532 (aarch64_expand_vec_series): New function.
13533 (aarch64_expand_sve_widened_duplicate): Likewise.
13534 (aarch64_expand_sve_const_vector): Likewise.
13535 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13536 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13537 into the register, rather than emitting a SET directly.
13538 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13539 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13540 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13541 (offset_9bit_signed_scaled_p): New functions.
13542 (aarch64_replicate_bitmask_imm): New function.
13543 (aarch64_bitmask_imm): Use it.
13544 (aarch64_cannot_force_const_mem): Reject expressions involving
13545 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13546 (aarch64_classify_index): Handle SVE indices, by requiring
13547 a plain register index with a scale that matches the element size.
13548 (aarch64_classify_address): Handle SVE addresses. Assert that
13549 the mode of the address is VOIDmode or an integer mode.
13550 Update call to aarch64_classify_symbol.
13551 (aarch64_classify_symbolic_expression): Update call to
13552 aarch64_classify_symbol.
13553 (aarch64_const_vec_all_in_range_p): New function.
13554 (aarch64_print_vector_float_operand): Likewise.
13555 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13556 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13557 and the FP immediates 1.0 and 0.5.
13558 (aarch64_print_address_internal): Handle SVE addresses.
13559 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13560 (aarch64_regno_regclass): Handle predicate registers.
13561 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13563 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13564 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13565 (aarch64_convert_sve_vector_bits): New function.
13566 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13567 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13568 rather than an rtx.
13569 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13570 Handle SVE vector and predicate modes. Accept VL-based constants
13571 that need only one temporary register, and VL offsets that require
13572 no temporary registers.
13573 (aarch64_conditional_register_usage): Mark the predicate registers
13574 as fixed if SVE isn't available.
13575 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13576 Return true for SVE vector and predicate modes.
13577 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13578 rather than an unsigned int. Handle SVE modes.
13579 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13581 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13583 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13584 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13585 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13586 (aarch64_sve_float_mul_immediate_p): New functions.
13587 (aarch64_sve_valid_immediate): New function.
13588 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13589 Explicitly reject structure modes. Check for INDEX constants.
13590 Handle PTRUE and PFALSE constants.
13591 (aarch64_check_zero_based_sve_index_immediate): New function.
13592 (aarch64_simd_imm_zero_p): Delete.
13593 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13594 vector modes. Accept constants in the range of CNT[BHWD].
13595 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13596 ask for an Advanced SIMD mode.
13597 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13598 (aarch64_simd_vector_alignment): Handle SVE predicates.
13599 (aarch64_vectorize_preferred_vector_alignment): New function.
13600 (aarch64_simd_vector_alignment_reachable): Use it instead of
13602 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13603 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13605 (MAX_VECT_LEN): Delete.
13606 (expand_vec_perm_d): Add a vec_flags field.
13607 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13608 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13609 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13611 (aarch64_evpc_rev): Rename to...
13612 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13613 (aarch64_evpc_rev_global): New function.
13614 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13615 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13617 (aarch64_evpc_sve_tbl): New function.
13618 (aarch64_expand_vec_perm_const_1): Update after rename of
13619 aarch64_evpc_rev. Handle SVE permutes too, trying
13620 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13621 than aarch64_evpc_tbl.
13622 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13623 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13624 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13625 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13626 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13627 (aarch64_expand_sve_vcond): New functions.
13628 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13629 of aarch64_vector_mode_p.
13630 (aarch64_dwarf_poly_indeterminate_value): New function.
13631 (aarch64_compute_pressure_classes): Likewise.
13632 (aarch64_can_change_mode_class): Likewise.
13633 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13634 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13635 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13636 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13637 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13638 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13639 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13640 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13642 (Dn, Dl, Dr): Accept const as well as const_vector.
13643 (Dz): Likewise. Compare against CONST0_RTX.
13644 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13645 of "vector" where appropriate.
13646 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13647 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13648 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13649 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13650 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13651 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13652 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13653 (v_int_equiv): Extend to SVE modes.
13654 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13656 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13657 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13658 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13659 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13660 (SVE_COND_FP_CMP): New int iterators.
13661 (perm_hilo): Handle the new unpack unspecs.
13662 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13664 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13665 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13666 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13667 (aarch64_equality_operator, aarch64_constant_vector_operand)
13668 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13669 (aarch64_sve_nonimmediate_operand): Likewise.
13670 (aarch64_sve_general_operand): Likewise.
13671 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13672 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13673 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13674 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13675 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13676 (aarch64_sve_float_arith_immediate): Likewise.
13677 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13678 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13679 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13680 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13681 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13682 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13683 (aarch64_sve_float_arith_operand): Likewise.
13684 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13685 (aarch64_sve_float_mul_operand): Likewise.
13686 (aarch64_sve_vec_perm_operand): Likewise.
13687 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13688 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13689 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13690 as well as const_vector.
13691 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13692 in file. Use CONST0_RTX and CONSTM1_RTX.
13693 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13694 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13695 Use aarch64_simd_imm_zero.
13696 * config/aarch64/aarch64-sve.md: New file.
13697 * config/aarch64/aarch64.md: Include it.
13698 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13699 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13700 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13701 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13702 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13703 (sve): New attribute.
13704 (enabled): Disable instructions with the sve attribute unless
13706 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13707 aarch64_expand_mov_immediate.
13708 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13709 CNT[BHSD] immediates.
13710 (movti): Split CONST_POLY_INT moves into two halves.
13711 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13712 Split additions that need a temporary here if the destination
13713 is the stack pointer.
13714 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13715 (*add<mode>3_poly_1): New instruction.
13716 (set_clobber_cc): New expander.
13718 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13720 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13721 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13722 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13723 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13724 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13725 Change innermode from fixed_mode_size to machine_mode.
13726 (simplify_subreg): Update call accordingly. Handle a constant-sized
13727 subreg of a variable-length CONST_VECTOR.
13729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13730 Alan Hayward <alan.hayward@arm.com>
13731 David Sherwood <david.sherwood@arm.com>
13733 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13734 (add_offset_to_base): New function, split out from...
13735 (create_mem_ref): ...here. When handling a scale other than 1,
13736 check first whether the address is valid without the offset.
13737 Add it into the base if so, leaving the index and scale as-is.
13739 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13742 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13743 fold_for_warn before checking if arg2 is INTEGER_CST.
13745 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13747 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13748 (store_multiple_operation): Delete.
13749 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13750 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13751 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13752 guarded by TARGET_STRING.
13753 (rs6000_output_load_multiple): Delete.
13754 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13755 OPTION_MASK_STRING / TARGET_STRING handling.
13756 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13757 (const rs6000_opt_masks) <"string">: Change mask to 0.
13758 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13759 (MASK_STRING): Delete.
13760 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13762 (load_multiple): Delete.
13769 (store_multiple): Delete.
13776 (movmemsi_8reg): Delete.
13777 (corresponding unnamed define_insn): Delete.
13778 (movmemsi_6reg): Delete.
13779 (corresponding unnamed define_insn): Delete.
13780 (movmemsi_4reg): Delete.
13781 (corresponding unnamed define_insn): Delete.
13782 (movmemsi_2reg): Delete.
13783 (corresponding unnamed define_insn): Delete.
13784 (movmemsi_1reg): Delete.
13785 (corresponding unnamed define_insn): Delete.
13786 * config/rs6000/rs6000.opt (mno-string): New.
13787 (mstring): Replace by deprecation warning stub.
13788 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13790 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13792 * regrename.c (regrename_do_replace): If replacing the same
13793 reg multiple times, try to reuse last created gen_raw_REG.
13796 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13797 main to workaround a bug in GDB.
13799 2018-01-12 Tom de Vries <tom@codesourcery.com>
13802 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13804 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13806 PR rtl-optimization/80481
13807 * ira-color.c (get_cap_member): New function.
13808 (allocnos_conflict_by_live_ranges_p): Use it.
13809 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13810 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13812 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13815 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13816 (*saddl_se_1): Ditto.
13817 (*ssubsi_1): Ditto.
13818 (*ssubl_se_1): Ditto.
13820 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13822 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13823 rather than wi::to_widest for DR_INITs.
13824 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13825 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13826 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13828 (vect_analyze_group_access_1): Note that here.
13830 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13832 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13833 polynomial type sizes.
13835 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13837 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13838 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13839 (gimple_add_tmp_var): Likewise.
13841 2018-01-12 Martin Liska <mliska@suse.cz>
13843 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13844 (gimple_alloc_sizes): Likewise.
13845 (dump_gimple_statistics): Use PRIu64 in printf format.
13846 * gimple.h: Change uint64_t to int.
13848 2018-01-12 Martin Liska <mliska@suse.cz>
13850 * tree-core.h: Use uint64_t instead of int.
13851 * tree.c (tree_node_counts): Likewise.
13852 (tree_node_sizes): Likewise.
13853 (dump_tree_statistics): Use PRIu64 in printf format.
13855 2018-01-12 Martin Liska <mliska@suse.cz>
13857 * Makefile.in: As qsort_chk is implemented in vec.c, add
13858 vec.o to linkage of gencfn-macros.
13859 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13860 passing the info to record_node_allocation_statistics.
13861 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13863 * ggc-common.c (struct ggc_usage): Add operator== and use
13864 it in operator< and compare function.
13865 * mem-stats.h (struct mem_usage): Likewise.
13866 * vec.c (struct vec_usage): Remove operator< and compare
13867 function. Can be simply inherited.
13869 2018-01-12 Martin Jambor <mjambor@suse.cz>
13872 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13873 * tree-ssa-math-opts.c: Include domwalk.h.
13874 (convert_mult_to_fma_1): New function.
13875 (fma_transformation_info): New type.
13876 (fma_deferring_state): Likewise.
13877 (cancel_fma_deferring): New function.
13878 (result_of_phi): Likewise.
13879 (last_fma_candidate_feeds_initial_phi): Likewise.
13880 (convert_mult_to_fma): Added deferring logic, split actual
13881 transformation to convert_mult_to_fma_1.
13882 (math_opts_dom_walker): New type.
13883 (math_opts_dom_walker::after_dom_children): New method, body moved
13884 here from pass_optimize_widening_mul::execute, added deferring logic
13886 (pass_optimize_widening_mul::execute): Moved most of code to
13887 math_opts_dom_walker::after_dom_children.
13888 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13889 * config/i386/i386.c (ix86_option_override_internal): Added
13890 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13892 2018-01-12 Richard Biener <rguenther@suse.de>
13895 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13896 inline instance vars.
13898 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13901 * config/rx/rx.c (rx_is_restricted_memory_address):
13902 Handle SUBREG case.
13904 2018-01-12 Richard Biener <rguenther@suse.de>
13906 PR tree-optimization/80846
13907 * target.def (split_reduction): New target hook.
13908 * targhooks.c (default_split_reduction): New function.
13909 * targhooks.h (default_split_reduction): Declare.
13910 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13911 target requests first reduce vectors by combining low and high
13913 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13914 (get_vectype_for_scalar_type_and_size): Export.
13915 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13916 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13917 * doc/tm.texi: Regenerate.
13918 * config/i386/i386.c (ix86_split_reduction): Implement
13919 TARGET_VECTORIZE_SPLIT_REDUCTION.
13921 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13924 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13925 in PIC mode except for TARGET_VXWORKS_RTP.
13926 * config/sparc/sparc.c: Include cfgrtl.h.
13927 (TARGET_INIT_PIC_REG): Define.
13928 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13929 (sparc_pic_register_p): New predicate.
13930 (sparc_legitimate_address_p): Use it.
13931 (sparc_legitimize_pic_address): Likewise.
13932 (sparc_delegitimize_address): Likewise.
13933 (sparc_mode_dependent_address_p): Likewise.
13934 (gen_load_pcrel_sym): Remove 4th parameter.
13935 (load_got_register): Adjust call to above. Remove obsolete stuff.
13936 (sparc_expand_prologue): Do not call load_got_register here.
13937 (sparc_flat_expand_prologue): Likewise.
