1 2018-06-18 Martin Sebor <msebor@redhat.com>
4 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
5 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6 Handle integer subtraction.
7 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10 2018-06-18 David Malcolm <dmalcolm@redhat.com>
12 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
13 param from rtx to rtx_insn *.
14 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
16 (frv_ifcvt_modify_insn): Likwise.
17 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
18 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
19 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
20 as_a <rtx_insn *> cast to local "unprotected_region" once
21 it's been established that it's not NULL or pc_rtx.
22 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
23 param "sethi" from rtx to rtx_insn *.
24 (nds32_group_float_insns): Likewise for param "insn".
25 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
27 (vax_output_int_subtract): Likewise.
28 * config/vax/vax.c (vax_output_int_add): Likewise for param
30 (vax_output_int_subtract): Likewise.
31 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
32 (emit_pattern_after): Likewise for param "after".
33 (emit_insn_after): Likewise.
34 (emit_jump_insn_after): Likewise.
35 (emit_call_insn_after): Likewise.
36 (emit_debug_insn_after): Likewise.
37 (emit_pattern_before): Likewise for param "before".
38 (emit_insn_before): Likewise.
39 (emit_jump_insn_before): Likewise.
40 * final.c (get_insn_template): Likewise for param "insn", removing
42 * output.h (get_insn_template): Likewise for 2nd param.
43 * rtl.h (emit_insn_before): Likewise.
44 (emit_jump_insn_before): Likewise.
45 (emit_debug_insn_before_noloc): Likewise.
46 (emit_insn_after): Likewise.
47 (emit_jump_insn_after): Likewise.
48 (emit_call_insn_after): Likewise.
49 (emit_debug_insn_after): Likewise.
50 (set_insn_deleted): Likewise for param.
52 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
55 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
56 floating point modes, so that IFmode is numerically greater than
57 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
58 to declare the ordering. This prevents IFmode from being
59 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
60 machine. Include rs6000-modes.h to share the fractional values
61 between genmodes* and the rest of the compiler.
65 * config/rs6000/rs6000-modes.h: New file.
66 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
67 meaning of rs6000_long_double_size so that 126..128 selects an
68 appropriate 128-bit floating point type.
69 (rs6000_option_override_internal): Likewise.
70 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
71 (TARGET_LONG_DOUBLE_128): Change the meaning of
72 rs6000_long_double_size so that 126..128 selects an appropriate
73 128-bit floating point type.
74 (LONG_DOUBLE_TYPE_SIZE): Update comment.
75 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
76 source and destination to match the standard usage.
77 (truncifkf2): Likewise.
78 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
79 ISA 2.07 to use an explicit clobber, instead of passing in a
81 (copysign<mode>3_soft): Likewise.
83 2018-06-18 David Malcolm <dmalcolm@redhat.com>
85 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
86 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
87 (vect_slp_analyze_instance_dependence): Likewise.
88 (vect_enhance_data_refs_alignment): Likewise.
89 (vect_analyze_data_refs_alignment): Likewise.
90 (vect_slp_analyze_and_verify_instance_alignment
91 (vect_analyze_data_ref_accesses): Likewise.
92 (vect_prune_runtime_alias_test_list): Likewise.
93 (vect_analyze_data_refs): Likewise.
94 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
95 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
96 (vect_analyze_scalar_cycles_1): Likewise.
97 (vect_get_loop_niters): Likewise.
98 (vect_analyze_loop_form_1): Likewise.
99 (vect_update_vf_for_slp): Likewise.
100 (vect_analyze_loop_operations): Likewise.
101 (vect_analyze_loop): Likewise.
102 (vectorizable_induction): Likewise.
103 (vect_transform_loop): Likewise.
104 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
105 * tree-vect-slp.c (vect_analyze_slp): Likewise.
106 (vect_make_slp_decision): Likewise.
107 (vect_detect_hybrid_slp): Likewise.
108 (vect_slp_analyze_operations): Likewise.
109 (vect_slp_bb): Likewise.
110 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
111 (vectorizable_bswap): Likewise.
112 (vectorizable_call): Likewise.
113 (vectorizable_simd_clone_call): Likewise.
114 (vectorizable_conversion): Likewise.
115 (vectorizable_assignment): Likewise.
116 (vectorizable_shift): Likewise.
117 (vectorizable_operation): Likewise.
118 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
120 2018-06-18 Martin Sebor <msebor@redhat.com>
122 PR tree-optimization/81384
123 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
124 * builtins.c (expand_builtin_strnlen): New function.
125 (expand_builtin): Call it.
126 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
127 * builtins.def (BUILT_IN_STRNLEN): New.
128 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
129 Warn for bounds in excess of maximum object size.
130 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
131 single-value ranges. Handle strnlen.
132 (handle_builtin_strlen): Handle strnlen.
133 (strlen_check_and_optimize_stmt): Same.
134 * doc/extend.texi (Other Builtins): Document strnlen.
136 2018-06-18 Maya Rashish <coypu@sdf.org>
138 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
139 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
140 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
142 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
144 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
146 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
148 * tree.c (escaped_string::escape): Replace cast to char * by
149 const_cast<char *> (unescaped).
151 2018-06-18 Nick Clifton <nickc@redhat.com>
154 * tree.c (escaped_string): New class. Converts an unescaped
155 string into its escaped equivalent.
156 (warn_deprecated_use): Use the new class to convert the
157 deprecation message, if present.
158 (test_escaped_strings): New self test.
159 (test_c_tests): Add test_escaped_strings.
160 * doc/extend.texi (deprecated): Add a note that the
161 deprecation message is affected by the -fmessage-length
162 option, and that control characters will be escaped.
163 (#pragma GCC error): Document this pragma.
164 (#pragma GCC warning): Likewise.
165 * doc/invoke.texi (-fmessage-length): Document this option's
166 effect on the #warning and #error preprocessor directives and
167 the deprecated attribute.
169 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
171 * tree.c (decl_value_expr_lookup): Revert latest change.
172 (decl_value_expr_insert): Likewise.
174 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
176 * gimplify.c (nonlocal_vlas): Delete.
177 (nonlocal_vla_vars): Likewise.
178 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
180 (gimplify_body): Do not create and destroy nonlocal_vlas.
181 * tree-nested.c: Include diagnostic.h.
182 (use_pointer_in_frame): Tweak.
183 (lookup_field_for_decl): Add assertion and declare the transformation.
184 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
185 internal error when the reference is in a wrong context. Do not
186 create a debug decl by default.
187 (note_nonlocal_block_vlas): Delete.
188 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
189 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
190 create a debug decl by default.
191 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
192 call is in a wrong context.
193 (fixup_vla_decls): New function.
194 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
195 debug variables were created.
196 * tree.c (decl_value_expr_lookup): Add checking assertion.
197 (decl_value_expr_insert): Likewise.
199 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
202 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
203 * tree-scalar-evolution.c (interpret_expr): Likewise.
204 (expression_expensive_p): Likewise.
205 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
206 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
207 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
208 (ssa_defined_by_minus_one_stmt_p): New.
210 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
213 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
214 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
215 * dojump.c (do_jump): Likewise.
216 * expr.c (expand_expr_real_2): Check operand type's sign.
217 * fold-const.c (const_unop): Handle ABSU_EXPR.
218 (fold_abs_const): Likewise.
219 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
220 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
221 (strip_sign_op_1): Likesise.
222 * match.pd: Add new pattern to generate ABSU_EXPR.
223 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
224 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
225 * tree-eh.c (operation_could_trap_helper_p): Likewise.
226 * tree-inline.c (estimate_operator_cost): Likewise.
227 * tree-pretty-print.c (dump_generic_node): Likewise.
228 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
229 * tree.def (ABSU_EXPR): New.
231 2018-06-16 Jakub Jelinek <jakub@redhat.com>
234 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
235 documented as preserved for backward compatibility only.
236 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
238 PR rtl-optimization/86108
239 * bb-reorder.c (create_forwarder_block): Renamed to ...
240 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
241 jump from new landing pad to the second part.
242 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
245 2018-06-15 Jakub Jelinek <jakub@redhat.com>
248 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
249 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
250 Only call store_expr for halves if the mode is the same.
253 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
254 Fix up comment formatting.
256 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
258 * typed-splay-tree.h (typed_splay_tree::remove): New function.
259 (typed_splay_tree::closure,
260 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
261 (typed_splay_tree::typed_splay_tree,
262 typed_splay_tree::operator =): Declared private.
263 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
264 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
265 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
266 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
267 typed_splay_tree::splay_tree_splay,
268 typed_splay_tree::splay_tree_foreach_helper,
269 typed_splay_tree::splay_tree_insert,
270 typed_splay_tree::splay_tree_remove,
271 typed_splay_tree::splay_tree_lookup,
272 typed_splay_tree::splay_tree_predecessor,
273 typed_splay_tree::splay_tree_successor,
274 typed_splay_tree::splay_tree_min,
275 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
276 (typed_splay_tree::root, typed_splay_tree::comp,
277 typed_splay_tree::delete_key,
278 typed_splay_tree::delete_value): New data members.
279 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
280 typed_splay_tree::remove.
282 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
284 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
285 -mginv and -mno-ginv to the assembler.
286 * config/mips/mips.opt (-mcrc): New option.
288 * doc/invoke.text (-mcrc): Document.
291 2018-06-15 Nick Clifton <nickc@redhat.com>
294 * tree.c (escaped_string): New class. Converts an unescaped
295 string into its escaped equivalent.
296 (warn_deprecated_use): Use the new class to convert the
297 deprecation message, if present.
298 (test_escaped_strings): New self test.
299 (test_c_tests): Add test_escaped_strings.
300 * doc/extend.texi (deprecated): Add a note that the
301 deprecation message is affected by the -fmessage-length
302 option, and that control characters will be escaped.
303 (#pragma GCC error): Document this pragma.
304 (#pragma GCC warning): Likewise.
305 * doc/invoke.texi (-fmessage-length): Document this option's
306 effect on the #warning and #error preprocessor directives and
307 the deprecated attribute.
309 2018-06-15 Richard Biener <rguenther@suse.de>
311 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
312 here, also noting vector size used.
313 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
314 size used in MSG_OPTIMIZED_LOCATIONS dump.
315 (pass_slp_vectorize::execute): Adjust.
317 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
320 * config/arc/arc.c (arc_return_address_register): Fix
323 2018-06-15 Richard Biener <rguenther@suse.de>
326 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
327 leave useless conversion stripping to force_gimple_operand_gsi.
328 (gimplify_build2): Likewise.
329 (gimplify_build1): Likewise.
331 2018-06-15 Richard Biener <rguenther@suse.de>
334 * tree-cfg.c (move_stmt_op): unshare invariant addresses
335 before adjusting their block.
337 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
339 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
340 multilibs for *-*-rtems*.
341 * config/riscv/t-rtems: New file.
343 2018-06-14 Jakub Jelinek <jakub@redhat.com>
346 * match.pd ((A +- CST1) +- CST2): Punt if last resort
347 unsigned_type_for returns NULL.
350 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
351 subregs of multi-word pseudos unless the float mode has word size.
353 2018-06-14 Richard Biener <rguenther@suse.de>
356 * tree-vect-generic.c (build_word_mode_vector_type): Remove
357 duplicate and harmful type_hash_canon.
358 * tree.c (type_hash_canon): Assert we didn't find ourselves.
360 2018-06-14 Richard Biener <rguenther@suse.de>
363 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
366 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
368 * config/rtems.h (STDINT_LONG32): Define.
370 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
371 Prachi Godbole <prachi.godbole@imgtec.com>
373 * config/mips/mips-cpus.def: Define P6600.
374 * config/mips/mips-tables.opt: Regenerate.
375 * config/mips/mips.c (mips_ucbranch_type): New enum.
376 (mips_rtx_cost_data): Add support for P6600.
377 (mips_issue_rate): Likewise.
378 (mips_multipass_dfa_lookahead): Likewise.
379 (mips_avoid_hazard): Likewise.
380 (mips_reorg_process_insns): Likewise.
381 (mips_classify_branch_p6600): New function.
382 * config/mips/mips.h (TUNE_P6600): New define.
383 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
384 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
385 * config/mips/mips.md: Include p6600.md.
386 (processor): Add p6600.
387 * config/mips/p6600.md: New file.
388 * doc/invoke.texi: Add p6600 to supported architectures.
390 2018-06-13 Martin Sebor <msebor@redhat.com>
392 PR tree-optimization/86114
393 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
395 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
397 2018-06-13 Richard Biener <rguenther@suse.de>
399 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
400 Properly set vector type of the intermediate stmt.
401 * tree-vect-stmts.c (vectorizable_operation): The destination
402 var always has vectype_out type.
404 2018-06-13 Jeff Law <law@redhat.com>
406 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
407 integer 0 for argument to print_rtl_with_bb.
408 (rl78_reorg): Likewise.
410 2018-06-13 David Malcolm <dmalcolm@redhat.com>
412 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
413 from rtx to rtx_insn *.
414 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
416 (add_sched_insns_for_speculation): Likewise for local "target",
417 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
418 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
419 from rtx_insn ** to rtx_code_label **.
420 (reorg_emit_nops): Likewise.
421 (c6x_reorg): Likewise for local "call_labels".
422 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
424 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
425 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
426 the loops over LABEL_REFS.
427 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
429 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
430 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
431 (split_branches): Strengthen local "olabel" from rtx to
432 rtx_insn *, adding a safe_as_a cast.
433 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
435 (add_insn_after): Likewise for first two params.
436 (add_insn_before): Likewise.
437 (remove_insn): Likewise for param.
438 (emit_pattern_before_noloc): Likewise for second and third params.
439 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
440 (emit_call_insn_before_noloc): Likewise.
441 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
443 (emit_barrier_before): Likewise.
444 (emit_label_before): Strengthen "label" param from "rtx" to
445 "rtx_code_label *". Strengthen "before" param from "rtx" to
447 (emit_insn_after_1): Strengthen "after" param from "rtx" to
449 (emit_pattern_after_noloc): Likewise.
450 (emit_insn_after_noloc): Likewise.
451 (emit_jump_insn_after_noloc): Likewise.
452 (emit_call_insn_after_noloc): Likewise.
453 (emit_debug_insn_after_noloc): Likewise.
454 (emit_barrier_after): Likewise.
455 (emit_label_after): Likewise for both params.
456 (emit_pattern_after_setloc): Likewise for "after" param. Convert
457 "loc" param from "int" to "location_t".
458 (emit_insn_after_setloc): Likewise.
459 (emit_jump_insn_after_setloc): Likewise.
460 (emit_call_insn_after_setloc): Likewise.
461 (emit_debug_insn_after_setloc): Likewise.
462 (emit_pattern_before_setloc): Likewise for "before" param. Convert
463 "loc" param from "int" to "location_t".
464 (emit_pattern_before): Convert NULL_RTX to NULL.
465 (emit_insn_before_setloc): Convert "loc" param from "int" to
467 (emit_jump_insn_before_setloc): Likewise.
468 (emit_call_insn_before_setloc): Likewise.
469 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
470 rtx_insn *. Convert "loc" param from "int" to "location_t".
471 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
472 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
473 Convert 3rd param from "int" to "location_t".
474 (emit_barrier_before, emit_barrier_after, next_real_insn):
475 Strengthen param from rtx to rtx_insn *.
476 (emit_label_before): Strengthen 1st param from "rtx" to
477 "rtx_code_label *". Strengthen 2nd param from "rtx" to
479 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
480 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
481 Strengthen 2nd param from "rtx" to "rtx_insn *".
482 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
483 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
484 Likewise. Convert 3rd param from "int" to "location_t".
485 (emit_label_after): Strengthen 1st param from "rtx" to
487 (next_real_insn, remove_insn): Strengthen param from "rtx" to
489 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
490 from "rtx" to "rtx_insn *".
492 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
494 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
495 bodies streamed in with -Q.
496 * dumpfile.c (dump_files): Add lto-stream-out dump file.
497 * dumpfile.h (tree_dump_index): Add lto_stream_out.
498 * gimple-streamer-out.c: Include gimple-pretty-print.h
499 (output_bb): Dump stmts streamed.
500 * lto-section-out.c: Include print-tree.h
501 (lto_begin_section): Dump sections created.
502 (lto_output_decl_index): Dump decl encoded.
503 * lto-streamer-out.c: Include print-tree.h
504 (create_output_block): Dump output block created.
505 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
506 (output_function): Dump function output.
507 (output_constructor): Dump constructor streamed.
508 (write_global_stream): Output indexes encoded.
509 (produce_asm_for_decls): Dump streams encoded.
510 * lto-streamer.c (streamer_dump_file): New global var.
511 * lto-streamer.h (streamer_dump_file): Declare.
512 * passes.c (ipa_write_summaries): Initialize streamer dump.
513 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
516 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
519 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
520 offsets for register save directives. Emit a second batch of save
521 directives, if need be, when the function accesses prior frames.
523 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
525 * config/arc/fpu.md (fmasf4): Force operand to register.
528 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
530 * config/arc/arc-protos.h (arc_pad_return): Remove.
531 * config/arc/arc.c (machine_function): Remove force_short_suffix
533 (arc_print_operand): Adjust printing of '&'.
534 (arc_verify_short): Remove conditional printing of short suffix.
535 (arc_final_prescan_insn): Remove reference to size_reason.
536 (pad_return): New function.
537 (arc_reorg): Call pad_return.
538 (arc_pad_return): Remove.
539 (arc_init_machine_status): Remove reference to force_short_suffix.
540 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
541 (attr length): When attribute iscompact is true force to 2
542 regardless; in the case of maybe check if we want to force the
543 instruction to have 4 bytes length.
544 (nopv): Change it to generate 4 byte long nop as well.
545 (blockage): New pattern.
546 (simple_return): Remove call to arc_pad_return.
547 (p_return_i): Likewise.
549 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
551 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
553 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
555 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
558 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
560 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
561 for ARC700 and ARCv2.
563 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
566 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
567 operands[2] instead of operands[1].
570 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
572 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
573 case, check whether the outer register overlaps an unallocatable
574 register, not just whether it fits the required class.
576 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
578 * poly-int.h (can_div_trunc_p): Add new overload in which all values
580 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
581 (memrefs_conflict_p): Likewise.
582 (init_alias_analysis): Likewise.
583 * cfgexpand.c (expand_debug_expr): Likewise.
584 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
585 * cse.c (fold_rtx): Likewise.
586 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
587 * expr.c (emit_block_move_hints): Likewise.
588 (clear_storage_hints, push_block, emit_push_insn): Likewise.
589 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
590 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
591 (emit_group_store): Likewise.
592 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
593 to read the PRE/POST_MODIFY increment.
594 * calls.c (store_one_arg): Use strip_offset.
595 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
597 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
599 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
600 (simplify_binary_operation_1): Extend CONST_INT handling to
602 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
603 than a HOST_WIDE_INT.
604 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
606 (adjust_mems, add_stores): Update accodingly.
607 (vt_canonicalize_addr): Track polynomial offsets.
608 (emit_note_insn_var_location): Likewise.
609 (vt_add_function_parameter): Likewise.
610 (vt_initialize): Likewise.
612 2018-06-12 Jeff Law <law@redhat.com>
614 * config.gcc (alpha*-*-freebsd*): Remove.
615 * config/alpha/freebsd.h: Remove.
617 2018-06-12 David Malcolm <dmalcolm@redhat.com>
620 * spellcheck-tree.c (levenshtein_distance): Rename to...
621 (get_edit_distance): ...this, and update for underlying renaming.
622 * spellcheck-tree.h (levenshtein_distance): Rename to...
623 (get_edit_distance): ...this.
624 * spellcheck.c (levenshtein_distance): Rename to...
625 (get_edit_distance): ...this. Convert from Levenshtein distance
626 to Damerau-Levenshtein distance by supporting transpositions of
627 adjacent characters. Rename "v1" to "v_next" and "v0" to
629 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
630 (selftest::test_edit_distance_unit_test_oneway): ...this, and
631 update for underlying renaming.