13938 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13939 (sparc_use_pseudo_pic_reg): New function.
13940 (sparc_init_pic_reg): Likewise.
13941 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13942 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13944 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13946 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13947 Add item for branch_cost.
13949 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13951 PR rtl-optimization/83565
13952 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13953 not extend the result to a larger mode for rotate operations.
13954 (num_sign_bit_copies1): Likewise.
13956 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13959 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13961 Use values-Xc.o for -pedantic.
13962 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13964 2018-01-12 Martin Liska <mliska@suse.cz>
13967 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13969 (possible_polymorphic_call_targets): Use it.
13970 (ipa_devirt): Likewise.
13972 2018-01-12 Martin Liska <mliska@suse.cz>
13974 * profile-count.h (enum profile_quality): Use 0 as invalid
13975 enum value of profile_quality.
13977 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13979 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13980 -mext-string options.
13982 2018-01-12 Richard Biener <rguenther@suse.de>
13984 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13985 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13986 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13988 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13990 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13992 * configure.ac (--with-long-double-format): Add support for the
13993 configuration option to change the default long double format on
13995 * config.gcc (powerpc*-linux*-*): Likewise.
13996 * configure: Regenerate.
13997 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13998 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13999 used without modification.
14001 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14003 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14004 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14005 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14006 MISC_BUILTIN_SPEC_BARRIER.
14007 (rs6000_init_builtins): Likewise.
14008 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14010 (speculation_barrier): New define_insn.
14011 * doc/extend.texi: Document __builtin_speculation_barrier.
14013 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14016 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14017 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14018 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14020 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14021 integral modes instead of "ss" and "sd".
14022 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14023 vectors with 32-bit and 64-bit elements.
14024 (vecdupssescalarmodesuffix): New mode attribute.
14025 (vec_dup<mode>): Use it.
14027 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14030 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14031 frame if argument is passed on stack.
14033 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14036 * ree.c (combine_reaching_defs): Optimize also
14037 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14038 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14040 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14042 PR middle-end/83189
14043 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14045 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14047 PR middle-end/83718
14048 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14049 after they are computed.
14051 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14053 PR tree-optimization/83695
14054 * gimple-loop-linterchange.cc
14055 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14056 reset cached scev information after interchange.
14057 (pass_linterchange::execute): Remove call to scev_reset_htab.
14059 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14061 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14062 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14063 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14064 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14065 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14066 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14067 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14068 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14069 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14070 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14071 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14072 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14073 (V_lane_reg): Likewise.
14074 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14076 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14077 (vfmal_lane_low<mode>_intrinsic,
14078 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14079 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14080 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14081 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14082 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14083 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14085 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14087 * config/arm/arm-cpus.in (fp16fml): New feature.
14088 (ALL_SIMD): Add fp16fml.
14089 (armv8.2-a): Add fp16fml as an option.
14090 (armv8.3-a): Likewise.
14091 (armv8.4-a): Add fp16fml as part of fp16.
14092 * config/arm/arm.h (TARGET_FP16FML): Define.
14093 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14095 * config/arm/arm-modes.def (V2HF): Define.
14096 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14097 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14098 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14099 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14100 vfmsl_low, vfmsl_high): New set of builtins.
14101 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14102 (vfml_op): New code attribute.
14103 (VFMLHALVES): New int iterator.
14104 (VFML, VFMLSEL): New mode attributes.
14105 (V_reg): Define mapping for V2HF.
14106 (V_hi, V_lo): New mode attributes.
14107 (VF_constraint): Likewise.
14108 (vfml_half, vfml_half_selector): New int attributes.
14109 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14111 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14112 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14114 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14115 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14116 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14117 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14119 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14120 Document new effective target and option set.
14122 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14124 * config/arm/arm-cpus.in (armv8_4): New feature.
14125 (ARMv8_4a): New fgroup.
14126 (armv8.4-a): New arch.
14127 * config/arm/arm-tables.opt: Regenerate.
14128 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14129 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14130 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14131 Add matching rules for -march=armv8.4-a and extensions.
14132 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14134 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14137 * config/rx/rx.md (BW): New mode attribute.
14138 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14140 2018-01-11 Richard Biener <rguenther@suse.de>
14142 PR tree-optimization/83435
14143 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14144 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14145 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14147 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14148 Alan Hayward <alan.hayward@arm.com>
14149 David Sherwood <david.sherwood@arm.com>
14151 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14153 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14154 (aarch64_print_address_internal): Use it to check for a zero offset.
14156 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14157 Alan Hayward <alan.hayward@arm.com>
14158 David Sherwood <david.sherwood@arm.com>
14160 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14161 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14162 Return a poly_int64 rather than a HOST_WIDE_INT.
14163 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14164 rather than a HOST_WIDE_INT.
14165 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14166 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14167 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14168 final_offset from HOST_WIDE_INT to poly_int64.
14169 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14170 to_constant when getting the number of units in an Advanced SIMD
14172 (aarch64_builtin_vectorized_function): Check for a constant number
14174 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14176 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14177 attribute instead of GET_MODE_NUNITS.
14178 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14179 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14180 GET_MODE_SIZE for fixed-size registers.
14181 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14182 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14183 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14184 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14185 (aarch64_print_operand, aarch64_print_address_internal)
14186 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14187 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14188 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14189 Handle polynomial GET_MODE_SIZE.
14190 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14191 wider than SImode without modification.
14192 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14193 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14194 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14195 passing and returning SVE modes.
14196 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14197 rather than GEN_INT.
14198 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14199 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14200 (aarch64_allocate_and_probe_stack_space): Likewise.
14201 (aarch64_layout_frame): Cope with polynomial offsets.
14202 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14203 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14204 polynomial offsets.
14205 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14206 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14207 poly_int64 rather than a HOST_WIDE_INT.
14208 (aarch64_get_separate_components, aarch64_process_components)
14209 (aarch64_expand_prologue, aarch64_expand_epilogue)
14210 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14211 (aarch64_anchor_offset): New function, split out from...
14212 (aarch64_legitimize_address): ...here.
14213 (aarch64_builtin_vectorization_cost): Handle polynomial
14214 TYPE_VECTOR_SUBPARTS.
14215 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14217 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14218 number of elements from the PARALLEL rather than the mode.
14219 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14220 rather than GET_MODE_BITSIZE.
14221 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14222 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14223 (aarch64_expand_vec_perm_const_1): Handle polynomial
14224 d->perm.length () and d->perm elements.
14225 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14226 Apply to_constant to d->perm elements.
14227 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14228 polynomial CONST_VECTOR_NUNITS.
14229 (aarch64_move_pointer): Take amount as a poly_int64 rather
14231 (aarch64_progress_pointer): Avoid temporary variable.
14232 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14233 the mode attribute instead of GET_MODE.
14235 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14236 Alan Hayward <alan.hayward@arm.com>
14237 David Sherwood <david.sherwood@arm.com>
14239 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14240 x exists before using it.
14241 (aarch64_add_constant_internal): Rename to...
14242 (aarch64_add_offset_1): ...this. Replace regnum with separate
14243 src and dest rtxes. Handle the case in which they're different,
14244 including when the offset is zero. Replace scratchreg with an rtx.
14245 Use 2 additions if there is no spare register into which we can
14246 move a 16-bit constant.
14247 (aarch64_add_constant): Delete.
14248 (aarch64_add_offset): Replace reg with separate src and dest
14249 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14250 Use aarch64_add_offset_1.
14251 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14252 an rtx rather than an int. Take the delta as a poly_int64
14253 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14254 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14255 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14256 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14257 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14258 and aarch64_add_sp.
14259 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14260 aarch64_add_constant.
14262 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14264 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14265 Use scalar_float_mode.
14267 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14269 * config/aarch64/aarch64-simd.md
14270 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14271 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14272 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14273 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14274 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14275 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14276 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14277 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14278 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14279 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14281 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14284 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14285 targ_options->x_arm_arch_string is non NULL.
14287 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14289 * config/aarch64/aarch64.h
14290 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14292 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14295 * expmed.c (emit_store_flag_force): Swap if const op0
14296 and change VOIDmode to mode of op0.
14298 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14300 PR rtl-optimization/83761
14301 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14302 than bytes to mode_for_size.
14304 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14306 PR middle-end/83189
14307 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14308 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14311 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14313 PR middle-end/83575
14314 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14315 when in layout mode.
14316 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14317 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14320 2018-01-10 Michael Collison <michael.collison@arm.com>
14322 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14323 * config/aarch64/aarch64-option-extension.def: Add
14324 AARCH64_OPT_EXTENSION of 'fp16fml'.
14325 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14326 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14327 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14328 * config/aarch64/constraints.md (Ui7): New constraint.
14329 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14330 (VFMLA_SEL_W): Ditto.
14333 (VFMLA16_LOW): New int iterator.
14334 (VFMLA16_HIGH): Ditto.
14335 (UNSPEC_FMLAL): New unspec.
14336 (UNSPEC_FMLSL): Ditto.
14337 (UNSPEC_FMLAL2): Ditto.
14338 (UNSPEC_FMLSL2): Ditto.
14339 (f16mac): New code attribute.
14340 * config/aarch64/aarch64-simd-builtins.def
14341 (aarch64_fmlal_lowv2sf): Ditto.
14342 (aarch64_fmlsl_lowv2sf): Ditto.
14343 (aarch64_fmlalq_lowv4sf): Ditto.
14344 (aarch64_fmlslq_lowv4sf): Ditto.
14345 (aarch64_fmlal_highv2sf): Ditto.
14346 (aarch64_fmlsl_highv2sf): Ditto.
14347 (aarch64_fmlalq_highv4sf): Ditto.
14348 (aarch64_fmlslq_highv4sf): Ditto.
14349 (aarch64_fmlal_lane_lowv2sf): Ditto.
14350 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14351 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14352 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14353 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14354 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14355 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14356 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14357 (aarch64_fmlal_lane_highv2sf): Ditto.
14358 (aarch64_fmlsl_lane_highv2sf): Ditto.
14359 (aarch64_fmlal_laneq_highv2sf): Ditto.
14360 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14361 (aarch64_fmlalq_lane_highv4sf): Ditto.
14362 (aarch64_fmlsl_lane_highv4sf): Ditto.
14363 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14364 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14365 * config/aarch64/aarch64-simd.md:
14366 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14367 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14368 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14369 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14370 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14371 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14372 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14373 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14374 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14375 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14376 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14377 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14378 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14379 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14380 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14381 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14382 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14383 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14384 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14385 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14386 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14387 (vfmlsl_low_u32): Ditto.
14388 (vfmlalq_low_u32): Ditto.
14389 (vfmlslq_low_u32): Ditto.
14390 (vfmlal_high_u32): Ditto.
14391 (vfmlsl_high_u32): Ditto.
14392 (vfmlalq_high_u32): Ditto.
14393 (vfmlslq_high_u32): Ditto.
14394 (vfmlal_lane_low_u32): Ditto.
14395 (vfmlsl_lane_low_u32): Ditto.
14396 (vfmlal_laneq_low_u32): Ditto.
14397 (vfmlsl_laneq_low_u32): Ditto.
14398 (vfmlalq_lane_low_u32): Ditto.
14399 (vfmlslq_lane_low_u32): Ditto.
14400 (vfmlalq_laneq_low_u32): Ditto.
14401 (vfmlslq_laneq_low_u32): Ditto.
14402 (vfmlal_lane_high_u32): Ditto.
14403 (vfmlsl_lane_high_u32): Ditto.
14404 (vfmlal_laneq_high_u32): Ditto.
14405 (vfmlsl_laneq_high_u32): Ditto.
14406 (vfmlalq_lane_high_u32): Ditto.
14407 (vfmlslq_lane_high_u32): Ditto.
14408 (vfmlalq_laneq_high_u32): Ditto.
14409 (vfmlslq_laneq_high_u32): Ditto.
14410 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14411 (AARCH64_FL_FOR_ARCH8_4): New.
14412 (AARCH64_ISA_F16FML): New ISA flag.
14413 (TARGET_F16FML): New feature flag for fp16fml.