632 (selftest::levenshtein_distance_unit_test): Rename to...
633 (selftest::test_get_edit_distance_unit): ...this, and update for
635 (selftest::test_find_closest_string): Add example from PR 69968
636 where transposition helps
637 (selftest::test_metric_conditions): Update for renaming.
638 (selftest::test_metric_conditions): Likewise.
639 (selftest::spellcheck_c_tests): Likewise.
640 * spellcheck.h (levenshtein_distance): Rename both overloads to...
641 (get_edit_distance): ...this.
642 (best_match::consider): Update for renaming.
644 2018-06-12 Martin Sebor <msebor@redhat.com>
646 PR tree-optimization/85259
647 * builtins.c (compute_objsize): Handle constant offsets.
648 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
649 true iff a warning has been issued.
650 * gimple.h (gimple_nonartificial_location): New function.
651 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
652 gimple_nonartificial_location and handle -Wno-system-headers.
653 (handle_builtin_stxncpy): Same.
655 2018-06-12 Martin Sebor <msebor@redhat.com>
658 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
660 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
662 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
663 BUILTIN_VEC_XST entries for pointer to double and long long.
665 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
668 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
670 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
673 2018-06-12 Martin Liska <mliska@suse.cz>
675 * doc/options.texi: Document IntegerRange.
677 2018-06-12 Martin Liska <mliska@suse.cz>
679 * config/i386/i386.opt: Make MPX-related options as Deprecated.
680 * opt-functions.awk: Handle Deprecated flag.
681 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
683 (read_cmdline_option): Report warning for a deprecated option.
684 * opts.h (struct cl_option): Add new field cl_deprecated.
685 (CL_ERR_DEPRECATED): New.
687 2018-06-12 Martin Liska <mliska@suse.cz>
689 * doc/options.texi: Document Deprecated option flag.
691 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
693 * config/arc/arc-arch.h (arc_extras): New enum.
694 (arc_cpu_t):Add field extra.
695 (arc_cpu_types): Consider the extras.
696 * config/arc/arc-cpus.def: Add extras info.
697 * config/arc/arc-opts.h (processor_type): Consider extra field.
698 * config/arc/arc.c (arc_override_options): Handle extra field.
700 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
702 * config/arc/arc-arch.h: Update ARC_OPTX macro.
703 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
705 * config/arc/arc.c (arc_init): Update pic warning.
706 (irq_range): Update irq range parsing warnings.
707 (arc_override_options): Update various warning messages.
708 (arc_handle_aux_attribute): Likewise.
710 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
712 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
714 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
716 * doc/sourcebuild.texi: Document usage of line number 0 in verify
717 compiler messages directives.
719 2018-06-12 Matthew Fortune <mfortune@gmail.com>
721 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
722 * config/mips/mips-tables.opt: Regenerate.
723 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
725 * doc/invoke.texi: Document -march=i6500.
727 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
729 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
730 (i6400_gpmul): Add cpu_unit.
731 (i6400_gpdiv): Likewise.
732 (i6400_msa_add_d): Update reservations.
733 (i6400_msa_int_add) Likewise.
734 (i6400_msa_short_logic3) Likewise.
735 (i6400_msa_short_logic2) Likewise.
736 (i6400_msa_short_logic) Likewise.
737 (i6400_msa_move) Likewise.
738 (i6400_msa_cmp) Likewise.
739 (i6400_msa_short_float2) Likewise.
740 (i6400_msa_div_d) Likewise.
741 (i6400_msa_long_logic1) Likewise.
742 (i6400_msa_long_logic2) Likewise.
743 (i6400_msa_mult) Likewise.
744 (i6400_msa_long_float2) Likewise.
745 (i6400_msa_long_float4) Likewise.
746 (i6400_msa_long_float5) Likewise.
747 (i6400_msa_long_float8) Likewise.
748 (i6400_fpu_fadd): Include frint type.
749 (i6400_fpu_store): New define_insn_reservation.
750 (i6400_fpu_load): Likewise.
751 (i6400_fpu_move): Likewise.
752 (i6400_fpu_fcmp): Likewise.
753 (i6400_fpu_fmadd): Likewise.
754 (i6400_int_mult): Include imul3nc type and update reservation.
755 (i6400_int_div): Include idiv3 type and update reservation.
756 (i6400_int_load): Update to check type not move_type.
757 (i6400_int_store): Likewise.
758 (i6400_int_prefetch): Set zero latency.
760 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
762 * gcc.c: Document new %@{...} sequence.
763 (LINK_COMMAND_SPEC): Use it for the -L switches.
764 (cpp_unique_options): Use it for the -I switches.
765 (at_file_argbuf): New global variable.
766 (in_at_file): Likewise.
767 (alloc_args): Create at_file_argbuf.
768 (clear_args): Truncate at_file_argbuf.
769 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
770 (open_at_file): New function.
771 (close_at_file): Likewise.
772 (create_at_file): Delete.
773 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
776 (validate_switches_from_spec): Deal with %@{...} sequence.
777 (validate_switches): Likewise.
778 (driver::finalize): Call clear_args.
780 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
782 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
784 2018-06-11 Martin Sebor <msebor@redhat.com>
786 * doc/invoke.texi (-Wall): List -Wc++17-compat.
787 (Wno-class-memaccess): Add @opindex.
788 (Wno-templates, Wno-multiple-inheritance): Same.
789 (Wno-virtual-inheritance, Wno-namespaces): Same.
790 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
791 (Wno-format-overflow, Wno-format-truncation): Same.
792 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
793 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
794 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
795 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
796 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
797 (Wno-misspelled-isr): Same.
799 2018-06-11 Martin Sebor <msebor@redhat.com>
801 * PR tree-optimization/86083
802 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
804 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
806 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
808 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
811 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
812 on the correct operand.
813 (*movdi_internal64): Ditto.
815 2018-06-11 Martin Liska <mliska@suse.cz>
817 PR tree-optimization/86089
818 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
820 2018-06-11 Julia Koval <julia.koval@intel.com>
822 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
823 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
824 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
826 2018-06-11 Olivier Hainque <hainque@adacore.com>
828 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
829 for Ada with strict dwarf2.
831 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
834 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
837 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
839 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
841 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
843 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
846 2018-06-08 Martin Liska <mliska@suse.cz>
848 * tree-cfg.h (debug_function): Fix argument type to match
851 2018-06-08 Martin Liska <mliska@suse.cz>
853 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
854 Remove usage of MPX-related (and removed) fields.
855 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
857 2018-06-08 David Malcolm <dmalcolm@redhat.com>
859 * cfg.c (debug): Use TDF_NONE rather than 0.
860 * cfghooks.c (debug): Likewise.
861 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
862 (struct dump_option_value_info): Convert to...
863 (struct kv_pair): ...this template type.
864 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
866 (optinfo_verbosity_options): Likewise.
867 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
869 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
870 than int for "optgroup_flags" param.
871 (dump_generic_expr_loc): Use dump_flags_t rather than int for
873 (dump_dec): Likewise.
874 (dump_finish): Use TDF_NONE rather than 0.
875 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
876 rather than int for "optgroup_flags" param. Use TDF_NONE rather
877 than 0. Update for change to option_ptr.
878 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
879 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
880 0. Update for changes to optinfo_verbosity_options and
882 (opt_info_switch_p): Convert optgroup_flags from int to
884 (dump_basic_block): Use dump_flags_t rather than int
885 for "dump_kind" param.
886 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
887 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
888 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
889 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
890 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
891 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
892 TDF_NONE): Convert from macros to...
893 (enum dump_flag): ...this new enum.
894 (dump_flags_t): Update to use enum.
895 (operator|, operator&, operator~, operator|=, operator&=):
896 Implement for dump_flags_t.
897 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
898 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
899 Convert from macros to...
900 (enum optgroup_flag): ...this new enum.
901 (optgroup_flags_t): New typedef.
902 (operator|, operator|=): Implement for optgroup_flags_t.
903 (struct dump_file_info): Convert field "alt_flags" to
904 dump_flags_t. Convert field "optgroup_flags" to
906 (dump_basic_block): Use dump_flags_t rather than int for param.
907 (dump_generic_expr_loc): Likewise.
908 (dump_dec): Likewise.
909 (dump_register): Convert param "optgroup_flags" to
911 (opt_info_enable_passes): Likewise.
912 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
914 * gimple-pretty-print.c (debug): Likewise.
915 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
916 (merged_store_group::apply_stores): Likewise.
917 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
918 * gimple.c (verify_gimple_pp): Likewise.
919 * graphite-poly.c (print_pbb_body): Likewise.
920 * passes.c (pass_manager::register_one_dump_file): Convert
921 local "optgroup_flags" to optgroup_flags_t.
922 * print-tree.c (print_node): Use TDF_NONE rather than 0.
924 (debug_body): Likewise.
925 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
927 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
929 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
930 (convert_mult_to_fma): Likewise.
931 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
932 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
933 * tree-vect-data-refs.c (dump_lower_bound): Convert param
934 "dump_kind" to dump_flags_t.
936 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
938 * config/rs6000/rs6000.c (min, max): Delete.
940 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
942 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
943 -mabi=spe and -mabi=no-spe.
945 2018-06-08 Martin Liska <mliska@suse.cz>
947 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
948 where we expect an existing summary.
950 2018-06-08 Martin Liska <mliska@suse.cz>
952 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
953 * ipa-inline.h (estimate_edge_growth): Likewise.
955 2018-06-08 Martin Liska <mliska@suse.cz>
957 * cgraph.c (function_version_hasher::hash): Use
958 cgraph_node::get_uid ().
959 (function_version_hasher::equal):
960 * cgraph.h (cgraph_node::get_uid): New method.
961 * ipa-inline.c (update_caller_keys): Use
962 cgraph_node::get_uid ().
963 (update_callee_keys): Likewise.
964 * ipa-utils.c (searchc): Likewise.
965 (ipa_reduced_postorder): Likewise.
966 * lto-cgraph.c (input_node): Likewise.
967 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
968 * symbol-summary.h (symtab_insertion): Likewise.
969 (symtab_removal): Likewise.
970 (symtab_duplication): Likewise.
971 * tree-pretty-print.c (dump_function_header): Likewise.
972 * tree-sra.c (convert_callers_for_node): Likewise.
974 2018-06-08 Martin Liska <mliska@suse.cz>
976 * cgraph.c (symbol_table::create_edge): Always assign a new
978 (symbol_table::free_edge): Do not recycle numbers.
979 * cgraph.h (cgraph_edge::get): New method.
980 * symbol-summary.h (symtab_removal): Use it.
981 (symtab_duplication): Likewise.
982 (call_summary::hashable_uid): Remove.
984 2018-06-08 Martin Liska <mliska@suse.cz>
986 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
987 (initialize_growth_caches): Remove.
988 (free_growth_caches): Likewise.
989 (do_estimate_edge_time): Use edge_growth_cache.
990 (do_estimate_edge_size): Likewise.
991 (do_estimate_edge_hints): Likewise.
992 * ipa-inline.c (reset_edge_caches): Likewise.
993 (recursive_inlining): Likewise.
994 (inline_small_functions): Likewise.
995 * ipa-inline.h (initialize_growth_caches): Remove.
996 (estimate_edge_size): Likewise.
997 (estimate_edge_time): Likewise.
998 (estimate_edge_hints): Likewise.
999 (reset_edge_growth_cache): Likewise.
1000 * symbol-summary.h (call_summary::remove): New method.
1002 2018-06-08 Martin Liska <mliska@suse.cz>
1004 * ipa-cp.c (class edge_clone_summary): New summary.
1005 (grow_edge_clone_vectors): Remove.
1006 (ipcp_edge_duplication_hook): Remove.
1007 (class edge_clone_summary_t): New call_summary class.
1008 (ipcp_edge_removal_hook): Remove.
1009 (edge_clone_summary_t::duplicate): New function.
1010 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1011 (create_specialized_node): Likewise.
1012 (ipcp_driver): Initialize edge_clone_summaries and do not
1015 2018-06-08 Martin Liska <mliska@suse.cz>
1017 * symbol-summary.h (get): New function.
1018 (call_summary::m_initialize_when_cloning): New class member.
1020 2018-06-08 Martin Liska <mliska@suse.cz>
1022 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1023 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1024 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1025 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1026 (remove_cgraph_node_from_order): Use the removed_nodes set.
1027 (do_per_function_toporder): Likwise.
1028 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1029 instead of summary_uid.
1030 (symtab_removal): Likewise.
1031 (symtab_duplication): Likewise.
1033 2018-06-08 Martin Liska <mliska@suse.cz>
1035 * ipa-cp.c (ipcp_store_bits_results): Use
1036 ipcp_transformation_sum.
1037 (ipcp_store_vr_results): Likewise.
1038 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1040 (ipcp_transformation_initialize): ... this.
1041 (ipa_set_node_agg_value_chain):
1042 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1043 (write_ipcp_transformation_info): Likewise.
1044 (read_ipcp_transformation_info): Likewise.
1045 (ipcp_update_bits): Likewise.
1046 (ipcp_update_vr): Likewise.
1047 (ipcp_transform_function): Likewise.
1048 * ipa-prop.h: Rename ipcp_transformation_summary to
1049 ipcp_transformation.
1050 (class ipcp_transformation_t): New function summary.
1051 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1052 (ipa_get_agg_replacements_for_node): Likewise.
1054 2018-06-08 Martin Liska <mliska@suse.cz>
1056 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1058 (class funct_state_summary_t): New function_summary class.
1059 (has_function_state): Remove.
1060 (get_function_state): Likewise.
1061 (set_function_state): Likewise.
1062 (add_new_function): Likewise.
1063 (funct_state_summary_t::insert): New function.
1064 (duplicate_node_data): Remove.
1065 (remove_node_data): Remove.
1066 (funct_state_summary_t::duplicate): New function.
1067 (register_hooks): Create new funct_state_summaries.
1068 (pure_const_generate_summary): Use it.
1069 (pure_const_write_summary): Likewise.
1070 (pure_const_read_summary): Likewise.
1071 (propagate_pure_const): Likewise.
1072 (propagate_nothrow): Likewise.
1073 (dump_malloc_lattice): Likewise.
1074 (propagate_malloc): Likewise.
1075 (execute): Do not register hooks, just remove summary
1077 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1080 2018-06-08 Martin Liska <mliska@suse.cz>
1082 * ipa-reference.c (remove_node_data): Remove.
1083 (duplicate_node_data): Likewise.
1084 (class ipa_ref_var_info_summary_t): New class.
1085 (class ipa_ref_opt_summary_t): Likewise.
1086 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1087 (get_reference_optimization_summary): Use
1088 ipa_ref_opt_sum_summaries.
1089 (set_reference_vars_info): Remove.
1090 (set_reference_optimization_summary): Likewise.
1091 (ipa_init): Create summaries.
1092 (init_function_info): Use function summary.
1093 (ipa_ref_opt_summary_t::duplicate): New function.
1094 (ipa_ref_opt_summary_t::remove): New function.
1095 (get_read_write_all_from_node): Fix GNU coding style.
1096 (propagate): Use function summary.
1097 (write_node_summary_p): Fix GNU coding style.
1098 (stream_out_bitmap): Likewise.
1099 (ipa_reference_read_optimization_summary): Use function summary.
1100 (ipa_reference_c_finalize): Do not release hooks.
1102 2018-06-08 Martin Liska <mliska@suse.cz>
1104 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1105 (analyze_function_body): Extract multiple calls of get_create.
1106 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1107 * ipa-inline.c (recursive_inlining): Use ::get method.
1108 * ipa-inline.h (estimate_edge_growth): Likewise.
1110 2018-06-08 Martin Liska <mliska@suse.cz>
1112 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1114 (hsa_function_summary::hsa_function_summary): Use the new enum
1116 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1117 * hsa-gen.c (hsa_get_host_function): Likewise.
1118 (get_brig_function_name): Likewise.
1119 * ipa-hsa.c (process_hsa_functions): Likewise.
1120 (ipa_hsa_write_summary): Likewise.
1121 * symbol-summary.h (symtab_duplication): Use ::get function/
1122 (get): New function.
1124 2018-06-08 Martin Liska <mliska@suse.cz>
1126 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1128 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1129 (hsa_register_kernel): Likewise.
1130 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1131 * hsa-gen.c (hsa_get_host_function): Likewise.
1132 (get_brig_function_name): Likewise.
1133 (generate_hsa): Likewise.
1134 (pass_gen_hsail::execute): Likewise.
1135 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1136 (devirtualization_time_bonus): Likewise.
1137 (ipcp_propagate_stage): Likewise.
1138 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1139 (edge_set_predicate): Likewise.
1140 (evaluate_conditions_for_known_args): Likewise.
1141 (evaluate_properties_for_edge): Likewise.
1142 (ipa_fn_summary::reset): Likewise.
1143 (ipa_fn_summary_t::duplicate): Likewise.
1144 (dump_ipa_call_summary): Likewise.
1145 (ipa_dump_fn_summary): Likewise.
1146 (analyze_function_body): Likewise.
1147 (compute_fn_summary): Likewise.
1148 (estimate_edge_devirt_benefit): Likewise.
1149 (estimate_edge_size_and_time): Likewise.
1150 (estimate_calls_size_and_time): Likewise.
1151 (estimate_node_size_and_time): Likewise.
1152 (inline_update_callee_summaries): Likewise.
1153 (remap_edge_change_prob): Likewise.
1154 (remap_edge_summaries): Likewise.
1155 (ipa_merge_fn_summary_after_inlining): Likewise.
1156 (ipa_update_overall_fn_summary): Likewise.
1157 (read_ipa_call_summary): Likewise.
1158 (inline_read_section): Likewise.
1159 (write_ipa_call_summary): Likewise.
1160 (ipa_fn_summary_write): Likewise.
1161 (ipa_free_fn_summary): Likewise.
1162 * ipa-hsa.c (process_hsa_functions): Likewise.
1163 (ipa_hsa_write_summary): Likewise.
1164 (ipa_hsa_read_section): Likewise.
1165 * ipa-icf.c (sem_function::merge): Likewise.
1166 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1167 (do_estimate_edge_time): Likewise.
1168 (estimate_size_after_inlining): Likewise.
1169 (estimate_growth): Likewise.
1170 (growth_likely_positive): Likewise.
1171 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1172 (inline_call): Likewise.
1173 * ipa-inline.c (caller_growth_limits): Likewise.
1174 (can_inline_edge_p): Likewise.
1175 (can_inline_edge_by_limits_p): Likewise.
1176 (compute_uninlined_call_time): Likewise.
1177 (compute_inlined_call_time): Likewise.
1178 (want_inline_small_function_p): Likewise.
1179 (edge_badness): Likewise.
1180 (update_caller_keys): Likewise.
1181 (update_callee_keys): Likewise.
1182 (recursive_inlining): Likewise.
1183 (inline_small_functions): Likewise.
1184 (inline_to_all_callers_1): Likewise.
1185 (dump_overall_stats): Likewise.
1186 (early_inline_small_functions): Likewise.
1187 (early_inliner): Likewise.
1188 * ipa-inline.h (estimate_edge_growth): Likewise.
1189 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1190 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1191 * ipa-prop.h (IPA_NODE_REF): Likewise.
1192 (IPA_EDGE_REF): Likewise.
1193 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1194 (propagate_malloc): Likewise.
1195 * ipa-split.c (execute_split_functions): Likewise.
1196 * symbol-summary.h: Rename get to get_create.
1198 (get_create): Likewise.
1199 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1201 2018-06-08 Martin Liska <mliska@suse.cz>
1203 * symbol-summary.h (release): Move definition out of class
1205 (symtab_removal): Likewise.
1206 (symtab_duplication): Likewise.
1208 2018-06-08 Martin Liska <mliska@suse.cz>
1210 * symbol-summary.h (function_summary): Move constructor
1211 implementation out of class declaration.
1212 (release): Likewise.
1213 (symtab_insertion): Likewise.
1214 (symtab_removal): Likewise.
1215 (symtab_duplication): Likewise.
1218 2018-06-08 Martin Liska <mliska@suse.cz>
1220 * Makefile.in: Remove support for MPX (macros, related functions,
1221 fields in cgraph_node, ...).
1222 * builtin-types.def (BT_BND): Likewise.
1223 (BT_FN_BND_CONST_PTR): Likewise.
1224 (BT_FN_CONST_PTR_BND): Likewise.
1225 (BT_FN_VOID_PTR_BND): Likewise.
1226 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1227 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1228 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1229 (expand_builtin_mempcpy_with_bounds): Likewise.
1230 (expand_builtin_memset_with_bounds): Likewise.
1231 (expand_builtin_memset_args): Likewise.
1232 (std_expand_builtin_va_start): Likewise.
1233 (expand_builtin): Likewise.
1234 (expand_builtin_with_bounds): Likewise.
1235 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1236 (DEF_LIB_BUILTIN_CHKP): Likewise.
1237 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1238 (DEF_CHKP_BUILTIN): Likewise.
1239 (BUILT_IN_MEMCPY): Likewise.
1240 (BUILT_IN_MEMMOVE): Likewise.
1241 (BUILT_IN_MEMPCPY): Likewise.
1242 (BUILT_IN_MEMSET): Likewise.
1243 (BUILT_IN_STPCPY): Likewise.
1244 (BUILT_IN_STRCAT): Likewise.
1245 (BUILT_IN_STRCHR): Likewise.
1246 (BUILT_IN_STRCPY): Likewise.
1247 (BUILT_IN_STRLEN): Likewise.
1248 (BUILT_IN_MEMCPY_CHK): Likewise.
1249 (BUILT_IN_MEMMOVE_CHK): Likewise.
1250 (BUILT_IN_MEMPCPY_CHK): Likewise.
1251 (BUILT_IN_MEMSET_CHK): Likewise.
1252 (BUILT_IN_STPCPY_CHK): Likewise.
1253 (BUILT_IN_STRCAT_CHK): Likewise.
1254 (BUILT_IN_STRCPY_CHK): Likewise.
1255 * calls.c (store_bounds): Likewise.
1256 (emit_call_1): Likewise.
1257 (special_function_p): Likewise.
1258 (maybe_warn_nonstring_arg): Likewise.
1259 (initialize_argument_information): Likewise.
1260 (finalize_must_preallocate): Likewise.
1261 (compute_argument_addresses): Likewise.
1262 (expand_call): Likewise.
1263 * cfgexpand.c (expand_call_stmt): Likewise.
1264 (expand_return): Likewise.
1265 (expand_gimple_stmt_1): Likewise.
1266 (pass_expand::execute): Likewise.
1267 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1268 (cgraph_node::remove): Likewise.
1269 (cgraph_node::dump): Likewise.
1270 (cgraph_node::verify_node): Likewise.
1271 * cgraph.h (chkp_function_instrumented_p): Likewise.
1272 (symtab_node::get_alias_target): Likewise.
1273 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1274 (cgraph_local_p): Likewise.
1275 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1276 (cgraph_edge::rebuild_references): Likewise.
1277 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1278 (walk_polymorphic_call_targets): Likewise.
1279 (cgraph_node::expand_thunk): Likewise.
1280 (symbol_table::output_weakrefs): Likewise.
1281 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1282 (ix86_handle_option): Likewise.
1283 * config/i386/constraints.md: Likewise.
1284 * config/i386/i386-builtin-types.def (BND): Likewise.
1288 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1289 (BDESC_FIRST): Likewise.
1291 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1292 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1293 * config/i386/i386.c (enum reg_class): Likewise.
1294 (ix86_target_string): Likewise.
1295 (ix86_option_override_internal): Likewise.
1296 (ix86_conditional_register_usage): Likewise.
1297 (ix86_valid_target_attribute_inner_p): Likewise.
1298 (ix86_set_indirect_branch_type): Likewise.
1299 (ix86_set_current_function): Likewise.
1300 (ix86_function_arg_regno_p): Likewise.
1301 (init_cumulative_args): Likewise.
1302 (ix86_function_arg_advance): Likewise.
1303 (ix86_function_arg): Likewise.
1304 (ix86_pass_by_reference): Likewise.
1305 (ix86_function_value_regno_p): Likewise.
1306 (ix86_function_value_1): Likewise.
1307 (ix86_function_value_bounds): Likewise.
1308 (ix86_return_in_memory): Likewise.
1309 (ix86_setup_incoming_vararg_bounds): Likewise.
1310 (ix86_va_start): Likewise.
1311 (indirect_thunk_need_prefix): Likewise.
1312 (print_reg): Likewise.
1313 (ix86_print_operand): Likewise.
1314 (ix86_expand_call): Likewise.
1315 (ix86_output_function_return): Likewise.
1316 (reg_encoded_number): Likewise.
1317 (BDESC_VERIFYS): Likewise.
1318 (ix86_init_mpx_builtins): Likewise.
1319 (ix86_init_builtins): Likewise.
1320 (ix86_emit_cmove): Likewise.
1321 (ix86_emit_move_max): Likewise.
1322 (ix86_expand_builtin): Likewise.
1323 (ix86_builtin_mpx_function): Likewise.
1324 (ix86_get_arg_address_for_bt): Likewise.
1325 (ix86_load_bounds): Likewise.
1326 (ix86_store_bounds): Likewise.
1327 (ix86_load_returned_bounds): Likewise.
1328 (ix86_store_returned_bounds): Likewise.
1329 (ix86_class_likely_spilled_p): Likewise.
1330 (ix86_hard_regno_mode_ok): Likewise.
1331 (x86_order_regs_for_local_alloc): Likewise.
1332 (ix86_mitigate_rop): Likewise.
1333 (ix86_bnd_prefixed_insn_p): Likewise.
1334 (ix86_mpx_bound_mode): Likewise.
1335 (ix86_make_bounds_constant): Likewise.
1336 (ix86_initialize_bounds): Likewise.
1337 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1338 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1339 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1340 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1341 (TARGET_CHKP_BOUND_MODE): Likewise.
1342 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1343 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1344 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1345 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1346 * config/i386/i386.h (TARGET_MPX): Likewise.
1347 (TARGET_MPX_P): Likewise.
1348 (VALID_BND_REG_MODE): Likewise.
1349 (FIRST_BND_REG): Likewise.
1350 (LAST_BND_REG): Likewise.
1351 (enum reg_class): Likewise.
1352 (BND_REG_P): Likewise.
1353 (BND_REGNO_P): Likewise.
1354 (BNDmode): Likewise.
1355 (ADJUST_INSN_LENGTH): Likewise.
1356 * config/i386/i386.md: Likewise.
1357 * config/i386/i386.opt: Likewise.
1358 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1359 (defined): Likewise.
1360 (LINK_MPX): Likewise.
1361 (MPX_SPEC): Likewise.
1362 (LIBMPX_SPEC): Likewise.
1363 (LIBMPXWRAPPERS_SPEC): Likewise.
1364 (CHKP_SPEC): Likewise.
1365 * config/i386/predicates.md: Likewise.
1366 * dbxout.c (dbxout_type): Likewise.
1367 * doc/extend.texi: Likewise.
1368 * doc/invoke.texi: Likewise.
1369 * doc/md.texi: Likewise.
1370 * doc/tm.texi: Likewise.
1371 * doc/tm.texi.in: Likewise.
1372 * dwarf2out.c (is_base_type): Likewise.
1373 (gen_formal_types_die): Likewise.
1374 (gen_subprogram_die): Likewise.
1375 (gen_type_die_with_usage): Likewise.
1376 (gen_decl_die): Likewise.
1377 (dwarf2out_late_global_decl): Likewise.
1378 * expr.c (expand_assignment): Likewise.
1379 (emit_storent_insn): Likewise.
1380 (store_expr_with_bounds): Likewise.
1381 (store_expr): Likewise.
1382 (expand_expr_real_1): Likewise.
1383 * expr.h (store_expr_with_bounds): Likewise.
1384 * function.c (use_register_for_decl): Likewise.
1385 (struct bounds_parm_data): Likewise.
1386 (assign_parms_augmented_arg_list): Likewise.
1387 (assign_parm_find_entry_rtl): Likewise.
1388 (assign_parm_is_stack_parm): Likewise.
1389 (assign_parm_load_bounds): Likewise.
1390 (assign_bounds): Likewise.
1391 (assign_parms): Likewise.
1392 (expand_function_start): Likewise.
1393 * gcc.c (CHKP_SPEC): Likewise.
1394 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1395 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1396 (wrestrict_dom_walker::check_call): Likewise.
1397 * gimple.c (gimple_build_call_from_tree): Likewise.
1398 * gimple.h (enum gf_mask): Likewise.
1399 (gimple_call_with_bounds_p): Likewise.
1400 (gimple_call_set_with_bounds): Likewise.
1401 * gimplify.c (gimplify_init_constructor): Likewise.
1402 * ipa-cp.c (initialize_node_lattices): Likewise.
1403 (propagate_constants_across_call): Likewise.
1404 (find_more_scalar_values_for_callers_subset): Likewise.
1405 * ipa-hsa.c (process_hsa_functions): Likewise.
1406 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1407 * ipa-icf.c (sem_function::merge): Likewise.
1408 * ipa-inline.c (early_inliner): Likewise.
1409 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1410 (warn_function_cold): Likewise.
1411 (propagate_pure_const): Likewise.
1412 * ipa-ref.h (enum GTY): Likewise.
1413 * ipa-split.c (find_retbnd): Likewise.
1414 (consider_split): Likewise.
1415 (split_function): Likewise.
1416 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1417 * ipa.c (walk_polymorphic_call_targets): Likewise.
1418 (symbol_table::remove_unreachable_nodes): Likewise.
1419 (process_references): Likewise.
1420 (cgraph_build_static_cdtor_1): Likewise.
1421 * lto-cgraph.c (lto_output_node): Likewise.
1422 (output_refs): Likewise.
1423 (compute_ltrans_boundary): Likewise.
1424 (input_overwrite_node): Likewise.
1425 (input_node): Likewise.
1426 (input_cgraph_1): Likewise.
1427 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1428 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1429 (class pass_chkp_instrumentation_passes): Likewise.
1430 (make_pass_chkp_instrumentation_passes): Likewise.
1431 * passes.def: Likewise.
1432 * rtl.h (struct GTY): Likewise.
1433 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1434 * stor-layout.c (layout_type): Likewise.
1435 * symtab.c: Likewise.
1436 * target.def: Likewise.
1437 * targhooks.c (default_chkp_bound_type): Likewise.
1438 (default_chkp_bound_mode): Likewise.
1439 (default_builtin_chkp_function): Likewise.
1440 (default_chkp_function_value_bounds): Likewise.
1441 (default_chkp_make_bounds_constant): Likewise.
1442 (default_chkp_initialize_bounds): Likewise.
1443 * targhooks.h (default_chkp_bound_type): Likewise.
1444 (default_chkp_bound_mode): Likewise.
1445 (default_builtin_chkp_function): Likewise.
1446 (default_chkp_function_value_bounds): Likewise.
1447 (default_chkp_make_bounds_constant): Likewise.
1448 (default_chkp_initialize_bounds): Likewise.
1449 * toplev.c (compile_file): Likewise.
1450 (process_options): Likewise.
1451 * tree-core.h (DEF_BUILTIN): Likewise.
1452 (DEF_BUILTIN_CHKP): Likewise.
1453 * tree-inline.c (declare_return_variable): Likewise.
1454 (remap_gimple_stmt): Likewise.
1455 (copy_bb): Likewise.
1456 (initialize_inlined_parameters): Likewise.
1457 (expand_call_inline): Likewise.
1458 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1459 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1460 (make_pass_ipa_chkp_produce_thunks): Likewise.
1461 (make_pass_chkp): Likewise.
1462 (make_pass_chkp_opt): Likewise.
1463 (make_pass_chkp_instrumentation_passes): Likewise.
1464 * tree-pretty-print.c (dump_generic_node): Likewise.
1465 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1466 * tree-ssa-dce.c (propagate_necessity): Likewise.
1467 (eliminate_unnecessary_stmts): Likewise.
1468 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1469 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1470 * tree-ssa-sccvn.h: Likewise.
1471 * tree-ssa-strlen.c (get_string_length): Likewise.
1472 (valid_builtin_call): Likewise.
1473 (adjust_last_stmt): Likewise.
1474 (handle_builtin_strchr): Likewise.
1475 (handle_builtin_strcpy): Likewise.
1476 (handle_builtin_stxncpy): Likewise.
1477 (handle_builtin_memcpy): Likewise.
1478 (handle_builtin_strcat): Likewise.
1479 (strlen_check_and_optimize_stmt): Likewise.
1480 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1481 * tree-streamer-in.c: Likewise.
1482 * tree-streamer.c (record_common_node): Likewise.
1483 * tree.c (tree_code_size): Likewise.
1484 (wide_int_to_tree_1): Likewise.
1485 (type_contains_placeholder_1): Likewise.
1486 (build_common_tree_nodes): Likewise.
1487 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1488 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1489 (POINTER_BOUNDS_P): Likewise.
1490 (BOUNDED_TYPE_P): Likewise.
1491 (BOUNDED_P): Likewise.
1492 (CALL_WITH_BOUNDS_P): Likewise.
1493 (pointer_bounds_type_node): Likewise.
1494 * value-prof.c (gimple_ic): Likewise.
1495 * var-tracking.c (vt_add_function_parameters): Likewise.
1496 * varasm.c (make_decl_rtl): Likewise.
1497 (assemble_start_function): Likewise.
1498 (output_constant): Likewise.
1499 (maybe_assemble_visibility): Likewise.
1500 * varpool.c (ctor_for_folding): Likewise.
1501 * chkp-builtins.def: Remove.
1502 * ipa-chkp.c: Remove.
1503 * ipa-chkp.h: Remove.
1504 * rtl-chkp.c: Remove.
1505 * rtl-chkp.h: Remove.
1506 * tree-chkp-opt.c: Remove.
1507 * tree-chkp.c: Remove.
1508 * tree-chkp.h: Remove.
1510 2018-06-07 Carl Love <cel@us.ibm.com>
1512 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1513 vextract_fp_from_shortl): Add BE support.
1515 2018-06-07 Paul Koning <ni1d@arrl.net>
1517 * compare-elim.c (try_merge_compare): Don't merge compare if
1518 address contains a side effect.
1519 (try_eliminate_compare): Likewise.
1521 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1523 * config.gcc: Support "tremont".
1524 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1525 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1527 * config/i386/i386.c (m_TREMONT): Define.
1528 (processor_target_table): Add "tremont".
1529 (PTA_TREMONT): Define.
1530 (ix86_lea_outperforms): Add TARGET_TREMONT.
1531 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1532 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1533 and M_INTEL_GOLDMONT_PLUS.
1534 (fold_builtin_cpu): Add "tremont".
1535 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1536 (ix86_option_override_internal): Add "tremont".
1537 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1538 (processor_type): Add PROCESSOR_TREMONT.
1539 * config/i386/x86-tune.def: Add m_TREMONT.
1540 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1542 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1544 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1545 symbol defined for msp430i* devices to be lower case.
1547 gcc/testsuite/gcc.target/msp430/
1548 * msp430i-device-symbol.c: New test.
1549 * msp430f-device-symbol.c: New test.
1550 * msp430.h: New test header file.
1552 2018-06-07 Richard Biener <rguenther@suse.de>
1554 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1555 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1556 Properly wrap signed arithmetic if overflow wraps.
1558 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1560 PR tree-optimization/69615
1561 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1562 of a cast from a same precision integral SSA_NAME in a bb dominated
1563 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1564 cast to utype if rhs2 has already a compatible type.
1566 2018-06-07 Richard Biener <rguenther@suse.de>
1568 PR tree-optimization/85935
1569 * graphite-scop-detection.c (find_params_in_bb): Analyze
1570 condition operands with respect to the correct loop. Assert
1571 the analysis doesn't fail.
1573 2018-06-04 Carl Love <cel@us.ibm.com>
1575 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1576 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1577 as it is slightly cheaper.
1578 (first_match_or_eos_index_<mode>):
1579 Calculate index using natural element order.
1580 (first_match_index_<mode>):
1581 Calculate index using natural element order.
1582 (first_match_or_eos_index_<mode>):
1583 Calculate index using natural order.
1584 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1585 for BE and LE modes.
1586 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1587 P9V_BUILTIN_VCLZLSBB_V16QI.
1588 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1591 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1593 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1594 indentation and line wrap for many prototypes. Add missing
1595 @smallexample directives around block of prototypes for vec_xl and
1598 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1600 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1601 track if we pass or return IEEE 128-bit floating point.
1602 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1603 C++ mangling that is compatible with GCC 8.1.
1604 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1605 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1606 floating point types.
1607 (rs6000_function_arg_advance_1): Likewise.
1608 (rs6000_mangle_type): Optionally generate mangled names that match
1609 what GCC 8.1 generated for IEEE 128-bit floating point types.
1610 (rs6000_globalize_decl_name): If we have an external function that
1611 passes or returns IEEE 128-bit types, generate a weak reference
1612 from the mangled name used in GCC 8.1 to the current mangled
1614 (rs6000_init_builtins): Make __ibm128 use the long double type if
1615 long double is IBM extended double. Make __float128 use the long
1616 double type if long double is IEEE 128-bit.
1619 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1620 macro for __ibm128 built-in functions.
1621 (PACK_IF): Add __ibm128 pack/unpack functions.
1622 (UNPACK_IF): Likewise.
1623 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1624 enable long double built-in functions if long double is IEEE
1625 128-bit floating point.
1626 (rs6000_invalid_builtin): Update long double built-in function
1628 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1629 functions, adjust the built-in function to use the long double
1630 built-in function if __ibm128 and long double are the same type.
1631 * doc/extend.texi (PowerPC builtins): Update documention for
1632 __builtin_{,un}pack_longdouble. Add documentation for
1633 __builtin_{,un}pack_ibm128.
1635 2018-06-06 Jim Wilson <jimw@sifive.com>
1637 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1638 (struct machine_function): New field interrupt_mode.
1639 (riscv_handle_type_attribute): New function. Add forward declaration.
1640 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1641 (riscv_expand_epilogue): Check interrupt_mode field.
1642 (riscv_set_current_function): Check interrupt attribute args and
1643 set interrupt_mode field.
1644 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1645 (riscv_sret, riscv_uret): New.
1646 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1647 new arguments to interrupt attribute.
1649 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1652 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1653 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1655 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1657 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1658 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1659 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1660 several redundant entries.
1662 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1664 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1665 type from "rtx" to "rtx_insn *".
1666 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1667 for local "call_insn", removing cast.
1668 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1670 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1672 PR tree-optimization/86066
1673 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1674 for BIT_INSERT_EXPR stores.
1676 2018-06-06 Richard Biener <rguenther@suse.de>
1678 PR tree-optimization/86062
1679 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1680 component refs ontop
1681 of to be offsetted base.
1683 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1685 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1686 to be static and remove check on interrupt attribute name.
1688 gcc/testsuite/gcc.target/msp430/
1689 * function-attributes-4.c: New test.
1690 * static-interrupts.c: New test.
1692 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1694 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1695 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1697 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1700 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1702 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1703 Remove second argument, change how error is called.
1704 (aarch64_layout_arg): Remove second argument from
1705 aarch64_err_no_fpadvsimd call.
1706 (aarch64_init_cumulative_args): Ditto.
1707 (aarch64_gimplify_va_arg_expr): Ditto.
1708 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1710 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1712 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1713 (*simple_return_indirect_internal<mode>): Rename from
1714 simple_return_indirect_internal. Use W mode iterator.
1715 (rstorssp): New expander.