14414 (doc/invoke.texi): Document new fp16fml option.
14416 2018-01-10 Michael Collison <michael.collison@arm.com>
14418 * config/aarch64/aarch64-builtins.c:
14419 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14421 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14422 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14423 (AARCH64_ISA_SHA3): New ISA flag.
14424 (TARGET_SHA3): New feature flag for sha3.
14425 * config/aarch64/iterators.md (sha512_op): New int attribute.
14426 (CRYPTO_SHA512): New int iterator.
14427 (UNSPEC_SHA512H): New unspec.
14428 (UNSPEC_SHA512H2): Ditto.
14429 (UNSPEC_SHA512SU0): Ditto.
14430 (UNSPEC_SHA512SU1): Ditto.
14431 * config/aarch64/aarch64-simd-builtins.def
14432 (aarch64_crypto_sha512hqv2di): New builtin.
14433 (aarch64_crypto_sha512h2qv2di): Ditto.
14434 (aarch64_crypto_sha512su0qv2di): Ditto.
14435 (aarch64_crypto_sha512su1qv2di): Ditto.
14436 (aarch64_eor3qv8hi): Ditto.
14437 (aarch64_rax1qv2di): Ditto.
14438 (aarch64_xarqv2di): Ditto.
14439 (aarch64_bcaxqv8hi): Ditto.
14440 * config/aarch64/aarch64-simd.md:
14441 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14442 (aarch64_crypto_sha512su0qv2di): Ditto.
14443 (aarch64_crypto_sha512su1qv2di): Ditto.
14444 (aarch64_eor3qv8hi): Ditto.
14445 (aarch64_rax1qv2di): Ditto.
14446 (aarch64_xarqv2di): Ditto.
14447 (aarch64_bcaxqv8hi): Ditto.
14448 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14449 (vsha512h2q_u64): Ditto.
14450 (vsha512su0q_u64): Ditto.
14451 (vsha512su1q_u64): Ditto.
14452 (veor3q_u16): Ditto.
14453 (vrax1q_u64): Ditto.
14454 (vxarq_u64): Ditto.
14455 (vbcaxq_u16): Ditto.
14456 * config/arm/types.md (crypto_sha512): New type attribute.
14457 (crypto_sha3): Ditto.
14458 (doc/invoke.texi): Document new sha3 option.
14460 2018-01-10 Michael Collison <michael.collison@arm.com>
14462 * config/aarch64/aarch64-builtins.c:
14463 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14464 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14465 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14466 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14467 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14468 (AARCH64_ISA_SM4): New ISA flag.
14469 (TARGET_SM4): New feature flag for sm4.
14470 * config/aarch64/aarch64-simd-builtins.def
14471 (aarch64_sm3ss1qv4si): Ditto.
14472 (aarch64_sm3tt1aq4si): Ditto.
14473 (aarch64_sm3tt1bq4si): Ditto.
14474 (aarch64_sm3tt2aq4si): Ditto.
14475 (aarch64_sm3tt2bq4si): Ditto.
14476 (aarch64_sm3partw1qv4si): Ditto.
14477 (aarch64_sm3partw2qv4si): Ditto.
14478 (aarch64_sm4eqv4si): Ditto.
14479 (aarch64_sm4ekeyqv4si): Ditto.
14480 * config/aarch64/aarch64-simd.md:
14481 (aarch64_sm3ss1qv4si): Ditto.
14482 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14483 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14484 (aarch64_sm4eqv4si): Ditto.
14485 (aarch64_sm4ekeyqv4si): Ditto.
14486 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14487 (sm3part_op): Ditto.
14488 (CRYPTO_SM3TT): Ditto.
14489 (CRYPTO_SM3PART): Ditto.
14490 (UNSPEC_SM3SS1): New unspec.
14491 (UNSPEC_SM3TT1A): Ditto.
14492 (UNSPEC_SM3TT1B): Ditto.
14493 (UNSPEC_SM3TT2A): Ditto.
14494 (UNSPEC_SM3TT2B): Ditto.
14495 (UNSPEC_SM3PARTW1): Ditto.
14496 (UNSPEC_SM3PARTW2): Ditto.
14497 (UNSPEC_SM4E): Ditto.
14498 (UNSPEC_SM4EKEY): Ditto.
14499 * config/aarch64/constraints.md (Ui2): New constraint.
14500 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14501 * config/arm/types.md (crypto_sm3): New type attribute.
14502 (crypto_sm4): Ditto.
14503 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14504 (vsm3tt1aq_u32): Ditto.
14505 (vsm3tt1bq_u32): Ditto.
14506 (vsm3tt2aq_u32): Ditto.
14507 (vsm3tt2bq_u32): Ditto.
14508 (vsm3partw1q_u32): Ditto.
14509 (vsm3partw2q_u32): Ditto.
14510 (vsm4eq_u32): Ditto.
14511 (vsm4ekeyq_u32): Ditto.
14512 (doc/invoke.texi): Document new sm4 option.
14514 2018-01-10 Michael Collison <michael.collison@arm.com>
14516 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14517 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14518 (AARCH64_FL_FOR_ARCH8_4): New.
14519 (AARCH64_FL_V8_4): New flag.
14520 (doc/invoke.texi): Document new armv8.4-a option.
14522 2018-01-10 Michael Collison <michael.collison@arm.com>
14524 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14525 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14526 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14527 * config/aarch64/aarch64-option-extension.def: Add
14528 AARCH64_OPT_EXTENSION of 'sha2'.
14529 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14530 (crypto): Disable sha2 and aes if crypto disabled.
14531 (crypto): Enable aes and sha2 if enabled.
14532 (simd): Disable sha2 and aes if simd disabled.
14533 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14535 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14536 (TARGET_SHA2): New feature flag for sha2.
14537 (TARGET_AES): New feature flag for aes.
14538 * config/aarch64/aarch64-simd.md:
14539 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14540 conditional on TARGET_AES.
14541 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14542 (aarch64_crypto_sha1hsi): Make pattern conditional
14544 (aarch64_crypto_sha1hv4si): Ditto.
14545 (aarch64_be_crypto_sha1hv4si): Ditto.
14546 (aarch64_crypto_sha1su1v4si): Ditto.
14547 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14548 (aarch64_crypto_sha1su0v4si): Ditto.
14549 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14550 (aarch64_crypto_sha256su0v4si): Ditto.
14551 (aarch64_crypto_sha256su1v4si): Ditto.
14552 (doc/invoke.texi): Document new aes and sha2 options.
14554 2018-01-10 Martin Sebor <msebor@redhat.com>
14556 PR tree-optimization/83781
14557 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14560 2018-01-11 Martin Sebor <msebor@gmail.com>
14561 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14563 PR tree-optimization/83501
14564 PR tree-optimization/81703
14566 * tree-ssa-strlen.c (get_string_cst): Rename...
14567 (get_string_len): ...to this. Handle global constants.
14568 (handle_char_store): Adjust.
14570 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14571 Jim Wilson <jimw@sifive.com>
14573 * config/riscv/riscv-protos.h (riscv_output_return): New.
14574 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14575 (riscv_attribute_table, riscv_output_return),
14576 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14577 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14578 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14580 (riscv_expand_prologue): Add early return for naked function.
14581 (riscv_expand_epilogue): Likewise.
14582 (riscv_function_ok_for_sibcall): Return false for naked function.
14583 (riscv_set_current_function): New.
14584 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14585 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14586 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14587 * doc/extend.texi (RISC-V Function Attributes): New.
14589 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14591 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14592 check for 128-bit long double before checking TCmode.
14593 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14594 128-bit long doubles before checking TFmode or TCmode.
14595 (FLOAT128_IBM_P): Likewise.
14597 2018-01-10 Martin Sebor <msebor@redhat.com>
14599 PR tree-optimization/83671
14600 * builtins.c (c_strlen): Unconditionally return zero for the empty
14602 Use -Warray-bounds for warnings.
14603 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14604 for non-constant array indices with COMPONENT_REF, arrays of
14605 arrays, and pointers to arrays.
14606 (gimple_fold_builtin_strlen): Determine and set length range for
14607 non-constant character arrays.
14609 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14611 PR middle-end/81897
14612 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14615 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14617 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14619 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14622 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14623 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14624 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14625 indexed_or_indirect_operand predicate.
14626 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14627 (*vsx_le_perm_load_v8hi): Likewise.
14628 (*vsx_le_perm_load_v16qi): Likewise.
14629 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14630 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14631 (*vsx_le_perm_store_v8hi): Likewise.
14632 (*vsx_le_perm_store_v16qi): Likewise.
14633 (eight unnamed splitters): Likewise.
14635 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14637 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14638 * config/rs6000/emmintrin.h: Likewise.
14639 * config/rs6000/mmintrin.h: Likewise.
14640 * config/rs6000/xmmintrin.h: Likewise.
14642 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14645 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14647 * tree.c (tree_nop_conversion): Return true for location wrapper
14649 (maybe_wrap_with_location): New function.
14650 (selftest::check_strip_nops): New function.
14651 (selftest::test_location_wrappers): New function.
14652 (selftest::tree_c_tests): Call it.
14653 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14654 (maybe_wrap_with_location): New decl.
14655 (EXPR_LOCATION_WRAPPER_P): New macro.
14656 (location_wrapper_p): New inline function.
14657 (tree_strip_any_location_wrapper): New inline function.
14659 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14662 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14663 stack_realign_offset for the largest alignment of stack slot
14665 (ix86_find_max_used_stack_alignment): New function.
14666 (ix86_finalize_stack_frame_flags): Use it. Set
14667 max_used_stack_alignment if we don't realign stack.
14668 * config/i386/i386.h (machine_function): Add
14669 max_used_stack_alignment.
14671 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14673 * config/arm/arm.opt (-mbranch-cost): New option.
14674 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14677 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14680 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14681 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14683 2018-01-10 Richard Biener <rguenther@suse.de>
14686 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14687 early out so it also covers the case where we have a non-NULL
14690 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14692 PR tree-optimization/83753
14693 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14694 for non-strided grouped accesses if the number of elements is 1.
14696 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14699 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14700 * i386.h (TARGET_USE_GATHER): Define.
14701 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14703 2018-01-10 Martin Liska <mliska@suse.cz>
14706 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14707 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14709 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14710 CLEANUP_NO_PARTITIONING is not set.
14712 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14714 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14715 for vectors, as a partial revert of r254296.
14716 * rtl.h (const_vec_p): Delete.
14717 (const_vec_duplicate_p): Don't test for vector CONSTs.
14718 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14719 * expmed.c (make_tree): Likewise.
14722 * common.md (E, F): Use CONSTANT_P instead of checking for
14724 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14725 checking for CONST_VECTOR.
14727 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14729 PR middle-end/83575
14730 * predict.c (force_edge_cold): Handle in more sane way edges
14731 with no prediction.
14733 2018-01-09 Carl Love <cel@us.ibm.com>
14735 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14737 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14738 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14739 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14740 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14741 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14742 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14743 * config/rs6000/rs6000-protos.h: Add extern defition for
14744 rs6000_generate_float2_double_code.
14745 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14747 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14748 (float2_v2df): Add define_expand.
14750 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14753 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14754 op_mode in the force_to_mode call.
14756 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14758 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14759 instead of checking each element individually.
14760 (aarch64_evpc_uzp): Likewise.
14761 (aarch64_evpc_zip): Likewise.
14762 (aarch64_evpc_ext): Likewise.
14763 (aarch64_evpc_rev): Likewise.
14764 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14765 instead of checking each element individually. Return true without
14767 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14768 whether all selected elements come from the same input, instead of
14769 checking each element individually. Remove calls to gen_rtx_REG,
14770 start_sequence and end_sequence and instead assert that no rtl is
14773 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14775 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14776 order of HIGH and CONST checks.
14778 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14780 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14781 if the destination isn't an SSA_NAME.
14783 2018-01-09 Richard Biener <rguenther@suse.de>
14785 PR tree-optimization/83668
14786 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14788 (canonicalize_loop_form): ... here, renamed from ...