1716 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1717 (clrssbsy): New expander.
1718 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1720 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1722 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1724 (cmse_check_pointed_object): Likewise.
1726 2018-06-05 Martin Liska <mliska@suse.cz>
1728 PR gcov-profile/47618
1729 * doc/invoke.texi: Document how -fprofile-dir format
1732 2018-06-05 Richard Biener <rguenther@suse.de>
1734 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1735 removal pretend DOM info isn't available so we do not update
1736 it and only remove edges, not dominated blocks. Actually free
1737 DOM info in case we removed something. Remove unreachable blocks.
1738 (mfb_keep_latches): Work with either DOM info or marked backedges.
1739 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1740 first. Mark backedges if DOM info isn't available.
1741 (Re-)compute DOM info after cleanup_control_flow_pre.
1743 2018-06-05 Richard Biener <rguenther@suse.de>
1745 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1746 (locus_discrim_hasher::hash): Adjust.
1747 (locus_discrim_hasher::equal): Likewise.
1748 (next_discriminator_for_locus): Work on line directly.
1749 (same_line_p): Pass in expanded locus1 as well.
1750 (assign_discriminators): Avoid redundant location expansions.
1752 2018-06-05 Richard Biener <rguenther@suse.de>
1754 PR tree-optimization/86046
1755 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1756 if required after clearing TREE_ADDRESSABLE.
1758 2018-06-05 Richard Biener <rguenther@suse.de>
1760 PR tree-optimization/86047
1761 * tree-ssa-loop.c (for_each_index): Glob handling of all
1762 decls and constants and really handle all of them.
1764 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1767 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1768 qualifier_void_pointer and qualifier_const_void_pointer.
1769 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1770 (arm_init_builtins): Handle the above.
1771 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1772 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1773 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1776 2018-06-05 Martin Liska <mliska@suse.cz>
1778 * auto-profile.c (read_autofdo_file): Do not use
1779 gcov_ctr_summary struct.
1780 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1781 * coverage.c (struct counts_entry): Likewise.
1782 (read_counts_file): Read just single summary entry.
1783 (get_coverage_counts): Use gcov_summary struct.
1784 * coverage.h (get_coverage_counts): Likewise.
1785 * gcov-dump.c (dump_working_sets): Likewise.
1786 (tag_summary): Dump just single summary.
1787 * gcov-io.c (gcov_write_summary): Write just histogram
1789 (gcov_read_summary): Read just single summary.
1790 (compute_working_sets): Use gcov_summary struct.
1791 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1792 of GCOV_COUNTERS_SUMMABLE.
1793 (GCOV_COUNTERS_SUMMABLE): Remove.
1794 (GCOV_FIRST_VALUE_COUNTER): Replace with
1795 GCOV_COUNTER_V_INTERVAL.
1796 (struct gcov_ctr_summary): Remove.
1797 (struct gcov_summary): Directly use fields of former
1799 (compute_working_sets): Use gcov_summary struct.
1800 * gcov.c (read_count_file): Do not use ctrs fields.
1801 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1803 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1805 * profile.c: Likewise.
1806 * profile.h: Likewise.
1808 2018-06-05 Martin Liska <mliska@suse.cz>
1810 PR gcov-profile/84846
1811 * gcov.c (output_lines): Print working directory only
1812 in intermediate format.
1814 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1816 * config/s390/s390-builtin-types.def: Add void function type.
1817 * config/s390/s390-builtins.def: Use the function type for the
1820 2018-06-04 Jim Wilson <jimw@sifive.com>
1822 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1824 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1825 and maybe_eh_return. Change regno to unsigned int. Use new args to
1826 handle EH_RETURN_DATA_REGNO registers properly.
1827 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1828 (riscv_expand_epilogue): Update comment. Change argument name and
1829 type. Update code to use new name and type. Pass new args to
1830 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1832 * config/riscv/riscv.md (NORMAL_RETURN): New.
1833 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1834 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1835 (eh_return): Call gen_eh_return_internal and emit barrier.
1836 (eh_return_internal): Call riscv_expand_epilogue.
1838 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1840 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1841 bit_insertion field and declare can_be_merged_into method.
1842 (merged_store_group::can_be_merged_into): New method.
1843 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1844 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1845 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1847 2018-06-04 Richard Biener <rguenther@suse.de>
1849 PR tree-optimization/85955
1850 * builtins.c (fold_builtin_sincos): Convert pointers to
1851 destination to appropriate type before dereferencing.
1853 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1855 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1857 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1859 * expr.c (expand_expr_real_1): Force the operand into memory if
1860 its TYPE_MODE is BLKmode and if there is no integer mode for
1861 the number of bits being extracted.
1863 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1867 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1868 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1870 2018-06-04 Richard Biener <rguenther@suse.de>
1872 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1873 (cleanup_tree_cfg_noloop): ... single caller. Do
1874 start_recording_case_labels later.
1876 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1878 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1879 to _IMMINTRIN_H_INCLUDED.
1880 * config/i386/pconfigintrin.h: Ditto.
1881 * config/i386/waitpkgintrin.h: Ditto.
1882 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1883 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1884 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1885 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1886 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1887 waitpkgintrin.h and cldemoteintrin.h.
1889 2018-06-04 Richard Biener <rguenther@suse.de>
1891 PR tree-optimization/86038
1892 * tracer.c (find_best_successor): Check probability for
1893 being initialized, bail out if not.
1895 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1898 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1899 of bits to ignore when comparing architectures.
1901 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1903 PR tree-optimization/69615
1904 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1905 maximum or minimum of the type, try to merge it also as if
1906 range1 is + [-, x - 1] or + [x + 1, -].
1909 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1911 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1913 PR tree-optimization/86034
1914 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1915 the unsigned bitfield type in a bit insertion sequence if it does not
1916 have a larger precision than the bitfield size.
1917 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1919 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1921 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1923 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1925 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1926 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1927 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1928 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1930 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1932 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1933 Disable -fdelete-null-pointer-checks for ELF toolchain.
1935 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1936 Kito Cheng <kito.cheng@gmail.com>
1938 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1939 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1940 (nds32*-*-*): Add glibc and uclibc conditions.
1941 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1942 (TARGET_EXCEPT_UNWIND_INFO): Define.
1943 * config/nds32/elf.h: New file.
1944 * config/nds32/linux.h: New file.
1945 * config/nds32/nds32-elf.opt: New file.
1946 * config/nds32/nds32-linux.opt: New file.
1947 * config/nds32/nds32-fp-as-gp.c
1948 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1949 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1951 (nds32_asm_file_end): Ditto.
1952 (nds32_print_operand): Ditto.
1953 (nds32_insert_attributes): Ditto.
1954 (nds32_init_libfuncs): New function.
1955 (TARGET_HAVE_TLS): Define.
1956 (TARGET_INIT_LIBFUNCS): Define.
1957 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1959 (TARGET_ELF): Apply different mcmodel setting.
1960 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1961 been migrated into elf.h and linux.h files.
1962 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1963 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1964 (mcmodel): The content has been migrated into nds32-elf.opt and
1965 nds32-linux.opt files.
1966 * config/nds32/t-elf: New file.
1967 * config/nds32/t-linux: New file.
1969 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1970 Shiva Chen <shiva0217@gmail.com>
1972 * config/nds32/constants.md (unspec_volatile_element): Add
1973 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1974 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1976 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1977 (make_pass_nds32_fp_as_gp): Declare.
1978 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1980 (nds32_asm_function_end_prologue): Remove unused asm output.
1981 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1982 (nds32_asm_file_start): Output necessary fp_as_gp information.
1983 (nds32_option_override): Adjust register usage.
1984 (nds32_expand_prologue): Consider fp_as_gp situation.
1985 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1986 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1989 (simple_return): Ditto.
1990 (omit_fp_begin): Output special directive for fp_as_gp.
1991 (omit_fp_end): Output special directive for fp_as_gp.
1992 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1993 mforbid-fp-as-gp): New options.
1995 2018-06-01 Mark Wielaard <mark@klomp.org>
1997 * dwarf2out.c (dwarf2out_finish): Remove generation of
1998 DW_AT_loclists_base.
2000 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2002 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2003 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2004 (struct merged_store_group): Add bit_insertion field.
2005 (dump_char_array): Use standard hexadecimal format.
2006 (merged_store_group::merged_store_group): Set bit_insertion to false.
2007 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2008 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2009 also print the mask in the dump file.
2010 (pass_store_merging::gate): Minor tweak.
2011 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2012 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2013 stores with INTEGER_CST stores.
2014 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2015 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2016 and use it throughout. Generate bit insertion sequences if need be.
2017 (pass_store_merging::process_store): Remove redundant condition.
2018 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2020 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2022 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2023 the 128-bit floating point types. Fix function comment.
2025 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2027 * config/aarch64/aarch64-simd.md
2028 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2030 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2033 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2035 PR tree-optimization/85989
2036 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2038 (backprop::intersect_uses): Check it when deciding whether this
2039 is a backedge reference.
2040 (backprop::process_block): Add each phi to m_visited_phis
2041 after visiting it, then clear it at the end.
2043 2018-06-01 Richard Biener <rguenther@suse.de>
2045 * tree-vectorizer.h (vect_dr_stmt): New function.
2046 (vect_get_load_cost): Adjust.
2047 (vect_get_store_cost): Likewise.
2048 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2049 Use vect_dr_stmt instead of DR_SMTT.
2050 (vect_record_base_alignments): Likewise.
2051 (vect_calculate_target_alignment): Likewise.
2052 (vect_compute_data_ref_alignment): Likewise and make static.
2053 (vect_update_misalignment_for_peel): Likewise.
2054 (vect_verify_datarefs_alignment): Likewise.
2055 (vector_alignment_reachable_p): Likewise.
2056 (vect_get_data_access_cost): Likewise. Pass down
2057 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2058 (vect_get_peeling_costs_all_drs): Likewise.
2059 (vect_peeling_hash_get_lowest_cost): Likewise.
2060 (vect_enhance_data_refs_alignment): Likewise.
2061 (vect_find_same_alignment_drs): Likewise.
2062 (vect_analyze_data_refs_alignment): Likewise.
2063 (vect_analyze_group_access_1): Likewise.
2064 (vect_analyze_group_access): Likewise.
2065 (vect_analyze_data_ref_access): Likewise.
2066 (vect_analyze_data_ref_accesses): Likewise.
2067 (vect_vfa_segment_size): Likewise.
2068 (vect_small_gap_p): Likewise.
2069 (vectorizable_with_step_bound_p): Likewise.
2070 (vect_prune_runtime_alias_test_list): Likewise.
2071 (vect_analyze_data_refs): Likewise.
2072 (vect_supportable_dr_alignment): Likewise.
2073 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2074 (vect_gen_prolog_loop_niters): Likewise.
2075 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2076 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2078 (vect_recog_mask_conversion_pattern): Likewise.
2079 (vect_try_gather_scatter_pattern): Likewise.
2080 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2081 to vect_get_store_cost.
2082 (vect_get_store_cost): Get stmt_info instead of DR.
2083 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2084 (vect_get_load_cost): Get stmt_info instead of DR.
2086 2018-06-01 Richard Biener <rguenther@suse.de>
2089 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2090 vars wrapped in MEM_REFs.
2092 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2094 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2095 Fix subreg tests so that we only return a choice between
2096 GENERAL_REGS and FP_REGS if the original classes included both.
2098 2018-06-01 Richard Biener <rguenther@suse.de>
2101 * tree-ssa-structalias.c (get_function_part_constraint):
2102 Handle NULL fi->decl.
2103 (find_func_aliases_for_call): Properly handle indirect
2104 fi from direct call.
2105 (find_func_clobbers): Likewise.
2106 (ipa_pta_execute): Likewise.
2107 (create_variable_info_for): For functions that are ifunc_resolver
2108 resolve to a varinfo that contains the result of the resolver call.
2109 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2112 2018-05-31 Michael Collison <michael.collison@arm.com>
2114 * config/aarch64/aarch64.md:
2115 (*fix_to_zero_extenddfdi2): New pattern.
2116 * gcc.target/aarch64/fix_extend1.c: New testcase.
2118 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2122 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2123 and BUILT_IN_STRNCMP_EQ.
2124 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2125 BUILT_IN_STRNCMP_EQ.
2126 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2127 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2128 (gimple_fold_builtin): Likewise.
2129 * tree-ssa-strlen.c (compute_string_length): New function.
2130 (determine_min_obsize): New function.
2131 (handle_builtin_string_cmp): New function to handle calls to
2132 string compare functions.
2133 (strlen_optimize_stmt): Add handling to builtin string compare
2135 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2136 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2137 * tree.c (build_common_builtin_nodes): Add new defines of
2138 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2140 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2143 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2144 functions with naked attribute.
2146 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2148 * config/i386/sse.md (avx_vec_concat<mode>):
2149 Substitute concat_tg_mode mode attribute with xtg_mode.
2150 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2151 (concat_tg_mode): Remove mode attribute.
2153 2018-05-31 Martin Sebor <msebor@redhat.com>
2156 * calls.c (alloc_max_size): Correct a logic error/typo.
2157 Treat excessive arguments as infinite. Warn for invalid arguments.
2158 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2160 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2163 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2164 and movx for Haswell.
2166 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2167 Cesar Philippidis <cesar@codesourcery.com>
2170 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2171 when emitting error on private/firstprivate reductions.
2172 * omp-low.c (lower_omp_target): Avoid reference-type processing
2173 on pointers for firstprivate clause.
2175 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2177 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2180 * config/aarch64/aarch64-simd.md
2181 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2182 (aarch64_ld1_x3_<mode>): Likewise
2183 (aarch64_st1x2<VALLDIF:mode>): Likewise
2184 (aarch64_st1_x2_<mode>): Likewise
2185 (aarch64_st1x3<VALLDIF:mode>): Likewise
2186 (aarch64_st1_x3_<mode>): Likewise
2187 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2188 (vld1_s8_x3): Likewise.
2189 (vld1_u16_x3): Likewise.
2190 (vld1_s16_x3): Likewise.
2191 (vld1_u32_x3): Likewise.
2192 (vld1_s32_x3): Likewise.
2193 (vld1_u64_x3): Likewise.
2194 (vld1_s64_x3): Likewise.
2195 (vld1_f16_x3): Likewise.
2196 (vld1_f32_x3): Likewise.
2197 (vld1_f64_x3): Likewise.
2198 (vld1_p8_x3): Likewise.
2199 (vld1_p16_x3): Likewise.
2200 (vld1_p64_x3): Likewise.
2201 (vld1q_u8_x3): Likewise.
2202 (vld1q_s8_x3): Likewise.
2203 (vld1q_u16_x3): Likewise.
2204 (vld1q_s16_x3): Likewise.
2205 (vld1q_u32_x3): Likewise.
2206 (vld1q_s32_x3): Likewise.
2207 (vld1q_u64_x3): Likewise.
2208 (vld1q_s64_x3): Likewise.
2209 (vld1q_f16_x3): Likewise.
2210 (vld1q_f32_x3): Likewise.
2211 (vld1q_f64_x3): Likewise.
2212 (vld1q_p8_x3): Likewise.
2213 (vld1q_p16_x3): Likewise.
2214 (vld1q_p64_x3): Likewise.
2215 (vst1_s64_x2): Likewise.
2216 (vst1_u64_x2): Likewise.
2217 (vst1_f64_x2): Likewise.
2218 (vst1_s8_x2): Likewise.
2219 (vst1_p8_x2): Likewise.
2220 (vst1_s16_x2): Likewise.
2221 (vst1_p16_x2): Likewise.
2222 (vst1_s32_x2): Likewise.
2223 (vst1_u8_x2): Likewise.
2224 (vst1_u16_x2): Likewise.
2225 (vst1_u32_x2): Likewise.
2226 (vst1_f16_x2): Likewise.
2227 (vst1_f32_x2): Likewise.
2228 (vst1_p64_x2): Likewise.
2229 (vst1q_s8_x2): Likewise.
2230 (vst1q_p8_x2): Likewise.
2231 (vst1q_s16_x2): Likewise.
2232 (vst1q_p16_x2): Likewise.
2233 (vst1q_s32_x2): Likewise.
2234 (vst1q_s64_x2): Likewise.
2235 (vst1q_u8_x2): Likewise.
2236 (vst1q_u16_x2): Likewise.
2237 (vst1q_u32_x2): Likewise.
2238 (vst1q_u64_x2): Likewise.
2239 (vst1q_f16_x2): Likewise.
2240 (vst1q_f32_x2): Likewise.
2241 (vst1q_f64_x2): Likewise.
2242 (vst1q_p64_x2): Likewise.
2243 (vst1_s64_x3): Likewise.
2244 (vst1_u64_x3): Likewise.
2245 (vst1_f64_x3): Likewise.
2246 (vst1_s8_x3): Likewise.
2247 (vst1_p8_x3): Likewise.
2248 (vst1_s16_x3): Likewise.
2249 (vst1_p16_x3): Likewise.
2250 (vst1_s32_x3): Likewise.
2251 (vst1_u8_x3): Likewise.
2252 (vst1_u16_x3): Likewise.
2253 (vst1_u32_x3): Likewise.
2254 (vst1_f16_x3): Likewise.
2255 (vst1_f32_x3): Likewise.
2256 (vst1_p64_x3): Likewise.
2257 (vst1q_s8_x3): Likewise.
2258 (vst1q_p8_x3): Likewise.
2259 (vst1q_s16_x3): Likewise.
2260 (vst1q_p16_x3): Likewise.
2261 (vst1q_s32_x3): Likewise.
2262 (vst1q_s64_x3): Likewise.
2263 (vst1q_u8_x3): Likewise.
2264 (vst1q_u16_x3): Likewise.
2265 (vst1q_u32_x3): Likewise.
2266 (vst1q_u64_x3): Likewise.
2267 (vst1q_f16_x3): Likewise.
2268 (vst1q_f32_x3): Likewise.
2269 (vst1q_f64_x3): Likewise.
2270 (vst1q_p64_x3): Likewise.
2272 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2274 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2275 user_label_prefix to name.
2277 * tree-core.h: Update comment about the format of NAME string
2278 passed to handler in attribute_spec.
2280 * config/msp430/msp430.md: Remove erroneous subreg expression from
2281 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2282 zero_extend{q,h}isi2.
2284 2018-05-30 Borislav Petkov <bp@suse.de>
2286 * doc/extend.texi: Document some architecture specific
2287 constraints and sort entries.
2289 2018-05-30 Martin Sebor <msebor@redhat.com>
2292 * builtins.c (expand_builtin_stpcpy_1): New function.
2293 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2294 only if the former succeeds.
2296 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2298 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2301 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2303 * doc/invoke.texi (-flinker-output): Document
2305 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2307 * passes.c (ipa_write_summaries): Only modify statements if body
2309 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2310 incrementally linking.
2311 (ipa_passes): Likewise.
2312 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2313 pass down resolution info.
2314 * common.opt (flag_incremental_link): Update info.
2315 * gcc.c (plugin specs): Turn flinker-output=* to
2316 -plugin-opt=-linker-output-known
2317 * toplev.c (compile_file): Also cut compilation when doing incremental
2319 * flag-types. (enum lto_partition_model): Add
2320 LTO_LINKER_OUTPUT_NOLTOREL.
2321 (invoke.texi): Add -flinker-output docs.
2322 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2323 link same way as WPA; do not stream in dead initializers.
2325 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2326 darf2out_register_external_decl): Support incremental link.
2328 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2330 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2332 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2334 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2335 it down to simple_object_copy_lto_debug_sections.
2336 (run_gcc): Determine incremental LTO link time and configure
2337 lto1 into non-wpa mode, disable renaming of debug sections.
2339 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2341 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2342 descriptions of various incorrectly documented functions.
2344 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2347 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2348 address check not strict.
2350 2018-05-30 Richard Biener <rguenther@suse.de>
2352 PR tree-optimization/85964
2353 * tracer.c (better_p): Drop initialized count check, we only
2354 call the function with initialized counts now.
2355 (find_best_successor): Do find a best edge if one
2356 has uninitialized count.
2357 (find_best_predecessor): Likewise. Do BB frequency check only
2358 if count is initialized.
2360 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2362 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2363 (aarch64_ldrstr_offset_compare): New.
2364 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2365 load/store orderings.
2366 (aarch64_gen_adjusted_ldpstp): Likewise.
2368 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2370 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2371 Check for subset of GENERAL_REGS and FP_REGS.
2372 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2375 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2377 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2378 and wi::to_poly_offset. Add the current offset and then check
2379 whether the sum fits, rather than using an unchecked addition of
2380 a checked term. Check for a shwi rather than a uhwi.
2381 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2382 (store_constructor): Use poly_int_tree_p.
2383 (expand_expr_real_1): Likewise.
2384 * function.c (assign_temp): Likewise.
2385 * fold-const.c (const_binop): Use poly_int_tree_p and
2387 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2389 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2390 to_poly_offset for MEM offsets.
2391 * ipa-icf.c (sem_variable::equals): Likewise.
2392 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2393 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2394 wi::to_poly_offset for BIT_FIELD_REF offsets.
2395 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2397 * var-tracking.c (emit_note_insn_var_location): Use
2398 tree_to_poly_uint64.
2400 2018-05-29 Jim Wilson <jimw@sifive.com>
2402 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2404 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2407 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2408 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2410 (sse4_1_round<mode>2): Use nonimmediate_operand
2411 for operand 1 predicate.
2413 2018-05-29 Martin Sebor <msebor@redhat.com>
2414 Richard Biener <rguenther@suse.de>
2417 * calls.c (get_size_range): Call determine_value_range instead
2418 of get_value_range..
2419 * tree-vrp.h (determine_value_range): Declared new function.
2420 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2422 2018-05-29 Richard Biener <rguenther@suse.de>
2424 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2425 sure to use non-pattern stmts for get_earlier_stmt arguments.
2426 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2427 called on pattern stmts.
2428 (get_later_stmt): Likewise.
2430 2018-05-29 Martin Liska <mliska@suse.cz>
2432 PR gcov-profile/85759
2433 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2436 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2438 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2440 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2444 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2445 VEC_PACK_FLOAT_EXPR): New tree codes.
2446 * tree-pretty-print.c (op_code_prio): Handle
2447 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2448 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2449 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2450 * tree-inline.c (estimate_operator_cost): Likewise.
2451 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2452 * fold-const.c (const_binop): Likewise.
2453 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2454 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2455 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2456 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2457 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2458 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2459 * expr.c (expand_expr_real_2): Likewise.
2460 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2461 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2462 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2464 * optabs.c (expand_widen_pattern_expr): For
2465 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2466 sign from result type rather than operand's type.
2467 (expand_binop_directly): For vec_packu_float_optab and
2468 vec_packs_float_optab allow result type to be different from operand's
2470 * optabs-tree.c (optab_for_tree_code): Handle
2471 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2472 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2473 * tree-vect-generic.c (expand_vector_operations_1): Handle
2474 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2475 VEC_PACK_FLOAT_EXPR.
2476 * tree-vect-stmts.c (supportable_widening_operation): Handle
2478 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2479 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2480 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2481 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2482 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2484 (vec_pack<floatprefix>_float_<mode>): New expander.
2485 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2487 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2488 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2489 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2490 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2491 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2493 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2494 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2495 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2496 VEC_PACK_FLOAT_EXPR): Document.
2498 2018-05-29 Richard Biener <rguenther@suse.de>
2500 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2502 (stmt_vec_info_vec): Make pointer.
2503 (init_stmt_vec_info_vec): Remove.
2504 (free_stmt_vec_info_vec): Likewise.
2505 (set_stmt_vec_info_vec): New function.
2506 (free_stmt_vec_infos): Likewise.
2507 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2508 (set_vinfo_for_stmt): Likewise.
2509 (get_earlier_stmt): Likewise.
2510 (get_later_stmt): Likewise.
2511 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2512 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2513 (vec_info::~vec_info): Free stmt_vec_infos.
2514 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2515 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2516 (pass_slp_vectorize::execute): Likewise.
2517 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2518 (free_stmt_vec_info_vec): Likewise.
2519 (set_stmt_vec_info_vec): New function.
2520 (free_stmt_vec_infos): Likewise.
2521 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2522 the global stmt_vec_info_vec.
2523 * tree-parloops.c (gather_scalar_reductions): Use
2524 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2527 2018-05-29 Richard Biener <rguenther@suse.de>
2529 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2531 2018-05-29 Martin Liska <mliska@suse.cz>
2532 David Malcolm <dmalcolm@redhat.com>
2534 * vec.c (test_reverse): New.
2535 (vec_c_tests): Add new test.
2536 * vec.h (vl_ptr>::reverse): New function.
2538 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2540 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2542 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2545 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2547 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2549 2018-05-28 Richard Biener <rguenther@suse.de>
2551 PR tree-optimization/85933
2552 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2553 look at stmts marked as vectorizable.
2555 2018-05-28 Richard Biener <rguenther@suse.de>
2557 PR tree-optimization/85934
2558 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2559 vector boolean check before scalar optimization.
2561 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2563 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2566 2018-05-28 Mark Wielaard <mark@klomp.org>
2568 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2569 if it is an expression containing a minus sign.
2571 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2573 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2575 2018-05-27 Paul Koning <ni1d@arrl.net>
2577 * config/pdp11/pdp11.md (truncsihi2): Remove.
2579 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2580 Chung-Ju Wu <jasonwucj@gmail.com>
2582 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2584 (unaligned_store_dw): Ditto.
2585 * config/nds32/nds32-memory-manipulation.c
2586 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2587 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2588 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2589 (emit_setmem_word_loop): Rename to ...
2590 (emit_setmem_doubleword_loop): ... this.
2591 (nds32_gen_dup_4_byte_to_word_value): New function.
2592 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2593 (nds32_expand_setmem_loop): Refine implementation.
2594 (nds32_expand_setmem_loop_v3m): Ditto.
2595 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2598 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2600 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2602 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2604 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2605 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2606 machine->attr_no_prologue_p.
2607 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2608 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2609 (nds32_expand_epilogue): Consider attr_naked_p.
2610 (nds32_expand_epilogue_v3pop): Likewise.
2611 (nds32_can_use_return_insn): Likewise.
2612 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2613 attr_no_prologue_p fields.
2614 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2616 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2619 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2621 * config/i386/sse.md
2622 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2624 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2626 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2628 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2630 (*<floatsuffix>floatv2div2sf2): Rename to ...
2631 (*float<floatunssuffix>v2div2sf2): ... this.
2632 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2633 (float<floatunssuffix>v2div2sf2_mask): ... this.
2634 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2635 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2636 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2638 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2640 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2642 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2644 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2646 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2648 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2649 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2650 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2651 gen_ufix_truncv8dfv8si2.
2652 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2653 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2654 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2655 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2656 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2657 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2658 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2659 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2661 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2665 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2667 2018-05-25 Jim Wilson <jimw@sifive.com>
2669 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2670 * config/riscv/riscv.c (struct machine_function): Add
2671 interrupt_handler_p and attribute_checked_p fields.
2672 (riscv_attribute_table): Add interrupt.
2673 (riscv_interrupt_type_p): New.
2674 (riscv_save_reg_p): Save extra regs for interrupt handler.
2675 (riscv_use_save_libcall): Return false for interrupt handler.
2676 (riscv_first_stack_step): Add forward declaration.
2677 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2678 for interrupt handler with large frame. Use it for saved reg list.
2679 (riscv_expand_prologue): Move flag_stack_usage_info support to
2680 eliminate duplication.
2681 (riscv_expand_epilogue): Generate mret for interrupt handler.
2682 (riscv_epilogue_uses): New.
2683 (riscv_can_use_return_insn): Return false for interrupt handler.
2684 (riscv_function_ok_for_sibcall): Likewise.
2685 (riscv_set_current_function): Add interrupt handler support.
2686 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2687 * config/riscv/riscv.md (UNSPECV_MRET): New.
2689 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2691 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2693 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2695 PR tree-optimization/85712
2696 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2697 this candidate has already been replaced in-situ by a copy.
2699 2018-05-25 Jason Merrill <jason@redhat.com>
2701 PR c++/80485 - inline function non-zero address.
2702 * symtab.c (nonzero_address): Check DECL_COMDAT.
2704 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2707 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2708 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2709 extension of SImode operation. Use const123_operand predicate.
2710 (*saddsi_1): Remove.
2711 (*saddl_se_1): Ditto.
2713 (*ssubl_se_1): Ditto.
2714 * config/alpha/predicates.md (const123_operand): New predicate.
2715 * config/alpha/constraints.md (P): Use IN_RANGE.
2717 2018-05-25 Richard Biener <rguenther@suse.de>
2719 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2721 (ref_maybe_used_by_stmt_p): Likewise.
2722 (stmt_may_clobber_ref_p): Likewise.
2723 (stmt_may_clobber_ref_p_1): Likewise.
2724 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2726 (ref_maybe_used_by_stmt_p): Likewise.
2727 (stmt_may_clobber_ref_p): Likewise.
2728 (stmt_may_clobber_ref_p_1): Likewise.
2729 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2730 the alias oracle to disambiguate DRs with stmts DR analysis
2732 (vect_analyze_data_refs): Do not give up on not analyzable
2733 DRs for BB vectorization. Remove code truncating the dataref
2736 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2739 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2740 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2741 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2743 2018-05-25 Richard Biener <rguenther@suse.de>
2745 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2746 function, combining stmt data ref gathering and fatal analysis
2748 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2749 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2750 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2751 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2752 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2754 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2756 PR tree-optimization/85720
2757 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2758 SCC if all partitions are builtins.
2759 (version_loop_by_alias_check): New parameter. Generate cancelable
2760 runtime alias check if all partitions are builtins.
2761 (distribute_loop): Update call to above function.
2763 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2765 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2766 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2767 (parm_default_def_partition_arg): Ditto.
2768 (set_parm_default_def_partition): Ditto.
2769 (get_parm_default_def_partitions): Ditto and make it static.
2770 (get_undefined_value_partitions): Ditto and make it static.
2771 (remove_ssa_form): Refactor call to init_var_map here.
2772 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2773 computation for loop region.
2774 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2775 (register_default_def): Delete.
2776 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2777 (parm_default_def_partition_arg): Ditto.
2778 (set_parm_default_def_partition): Ditto.
2779 (get_parm_default_def_partitions): Ditto and make it static.
2780 (get_undefined_value_partitions): Ditto and make it static.
2781 (coalesce_with_default, coalesce_with_default): Update comment.
2782 (create_coalesce_list_for_region): New func factored out from
2783 create_outofssa_var_map.
2784 (populate_coalesce_list_for_outofssa): New func factored out from
2785 create_outofssa_var_map and coalesce_ssa_name.
2786 (create_outofssa_var_map): Delete.
2787 (coalesce_ssa_name): Refactor to support live range computation.
2788 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2789 (get_parm_default_def_partitions): Delete.
2790 (get_undefined_value_partitions): Ditto.
2791 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2792 computation for loop region.
2793 (new_tree_live_info, loe_visit_block): Ditto.
2794 (live_worklist, set_var_live_on_entry): Ditto.
2795 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2796 * tree-ssa-live.h (struct _var_map): New fields.
2797 (init_var_map): Change decl.
2798 (region_contains_p): New.
2800 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2802 * tree-ssa-live.h (live_merge_and_clear): Delete.
2804 2018-05-25 Richard Biener <rguenther@suse.de>
2807 * tree-dump.c (dequeue_and_dump): Remove access to removed
2808 operand 2 of a SWITCH_EXPR.
2810 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2812 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2813 multiplication and division.
2814 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2815 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2816 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2817 (cond_udiv_optab, cond_umod_optab): New optabs.
2818 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2819 (IFN_COND_RDIV): New internal functions.
2820 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2821 TRUNC_MOD_EXPR and RDIV_EXPR.
2822 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2823 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2825 (SVE_INT_BINARY): Include mult.
2826 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2827 (optab, sve_int_op): Handle mult.
2828 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2830 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2831 for SVE_INT_BINARY_SD.
2833 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2835 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2836 (optab, sve_int_op): Handle div and udiv.
2837 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2838 for SVE_INT_BINARY_SD.
2839 (*<optab><mode>3): New insn for the same.
2841 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2843 * tree-vect-patterns.c: Include predict.h.
2844 (vect_recog_divmod_pattern): Restrict check for division support
2845 to when optimizing for size.
2847 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2849 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2850 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2851 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2852 (gimple_match_op::set_op): Likewise.
2853 (gimple_resimplify4): Declare.
2854 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2855 (expr::gen_transform): Likewise.
2856 (decision_tree::gen): Generate a simplification routine for 4 operands.
2857 * gimple-match-head.c (gimple_simplify): Add an overload for
2858 4 operands. In the top-level function, handle up to 4 call
2859 arguments and call gimple_resimplify4.
2860 (gimple_resimplify4): New function.
2861 (build_call_internal): Pass a fourth operand.
2862 (maybe_push_to_seq): Likewise.
2863 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2864 Fold VEC_COND_EXPRs of an operation and a default value into
2865 an IFN_COND_* function if possible.
2866 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2868 (SVE_COND_FP_BINARY): Include them.
2869 (optab, sve_fp_op): Handle them.
2870 (SVE_INT_BINARY_REV): New code iterator.
2871 (SVE_COND_FP_BINARY_REV): New int iterator.
2872 (commutative): New int attribute.
2873 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2875 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2877 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2878 (*cond_<optab><mode>): New patterns for reversed operands.
2880 2018-05-25 Richard Biener <rguenther@suse.de>
2882 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2883 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2884 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2885 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2886 * tree-vect-data-refs.c (everywhere): Adjust users.
2887 * tree-vect-loop.c (everywhere): Likewise.
2888 * tree-vect-slp.c (everywhere): Likewise.
2889 * tree-vect-stmts.c (everywhere): Likewise.
2890 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2892 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2894 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2896 (gcc_cv_as_section_exclude): ... this.
2897 Try Solaris as #exclude syntax.
2898 * configure: Regenerate.
2899 * config.in: Regenerate.
2900 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2902 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2903 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2905 * varasm.c (default_elf_asm_named_section): Don't check if
2906 HAVE_GAS_SECTION_EXCLUDE is defined.
2908 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2910 * doc/md.texi: Update the documentation of the cond_* optabs
2911 to mention the new final operand. Fix GET_MODE_NUNITS call.
2912 Describe the scalar case too.
2913 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2914 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2916 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2917 (get_conditional_internal_fn): Update comment.
2918 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2919 accumulator value as a final argument to conditional functions.
2920 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2921 a define_expand and add an "else" operand. Assert for now that
2922 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2923 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2924 (*cond_<optab><mode>): New patterns.
2925 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2926 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2927 (UNSPEC_COND_EOR): Delete.
2928 (optab): Remove associated mappings.
2929 (SVE_INT_BINARY): New code iterator.
2930 (sve_int_op): Remove int attribute and add "minus" to the code
2932 (SVE_COND_INT_OP): Delete.
2933 (SVE_COND_FP_OP): Rename to...
2934 (SVE_COND_FP_BINARY): ...this.
2936 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2938 * optabs.c (can_reuse_operands_p): New function.
2939 (maybe_legitimize_operands): Try to reuse the results for
2942 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2944 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2945 Add {q} suffix to insn mnemonic.
2947 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2949 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2950 (msp430_warn_func_return): New.
2952 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2954 * fold-const.c (tree_nonzero_bits): New function.
2955 * fold-const.h (tree_nonzero_bits): Likewise.
2956 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2957 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2959 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2963 * varasm.c (assemble_alias): Check ifunc_resolver only on
2966 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2969 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2970 when memory input operand is handled.
2972 2018-05-24 Luis Machado <luis.machado@linaro.org>
2974 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2976 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2978 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2980 * match.pd: Delay FMA folds until after vectorization.
2982 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2985 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2986 address check not strict.
2988 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2990 * gimple-match.h (gimple_match_op): New class.
2991 (mprts_hook): Replace parameters with a gimple_match_op *.
2992 (maybe_build_generic_op): Likewise.
2993 (gimple_simplified_result_is_gimple_val): Replace parameters with
2994 a const gimple_match_op *.
2995 (gimple_simplify): Replace code_helper * and tree * parameters with
2996 a gimple_match_op * parameter.
2997 (gimple_resimplify1): Replace code_helper *, tree and tree *
2998 parameters with a gimple_match_op * parameter.
2999 (gimple_resimplify2): Likewise.
3000 (gimple_resimplify3): Likewise.
3001 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3002 parameters with a gimple_match_op * parameter.
3003 * gimple-match-head.c (gimple_simplify): Change prototypes of
3004 auto-generated functions to take a gimple_match_op * instead of
3005 separate code_helper * and tree * parameters. Make the same
3006 change in the top-level overload and update calls to the
3007 gimple_resimplify routines. Update calls to the auto-generated
3008 functions and to maybe_push_res_to_seq in the publicly-facing
3009 operation-specific gimple_simplify overloads.
3010 (gimple_match_op::MAX_NUM_OPS): Define.
3011 (gimple_resimplify1): Replace rcode and ops with a single res_op
3012 parameter. Update call to gimple_simplify.
3013 (gimple_resimplify2): Likewise.
3014 (gimple_resimplify3): Likewise.
3015 (mprts_hook): Replace parameters with a gimple_match_op *.
3016 (maybe_build_generic_op): Likewise.
3017 (build_call_internal): Replace type, nargs and ops with
3018 a gimple_match_op *.
3019 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3020 with a single gimple_match_op *. Update calls to mprts_hook,
3021 build_call_internal and gimple_simplified_result_is_gimple_val.
3022 Factor out code that is common to the tree_code and combined_fn cases.
3023 * genmatch.c (expr::gen_transform): Replace tem_code and
3024 tem_ops with a gimple_match_op called tem_op. Update calls
3025 to the gimple_resimplify functions and maybe_push_res_to_seq.
3026 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3027 res_ops. Update call to the gimple_resimplify functions.
3028 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3029 (decision_tree::gen): Make the functions take a gimple_match_op *
3030 called res_op instead of separate res_code and res_ops parameters.
3031 Update call accordingly.
3032 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3033 and ops with a single res_op parameter. Update calls to
3034 maybe_build_generic_op and maybe_push_res_to_seq.
3035 (fold_stmt_1): Update calls to gimple_simplify and
3036 replace_stmt_with_simplification.
3037 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3038 and gimple_simplified_result_is_gimple_val.
3039 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3041 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3042 with a gimple_match_op *.
3043 (vn_nary_build_or_lookup): Likewise. Update call to
3044 vn_nary_build_or_lookup_1.
3045 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3046 gimple_match_op *. Update calls to the gimple_resimplify routines
3047 and to gimple_simplified_result_is_gimple_val.
3048 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3049 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3050 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3051 (visit_nary_op): Likewise.
3052 (visit_reference_op_load): Likewise.
3054 2018-05-23 Luis Machado <luis.machado@linaro.org>
3056 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3057 modifier for printing the step amount.
3059 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3062 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3065 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3067 * doc/sourcebuild.texi (Endianness): New subsubsection.
3069 2018-05-23 Luis Machado <luis.machado@linaro.org>
3071 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3072 <prefetch_dynamic_strides>: New const bool field.
3073 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3074 prefetch_dynamic_strides.
3075 (exynosm1_prefetch_tune): Likewise.
3076 (thunderxt88_prefetch_tune): Likewise.
3077 (thunderx_prefetch_tune): Likewise.
3078 (thunderx2t99_prefetch_tune): Likewise.
3079 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3081 (aarch64_override_options_internal): Update to set
3082 PARAM_PREFETCH_DYNAMIC_STRIDES.
3083 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3084 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3085 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3086 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3087 prefetch-dynamic-strides setting.