14789 (canonicalize_loop_closed_ssa_form): ... this and amended to
14790 swap successor edges for loop exit blocks to make us use
14791 the RPO order we need for initial schedule generation.
14793 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14795 PR tree-optimization/64811
14796 * match.pd: When optimizing comparisons with Inf, avoid
14797 introducing or losing exceptions from comparisons with NaN.
14799 2018-01-09 Martin Liska <mliska@suse.cz>
14802 * asan.c (shadow_mem_size): Add gcc_assert.
14804 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14806 Don't save registers in main().
14809 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14810 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14811 * config/avr/avr.c (avr_set_current_function): Don't error if
14812 naked, OS_task or OS_main are specified at the same time.
14813 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14815 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14817 * common/config/avr/avr-common.c (avr_option_optimization_table):
14818 Switch on -mmain-is-OS_task for optimizing compilations.
14820 2018-01-09 Richard Biener <rguenther@suse.de>
14822 PR tree-optimization/83572
14823 * graphite.c: Include cfganal.h.
14824 (graphite_transform_loops): Connect infinite loops to exit
14825 and remove fake edges at the end.
14827 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14829 * ipa-inline.c (edge_badness): Revert accidental checkin.
14831 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14834 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14835 symbols; not inline clones.
14837 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14840 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14841 hard registers. Formatting fixes.
14843 PR preprocessor/83722
14844 * gcc.c (try_generate_repro): Pass
14845 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14846 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14849 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14850 Kito Cheng <kito.cheng@gmail.com>
14852 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14853 (riscv_leaf_function_p): Delete.
14854 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14856 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14858 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14860 (do_ifelse): New function.
14861 (do_isel): New function.
14862 (do_sub3): New function.
14863 (do_add3): New function.
14864 (do_load_mask_compare): New function.
14865 (do_overlap_load_compare): New function.
14866 (expand_compare_loop): New function.
14867 (expand_block_compare): Call expand_compare_loop() when appropriate.
14868 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14869 option description.
14870 (-mblock-compare-inline-loop-limit): New option.
14872 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14875 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14876 Reverse order of second and third operands in first alternative.
14877 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14878 of first and second elements in UNSPEC_VPERMR vector.
14879 (altivec_expand_vec_perm_le): Likewise.
14881 2018-01-08 Jeff Law <law@redhat.com>
14883 PR rtl-optimizatin/81308
14884 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14885 (process_switch): If group_case_labels makes a change, then set
14887 (pass_convert_switch::execute): If a switch is converted, then
14888 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14890 PR rtl-optimization/81308
14891 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14894 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14896 PR target/83663 - Revert r255946
14897 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14898 generation for cases where splatting a value is not useful.
14899 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14900 across a vec_duplicate and a paradoxical subreg forming a vector
14901 mode to a vec_concat.
14903 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14905 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14906 -march=armv8.3-a variants.
14907 * config/arm/t-multilib: Likewise.
14908 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14910 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14912 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14914 (cceq_ior_compare_complement): Give it a name so I can use it, and
14915 change boolean_or_operator predicate to boolean_operator so it can
14916 be used to generate a crand.
14917 (eqne): New code iterator.
14918 (bd/bd_neg): New code_attrs.
14919 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14920 a single define_insn.
14921 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14922 decrement (bdnzt/bdnzf/bdzt/bdzf).
14923 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14924 with the new names of the branch decrement patterns, and added the
14925 names of the branch decrement conditional patterns.
14927 2018-01-08 Richard Biener <rguenther@suse.de>
14929 PR tree-optimization/83563
14930 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14933 2018-01-08 Richard Biener <rguenther@suse.de>
14935 PR middle-end/83713
14936 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14938 2018-01-08 Richard Biener <rguenther@suse.de>
14940 PR tree-optimization/83685
14941 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14942 references to abnormals.
14944 2018-01-08 Richard Biener <rguenther@suse.de>
14947 * dwarf2out.c (output_indirect_strings): Handle empty
14948 skeleton_debug_str_hash.
14949 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14951 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14953 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14954 (emit_store_direct): Likewise.
14955 (arc_trampoline_adjust_address): Likewise.
14956 (arc_asm_trampoline_template): New function.
14957 (arc_initialize_trampoline): Use asm_trampoline_template.
14958 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14959 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14960 * config/arc/arc.md (flush_icache): Delete pattern.
14962 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14964 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14965 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14968 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14971 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14972 by not USED_FOR_TARGET.
14973 (make_pass_resolve_sw_modes): Likewise.
14975 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14977 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14980 2018-01-08 Richard Biener <rguenther@suse.de>
14982 PR middle-end/83580
14983 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14985 2018-01-08 Richard Biener <rguenther@suse.de>
14987 PR middle-end/83517
14988 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14990 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14992 PR middle-end/81897
14993 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14994 basic blocks with a small number of successors.
14995 (convert_control_dep_chain_into_preds): Improve handling of
14997 (dump_predicates): Split apart into...
14998 (dump_pred_chain): ...here...
14999 (dump_pred_info): ...and here.
15000 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15001 (can_chain_union_be_invalidated_p): Improve check for invalidation
15003 (uninit_uses_cannot_happen): Avoid unnecessary if
15004 convert_control_dep_chain_into_preds yielded nothing.
15006 2018-01-06 Martin Sebor <msebor@redhat.com>
15008 PR tree-optimization/83640
15009 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15010 subtracting negative offset from size.
15011 (builtin_access::overlap): Adjust offset bounds of the access to fall
15012 within the size of the object if possible.
15014 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15016 PR rtl-optimization/83699
15017 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15018 extract_bit_field_as_subreg to cases in which the extracted
15019 value is also a vector.
15021 * lra-constraints.c (process_alt_operands): Test for the equivalence
15022 substitutions when detecting a possible reload cycle.
15024 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15027 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15028 by default if flag_selective_schedling{,2}. Formatting fixes.
15030 PR rtl-optimization/83682
15031 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15032 if it has non-VECTOR_MODE element mode.
15033 (vec_duplicate_p): Likewise.
15035 PR middle-end/83694
15036 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15037 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15039 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15042 * config/i386/i386-builtin.def
15043 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15044 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15045 Require also OPTION_MASK_ISA_AVX512F in addition to
15046 OPTION_MASK_ISA_GFNI.
15047 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15048 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15049 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15050 to OPTION_MASK_ISA_GFNI.
15051 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15052 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15053 OPTION_MASK_ISA_AVX512BW.
15054 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15055 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15056 addition to OPTION_MASK_ISA_GFNI.
15057 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15058 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15059 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15060 to OPTION_MASK_ISA_GFNI.
15061 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15062 a requirement for all ISAs rather than any of them with a few
15064 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15066 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15067 bitmasks to be enabled with 3 exceptions, instead of requiring any
15068 enabled ISA with lots of exceptions.
15069 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15070 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15071 Change avx512bw in isa attribute to avx512f.
15072 * config/i386/sgxintrin.h: Add license boilerplate.
15073 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15074 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15075 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15076 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15078 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15079 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15080 temporarily sse2 rather than sse if not enabled already.
15083 * config/i386/sse.md (VI248_VLBW): Rename to ...
15084 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15085 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15086 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15087 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15088 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15089 mode iterator instead of VI248_VLBW.
15091 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15093 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15094 (record_modified): Skip clobbers; add debug output.
15095 (param_change_prob): Use sreal frequencies.
15097 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15099 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15100 punt for user-aligned variables.
15102 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15104 * tree-chrec.c (chrec_contains_symbols): Return true for
15107 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15110 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15111 of (x|y) == x for BICS pattern.
15113 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15115 PR tree-optimization/83605
15116 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15117 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15120 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15122 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15123 * config/epiphany/rtems.h: New file.
15125 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15126 Uros Bizjak <ubizjak@gmail.com>
15129 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15130 QIreg_operand instead of register_operand predicate.
15131 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15132 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15133 comments instead of -fmitigate[-_]rop.
15135 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15138 * cgraphunit.c (symbol_table::compile): Switch to text_section
15139 before calling assembly_start debug hook.
15140 * run-rtl-passes.c (run_rtl_passes): Likewise.
15143 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15145 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15146 range_int_cst_p rather than !symbolic_range_p before calling
15147 extract_range_from_multiplicative_op_1.
15149 2018-01-04 Jeff Law <law@redhat.com>
15151 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15152 redundant test in assertion.
15154 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15156 * doc/rtl.texi: Document machine_mode wrapper classes.
15158 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15160 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15161 using tree_to_uhwi.
15163 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15165 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15166 the VEC_PERM_EXPR fold to fail.
15168 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15171 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15172 to switched_sections.
15174 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15177 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15178 test for d.testing.
15180 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15183 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15184 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15186 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15189 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15190 is BLKmode and bitpos not zero or mode change is needed.
15192 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15195 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15198 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15201 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15202 instead of MULT rtx. Update all corresponding splitters.
15203 (*saddl_se): Ditto.
15204 (*ssub<modesuffix>): Ditto.
15205 (*ssubl_se): Ditto.
15206 (*cmp_sadd_di): Update split patterns.
15207 (*cmp_sadd_si): Ditto.
15208 (*cmp_sadd_sidi): Ditto.
15209 (*cmp_ssub_di): Ditto.
15210 (*cmp_ssub_si): Ditto.
15211 (*cmp_ssub_sidi): Ditto.
15212 * config/alpha/predicates.md (const23_operand): New predicate.
15213 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15214 Look for ASHIFT, not MULT inner operand.
15215 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15217 2018-01-04 Martin Liska <mliska@suse.cz>
15219 PR gcov-profile/83669
15220 * gcov.c (output_intermediate_file): Add version to intermediate
15222 * doc/gcov.texi: Document new field 'version' in intermediate
15223 file format. Fix location of '-k' option of gcov command.
15225 2018-01-04 Martin Liska <mliska@suse.cz>
15228 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15230 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15232 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15234 2018-01-03 Martin Sebor <msebor@redhat.com>
15236 PR tree-optimization/83655
15237 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15238 checking calls with invalid arguments.
15240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15242 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15243 (vectorizable_mask_load_store): Delete.
15244 (vectorizable_call): Return false for masked loads and stores.
15245 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15246 instead of gimple_assign_rhs1.
15247 (vectorizable_load): Handle IFN_MASK_LOAD.
15248 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15252 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15254 (vectorizable_mask_load_store): ...here.
15255 (vectorizable_load): ...and here.
15257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15259 * tree-vect-stmts.c (vect_build_all_ones_mask)
15260 (vect_build_zero_merge_argument): New functions, split out from...
15261 (vectorizable_load): ...here.
15263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15265 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15267 (vectorizable_mask_load_store): ...here.
15268 (vectorizable_store): ...and here.
15270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15272 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15274 (vectorizable_mask_load_store): ...here.
15276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15278 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15279 (vect_model_store_cost): Take a vec_load_store_type instead of a
15281 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15282 (vect_model_store_cost): Take a vec_load_store_type instead of a
15284 (vectorizable_mask_load_store): Update accordingly.
15285 (vectorizable_store): Likewise.
15286 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15290 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15291 IFN_MASK_LOAD calls here rather than...
15292 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15295 Alan Hayward <alan.hayward@arm.com>
15296 David Sherwood <david.sherwood@arm.com>
15298 * expmed.c (extract_bit_field_1): For vector extracts,
15299 fall back to extract_bit_field_as_subreg if vec_extract
15302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15303 Alan Hayward <alan.hayward@arm.com>
15304 David Sherwood <david.sherwood@arm.com>
15306 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15307 they are variable or constant sized.
15308 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15309 slots for constant-sized data.
15311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15312 Alan Hayward <alan.hayward@arm.com>
15313 David Sherwood <david.sherwood@arm.com>
15315 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15316 handling COND_EXPRs with boolean comparisons, try to find a better
15317 basis for the mask type than the boolean itself.
15319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15321 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15322 is calculated and how it can be overridden.
15323 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15324 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15326 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15330 Alan Hayward <alan.hayward@arm.com>
15331 David Sherwood <david.sherwood@arm.com>
15333 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15334 Remove the mode argument.