3089 2018-05-23 Luis Machado <luis.machado@linaro.org>
3091 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3092 <minimum_stride>: New const int field.
3093 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3094 minimum_stride field defaulting to -1.
3095 (exynosm1_prefetch_tune): Likewise.
3096 (thunderxt88_prefetch_tune): Likewise.
3097 (thunderx_prefetch_tune): Likewise.
3098 (thunderx2t99_prefetch_tune): Likewise.
3099 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3100 <default_opt_level>: Set to 3.
3101 (aarch64_override_options_internal): Update to set
3102 PARAM_PREFETCH_MINIMUM_STRIDE.
3103 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3104 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3105 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3106 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3107 stride is constant and is below the minimum stride threshold.
3109 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3111 * config/arm/arm-cpus.in (mode26): Delete.
3112 (armv4): Delete mode26 reference.
3113 * config/arm/arm.c (arm_configure_build_target): Delete use of
3116 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3118 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3120 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3121 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3123 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3124 (floatunsdidf2): Ditto.
3126 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3128 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3129 (fixuns_trunc<mode>si2_avx512f): Ditto.
3130 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3131 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3132 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3134 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3136 PR rtl-optimization/79985
3137 * df-scan.c (df_insn_refs_collect): Remove special case for
3138 global registers and asm statements.
3140 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3142 * extend.texi (Global Register Variables): Rewrite the bullet list.
3143 Note that the register is available for allocation. Note that access
3144 via inline asm must use constraints. Add note about async-signal
3145 handlers. Remove paragraph about automagic register selection.
3147 2018-05-23 Richard Biener <rguenther@suse.de>
3149 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3150 of fixed offset from memset VN.
3152 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3154 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3156 (alloc_cand_and_find_basis): Initialize first_interp field.
3157 (slsr_process_mul): Modify first_interp field.
3158 (slsr_process_add): Likewise.
3159 (slsr_process_cast): Modify first_interp field for each new
3161 (slsr_process_copy): Likewise.
3162 (dump_candidate): Dump first_interp field.
3163 (replace_mult_candidate): Process all interpretations, not just
3165 (replace_rhs_if_not_dup): Likewise.
3166 (replace_one_candidate): Likewise.
3168 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3170 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3172 (aarch64_needs_frame_chain): New function.
3173 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3175 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3178 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3179 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3180 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3181 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3182 as true for strict-align.
3183 (aarch64_can_inline_p): Perform checks even when callee has no
3184 attributes to check for strict alignment.
3185 * doc/extend.texi (AArch64 Function Attributes): Document
3187 * doc/invoke.texi: (AArch64 Options): Likewise.
3189 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3191 PR tree-optimization/85853
3192 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3193 the handling of the root of the node to...
3194 (vect_slp_analyze_node_operations_1): ...this new function,
3195 and run the whole thing with the child nodes' def types
3196 set according to their SLP node's def type.
3198 2018-05-23 Richard Biener <rguenther@suse.de>
3201 * tree-data-ref.c (create_runtime_alias_checks): Defer
3202 and ignore overflow warnings.
3204 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3206 PR tree-optimization/85822
3207 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3210 2018-05-23 Richard Biener <rguenther@suse.de>
3212 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3213 memset constants via native_interpret_expr.
3215 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3218 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3220 (cgraph_node::create_alias): Likewise.
3221 (cgraph_node::get_availability): Check ifunc_resolver instead
3222 of looking up ifunc attribute.
3223 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3224 * varasm.c (do_assemble_alias): Likewise.
3225 (assemble_alias): Likewise.
3226 (default_binds_local_p_3): Likewise.
3227 * cgraph.h (cgraph_node): Add ifunc_resolver.
3228 (cgraph_node::only_called_directly_or_aliased_p): Return false
3230 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3232 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3233 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3234 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3235 instead of looking up ifunc attribute.
3237 2018-05-22 Luis Machado <luis.machado@linaro.org>
3239 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3241 2018-05-22 Martin Sebor <msebor@redhat.com>
3244 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3245 only when expasion succeeds.
3246 (expand_builtin_strcmp): Same.
3247 (expand_builtin_strncmp): Same.
3249 2018-05-22 Martin Sebor <msebor@redhat.com>
3251 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3253 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3254 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3256 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3257 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3259 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3260 Add check for legitimate_address.
3261 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3262 (aarch64_swap_ldrstr_operands): New.
3263 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3266 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3267 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3269 * config/aarch64/aarch64.md: New patterns to generate stp
3271 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3272 single words and double words.
3273 (load_pair_sw, load_pair_dw): Likewise.
3274 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3276 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3278 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3279 for different mode ldpstp and add peephole for merged zero stores.
3281 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3283 (aarch64_gen_store_pair): Rename calls to match new patterns.
3284 (aarch64_gen_load_pair): Rename calls to match new patterns.
3285 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3286 (load_pair<DREG:mode><DREG2:mode>): ... This.
3287 (store_pair<mode>): Rename to...
3288 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3289 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3291 (V_INT_EQUIV): Handle SImode.
3292 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3295 2018-05-22 Martin Sebor <msebor@redhat.com>
3298 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3299 or ajust the presumed bound on an operation to avoid unnecessary
3302 2018-05-22 Martin Sebor <msebor@redhat.com>
3304 PR tree-optimization/85826
3305 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3306 assuming that a DECL necesarily has a constant size.
3308 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3311 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3313 2018-05-22 Richard Biener <rguenther@suse.de>
3315 PR tree-optimization/85834
3316 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3317 non-constant and non-zero memset arguments.
3319 2018-05-22 Martin Liska <mliska@suse.cz>
3322 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3324 2018-05-22 Richard Biener <rguenther@suse.de>
3326 PR tree-optimization/85863
3327 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3328 comparisons when vectype is specified.
3329 (vectorizable_condition): Do not specify vectype for
3330 vect_is_simple_cond when SLP vectorizing.
3332 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3335 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3336 define __ibm128 as long double.
3337 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3338 as a distinct type when IEEE 128-bit support is enabled.
3339 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3340 128-bit types to use the correct functions.
3341 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3342 convert between 128-bit floating point types that have different
3343 modes but the same representation, instead of using gen_lowpart to
3345 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3347 (IFKF_reg): New attributes to give the register constraints for
3349 (extend<mode>tf2_internal): New insns to mark an explicit
3350 conversion between 128-bit floating point types that have a
3351 different mode but share the same representation.
3353 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3355 PR tree-optimization/85814
3356 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3357 a null return from get_strinfo when unsharing the next
3358 strinfo in the chain.
3360 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3363 * varasm.c (weak_finish): Clean up weak_decls.
3365 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3367 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3368 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3369 UNSPEC_UADALP values.
3370 * config/aarch64/iterators.md (ABAL): New int iterator.
3373 (sur): Add mappings for the above.
3374 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3376 (aarch64_<sur>abal<mode>_4): Likewise.
3377 (aarch64_<sur>adalp<mode>_3): Likewise.
3378 (<sur>sadv16qi): New define_expand.
3380 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3382 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3383 (*movdf_internal): Ditto.
3384 (*rcpsf2_sse): Ditto.
3385 (*rsqrtsf2_sse): Ditto.
3386 (*sqrt<mode>2_sse): Ditto.
3388 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3390 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3392 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3393 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3394 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3395 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3397 * config/aarch64/arm_neon.h: Likewise.
3398 * config/aarch64/iterators.md (VQ_I): New.
3400 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3402 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3404 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3407 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3409 * config/nds32/constraints.md (S): New constraint.
3410 * config/nds32/nds32.md (call_internal): Use constraint S.
3411 (call_value_internal): Likewise.
3412 (sibcall_internal): Likewise.
3413 (sibcall_value_internal): Likewise.
3415 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3416 Chung-Ju Wu <jasonwucj@gmail.com>
3418 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3421 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3422 Chung-Ju Wu <jasonwucj@gmail.com>
3424 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3425 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3426 (nds32_rtx_costs_impl): Simplify.
3427 (nds32_address_cost_impl): Simplify.
3428 (nds32_init_rtx_costs): New function.
3429 (nds32_rtx_costs_speed_prefer): Likewise.
3430 (nds32_rtx_costs_size_prefer): Likewise.
3431 (nds32_address_cost_speed_prefer): Likewise.
3432 (nds32_address_cost_speed_fwprop): Likewise.
3433 (nds32_address_cost_size_prefer): Likewise.
3434 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3435 * config/nds32/nds32.c (nds32_option_override): Use
3436 nds32_init_rtx_costs function.
3438 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3440 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3441 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3442 (TARGET_PIPELINE_N8): Likewise.
3443 (TARGET_PIPELINE_N10): Likewise.
3444 (TARGET_PIPELINE_N13): Likewise.
3445 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3447 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3449 * config/nds32/nds32-fpu.md: Update copyright year.
3451 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3453 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3455 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3457 * config/nds32/nds32.c
3458 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3459 * config/nds32/nds32.opt (minline-asm-r15): New option.
3461 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3463 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3465 * config/nds32/nds32.md (abssi2): New pattern.
3467 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3469 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3470 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3471 Merge insn pattern from sse_cvtsi2ss<round_name> and
3472 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3473 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3474 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3475 using SWI48 mode iterator.
3476 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3477 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3478 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3479 pattern from sse_cvttss2si<round_saeonly_name>
3480 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3481 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3482 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3483 using SWI48 mode iterator.
3484 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3485 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3486 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3487 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3488 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3489 using SWI48 mode iterator.
3490 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3491 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3492 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3493 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3494 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3495 SWI48 mode iterator.
3496 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3497 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3498 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3499 pattern from sse_cvttsd2si<round_saeonly_name>
3500 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3502 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3504 * config/nds32/nds32-md-auxiliary.c
3505 (nds32_valid_smw_lwm_base_p): Refine.
3506 (nds32_output_smw_single_word): Refine.
3507 (nds32_output_smw_double_word): New.
3508 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3510 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3512 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3513 (nds32_output_stack_pop): Refine.
3514 (nds32_expand_unaligned_load): Refine.
3515 (nds32_expand_unaligned_store): Refine.
3517 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3518 Chung-Ju Wu <jasonwucj@gmail.com>
3520 * config/nds32/constants.md: Add TP_REGNUM constant.
3521 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3522 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3524 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3525 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3526 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3527 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3528 and PIC code generation.
3529 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3531 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3533 * config/nds32/nds32.md: Support TLS and PIC.
3534 * config/nds32/nds32.c: Support TLS and PIC.
3535 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3536 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3539 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3541 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3542 mode with E_ prefix.
3544 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3545 Chung-Ju Wu <jasonwucj@gmail.com>
3547 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3548 * config/nds32/nds32-md-auxiliary.c
3549 (symbolic_reference_mentioned_p): New.
3550 (nds32_legitimize_ict_address): New.
3551 (nds32_expand_ict_move): New.
3552 (nds32_indirect_call_referenced_p): New.
3553 (nds32_symbol_binds_local_p): Delete.
3554 (nds32_long_call_p): Modify.
3555 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3556 * config/nds32/nds32-protos.h
3557 (symbolic_reference_mentioned_p): Declare.
3558 (nds32_legitimize_ict_address): Declare.
3559 (nds32_expand_ict_move): Declare.
3560 (nds32_indirect_call_referenced_p): Declare.
3561 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3562 (nds32_relax_group): Use nds32_ict_const_p as condition.
3563 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3564 (nds32_asm_file_start): Output ict_model directive in asm code.
3565 (nds32_legitimate_address_p): Consider indirect call.
3566 (nds32_print_operand): Consider indirect call.
3567 (nds32_print_operand_address): Consider indirect call.
3568 (nds32_insert_attributes): Handle "indirect_call" attribute.
3569 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3570 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3571 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3572 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3573 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3574 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3575 (TARGET_ICT_MODEL_SMALL): Define.
3576 (TARGET_ICT_MODEL_LARGE): Define.
3577 * config/nds32/nds32.md (movsi): Consider ict model.
3578 (call, call_value): Consider ict model.
3579 (sibcall, sibcall_value): Consider ict model.
3580 * config/nds32/nds32.opt (mict-model): New option.
3581 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3584 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3585 Monk Chiang <sh.chiang04@gmail.com>
3586 Jim Wilson <jimw@sifive.com>
3588 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3589 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3590 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3591 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3592 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3593 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3594 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3595 compute save_libcall_adjustment properly.
3596 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3597 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3598 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3599 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3600 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3601 (ABI_SPEC): Handle mabi=ilp32e.
3602 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3603 (RVE): Add RVE mask.
3604 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3605 <-march>: Add rv32e as an example.
3607 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3610 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3611 (intra_create_variable_infos): Handle C++ constructors.
3613 2018-05-18 Martin Liska <mliska@suse.cz>
3615 * passes.def: Remove a redundant pass.
3617 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3620 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3622 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3624 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3626 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3628 (armv2, armv2a, armv3, armv3m): Delete architectures.
3629 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3630 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3631 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3633 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3634 (*mulsidi3adddi): Likewise.
3635 (mulsidi3): Likewise.
3636 (*mulsidi3_nov6): Likewise.
3637 (umulsidi3): Likewise.
3638 (umulsidi3_nov6): Likewise.
3639 (umaddsidi4): Likewise.
3640 (*umulsidi3adddi): Likewise.
3641 (smulsi3_highpart): Likewise.
3642 (*smulsi3_highpart_nov6): Likewise.
3643 (umulsi3_highpart): Likewise.
3644 (*umulsi3_highpart_nov6): Likewise.
3645 * config/arm/arm.h (arm_arch3m): Delete.
3646 * config/arm/arm.c (arm_arch3m): Delete.
3647 (arm_option_override_internal): Update armv3-related comment.
3648 (arm_configure_build_target): Delete use of isa_bit_mode32.
3649 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3650 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3651 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3653 (mulusa3): Likewise.
3654 * config/arm/arm-protos.h (arm_arch3m): Delete.
3655 * config/arm/arm-tables.opt: Regenerate.
3656 * config/arm/arm-tune.md: Likewise.
3657 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3658 deleted architectures.
3660 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3662 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3663 (armv5t, armv5te): New features.
3664 (ARMv5, ARMv5e): Delete fgroups.
3665 (ARMv5t, ARMv5te): Adjust for above changes.
3667 (armv5, armv5e): Delete arches.
3668 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3670 (*call_reg_arm): Likewise.
3671 (*call_value_reg_armv5): Likewise.
3672 (*call_value_reg_arm): Likewise.
3673 (*call_symbol): Likewise.
3674 (*call_value_symbol): Likewise.
3675 (*sibcall_insn): Likewise.
3676 (*sibcall_value_insn): Likewise.
3678 (prefetch): Likewise.
3679 (define_split and define_peephole2 dependent on arm_arch5):
3681 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3683 (TARGET_ARM_QBIT): Likewise.
3684 (TARGET_DSP_MULTIPLY): Likewise.
3685 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3686 (arm_arch5, arm_arch5e): Delete.
3687 (arm_arch5t, arm_arch5te): Declare.
3688 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3689 (arm_arch5t): Declare.
3690 (arm_option_reconfigure_globals): Update for the above.
3691 (arm_options_perform_arch_sanity_checks): Update comment, replace
3692 use of arm_arch5 with arm_arch5t.
3693 (use_return_insn): Likewise.
3694 (arm_emit_call_insn): Likewise.
3695 (output_return_instruction): Likewise.
3696 (arm_final_prescan_insn): Likewise.
3697 (arm_coproc_builtin_available): Likewise.
3698 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3699 arm_arch5e with arm_arch5t and arm_arch5te.
3700 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3701 (arm_arch5t, arm_arch5te): Declare.
3702 * config/arm/arm-tables.opt: Regenerate.
3703 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3704 * config/arm/t-multilib: Likewise.
3705 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3706 instead of arm_arch5.
3707 (*call_reg_thumb1): Likewise.
3708 (*call_value_reg_thumb1_v5): Likewise.
3709 (*call_value_reg_thumb1): Likewise.
3710 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3712 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3714 2018-05-18 Martin Liska <mliska@suse.cz>
3716 PR gcov-profile/84846
3717 * doc/gcov.texi: Document -t option of gcov tool.
3719 2018-05-18 Martin Liska <mliska@suse.cz>
3721 PR gcov-profile/84846
3722 * gcov.c (print_usage): Add new -t option.
3723 (process_args): Handle the option.
3724 (generate_results): Use stdout as output when requested by
3727 2018-05-18 Martin Liska <mliska@suse.cz>
3729 PR gcov-profile/84846
3730 * coverage.c (coverage_init): Write PWD to .gcno file.
3731 * doc/gcov.texi: Document how working directory is printed.
3732 * gcov-dump.c (dump_gcov_file): Print PWD.
3733 * gcov.c (output_intermediate_file): Likewise.
3734 (read_graph_file): Read PWD string.
3735 (output_lines): Print PWD.
3737 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3740 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3741 for retval and return false if all args to phi are zero.
3743 2018-05-18 Richard Biener <rguenther@suse.de>
3745 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3747 (evrp_dom_walker::before_dom_children): Call it.
3749 2018-05-18 Richard Biener <rguenther@suse.de>
3751 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3752 results when processing array refs with variable index.
3754 2018-05-18 Toon Moene <toon@moene.org>
3756 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3757 directly after that of -floop-interchange. Indicate that both
3758 options are enabled by default when specifying -O3.
3760 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3762 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3763 iterator. Delete separate integer-mode vec_set<mode> expander.
3764 (aarch64_simd_vec_setv2di): Delete.
3765 (vec_setv2di): Delete.
3766 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3767 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3768 the "w, r" alternative.
3770 2018-05-18 Martin Liska <mliska@suse.cz>
3772 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3773 * tree-pass.h (make_pass_lower_switch_O0): New function.
3774 * tree-switch-conversion.c (node_has_low_bound): Remove.
3775 (node_has_high_bound): Likewise.
3776 (node_is_bounded): Likewise.
3777 (class pass_lower_switch): Make it a template type and create
3779 (pass_lower_switch::execute): Add template argument.
3780 (make_pass_lower_switch): New function.
3781 (make_pass_lower_switch_O0): New function.
3782 (do_jump_if_equal): Remove.
3783 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3784 all the hard work to tree optimization passes.
3786 2018-05-18 Martin Liska <mliska@suse.cz>
3788 * dbgcnt.c (limit_low): Renamed from limit.
3789 (limit_high): New variable.
3790 (dbg_cnt_is_enabled): Check for upper limit.
3791 (dbg_cnt): Adjust dumping.
3792 (dbg_cnt_set_limit_by_index): Add new argument for high
3794 (dbg_cnt_set_limit_by_name): Likewise.
3795 (dbg_cnt_process_single_pair): Parse new format.
3796 (dbg_cnt_process_opt): Use strtok.
3797 (dbg_cnt_list_all_counters): Remove 'value' and add
3799 * doc/invoke.texi: Document changes.
3801 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3803 * doc/sourcebuild.texi (scalar_all_fma): Document.
3804 * tree.def (FMA_EXPR): Delete.
3805 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3806 * internal-fn.c (ternary_direct): New macro.
3807 (expand_ternary_optab_fn): Likewise.
3808 (direct_ternary_optab_supported_p): Likewise.
3809 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3810 * builtins.c (fold_builtin_fma): Delete.
3811 (fold_builtin_3): Don't call it.
3812 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3813 * expr.c (expand_expr_real_2): Likewise.
3814 * fold-const.c (operand_equal_p): Likewise.
3815 (fold_ternary_loc): Likewise.
3816 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3817 * gimple.c (DEFTREECODE): Likewise.
3818 * gimplify.c (gimplify_expr): Likewise.
3819 * optabs-tree.c (optab_for_tree_code): Likewise.
3820 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3821 * tree-eh.c (operation_could_trap_p): Likewise.
3822 (stmt_could_throw_1_p): Likewise.