15335 (aarch64_simd_valid_immediate): Remove the mode and inverse
15337 * config/aarch64/iterators.md (bitsize): New iterator.
15338 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15339 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15340 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15341 aarch64_simd_valid_immediate.
15342 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15343 (aarch64_reg_or_bic_imm): Likewise.
15344 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15345 with an insn_type enum and msl with a modifier_type enum.
15346 Replace element_width with a scalar_mode. Change the shift
15347 to unsigned int. Add constructors for scalar_float_mode and
15348 scalar_int_mode elements.
15349 (aarch64_vect_float_const_representable_p): Delete.
15350 (aarch64_can_const_movi_rtx_p)
15351 (aarch64_simd_scalar_immediate_valid_for_move)
15352 (aarch64_simd_make_constant): Update call to
15353 aarch64_simd_valid_immediate.
15354 (aarch64_advsimd_valid_immediate_hs): New function.
15355 (aarch64_advsimd_valid_immediate): Likewise.
15356 (aarch64_simd_valid_immediate): Remove mode and inverse
15357 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15358 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15359 and aarch64_float_const_representable_p on the result.
15360 (aarch64_output_simd_mov_immediate): Remove mode argument.
15361 Update call to aarch64_simd_valid_immediate and use of
15362 simd_immediate_info.
15363 (aarch64_output_scalar_simd_mov_immediate): Update call
15366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15367 Alan Hayward <alan.hayward@arm.com>
15368 David Sherwood <david.sherwood@arm.com>
15370 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15371 (mode_nunits): Likewise CONST_MODE_NUNITS.
15372 * machmode.def (ADJUST_NUNITS): Document.
15373 * genmodes.c (mode_data::need_nunits_adj): New field.
15374 (blank_mode): Update accordingly.
15375 (adj_nunits): New variable.
15376 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15378 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15379 listed in adj_nunits.
15380 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15381 listed in adj_nunits. Don't emit case statements for such modes.
15382 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15383 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15384 nothing if adj_nunits is nonnull.
15385 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15386 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15387 (emit_mode_fbit): Update use of print_maybe_const_decl.
15388 (emit_move_size): Likewise. Treat the array as non-const
15390 (emit_mode_adjustments): Handle adj_nunits.
15392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15394 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15395 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15396 (VECTOR_MODES): Use it.
15397 (make_vector_modes): Take the prefix as an argument.
15399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15400 Alan Hayward <alan.hayward@arm.com>
15401 David Sherwood <david.sherwood@arm.com>
15403 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15404 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15405 for MODE_VECTOR_BOOL.
15406 * machmode.def (VECTOR_BOOL_MODE): Document.
15407 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15408 (make_vector_bool_mode): New function.
15409 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15411 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15412 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15414 * stor-layout.c (int_mode_for_mode): Likewise.
15415 * tree.c (build_vector_type_for_mode): Likewise.
15416 * varasm.c (output_constant_pool_2): Likewise.
15417 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15418 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15419 for MODE_VECTOR_BOOL.
15420 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15421 of mode class checks.
15422 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15423 instead of a list of mode class checks.
15424 (expand_vector_scalar_condition): Likewise.
15425 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15428 Alan Hayward <alan.hayward@arm.com>
15429 David Sherwood <david.sherwood@arm.com>
15431 * machmode.h (mode_size): Change from unsigned short to
15433 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15434 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15435 or if measurement_type is not polynomial.
15436 (fixed_size_mode::includes_p): Check for constant-sized modes.
15437 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15438 return a poly_uint16 rather than an unsigned short.
15439 (emit_mode_size): Change the type of mode_size from unsigned short
15440 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15441 (emit_mode_adjustments): Cope with polynomial vector sizes.
15442 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15444 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15446 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15447 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15448 * caller-save.c (setup_save_areas): Likewise.
15449 (replace_reg_with_saved_mem): Likewise.
15450 * calls.c (emit_library_call_value_1): Likewise.
15451 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15452 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15453 (gen_lowpart_for_combine): Likewise.
15454 * convert.c (convert_to_integer_1): Likewise.
15455 * cse.c (equiv_constant, cse_insn): Likewise.
15456 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15457 (cselib_subst_to_values): Likewise.
15458 * dce.c (word_dce_process_block): Likewise.
15459 * df-problems.c (df_word_lr_mark_ref): Likewise.
15460 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15461 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15462 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15463 (rtl_for_decl_location): Likewise.
15464 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15465 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15466 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15467 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15468 (expand_expr_real_1): Likewise.
15469 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15470 (pad_below): Likewise.
15471 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15472 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15473 * ira.c (get_subreg_tracking_sizes): Likewise.
15474 * ira-build.c (ira_create_allocno_objects): Likewise.
15475 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15476 (ira_sort_regnos_for_alter_reg): Likewise.
15477 * ira-costs.c (record_operand_costs): Likewise.
15478 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15479 (resolve_simple_move): Likewise.
15480 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15481 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15482 (lra_constraints): Likewise.
15483 (CONST_POOL_OK_P): Reject variable-sized modes.
15484 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15485 (add_pseudo_to_slot, lra_spill): Likewise.
15486 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15487 * optabs-query.c (get_best_extraction_insn): Likewise.
15488 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15489 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15490 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15491 * recog.c (offsettable_address_addr_space_p): Likewise.
15492 * regcprop.c (maybe_mode_change): Likewise.
15493 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15494 * regrename.c (build_def_use): Likewise.
15495 * regstat.c (dump_reg_info): Likewise.
15496 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15497 (find_reloads, find_reloads_subreg_address): Likewise.
15498 * reload1.c (eliminate_regs_1): Likewise.
15499 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15500 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15501 (simplify_binary_operation_1, simplify_subreg): Likewise.
15502 * targhooks.c (default_function_arg_padding): Likewise.
15503 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15504 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15505 (verify_gimple_assign_ternary): Likewise.
15506 * tree-inline.c (estimate_move_cost): Likewise.
15507 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15508 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15509 (get_address_cost_ainc): Likewise.
15510 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15511 (vect_supportable_dr_alignment): Likewise.
15512 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15513 (vectorizable_reduction): Likewise.
15514 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15515 (vectorizable_operation, vectorizable_load): Likewise.
15516 * tree.c (build_same_sized_truth_vector_type): Likewise.
15517 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15518 * var-tracking.c (emit_note_insn_var_location): Likewise.
15519 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15520 (ADDR_VEC_ALIGN): Likewise.
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_to_bits): Return a poly_uint16 rather than an
15528 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15529 or if measurement_type is polynomial.
15530 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15531 * combine.c (make_extraction): Likewise.
15532 * dse.c (find_shift_sequence): Likewise.
15533 * dwarf2out.c (mem_loc_descriptor): Likewise.
15534 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15535 (extract_bit_field, extract_low_bits): Likewise.
15536 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15537 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15538 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15539 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15540 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15541 * reload.c (find_reloads): Likewise.
15542 * reload1.c (alter_reg): Likewise.
15543 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15544 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15545 * tree-if-conv.c (predicate_mem_writes): Likewise.
15546 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15547 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15548 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15549 * valtrack.c (dead_debug_insert_temp): Likewise.
15550 * varasm.c (mergeable_constant_section): Likewise.
15551 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15554 Alan Hayward <alan.hayward@arm.com>
15555 David Sherwood <david.sherwood@arm.com>
15557 * expr.c (expand_assignment): Cope with polynomial mode sizes
15558 when assigning to a CONCAT.
15560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15561 Alan Hayward <alan.hayward@arm.com>
15562 David Sherwood <david.sherwood@arm.com>
15564 * machmode.h (mode_precision): Change from unsigned short to
15566 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15568 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15569 or if measurement_type is not polynomial.
15570 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15571 in which the mode is already known to be a scalar_int_mode.
15572 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15573 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15575 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15576 for GET_MODE_PRECISION.
15577 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15578 for GET_MODE_PRECISION.
15579 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15581 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15582 (expand_field_assignment, make_extraction): Likewise.
15583 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15584 (get_last_value): Likewise.
15585 * convert.c (convert_to_integer_1): Likewise.
15586 * cse.c (cse_insn): Likewise.
15587 * expr.c (expand_expr_real_1): Likewise.
15588 * lra-constraints.c (simplify_operand_subreg): Likewise.
15589 * optabs-query.c (can_atomic_load_p): Likewise.
15590 * optabs.c (expand_atomic_load): Likewise.
15591 (expand_atomic_store): Likewise.
15592 * ree.c (combine_reaching_defs): Likewise.
15593 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15594 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15595 * tree.h (type_has_mode_precision_p): Likewise.
15596 * ubsan.c (instrument_si_overflow): Likewise.
15598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15599 Alan Hayward <alan.hayward@arm.com>
15600 David Sherwood <david.sherwood@arm.com>
15602 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15603 polynomial numbers of units.
15604 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15605 (valid_vector_subparts_p): New function.
15606 (build_vector_type): Remove temporary shim and take the number
15607 of units as a poly_uint64 rather than an int.
15608 (build_opaque_vector_type): Take the number of units as a
15609 poly_uint64 rather than an int.
15610 * tree.c (build_vector_from_ctor): Handle polynomial
15611 TYPE_VECTOR_SUBPARTS.
15612 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15613 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15614 (build_vector_from_val): If the number of units is variable,
15615 use build_vec_duplicate_cst for constant operands and
15616 VEC_DUPLICATE_EXPR otherwise.
15617 (make_vector_type): Remove temporary is_constant ().
15618 (build_vector_type, build_opaque_vector_type): Take the number of
15619 units as a poly_uint64 rather than an int.
15620 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15622 * cfgexpand.c (expand_debug_expr): Likewise.
15623 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15624 (store_constructor, expand_expr_real_1): Likewise.
15625 (const_scalar_mask_from_tree): Likewise.
15626 * fold-const-call.c (fold_const_reduction): Likewise.
15627 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15628 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15629 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15630 (fold_relational_const): Likewise.
15631 (native_interpret_vector): Likewise. Change the size from an
15632 int to an unsigned int.
15633 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15634 TYPE_VECTOR_SUBPARTS.
15635 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15636 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15637 duplicating a non-constant operand into a variable-length vector.
15638 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15639 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15640 * ipa-icf.c (sem_variable::equals): Likewise.
15641 * match.pd: Likewise.
15642 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15643 * print-tree.c (print_node): Likewise.
15644 * stor-layout.c (layout_type): Likewise.
15645 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15646 * tree-cfg.c (verify_gimple_comparison): Likewise.
15647 (verify_gimple_assign_binary): Likewise.
15648 (verify_gimple_assign_ternary): Likewise.
15649 (verify_gimple_assign_single): Likewise.
15650 * tree-pretty-print.c (dump_generic_node): Likewise.
15651 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15652 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15653 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15654 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15655 (vect_shift_permute_load_chain): Likewise.
15656 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15657 (expand_vector_condition, optimize_vector_constructor): Likewise.
15658 (lower_vec_perm, get_compute_type): Likewise.
15659 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15660 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15661 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15662 (vect_recog_mask_conversion_pattern): Likewise.
15663 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15664 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15665 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15666 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15667 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15668 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15669 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15670 (supportable_widening_operation): Likewise.
15671 (supportable_narrowing_operation): Likewise.
15672 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15674 * varasm.c (output_constant): Likewise.
15676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15677 Alan Hayward <alan.hayward@arm.com>
15678 David Sherwood <david.sherwood@arm.com>
15680 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15681 so that both the length == 3 and length != 3 cases set up their
15682 own permute vectors. Add comments explaining why we know the
15683 number of elements is constant.
15684 (vect_permute_load_chain): Likewise.
15686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15687 Alan Hayward <alan.hayward@arm.com>
15688 David Sherwood <david.sherwood@arm.com>
15690 * machmode.h (mode_nunits): Change from unsigned char to
15692 (ONLY_FIXED_SIZE_MODES): New macro.
15693 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15694 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15695 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15697 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15698 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15699 or if measurement_type is not polynomial.