3823 * tree-inline.c (estimate_operator_cost): Likewise.
3824 * tree-pretty-print.c (dump_generic_node): Likewise.
3825 (op_code_prio): Likewise.
3826 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3827 * tree-ssa-operands.c (get_expr_operands): Likewise.
3828 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3829 * fold-const-call.h (fold_fma): Delete.
3830 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3831 CFN_FNMA and CFN_FNMS.
3833 * genmatch.c (combined_fn): New enum.
3834 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3835 (commutative_op): New function.
3836 (commutate): Use it. Handle more than 2 operands.
3837 (dt_operand::gen_gimple_expr): Use commutative_op.
3838 (parser::parse_expr): Allow :c to be used with non-binary
3839 operators if the commutative operand is known.
3840 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3841 CFN_FMS, CFN_FNMA and CFN_FNMS.
3842 (backprop::process_assign_use): Remove FMA_EXPR handling.
3843 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3844 (gen_hsa_fma): New function.
3845 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3846 IFN_FNMA and IFN_FNMS.
3847 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3848 * gimple-fold.h (follow_all_ssa_edges): Declare.
3849 * gimple-fold.c (follow_all_ssa_edges): New function.
3850 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3851 gimple_build interface and use follow_all_ssa_edges to fold the result.
3852 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3853 instead of checking for optabs directly.
3854 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3855 rather than FMA_EXPRs.
3856 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3857 call to IFN_FMA instead of an FMA_EXPR.
3859 2018-05-17 Jim Wilson <jimw@sifive.com>
3861 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3862 sign extended promoted subreg, and the range does not have the sign bit
3863 set, then do a sign extend.
3865 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3866 test, check for sign extended subreg and/or constant operands, and
3867 do a sign extend in that case.
3869 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3871 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3872 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3874 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3875 Change logics_shift_reg to logics_shift_imm.
3876 (thunderx2t99_fp_loadpair_basic): Delete.
3877 (thunderx2t99_fp_storepair_basic): Delete.
3878 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3879 (thunderx2t99_asimd_polynomial): Delete.
3880 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3881 and neon_fp_mul_d_scalar_q.
3882 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3883 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3884 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3885 (thunderx2t99_asimd_lut): Add missing tbl types.
3886 (thunderx2t99_asimd_ext): Delete.
3887 (thunderx2t99_asimd_load1_1_mult): Delete.
3888 (thunderx2t99_asimd_load1_2_mult): Delete.
3889 (thunderx2t99_asimd_load1_ldp): New.
3890 (thunderx2t99_asimd_load1): New.
3891 (thunderx2t99_asimd_load2): Add missing *load2* types.
3892 (thunderx2t99_asimd_load3): New.
3893 (thunderx2t99_asimd_load4): New.
3894 (thunderx2t99_asimd_store1_1_mult): Delete.
3895 (thunderx2t99_asimd_store1_2_mult): Delete.
3896 (thunderx2t99_asimd_store2_mult): Delete.
3897 (thunderx2t99_asimd_store2_onelane): Delete.
3898 (thunderx2t99_asimd_store_stp): New.
3899 (thunderx2t99_asimd_store1): New.
3900 (thunderx2t99_asimd_store2): New.
3901 (thunderx2t99_asimd_store3): New.
3902 (thunderx2t99_asimd_store4): New.
3904 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3906 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3907 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3909 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3910 Segher Boessenkool <segher@kernel.crashing.org>
3913 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3916 2018-05-17 Richard Biener <rguenther@suse.de>
3918 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3919 for pruning loop and prune defs feeding only already visited PHIs.
3921 2018-05-17 Richard Biener <rguenther@suse.de>
3923 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3925 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3926 Richard Biener <rguenther@suse.de>
3928 PR tree-optimization/85793
3929 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3930 for VMAT_ELEMENTWISE.
3932 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3934 * internal-fn.h (lookup_internal_fn): Declare
3935 * internal-fn.c (lookup_internal_fn): New function.
3936 * gimple.c (gimple_build_call_from_tree): Handle calls to
3938 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3939 internal function names.
3940 * tree-pretty-print.c (dump_generic_node): Likewise.
3941 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3943 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3945 * gimple-fold.h (gimple_build): Make the function forms take
3946 combined_fn rather than built_in_function.
3947 (gimple_simplify): Likewise.
3948 * gimple-match-head.c (gimple_simplify): Likewise.
3949 * gimple-fold.c (gimple_build): Likewise.
3950 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3951 rather than gimple_build_call_internal.
3952 (get_initial_defs_for_reduction): Likewise.
3953 (vect_create_epilog_for_reduction): Likewise.
3954 (vectorizable_live_operation): Likewise.
3956 2018-05-17 Martin Liska <mliska@suse.cz>
3958 * gimple-ssa-sprintf.c (format_directive): Do not use
3959 space in between 'G_' and '('.
3961 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3964 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3965 even if the mask is not all ones.
3968 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3970 (ix86_gimple_fold_builtin): Likewise.
3973 * config/i386/i386.c: Include tree-vector-builder.h.
3974 (ix86_vector_shift_count): New function.
3975 (ix86_fold_builtin): Fold shift builtins by scalar count.
3976 (ix86_gimple_fold_builtin): Likewise.
3978 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3979 _mm512_setzero): New intrinsics.
3981 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3982 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3984 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3985 code generation for cases where splatting a value is not useful.
3986 * simplify-rtx.c (simplify_ternary_operation): Simplify
3987 vec_merge across a vec_duplicate and a paradoxical subreg forming
3988 a vector mode to a vec_concat.
3990 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3992 * config.gcc: Support "goldmont-plus".
3993 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3995 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3996 PROCESSOR_GOLDMONT_PLUS.
3997 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3998 (processor_target_table): Add "goldmont-plus".
3999 (PTA_GOLDMONT_PLUS): Define.
4000 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4001 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4002 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4003 (fold_builtin_cpu): Add "goldmont-plus".
4004 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4005 (ix86_option_override_internal): Add "goldmont-plus".
4006 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4007 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4008 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4009 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4011 2018-05-17 Richard Biener <rguenther@suse.de>
4013 PR tree-optimization/85757
4014 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4015 remove defs that only feed that PHI from further processing.
4017 2018-05-16 Jim Wilson <jimw@sifive.com>
4019 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4021 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4023 2018-05-16 Mark Wielaard <mark@klomp.org>
4025 * dwarf2out.c (count_index_strings): New function.
4026 (output_indirect_strings): Call count_index_strings and generate
4027 header for dwarf_version >= 5.
4029 2018-05-16 Mark Wielaard <mark@klomp.org>
4031 * dwarf2out.c (dwarf_FORM): New function.
4032 (set_indirect_string): Use dwarf_FORM.
4033 (reset_indirect_string): Likewise.
4034 (size_of_die): Likewise.
4035 (value_format): Likewise.
4036 (output_die): Likewise.
4037 (add_skeleton_AT_string): Likewise.
4038 (output_macinfo_op): Likewise.
4039 (index_string): Likewise.
4040 (output_index_string_offset): Likewise.
4041 (output_index_string): Likewise.
4042 (count_index_strings): Likewise.
4044 2018-05-16 Carl Love <cel@us.ibm.com>
4046 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4047 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4049 2018-05-16 Martin Jambor <mjambor@suse.cz>
4051 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4052 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4054 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4056 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4057 (fnma<mode>4): Likewise.
4058 (fms<mode>4): Likewise.
4059 (fnms<mode>4): Likewise.
4060 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4061 (aarch64_fnma<mode>4): Likewise.
4062 (aarch64_fms<mode>4): Likewise.
4063 (aarch64_fnms<mode>4): Likewise.
4064 (aarch64_fnmadd<mode>4): Likewise.
4066 2018-05-16 Jason Merrill <jason@redhat.com>
4068 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4070 2018-05-16 Richard Biener <rguenther@suse.de>
4072 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4073 (dump_stmt_cost): Declare.
4074 (add_stmt_cost): Dump cost we add.
4075 (add_stmt_costs): New function.
4076 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4078 (vect_analyze_stmt): Adjust prototype.
4079 (vectorizable_condition): Likewise.
4080 (vectorizable_live_operation): Likewise.
4081 (vectorizable_reduction): Likewise.
4082 (vectorizable_induction): Likewise.
4083 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4084 cost vector to pass to vectorizable_ and record afterwards.
4085 (vect_model_reduction_cost): Take cost vector argument and adjust.
4086 (vect_model_induction_cost): Likewise.
4087 (vectorizable_reduction): Likewise.
4088 (vectorizable_induction): Likewise.
4089 (vectorizable_live_operation): Likewise.
4090 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4091 SLP_TREE_NUMBER_OF_VEC_STMTS.
4092 (vect_analyze_slp_cost_1): Remove.
4093 (vect_analyze_slp_cost): Likewise.
4094 (vect_slp_analyze_node_operations): Take visited args and
4095 a target cost vector. Avoid processing already visited stmt sets.
4096 (vect_slp_analyze_operations): Use a local cost vector to gather
4097 costs and register those of non-discarded instances.
4098 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4099 (vect_schedule_slp_instance): Remove copying of
4100 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4102 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4103 adding cost. Record cost entry location.
4104 (vect_prologue_cost_for_slp_op): Function to compute cost of
4105 a constant or invariant generated for SLP vect in the prologue,
4106 split out from vect_analyze_slp_cost_1.
4107 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4108 (vect_model_promotion_demotion_cost): Likewise.
4109 (vect_model_store_cost): Likewise, make static.
4110 (vect_model_load_cost): Likewise.
4111 (vectorizable_bswap): Add cost vector arg and adjust.
4112 (vectorizable_call): Likewise.
4113 (vectorizable_simd_clone_call): Likewise.
4114 (vectorizable_conversion): Likewise.
4115 (vectorizable_assignment): Likewise.
4116 (vectorizable_shift): Likewise.
4117 (vectorizable_operation): Likewise.
4118 (vectorizable_store): Likewise.
4119 (vectorizable_load): Likewise.
4120 (vectorizable_condition): Likewise.
4121 (vectorizable_comparison): Likewise.
4122 (can_vectorize_live_stmts): Likewise.
4123 (vect_analyze_stmt): Likewise.
4124 (vect_transform_stmt): Adjust calls to vectorizable_*.
4125 * tree-vectorizer.c: Include gimple-pretty-print.h.
4126 (dump_stmt_cost): New function.
4128 2018-05-16 Richard Biener <rguenther@suse.de>
4130 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4131 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4132 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4133 (check_name): New callback.
4134 (dse_classify_store): Track cycles via a visited bitmap of PHI
4135 defs and simplify handling of in-loop and across loop dead stores
4136 and properly fail for loop-variant refs. Handle byte-tracking with
4137 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4140 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4142 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4143 (vect_get_mask_type_for_stmt): Likewise.
4144 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4146 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4147 to determine the statement's vector type and the vector type that
4148 should be used for calculating nunits. Deal with cases in which
4149 the type has to be deferred.
4150 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4151 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4152 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4153 (vect_determine_vf_for_stmt): New functions, split out from...
4154 (vect_determine_vectorization_factor): ...here.
4155 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4156 (vect_get_mask_type_for_stmt): New functions, split out from
4157 vect_determine_vectorization_factor.
4159 2018-05-16 Richard Biener <rguenther@suse.de>
4161 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4162 verify the [VEC_]COND_EXPR embedded comparison.
4164 2018-05-15 Martin Sebor <msebor@redhat.com>
4166 PR tree-optimization/85753
4167 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4168 RECORD_TYPE in addition to ARRAY_TYPE.
4170 2018-05-15 Martin Sebor <msebor@redhat.com>
4173 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4175 2018-05-15 Richard Biener <rguenther@suse.de>
4177 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4178 add by_clobber_p one. Change algorithm to collect all defs
4179 representing uses we need to walk and try reducing them to
4180 a single one before failing.
4181 (dse_dom_walker::dse_optimize_stmt): Adjust.
4183 2018-05-13 Mark Wielaard <mark@klomp.org>
4185 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4186 (size_of_loc_descr): Likewise.
4187 (output_loc_operands): Likewise.
4188 (output_loc_operands_raw): Likewise.
4189 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4190 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4191 (hash_loc_operands): Likewise.
4192 (compare_loc_operands): Likewise.
4194 2018-05-14 Mark Wielaard <mark@klomp.org>
4196 * dwarf2out.c (count_index_addrs): New function.
4197 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4199 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4201 PR tree-optimization/83648
4202 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4203 return value as malloc candidate.
4205 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4208 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4209 param as true in call to suggest_attribute.
4211 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4214 -mreadonly-in-sdata.
4216 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4218 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4220 (aarch64_crypto_aesd_fused): Likewise.
4222 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4224 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4225 (movsi_aarch64): Likewise.
4226 (load_pairsi): Likewise.
4227 (load_pairdi): Likewise.
4228 (store_pairsi): Likewise.
4229 (store_pairdi): Likewise.
4230 (load_pairsf): Likewise.
4231 (load_pairdf): Likewise.
4232 (store_pairsf): Likewise.
4233 (store_pairdf): Likewise.
4234 (zero_extend): Likewise.
4235 (trunc): Swap alternatives.
4236 (fcvt_target): Add '?' to prefer w over r.
4238 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4241 * config/i386/i386.md: Disallow non-commutative arithmetics in
4242 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4243 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4244 in the peephole2 before it.
4246 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4248 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4249 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4250 (ix86_handle_option): Handle -mcldemote.
4251 * config.gcc: New header.
4252 * config/i386/cldemoteintrin.h: New file.
4253 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4254 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4256 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4257 OPTION_MASK_ISA_CLDEMOTE.
4258 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4259 (ix86_valid_target_attribute_inner_p): Ditto.
4260 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4261 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4262 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4263 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4264 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4266 * config/i386/i386.opt: Add -mcldemote.
4267 * config/i386/x86intrin.h: New header.
4268 * doc/invoke.texi: Add -mcldemote.
4270 2018-05-14 Richard Biener <rguenther@suse.de>
4272 * doc/match-and-simplify.texi: Adjust :s documentation.
4274 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4276 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4277 intended memcpy size.
4278 (REORDER_45): Likewise.
4280 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4282 * sort.cc: New file.
4283 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4284 * vec.c (qsort_chk): Use gcc_qsort.
4285 * Makefile.in (OBJS-libcommon): Add sort.o.
4286 (build/sort.o): New target. Use it...
4287 (BUILD_RTL): ... here, and...
4288 (build/gencfn-macros): ... here, and...
4289 (build/genmatch): ... here.
4291 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4292 Chung-Ju Wu <jasonwucj@gmail.com>
4294 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4295 * config/nds32/nds32-graywolf.md: New file.
4296 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4299 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4300 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4301 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4302 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4303 * config/nds32/pipelines.md: Include n15 settings.
4305 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4306 Chung-Ju Wu <jasonwucj@gmail.com>
4308 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4309 * config/nds32/nds32-n13.md: New file.
4310 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4311 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4313 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4314 * config/nds32/nds32.md (pipeline_model): Add n13.
4315 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4316 * config/nds32/pipelines.md: Include n13 settings.
4318 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4319 Chung-Ju Wu <jasonwucj@gmail.com>
4321 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4322 * config/nds32/nds32-n10.md: New file.
4323 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4324 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4326 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4327 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4328 * config/nds32/nds32.md (pipeline_model): Add n10.
4329 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4330 * config/nds32/pipelines.md: Include n10 settings.
4332 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4333 Kito Cheng <kito.cheng@gmail.com>
4334 Chung-Ju Wu <jasonwucj@gmail.com>
4336 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4337 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4338 Add enum values for DSP extension instructions.
4339 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4341 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4342 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4344 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4345 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4346 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4347 * config/nds32/nds32-intrinsic.md: Likewise.
4348 * config/nds32/nds32_intrinsic.h: Likewise.
4349 * config/nds32/nds32-md-auxiliary.c: Likewise.
4350 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4351 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4352 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4353 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4354 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4355 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4356 TYPE_DMAC in switch statement.
4357 * config/nds32/nds32.c: New checking and implementation for DSP
4358 extension instructions.
4359 * config/nds32/nds32.h: Likewise.
4360 * config/nds32/nds32.md: Likewise.
4361 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4362 * config/nds32/predicates.md: Implement new predicates for DSP
4365 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4367 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4368 Reformat alternatives and attributes so it is easier to identify
4369 which constraints/attributes go with which instruction.
4370 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4371 (mov<mode>_softfloat32, FMOVE64): Likewise.
4372 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4373 (mov<mode>_softfloat64, FMOVE64): Likewise.
4375 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4377 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4379 (Basic PowerPC Built-in Functions): The new name of the
4380 subsection previously known as "PowerPC Built-in Functions".
4381 (Basic PowerPC Built-in Functions Available on all Configurations):
4383 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4384 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4385 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4386 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4388 2018-05-11 Martin Jambor <mjambor@suse.cz>
4391 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4394 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4397 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4399 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4401 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4402 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4403 (ix86_handle_option): Handle -mwaitpkg.
4404 * config.gcc: New header.
4405 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4406 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4407 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4409 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4410 OPTION_MASK_ISA_WAITPKG.
4411 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4412 (ix86_option_override_internal): Add PTA_WAITPKG.
4413 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4414 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4415 IX86_BUILTIN_TPAUSE.
4416 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4417 __builtin_ia32_umwait and __builtin_ia32_tpause.
4418 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4419 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4420 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4421 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4422 UNSPECV_TPAUSE): New.
4423 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4424 * config/i386/i386.opt: Add -mwaitpkg.
4425 * config/i386/waitpkgintrin.h: New file.
4426 * config/i386/x86intrin.h: New header.
4427 * doc/invoke.texi: Add -mwaitpkg.
4429 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4432 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4434 (cortex-m0): Use armv6s-m isa.
4435 (cortex-m0plus): Likewise.
4436 (cortex-m1): Likewise.
4437 (cortex-m0.small-multiply): Likewise.
4438 (cortex-m0plus.small-multiply): Likewise.
4439 (cortex-m1.small-multiply): Likewise.
4441 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4442 Jakub Jelinek <jakub@redhat.com>
4444 PR tree-optimization/85692
4445 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4446 source permute as well.
4448 2018-05-11 Martin Liska <mliska@suse.cz>
4451 * doc/extend.texi: Document LLVM style format for no_sanitize
4454 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4456 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4457 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4458 (mode_supports_vsx_dform_quad): Likewise.
4459 (mode_supports_vmx_dform): Move these functions to be next to the
4460 other mode_supports functions.
4461 (mode_supports_dq_form): Likewise.
4462 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4463 mode_supports_dq_form.
4464 (reg_offset_addressing_ok_p): Likewise.
4465 (offsettable_ok_by_alignment): Likewise.
4466 (rs6000_legitimate_offset_address_p): Likewise.
4467 (legitimate_lo_sum_address_p): Likewise.
4468 (rs6000_legitimize_address): Likewise.
4469 (rs6000_legitimize_reload_address): Likewise.
4470 (rs6000_secondary_reload_inner): Likewise.
4471 (rs6000_preferred_reload_class): Likewise.
4472 (rs6000_output_move_128bit): Likewise.
4474 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4476 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4477 Generate SImode target register for null target.
4478 <case IX86_BUILTIN_XGETBV>: Ditto.
4479 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4480 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4482 2018-05-10 Carl Love <cel@us.ibm.com>
4484 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4485 dcbtt and dcbtstt if operands[2] is 0.
4487 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4490 * config/i386/sse.md (usadv64qi): New expander.
4492 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4494 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4495 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4496 -maltivec=be support.
4497 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4498 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4499 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4500 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4501 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4502 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4503 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4504 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4505 altivec_vsumsws): Adjust.
4506 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4507 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4508 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4510 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4511 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4512 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4513 (altivec_lve<VI_char>x): Delete expand.
4514 (*altivec_lve<VI_char>x_internal): Rename to...
4515 (altivec_lve<VI_char>x): ... this.