15700 * genmodes.c (ZERO_COEFFS): New macro.
15701 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15703 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15704 Use ZERO_COEFFS when emitting initializers.
15705 * data-streamer.h (bp_pack_poly_value): New function.
15706 (bp_unpack_poly_value): Likewise.
15707 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15708 for GET_MODE_NUNITS.
15709 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15710 for GET_MODE_NUNITS.
15711 * tree.c (make_vector_type): Remove temporary shim and make
15712 the real function take the number of units as a poly_uint64
15713 rather than an int.
15714 (build_vector_type_for_mode): Handle polynomial nunits.
15715 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15716 * emit-rtl.c (const_vec_series_p_1): Likewise.
15717 (gen_rtx_CONST_VECTOR): Likewise.
15718 * fold-const.c (test_vec_duplicate_folding): Likewise.
15719 * genrecog.c (validate_pattern): Likewise.
15720 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15721 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15722 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15723 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15724 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15725 * rtlanal.c (subreg_get_info): Likewise.
15726 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15727 (vect_grouped_load_supported): Likewise.
15728 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15729 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15730 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15731 (simplify_const_unary_operation, simplify_binary_operation_1)
15732 (simplify_const_binary_operation, simplify_ternary_operation)
15733 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15734 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15735 instead of CONST_VECTOR_NUNITS.
15736 * varasm.c (output_constant_pool_2): Likewise.
15737 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15738 explicit-encoded elements in the XVEC for variable-length vectors.
15740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15742 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15745 Alan Hayward <alan.hayward@arm.com>
15746 David Sherwood <david.sherwood@arm.com>
15748 * coretypes.h (fixed_size_mode): Declare.
15749 (fixed_size_mode_pod): New typedef.
15750 * builtins.h (target_builtins::x_apply_args_mode)
15751 (target_builtins::x_apply_result_mode): Change type to
15752 fixed_size_mode_pod.
15753 * builtins.c (apply_args_size, apply_result_size, result_vector)
15754 (expand_builtin_apply_args_1, expand_builtin_apply)
15755 (expand_builtin_return): Update accordingly.
15757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15759 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15760 * cselib.c (cselib_hash_rtx): Likewise.
15761 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15762 CONST_VECTOR encoding.
15764 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15765 Jeff Law <law@redhat.com>
15768 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15769 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15770 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15771 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15774 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15775 explicitly probe *sp in a noreturn function if there were any callee
15776 register saves or frame pointer is needed.
15778 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15781 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15782 BLKmode for ternary, binary or unary expressions.
15785 * var-tracking.c (delete_vta_debug_insn): New inline function.
15786 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15787 insns from get_insns () to NULL instead of each bb separately.
15788 Use delete_vta_debug_insn. No longer static.
15789 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15790 delete_vta_debug_insns callers.
15791 * rtl.h (delete_vta_debug_insns): Declare.
15792 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15793 instead of variable_tracking_main.
15795 2018-01-03 Martin Sebor <msebor@redhat.com>
15797 PR tree-optimization/83603
15798 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15799 arguments past the endof the argument list in functions declared
15800 without a prototype.
15801 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15802 Avoid checking when arguments are null.
15804 2018-01-03 Martin Sebor <msebor@redhat.com>
15807 * doc/extend.texi (attribute const): Fix a typo.
15808 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15809 issuing -Wsuggest-attribute for void functions.
15811 2018-01-03 Martin Sebor <msebor@redhat.com>
15813 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15814 offset_int::from instead of wide_int::to_shwi.
15815 (maybe_diag_overlap): Remove assertion.
15816 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15817 * gimple-ssa-sprintf.c (format_directive): Same.
15818 (parse_directive): Same.
15819 (sprintf_dom_walker::compute_format_length): Same.
15820 (try_substitute_return_value): Same.
15822 2018-01-03 Jeff Law <law@redhat.com>
15824 PR middle-end/83654
15825 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15826 non-constant residual for zero at runtime and avoid probing in
15827 that case. Reorganize code for trailing problem to mirror handling
15830 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15832 PR tree-optimization/83501
15833 * tree-ssa-strlen.c (get_string_cst): New.
15834 (handle_char_store): Call get_string_cst.
15836 2018-01-03 Martin Liska <mliska@suse.cz>
15838 PR tree-optimization/83593
15839 * tree-ssa-strlen.c: Include tree-cfg.h.
15840 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15841 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15842 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15844 (strlen_dom_walker::before_dom_children): Call
15845 gimple_purge_dead_eh_edges. Dump tranformation with details
15847 (strlen_dom_walker::before_dom_children): Update call by adding
15848 new argument cleanup_eh.
15849 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15851 2018-01-03 Martin Liska <mliska@suse.cz>
15854 * cif-code.def (VARIADIC_THUNK): New enum value.
15855 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15858 2018-01-03 Jan Beulich <jbeulich@suse.com>
15860 * sse.md (mov<mode>_internal): Tighten condition for when to use
15861 vmovdqu<ssescalarsize> for TI and OI modes.
15863 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15865 Update copyright years.
15867 2018-01-03 Martin Liska <mliska@suse.cz>
15870 * ipa-visibility.c (function_and_variable_visibility): Skip
15871 functions with noipa attribure.
15873 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15875 * gcc.c (process_command): Update copyright notice dates.
15876 * gcov-dump.c (print_version): Ditto.
15877 * gcov.c (print_version): Ditto.
15878 * gcov-tool.c (print_version): Ditto.
15879 * gengtype.c (create_file): Ditto.
15880 * doc/cpp.texi: Bump @copying's copyright year.
15881 * doc/cppinternals.texi: Ditto.
15882 * doc/gcc.texi: Ditto.
15883 * doc/gccint.texi: Ditto.
15884 * doc/gcov.texi: Ditto.
15885 * doc/install.texi: Ditto.
15886 * doc/invoke.texi: Ditto.
15888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15890 * vector-builder.h (vector_builder::m_full_nelts): Change from
15891 unsigned int to poly_uint64.
15892 (vector_builder::full_nelts): Update prototype accordingly.
15893 (vector_builder::new_vector): Likewise.
15894 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15895 (vector_builder::operator ==): Likewise.
15896 (vector_builder::finalize): Likewise.
15897 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15898 Take the number of elements as a poly_uint64 rather than an
15900 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15901 from unsigned int to poly_uint64.
15902 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15903 (vec_perm_indices::new_vector): Likewise.
15904 (vec_perm_indices::length): Likewise.
15905 (vec_perm_indices::nelts_per_input): Likewise.
15906 (vec_perm_indices::input_nelts): Likewise.
15907 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15908 number of elements per input as a poly_uint64 rather than an
15909 unsigned int. Use the original encoding for variable-length
15910 vectors, rather than clamping each individual element.
15911 For the second and subsequent elements in each pattern,
15912 clamp the step and base before clamping their sum.
15913 (vec_perm_indices::series_p): Handle polynomial element counts.
15914 (vec_perm_indices::all_in_range_p): Likewise.
15915 (vec_perm_indices_to_tree): Likewise.
15916 (vec_perm_indices_to_rtx): Likewise.
15917 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15918 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15919 (tree_vector_builder::new_binary_operation): Handle polynomial
15920 element counts. Return false if we need to know the number
15921 of elements at compile time.
15922 * fold-const.c (fold_vec_perm): Punt if the number of elements
15923 isn't known at compile time.
15925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15927 * vec-perm-indices.h (vec_perm_builder): Change element type
15928 from HOST_WIDE_INT to poly_int64.
15929 (vec_perm_indices::element_type): Update accordingly.
15930 (vec_perm_indices::clamp): Handle polynomial element_types.
15931 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15932 (vec_perm_indices::all_in_range_p): Likewise.
15933 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15935 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15936 polynomial vec_perm_indices element types.
15937 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15938 * fold-const.c (fold_vec_perm): Likewise.
15939 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15940 * tree-vect-generic.c (lower_vec_perm): Likewise.
15941 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15942 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15943 element type to HOST_WIDE_INT.
15945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15946 Alan Hayward <alan.hayward@arm.com>
15947 David Sherwood <david.sherwood@arm.com>
15949 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15950 rather than an int. Use plus_constant.
15951 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15952 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15955 Alan Hayward <alan.hayward@arm.com>
15956 David Sherwood <david.sherwood@arm.com>
15958 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15959 a HOST_WIDE_INT to a poly_int64.
15961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15962 Alan Hayward <alan.hayward@arm.com>
15963 David Sherwood <david.sherwood@arm.com>
15965 * calls.c (load_register_parameters): Cope with polynomial
15966 mode sizes. Require a constant size for BLKmode parameters
15967 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15968 forces a parameter to be padded at the lsb end in order to
15969 fill a complete number of words, require the parameter size
15970 to be ordered wrt UNITS_PER_WORD.
15972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15973 Alan Hayward <alan.hayward@arm.com>
15974 David Sherwood <david.sherwood@arm.com>
15976 * reload1.c (spill_stack_slot_width): Change element type
15977 from unsigned int to poly_uint64_pod.
15978 (alter_reg): Treat mode sizes as polynomial.
15980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15981 Alan Hayward <alan.hayward@arm.com>
15982 David Sherwood <david.sherwood@arm.com>
15984 * reload.c (complex_word_subreg_p): New function.
15985 (reload_inner_reg_of_subreg, push_reload): Use it.
15987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15988 Alan Hayward <alan.hayward@arm.com>
15989 David Sherwood <david.sherwood@arm.com>
15991 * lra-constraints.c (process_alt_operands): Reject matched
15992 operands whose sizes aren't ordered.
15993 (match_reload): Refer to this check here.
15995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15996 Alan Hayward <alan.hayward@arm.com>
15997 David Sherwood <david.sherwood@arm.com>
15999 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16000 that the mode size is in the set {1, 2, 4, 8, 16}.
16002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16003 Alan Hayward <alan.hayward@arm.com>
16004 David Sherwood <david.sherwood@arm.com>
16006 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16007 Use plus_constant instead of gen_rtx_PLUS.
16009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16010 Alan Hayward <alan.hayward@arm.com>
16011 David Sherwood <david.sherwood@arm.com>
16013 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16014 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16015 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16016 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16017 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16018 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16019 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16020 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16021 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16022 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16024 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16025 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16026 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16027 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16028 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16029 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16030 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16031 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16032 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16033 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16035 * expr.c (emit_move_resolve_push): Treat the input and result
16036 of PUSH_ROUNDING as a poly_int64.
16037 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16038 (emit_push_insn): Likewise.
16039 * lra-eliminations.c (mark_not_eliminable): Likewise.
16040 * recog.c (push_operand): Likewise.
16041 * reload1.c (elimination_effects): Likewise.
16042 * rtlanal.c (nonzero_bits1): Likewise.
16043 * calls.c (store_one_arg): Likewise. Require the padding to be
16044 known at compile time.
16046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16047 Alan Hayward <alan.hayward@arm.com>
16048 David Sherwood <david.sherwood@arm.com>
16050 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16051 Use plus_constant instead of gen_rtx_PLUS.
16053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16054 Alan Hayward <alan.hayward@arm.com>
16055 David Sherwood <david.sherwood@arm.com>
16057 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16058 rather than an int.
16060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16061 Alan Hayward <alan.hayward@arm.com>
16062 David Sherwood <david.sherwood@arm.com>
16064 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16065 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16066 via stack temporaries. Treat the mode size as polynomial too.
16068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16069 Alan Hayward <alan.hayward@arm.com>
16070 David Sherwood <david.sherwood@arm.com>
16072 * expr.c (expand_expr_real_2): When handling conversions involving
16073 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16074 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16075 as a poly_uint64 too.
16077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16078 Alan Hayward <alan.hayward@arm.com>
16079 David Sherwood <david.sherwood@arm.com>
16081 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16084 Alan Hayward <alan.hayward@arm.com>
16085 David Sherwood <david.sherwood@arm.com>
16087 * combine.c (can_change_dest_mode): Handle polynomial
16088 REGMODE_NATURAL_SIZE.
16089 * expmed.c (store_bit_field_1): Likewise.
16090 * expr.c (store_constructor): Likewise.
16091 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16092 and polynomial REGMODE_NATURAL_SIZE.
16093 (gen_lowpart_common): Likewise.
16094 * reginfo.c (record_subregs_of_mode): Likewise.
16095 * rtlanal.c (read_modify_subreg_p): Likewise.
16097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16098 Alan Hayward <alan.hayward@arm.com>
16099 David Sherwood <david.sherwood@arm.com>
16101 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16102 numbers of elements.
16104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16105 Alan Hayward <alan.hayward@arm.com>
16106 David Sherwood <david.sherwood@arm.com>
16108 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16115 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
16122 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16123 (ipa_simd_modify_function_body): Likewise.
16125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16126 Alan Hayward <alan.hayward@arm.com>
16127 David Sherwood <david.sherwood@arm.com>
16129 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16130 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16131 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16132 (expand_vector_condition, vector_element): Likewise.
16133 (subparts_gt): New function.
16134 (get_compute_type): Use subparts_gt.
16135 (count_type_subparts): Delete.
16136 (expand_vector_operations_1): Use subparts_gt instead of
16137 count_type_subparts.
16139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16140 Alan Hayward <alan.hayward@arm.com>
16141 David Sherwood <david.sherwood@arm.com>
16143 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16144 (vect_compile_time_alias): ...this new function. Do the calculation
16145 on poly_ints rather than trees.
16146 (vect_prune_runtime_alias_test_list): Update call accordingly.
16148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16149 Alan Hayward <alan.hayward@arm.com>
16150 David Sherwood <david.sherwood@arm.com>
16152 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16154 (vect_schedule_slp_instance): Likewise.
16156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16157 Alan Hayward <alan.hayward@arm.com>
16158 David Sherwood <david.sherwood@arm.com>
16160 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16161 constant and extern definitions for variable-length vectors.
16162 (vect_get_constant_vectors): Note that the number of units
16163 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16170 of units as polynomial. Choose between WIDE and NARROW based
16173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16174 Alan Hayward <alan.hayward@arm.com>
16175 David Sherwood <david.sherwood@arm.com>
16177 * tree-vect-stmts.c (simd_clone_subparts): New function.
16178 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16185 vectors as polynomial. Use build_index_vector for
16186 IFN_GOMP_SIMD_LANE.
16188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16189 Alan Hayward <alan.hayward@arm.com>
16190 David Sherwood <david.sherwood@arm.com>
16192 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16193 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16194 for variable-length vectors.
16195 (vectorizable_mask_load_store): Treat the number of units as
16196 polynomial, asserting that it is constant if the condition has
16197 already been enforced.
16198 (vectorizable_store, vectorizable_load): Likewise.
16200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16201 Alan Hayward <alan.hayward@arm.com>
16202 David Sherwood <david.sherwood@arm.com>
16204 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16205 of units as polynomial. Punt if we can't tell at compile time
16206 which vector contains the final result.
16208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16209 Alan Hayward <alan.hayward@arm.com>
16210 David Sherwood <david.sherwood@arm.com>
16212 * tree-vect-loop.c (vectorizable_induction): Treat the number
16213 of units as polynomial. Punt on SLP inductions. Use an integer
16214 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16215 cast of such a series for variable-length floating-point
16218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16219 Alan Hayward <alan.hayward@arm.com>
16220 David Sherwood <david.sherwood@arm.com>
16222 * tree.h (build_index_vector): Declare.
16223 * tree.c (build_index_vector): New function.
16224 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16225 of units as polynomial, forcibly converting it to a constant if
16226 vectorizable_reduction has already enforced the condition.
16227 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16228 to create a {1,2,3,...} vector.
16229 (vectorizable_reduction): Treat the number of units as polynomial.
16230 Choose vectype_in based on the largest scalar element size rather
16231 than the smallest number of units. Enforce the restrictions
16234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16235 Alan Hayward <alan.hayward@arm.com>
16236 David Sherwood <david.sherwood@arm.com>
16238 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16239 number of units as polynomial.
16241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16242 Alan Hayward <alan.hayward@arm.com>
16243 David Sherwood <david.sherwood@arm.com>
16245 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16246 * target.def (autovectorize_vector_sizes): Return the vector sizes
16247 by pointer, using vector_sizes rather than a bitmask.
16248 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16249 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16250 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16252 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16253 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16254 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16255 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16256 * omp-general.c (omp_max_vf): Likewise.
16257 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16258 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16259 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16260 * tree-vect-slp.c (vect_slp_bb): Likewise.
16261 * doc/tm.texi: Regenerate.
16262 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16264 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16265 the vector size as a poly_uint64 rather than an unsigned int.
16266 (current_vector_size): Change from an unsigned int to a poly_uint64.
16267 (get_vectype_for_scalar_type): Update accordingly.
16268 * tree.h (build_truth_vector_type): Take the size and number of
16269 units as a poly_uint64 rather than an unsigned int.
16270 (build_vector_type): Add a temporary overload that takes
16271 the number of units as a poly_uint64 rather than an unsigned int.
16272 * tree.c (make_vector_type): Likewise.
16273 (build_truth_vector_type): Take the number of units as a poly_uint64
16274 rather than an unsigned int.
16276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16277 Alan Hayward <alan.hayward@arm.com>
16278 David Sherwood <david.sherwood@arm.com>
16280 * target.def (get_mask_mode): Take the number of units and length
16281 as poly_uint64s rather than unsigned ints.
16282 * targhooks.h (default_get_mask_mode): Update accordingly.
16283 * targhooks.c (default_get_mask_mode): Likewise.
16284 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16285 * doc/tm.texi: Regenerate.
16287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16288 Alan Hayward <alan.hayward@arm.com>
16289 David Sherwood <david.sherwood@arm.com>
16291 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16292 * omp-general.c (omp_max_vf): Likewise.
16293 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16294 (expand_omp_simd): Handle polynomial safelen.
16295 * omp-low.c (omplow_simd_context): Add a default constructor.
16296 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16297 (lower_rec_simd_input_clauses): Update accordingly.
16298 (lower_rec_input_clauses): Likewise.
16300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16301 Alan Hayward <alan.hayward@arm.com>
16302 David Sherwood <david.sherwood@arm.com>
16304 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16305 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16306 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16307 (vect_analyze_slp_cost): Likewise.
16308 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16309 (vect_model_load_cost): Likewise.
16311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16312 Alan Hayward <alan.hayward@arm.com>
16313 David Sherwood <david.sherwood@arm.com>
16315 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16316 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16317 from an unsigned int * to a poly_uint64_pod *.
16318 (calculate_unrolling_factor): New function.
16319 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
16326 from an unsigned int to a poly_uint64.
16327 (_loop_vec_info::slp_unrolling_factor): Likewise.
16328 (_loop_vec_info::vectorization_factor): Change from an int
16330 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16331 (vect_get_num_vectors): New function.
16332 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16333 (vect_get_num_copies): Use vect_get_num_vectors.
16334 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16335 to an unsigned int *.
16336 (vect_analyze_data_refs): Change min_vf from an int * to a
16338 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16339 than an unsigned HOST_WIDE_INT.
16340 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16341 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16342 to an unsigned int *.
16343 (vect_analyze_data_ref_dependences): Likewise.
16344 (vect_compute_data_ref_alignment): Handle polynomial vf.
16345 (vect_enhance_data_refs_alignment): Likewise.
16346 (vect_prune_runtime_alias_test_list): Likewise.
16347 (vect_shift_permute_load_chain): Likewise.
16348 (vect_supportable_dr_alignment): Likewise.
16349 (dependence_distance_ge_vf): Take the vectorization factor as a
16350 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16351 (vect_analyze_data_refs): Change min_vf from an int * to a
16353 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16354 vfm1 as a poly_uint64 rather than an int. Make the same change
16355 for the returned bound_scalar.
16356 (vect_gen_vector_loop_niters): Handle polynomial vf.
16357 (vect_do_peeling): Likewise. Update call to
16358 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16359 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16361 * tree-vect-loop.c (vect_determine_vectorization_factor)
16362 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16363 (vect_get_known_peeling_cost): Likewise.
16364 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16365 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16366 (vect_transform_loop): Likewise. Use the lowest possible VF when
16367 updating the upper bounds of the loop.
16368 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16369 rather than an int.
16370 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16371 polynomial unroll factors.
16372 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16373 (vect_make_slp_decision): Likewise.
16374 (vect_supported_load_permutation_p): Likewise, and polynomial
16376 (vect_analyze_slp_cost): Handle polynomial vf.
16377 (vect_slp_analyze_node_operations): Likewise.
16378 (vect_slp_analyze_bb_1): Likewise.
16379 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16380 than an unsigned HOST_WIDE_INT.
16381 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16382 (vectorizable_load): Handle polynomial vf.
16383 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16385 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16388 Alan Hayward <alan.hayward@arm.com>
16389 David Sherwood <david.sherwood@arm.com>
16391 * match.pd: Handle bit operations involving three constants
16392 and try to fold one pair.
16394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16396 * tree-vect-loop-manip.c: Include gimple-fold.h.
16397 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16398 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16399 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16400 Add a path that uses a step of VF instead of 1, but disable it
16402 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16403 and niters_no_overflow parameters. Update calls to
16404 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16405 Create a new SSA name if the latter choses to use a ste other
16406 than zero, and return it via niters_vector_mult_vf_var.
16407 * tree-vect-loop.c (vect_transform_loop): Update calls to
16408 vect_do_peeling, vect_gen_vector_loop_niters and
16409 slpeel_make_loop_iterate_ntimes.
16410 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16411 (vect_gen_vector_loop_niters): Update declarations after above changes.
16413 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16415 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16416 128-bit round to integer instructions.
16417 (ceil<mode>2): Likewise.
16418 (btrunc<mode>2): Likewise.
16419 (round<mode>2): Likewise.
16421 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16423 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16424 unaligned VSX load/store on P8/P9.
16425 (expand_block_clear): Allow the use of unaligned VSX
16426 load/store on P8/P9.
16428 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16430 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16432 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16433 swap associated with both a load and a store.
16435 2018-01-02 Andrew Waterman <andrew@sifive.com>
16437 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16438 * config/riscv/riscv.md (clear_cache): Use it.
16440 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16442 * web.c: Remove out-of-date comment.
16444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16446 * expr.c (fixup_args_size_notes): Check that any existing
16447 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16448 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16449 (emit_single_push_insn): ...here.
16451 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16453 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16454 (const_vector_encoded_nelts): New function.
16455 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16456 (const_vector_int_elt, const_vector_elt): Declare.
16457 * emit-rtl.c (const_vector_int_elt_1): New function.
16458 (const_vector_elt): Likewise.
16459 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16460 of CONST_VECTOR_ELT.
16462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16464 * expr.c: Include rtx-vector-builder.h.
16465 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16466 directly on the tree encoding.
16467 (const_vector_from_tree): Likewise.
16468 * optabs.c: Include rtx-vector-builder.h.
16469 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16470 sequence of "u" values.
16471 * vec-perm-indices.c: Include rtx-vector-builder.h.
16472 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16473 directly on the vec_perm_indices encoding.
16475 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16477 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16478 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16479 * rtx-vector-builder.h: New file.
16480 * rtx-vector-builder.c: Likewise.
16481 * rtl.h (rtx_def::u2): Add a const_vector field.
16482 (CONST_VECTOR_NPATTERNS): New macro.
16483 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16484 (CONST_VECTOR_DUPLICATE_P): Likewise.
16485 (CONST_VECTOR_STEPPED_P): Likewise.
16486 (CONST_VECTOR_ENCODED_ELT): Likewise.
16487 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16488 (unwrap_const_vec_duplicate): Likewise.
16489 (const_vec_series_p): Check for a non-duplicated vector encoding.
16490 Say that the function only returns true for integer vectors.
16491 * emit-rtl.c: Include rtx-vector-builder.h.
16492 (gen_const_vec_duplicate_1): Delete.
16493 (gen_const_vector): Call gen_const_vec_duplicate instead of
16494 gen_const_vec_duplicate_1.
16495 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16496 (gen_const_vec_duplicate): Use rtx_vector_builder.
16497 (gen_const_vec_series): Likewise.
16498 (gen_rtx_CONST_VECTOR): Likewise.
16499 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16500 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16501 Build a new vector rather than modifying a CONST_VECTOR in-place.
16502 (handle_special_swappables): Update call accordingly.
16503 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16504 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16505 Build a new vector rather than modifying a CONST_VECTOR in-place.
16506 (handle_special_swappables): Update call accordingly.
16508 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16510 * simplify-rtx.c (simplify_const_binary_operation): Use
16511 CONST_VECTOR_ELT instead of XVECEXP.
16513 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16515 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16516 the selector elements to be different from the data elements
16517 if the selector is a VECTOR_CST.
16518 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16519 ssizetype for the selector.
16521 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16523 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16524 before testing each element individually.
16525 * tree-vect-generic.c (lower_vec_perm): Likewise.
16527 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16529 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16530 * selftest-run-tests.c (selftest::run_tests): Call it.
16531 * vector-builder.h (vector_builder::operator ==): New function.
16532 (vector_builder::operator !=): Likewise.
16533 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16534 (vec_perm_indices::all_from_input_p): New function.
16535 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16536 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16537 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16538 instead of reading the VECTOR_CST directly. Detect whether both
16539 vector inputs are the same before constructing the vec_perm_indices,
16540 and update the number of inputs argument accordingly. Use the
16541 utility functions added above. Only construct sel2 if we need to.
16543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16545 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16546 the broadcast of the low byte.
16547 (expand_mult_highpart): Use an explicit encoding for the permutes.
16548 * optabs-query.c (can_mult_highpart_p): Likewise.
16549 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16550 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16551 (vectorizable_bswap): Likewise.
16552 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16553 explicit encoding for the power-of-2 permutes.
16554 (vect_permute_store_chain): Likewise.
16555 (vect_grouped_load_supported): Likewise.
16556 (vect_permute_load_chain): Likewise.
16558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16560 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16561 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16562 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16563 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16564 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16565 (vect_gen_perm_mask_any): Likewise.
16567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16569 * int-vector-builder.h: New file.
16570 * vec-perm-indices.h: Include int-vector-builder.h.
16571 (vec_perm_indices): Redefine as an int_vector_builder.
16572 (auto_vec_perm_indices): Delete.
16573 (vec_perm_builder): Redefine as a stand-alone class.
16574 (vec_perm_indices::vec_perm_indices): New function.
16575 (vec_perm_indices::clamp): Likewise.
16576 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16577 (vec_perm_indices::new_vector): New function.
16578 (vec_perm_indices::new_expanded_vector): Update for new
16579 vec_perm_indices class.
16580 (vec_perm_indices::rotate_inputs): New function.
16581 (vec_perm_indices::all_in_range_p): Operate directly on the
16582 encoded form, without computing elided elements.
16583 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16584 encoding. Update for new vec_perm_indices class.
16585 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16586 the given vec_perm_builder.
16587 (expand_vec_perm_var): Update vec_perm_builder constructor.
16588 (expand_mult_highpart): Use vec_perm_builder instead of
16589 auto_vec_perm_indices.
16590 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16591 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16592 or double series encoding as appropriate.
16593 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16594 vec_perm_indices instead of auto_vec_perm_indices.
16595 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16596 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16597 (vect_permute_store_chain): Likewise.
16598 (vect_grouped_load_supported): Likewise.
16599 (vect_permute_load_chain): Likewise.
16600 (vect_shift_permute_load_chain): Likewise.
16601 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16602 (vect_transform_slp_perm_load): Likewise.
16603 (vect_schedule_slp_instance): Likewise.
16604 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16605 (vectorizable_mask_load_store): Likewise.
16606 (vectorizable_bswap): Likewise.
16607 (vectorizable_store): Likewise.
16608 (vectorizable_load): Likewise.
16609 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16610 vec_perm_indices instead of auto_vec_perm_indices. Use
16611 tree_to_vec_perm_builder to read the vector from a tree.
16612 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16613 vec_perm_builder instead of a vec_perm_indices.
16614 (have_whole_vector_shift): Use vec_perm_builder and
16615 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16616 truncation to calc_vec_perm_mask_for_shift.
16617 (vect_create_epilog_for_reduction): Likewise.
16618 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16619 from auto_vec_perm_indices to vec_perm_indices.
16620 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16621 instead of changing individual elements.
16622 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16623 the vector in d.perm.
16624 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16625 from auto_vec_perm_indices to vec_perm_indices.
16626 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16627 instead of changing individual elements.
16628 (arm_vectorize_vec_perm_const): Use new_vector to install
16629 the vector in d.perm.
16630 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16631 Update vec_perm_builder constructor.
16632 (rs6000_expand_interleave): Likewise.
16633 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16634 (rs6000_expand_interleave): Likewise.
16636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16638 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16639 to qimode could truncate the indices.
16640 * optabs.c (expand_vec_perm_var): Likewise.
16642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16644 * Makefile.in (OBJS): Add vec-perm-indices.o.
16645 * vec-perm-indices.h: New file.
16646 * vec-perm-indices.c: Likewise.
16647 * target.h (vec_perm_indices): Replace with a forward class
16649 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16650 * optabs.h: Include vec-perm-indices.h.
16651 (expand_vec_perm): Delete.
16652 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16653 (expand_vec_perm_const): Declare.
16654 * target.def (vec_perm_const_ok): Replace with...
16655 (vec_perm_const): ...this new hook.
16656 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16657 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16658 * doc/tm.texi: Regenerate.
16659 * optabs.def (vec_perm_const): Delete.
16660 * doc/md.texi (vec_perm_const): Likewise.
16661 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16662 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16663 expand_vec_perm for constant permutation vectors. Assert that
16664 the mode of variable permutation vectors is the integer equivalent
16665 of the mode that is being permuted.
16666 * optabs-query.h (selector_fits_mode_p): Declare.
16667 * optabs-query.c: Include vec-perm-indices.h.
16668 (selector_fits_mode_p): New function.
16669 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16670 is defined, instead of checking whether the vec_perm_const_optab
16671 exists. Use targetm.vectorize.vec_perm_const instead of
16672 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16673 fit in the vector mode before using a variable permute.
16674 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16675 vec_perm_indices instead of an rtx.
16676 (expand_vec_perm): Replace with...
16677 (expand_vec_perm_const): ...this new function. Take the selector
16678 as a vec_perm_indices rather than an rtx. Also take the mode of
16679 the selector. Update call to shift_amt_for_vec_perm_mask.
16680 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16681 Use vec_perm_indices::new_expanded_vector to expand the original
16682 selector into bytes. Check whether the indices fit in the vector
16683 mode before using a variable permute.
16684 (expand_vec_perm_var): Make global.
16685 (expand_mult_highpart): Use expand_vec_perm_const.
16686 * fold-const.c: Includes vec-perm-indices.h.
16687 * tree-ssa-forwprop.c: Likewise.
16688 * tree-vect-data-refs.c: Likewise.
16689 * tree-vect-generic.c: Likewise.
16690 * tree-vect-loop.c: Likewise.
16691 * tree-vect-slp.c: Likewise.
16692 * tree-vect-stmts.c: Likewise.
16693 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16695 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16696 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16697 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16698 (aarch64_vectorize_vec_perm_const): ...this new function.
16699 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16700 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16701 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16702 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16703 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16704 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16705 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16707 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16708 check for NEON modes.
16709 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16710 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16711 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16712 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16714 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16715 the old VEC_PERM_CONST conditions.
16716 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16717 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16718 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16719 (ia64_vectorize_vec_perm_const_ok): Merge into...
16720 (ia64_vectorize_vec_perm_const): ...this new function.
16721 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16722 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16723 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16724 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16725 * config/mips/mips.c (mips_expand_vec_perm_const)
16726 (mips_vectorize_vec_perm_const_ok): Merge into...
16727 (mips_vectorize_vec_perm_const): ...this new function.
16728 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16729 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16730 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16731 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16732 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16733 (rs6000_expand_vec_perm_const): Delete.
16734 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16736 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16737 (altivec_expand_vec_perm_const_le): Take each operand individually.
16738 Operate on constant selectors rather than rtxes.
16739 (altivec_expand_vec_perm_const): Likewise. Update call to
16740 altivec_expand_vec_perm_const_le.
16741 (rs6000_expand_vec_perm_const): Delete.
16742 (rs6000_vectorize_vec_perm_const_ok): Delete.
16743 (rs6000_vectorize_vec_perm_const): New function.
16744 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16745 an element count and rtx array.
16746 (rs6000_expand_extract_even): Update call accordingly.
16747 (rs6000_expand_interleave): Likewise.
16748 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16749 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16750 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16751 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16752 (rs6000_expand_vec_perm_const): Delete.
16753 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16754 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16755 (altivec_expand_vec_perm_const_le): Take each operand individually.
16756 Operate on constant selectors rather than rtxes.
16757 (altivec_expand_vec_perm_const): Likewise. Update call to
16758 altivec_expand_vec_perm_const_le.
16759 (rs6000_expand_vec_perm_const): Delete.
16760 (rs6000_vectorize_vec_perm_const_ok): Delete.
16761 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16762 reference to the SPE evmerge intructions.
16763 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16764 an element count and rtx array.
16765 (rs6000_expand_extract_even): Update call accordingly.
16766 (rs6000_expand_interleave): Likewise.
16767 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16768 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16770 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16772 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16774 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16775 vector mode and that that mode matches the mode of the data
16777 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16778 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16779 directly using expand_vec_perm_1 when forcing selectors into
16781 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16785 * optabs-query.h (can_vec_perm_p): Delete.
16786 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16787 * optabs-query.c (can_vec_perm_p): Split into...
16788 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16789 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16790 particular selector is valid.
16791 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16792 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16793 (vect_grouped_load_supported): Likewise.
16794 (vect_shift_permute_load_chain): Likewise.
16795 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16796 (vect_transform_slp_perm_load): Likewise.
16797 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16798 (vectorizable_bswap): Likewise.
16799 (vect_gen_perm_mask_checked): Likewise.
16800 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16801 implementations of variable permutation vectors into account
16802 when deciding which selector to use.
16803 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16804 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16805 with a false third argument.
16806 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16807 to test whether the constant selector is valid and can_vec_perm_var_p
16808 to test whether a variable selector is valid.
16810 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16812 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16813 * optabs-query.c (can_vec_perm_p): Likewise.
16814 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16815 instead of vec_perm_indices.
16816 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16817 (vect_gen_perm_mask_checked): Likewise,
16818 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16819 (vect_gen_perm_mask_checked): Likewise,
16821 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16823 * optabs-query.h (qimode_for_vec_perm): Declare.
16824 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16825 (qimode_for_vec_perm): ...this new function.
16826 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16828 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16830 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16831 does not have a conditional at the top.
16833 2018-01-02 Richard Biener <rguenther@suse.de>
16835 * ipa-inline.c (big_speedup_p): Fix expression.
16837 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16840 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16843 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16847 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16848 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16849 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16850 cond_taken_branch_cost 3->4.
16852 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16854 PR tree-optimization/83581
16855 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16856 TODO_cleanup_cfg if any changes have been made.
16858 PR middle-end/83608
16859 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16860 convert_modes if target mode has the right side, but different mode
16863 PR middle-end/83609
16864 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16865 last argument when extracting from CONCAT. If either from_real or
16866 from_imag is NULL, use expansion through memory. If result is not
16867 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16868 the parts directly to inner mode, if even that fails, use expansion
16871 PR middle-end/83623
16872 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16873 check for bswap in mode rather than HImode and use that in expand_unop
16876 Copyright (C) 2018 Free Software Foundation, Inc.
16878 Copying and distribution of this file, with or without modification,
16879 are permitted in any medium without royalty provided the copyright
16880 notice and this notice are preserved.