4516 (altivec_lvxl_<mode>): Delete expand.
4517 (*altivec_lvxl_<mode>_internal): Rename to ...
4518 (altivec_lvxl_<mode>): ... this.
4519 (altivec_stvxl_<mode>): Delete expand.
4520 (*altivec_stvxl_<mode>_internal): Rename to ...
4521 (altivec_stvxl_<mode>): ... this.
4522 (altivec_stve<VI_char>x): Delete expand.
4523 (*altivec_stve<VI_char>x_internal): Rename to ...
4524 (altivec_stve<VI_char>x): ... this.
4525 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4526 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4527 reduc_plus_scal_<mode>): Adjust.
4528 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4530 (rs6000_cpu_cpp_builtins): Adjust.
4531 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4532 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4533 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4534 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4535 -maltivec=be support.
4536 (rs6000_split_vec_extract_var): Adjust.
4537 (rs6000_split_v4si_init): Adjust.
4538 (swap_selector_for_mode): Delete.
4539 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4540 altivec_expand_stvex_be): Delete.
4541 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4542 -maltivec=be support.
4543 (rs6000_gimple_fold_builtin): Ditto.
4544 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4546 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4547 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4548 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4549 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4550 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4551 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4552 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4553 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4554 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4555 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4556 anonymous split): Adjust.
4557 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4558 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4560 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4562 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4563 when --with-gxx-include-dir is also specified.
4564 * configure: Regenerate.
4566 2018-05-09 Jim Wilson <jimw@sifive.com>
4569 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4570 * config/riscv/t-withmultilib: New.
4571 * config/riscv/withmultilib.h: New.
4572 * doc/install.texi: Document RISC-V --with-multilib-list support.
4574 2018-05-09 Richard Biener <rguenther@suse.de>
4576 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4578 (vect_bb_vectorization_profitable_p): Adjust. Compute
4579 actual scalar cost using the cost vector and the add_stmt_cost
4582 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4584 PR rtl-optimization/85645
4585 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4586 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4588 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4590 PR rtl-optimization/85645
4591 * shrink-wrap.c (spread_components): Return a boolean saying if
4592 anything was changed.
4593 (try_shrink_wrapping_separate): Iterate spread_components until
4594 nothing changes anymore.
4596 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4598 PR rtl-optimization/85645
4599 * regrename.c (build_def_use): Also kill the chains that include the
4600 destination of a REG_CFA_REGISTER note.
4602 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4604 PR rtl-optimization/85645
4605 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4606 insn that has a REG_CFA_REGISTER note.
4608 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4610 * cfgexpand.c (expand_clobber): New function.
4611 (expand_gimple_stmt_1): Use it.
4612 * tree-vect-stmts.c (vect_clobber_variable): New function,
4614 (vectorizable_simd_clone_call): ...here.
4615 (vectorizable_store): Emit a clobber either side of an
4616 IFN_STORE_LANES sequence.
4617 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4619 2018-05-09 Tom de Vries <tom@codesourcery.com>
4622 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4623 (define_insn "trap_if_false"): Add exit after trap.
4625 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4627 PR rtl-optimization/85638
4628 * bb-reorder.c: Include common/common-target.h.
4629 (create_forwarder_block): New function extracted from...
4630 (fix_up_crossing_landing_pad): ...here. Rename into...
4631 (dw2_fix_up_crossing_landing_pad): ...this.
4632 (sjlj_fix_up_crossing_landing_pad): New function.
4633 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4634 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4635 from both partitions and exit the loop after one iteration.
4637 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4640 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4642 (Basic PowerPC Built-in Functions): The new name of the
4643 subsection previously known as "PowerPC Built-in Functions".
4644 (Basic PowerPC Built-in Functions Available on all Configurations):
4646 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4648 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4650 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4652 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4655 2018-05-08 Jim Wilson <jimw@sifive.com>
4657 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4658 (LD_EMUL_SUFFIX): New.
4659 (LINK_SPEC): Use it.
4661 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4663 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4665 (Basic PowerPC Built-in Functions): The new name of the
4666 subsection previously known as "PowerPC Built-in Functions".
4667 (Basic PowerPC Built-in Functions Available on all Configurations):
4669 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4671 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4673 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4675 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4678 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4681 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4682 after cmpelim optimization.
4684 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4686 * config.gcc: Support "goldmont".
4687 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4688 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4690 * config/i386/i386.c (m_GOLDMONT): Define.
4691 (processor_target_table): Add "goldmont".
4692 (PTA_GOLDMONT): Define.
4693 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4694 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4695 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4696 (fold_builtin_cpu): Add "goldmont".
4697 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4698 (ix86_option_override_internal): Add "goldmont".
4699 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4700 (processor_type): Add PROCESSOR_GOLDMONT.
4701 * config/i386/i386.md: Add CPU "glm".
4702 * config/i386/glm.md: New file.
4703 * config/i386/x86-tune.def: Add m_GOLDMONT.
4704 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4706 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4709 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4711 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4712 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4713 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4716 * config/i386/i386.c (ix86_fold_builtin): Handle
4717 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4720 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4721 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4723 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4726 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4727 (check_arch): Likewise.
4728 (check_fpu): Return the result rather than printing it.
4729 (end arch): Fix operator precedence.
4730 (end cpu): Likewise.
4731 (END): Print the result from check_fpu.
4733 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4734 Alan Hayward <alan.hayward@arm.com>
4735 David Sherwood <david.sherwood@arm.com>
4737 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4738 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4739 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4740 (*fcmuo<mode>_and): New patterns.
4742 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4744 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4745 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4746 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4747 (cmp_op, sve_imm_con): New code attributes.
4748 (SVE_COND_INT_CMP, imm_con): Delete.
4749 (cmp_op): Remove above unspecs from int attribute.
4750 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4752 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4753 comparison-specific unspecs.
4754 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4755 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4756 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4757 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4758 (*vec_fcm<cmp_op><mode>): Rename to...
4759 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4760 (*vec_fcmuo<mode>): Rename to...
4761 (*fcmuo<mode>): ...this and adjust likewise.
4762 (*pred_fcm<cmp_op><mode>): New pattern.
4763 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4764 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4766 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4768 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4769 (aarch64_emit_sve_predicated_cond): New function.
4770 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4771 (aarch64_emit_unspec_cond_or): Replace with...
4772 (aarch64_emit_sve_or_conds): ...this new function. Use
4773 aarch64_emit_sve_ptrue_op for the individual comparisons and
4774 aarch64_emit_binop to OR them together.
4775 (aarch64_emit_inverted_unspec_cond): Replace with...
4776 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4777 aarch64_emit_sve_ptrue_op for the comparison and
4778 aarch64_emit_unop to invert the result.
4779 (aarch64_expand_sve_vec_cmp_float): Update after the above
4780 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4782 2018-05-07 Nathan Sidwell <nathan@acm.org>
4784 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4785 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4786 (Backwards Compatibility): Likewise.
4788 2018-05-07 Luis Machado <luis.machado@linaro.org>
4792 2018-05-07 Luis Machado <luis.machado@linaro.org>
4794 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4795 <prefetch_dynamic_strides>: New const bool field.
4796 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4797 prefetch_dynamic_strides.
4798 (exynosm1_prefetch_tune): Likewise.
4799 (thunderxt88_prefetch_tune): Likewise.
4800 (thunderx_prefetch_tune): Likewise.
4801 (thunderx2t99_prefetch_tune): Likewise.
4802 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4804 (aarch64_override_options_internal): Update to set
4805 PARAM_PREFETCH_DYNAMIC_STRIDES.
4806 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4807 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4808 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4809 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4810 prefetch-dynamic-strides setting.
4812 2018-05-07 Luis Machado <luis.machado@linaro.org>
4814 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4815 <minimum_stride>: New const int field.
4816 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4817 minimum_stride field.
4818 (exynosm1_prefetch_tune): Likewise.
4819 (thunderxt88_prefetch_tune): Likewise.
4820 (thunderx_prefetch_tune): Likewise.
4821 (thunderx2t99_prefetch_tune): Likewise.
4822 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4823 (aarch64_override_options_internal): Update to set
4824 PARAM_PREFETCH_MINIMUM_STRIDE.
4825 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4826 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4827 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4828 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4829 stride is constant and is below the minimum stride threshold.
4831 2018-05-07 Luis Machado <luis.machado@linaro.org>
4833 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4836 2018-05-07 Luis Machado <luis.machado@linaro.org>
4838 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4839 <prefetch_dynamic_strides>: New const bool field.
4840 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4841 prefetch_dynamic_strides.
4842 (exynosm1_prefetch_tune): Likewise.
4843 (thunderxt88_prefetch_tune): Likewise.
4844 (thunderx_prefetch_tune): Likewise.
4845 (thunderx2t99_prefetch_tune): Likewise.
4846 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4848 (aarch64_override_options_internal): Update to set
4849 PARAM_PREFETCH_DYNAMIC_STRIDES.
4850 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4851 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4852 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4853 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4854 prefetch-dynamic-strides setting.
4856 2018-05-07 Luis Machado <luis.machado@linaro.org>
4858 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4859 <minimum_stride>: New const int field.
4860 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4861 minimum_stride field.
4862 (exynosm1_prefetch_tune): Likewise.
4863 (thunderxt88_prefetch_tune): Likewise.
4864 (thunderx_prefetch_tune): Likewise.
4865 (thunderx2t99_prefetch_tune): Likewise.
4866 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4867 (aarch64_override_options_internal): Update to set
4868 PARAM_PREFETCH_MINIMUM_STRIDE.
4869 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4870 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4871 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4872 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4873 stride is constant and is below the minimum stride threshold.
4875 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4878 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4879 the type is addressable. Don't force op into register if it has
4882 2018-05-05 Roland McGrath <mcgrathr@google.com>
4885 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4886 any section for which we don't know a specific type it should have,
4887 regardless of name. Previously this was done only for the exact
4888 names ".init_array", ".fini_array", and ".preinit_array".
4889 (default_elf_asm_named_section): Add comment about
4890 relationship with default_section_type_flags and SECTION_NOTYPE.
4891 (get_section): Don't consider it a type conflict if one side has
4892 SECTION_NOTYPE and the other doesn't, as long as neither has the
4893 SECTION_BSS et al used in the default_section_type_flags logic.
4895 2018-05-05 Tom de Vries <tom@codesourcery.com>
4898 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4899 (workaround_barsyncs): New function.
4900 (nvptx_reorg): Use workaround_barsyncs.
4901 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4902 (define_expand "nvptx_membar_cta"): New define_expand.
4903 (define_insn "*nvptx_membar_cta"): New insn.
4905 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4907 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4908 To improve optimization opportunities.
4909 * builtin-types.def: The new needed builtin types for the above.
4911 2018-05-04 Richard Biener <rguenther@suse.de>
4913 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4914 * gimple-ssa-store-merging.c
4915 (imm_store_chain_info::output_merged_store): Remove redundant create,
4916 release split_store vector contents on failure.
4917 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4918 scalar stmt vector on cache hit.
4920 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4922 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4924 * config.gcc (powerpc-xilinx-eabi*): Remove.
4925 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4927 (fusion_addis_mem_combo_load): Ditto.
4928 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4930 (rs6000_cpu_cpp_builtins): Ditto.
4931 * config/rs6000/rs6000-linux.c
4932 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4933 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4934 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4936 (rs6000_setup_reg_addr_masks): Ditto.
4937 (rs6000_init_hard_regno_mode_ok): Ditto.
4938 (rs6000_option_override_internal): Ditto.
4939 (legitimate_lo_sum_address_p): Ditto.
4940 (rs6000_legitimize_address): Ditto.
4941 (rs6000_legitimize_reload_address): Ditto.
4942 (rs6000_legitimate_address_p): Ditto.
4943 (abi_v4_pass_in_fpr): Ditto.
4944 (setup_incoming_varargs): Ditto.
4945 (rs6000_gimplify_va_arg): Ditto.
4946 (rs6000_split_multireg_move): Ditto.
4947 (rs6000_savres_strategy): Ditto.
4948 (rs6000_emit_prologue_components): Ditto.
4949 (rs6000_emit_epilogue_components): Ditto.
4950 (rs6000_emit_prologue): Ditto.
4951 (rs6000_emit_epilogue): Ditto.
4952 (rs6000_elf_file_end): Ditto.
4953 (rs6000_function_value): Ditto.
4954 (rs6000_libcall_value): Ditto.
4955 * config/rs6000/rs6000.h: Ditto.
4956 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4957 (TARGET_MINMAX): ... this. New.
4958 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4959 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4960 (*movsi_internal1_single): Delete.
4961 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4962 mfpu=, mxilinx-fpu): Delete.
4963 * config/rs6000/singlefp.h: Delete.
4964 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4965 * config/rs6000/t-rs6000: Ditto.
4966 * config/rs6000/t-xilinx: Delete.
4967 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4968 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4969 (VStype_simple): Delete.
4970 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4971 * config/rs6000/xfpu.h: Delete.
4972 * config/rs6000/xfpu.md: Delete.
4973 * config/rs6000/xilinx.h: Delete.
4974 * config/rs6000/xilinx.opt: Delete.
4975 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4976 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4978 2018-05-04 Tom de Vries <tom@codesourcery.com>
4981 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4984 2018-05-04 Richard Biener <rguenther@suse.de>
4987 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4988 (expand_complex_div_wide): Likewise.
4989 (expand_complex_operations_1): Likewise.
4990 (expand_complex_libcall): Preserve EH info of the original stmt.
4991 (tree_lower_complex): Handle removed blocks.
4992 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4993 on complex multiplication and division libcall builtins.
4995 2018-05-04 Richard Biener <rguenther@suse.de>
4998 * fold-const.c (negate_expr_p): Restrict negation of operand
4999 zero of a division to when we know that can happen without
5001 (fold_negate_expr_1): Likewise.
5003 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5006 * real.h (real_nextafter): Declare.
5007 * real.c (real_nextafter): New function.
5008 * fold-const-call.c (fold_const_nextafter): New function.
5009 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5010 CASE_CFN_NEXTTOWARD.
5011 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5012 even when arg1_mode is different from arg0_mode.
5014 2018-05-03 Nathan Sidwell <nathan@acm.org>
5016 * doc/extend.texi (Deprecated Features): Remove
5018 (Backwards Compatibility): Likewise.
5019 * doc/invoke.texi (C++ Language Options): Likewise.
5020 (C++ Dialect Options): Likewise.
5022 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5025 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5026 _mm512_mask_mullox_epi64): New intrinsics.
5028 2018-05-03 Tom de Vries <tom@codesourcery.com>
5031 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5032 dump files): Add offload-tree.
5034 2018-05-03 Richard Biener <rguenther@suse.de>
5036 PR tree-optimization/85615
5037 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5038 to loops not nested in BBs loop father to avoid creating multi-entry
5041 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5043 PR tree-optimization/70291
5044 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5045 arguments. Change return type to tree. Emit libcall as a new
5046 statement rather than replacing existing one when inplace_p is true.
5047 (expand_complex_multiplication_components): New function.
5048 (expand_complex_multiplication): Expand floating-point complex
5049 multiplication using the above.
5050 (expand_complex_division): Rename inner_type parameter to type.
5051 Update expand_complex_libcall call-site.
5052 (expand_complex_operations_1): Update expand_complex_multiplication
5053 and expand_complex_division call-sites.
5055 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5058 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5059 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5060 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5061 the highest significant bit of the shift count mask is clear. In
5062 check whether and[sq]i3 is needed verify that all significant bits
5063 of the shift count other than the highest are set.
5065 2018-05-02 Tom de Vries <tom@codesourcery.com>
5068 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5069 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5070 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5071 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5072 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5073 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5074 __builtin_goacc_parlevel_size.
5076 2018-05-02 Richard Biener <rguenther@suse.de>
5078 PR tree-optimization/85597
5079 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5080 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5083 2018-05-02 Tom de Vries <tom@codesourcery.com>
5086 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5087 dump files): Add ltrans-tree.
5089 2018-05-02 Tom de Vries <tom@codesourcery.com>
5092 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5093 dump files): Add wpa-ipa.
5095 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5097 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5098 powerpc*-*-linux*paired* target.
5099 * config/rs6000/750cl.h: Delete.
5100 * config/rs6000/paired.h: Delete.
5101 * config/rs6000/paired.md: Delete.
5102 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5104 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5105 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5106 comment. Remove paired float support.
5107 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5108 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5110 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5111 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5113 * config/rs6000/rs6000.c: Remove paired float support.
5114 (paired_expand_vector_init, paired_expand_vector_move,
5115 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5116 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5117 paired_expand_builtin, paired_expand_predicate_builtin,
5118 paired_init_builtins): Delete.
5119 * config/rs6000/rs6000.h: Remove paired float support.
5120 * config/rs6000/rs6000.md: Remove paired float support.
5121 (move_from_CR_ov_bit): Delete.
5122 * config/rs6000/rs6000.opt (mpaired): Delete.
5123 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5124 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5126 2018-05-02 Richard Biener <rguenther@suse.de>
5129 * gimplify.c (gimplify_save_expr): When in SSA form allow
5130 SAVE_EXPRs to compute to SSA vars.
5132 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5135 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5136 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5137 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5138 clobber operands[2], instead use a new pseudo. Formatting fixes.
5140 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5142 PR tree-optimization/85586
5143 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5144 exit early for statements in the same group if the accesses are
5147 2018-05-02 Tom de Vries <tom@codesourcery.com>
5150 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5153 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5155 PR tree-optimization/85143
5156 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5158 2018-05-01 Tom de Vries <tom@codesourcery.com>
5161 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5162 not found" error message.
5164 2018-05-01 Tom de Vries <tom@codesourcery.com>
5167 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5168 * vec.c (test_ordered_remove_if): New function.
5169 (vec_c_tests): Call test_ordered_remove_if.
5170 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5171 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5172 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5173 VEC_ORDERED_REMOVE_IF.
5175 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5177 PR tree-optimization/82665
5178 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5179 pointer subtraction where arguments come from a memchr call.
5181 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5183 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5184 --push-state --as-needed and --pop-state instead of --as-needed and
5185 --no-as-needed if ld supports it.
5186 * configure: Regenerated.
5189 * doc/install.texi2html: Replace _002d with - and _002a with * in
5190 generated html files using sed.
5192 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5195 * gcc-rich-location.c (blank_line_before_p): New function.
5196 (use_new_line): New function.
5197 (gcc_rich_location::add_fixit_insert_formatted): New function.
5198 * gcc-rich-location.h
5199 (gcc_rich_location::add_fixit_insert_formatted): New function.
5201 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5203 * selftest.c (assert_streq): Rename "expected" and "actual" to
5204 "val1" and "val2". Extend NULL-handling to cover both inputs
5205 symmetrically, while still requiring both to be non-NULL for a pass.
5206 * selftest.h (assert_streq): Rename "expected" and "actual" to
5208 (ASSERT_EQ): Likewise.
5209 (ASSERT_EQ_AT): Likewise.
5210 (ASSERT_KNOWN_EQ): Likewise.
5211 (ASSERT_KNOWN_EQ_AT): Likewise.
5212 (ASSERT_NE): Likewise.
5213 (ASSERT_MAYBE_NE): Likewise.
5214 (ASSERT_MAYBE_NE_AT): Likewise.
5215 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5216 the assertion to pass.
5217 (ASSERT_STREQ_AT): Likewise.
5219 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5221 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5222 interaction with -pie.
5224 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5226 * selftest.h: Fix alphabetization of per-source-file selftest
5229 2018-04-30 Jason Merrill <jason@redhat.com>
5231 PR c++/61982 - dead stores to destroyed objects.
5232 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5235 2018-04-30 Jason Merrill <jason@redhat.com>
5237 * tree.c (build_clobber): New.
5238 * tree.h: Declare it.
5239 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5241 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5243 * diagnostic-show-locus.c (layout::layout): Update for
5244 location_get_source_line returning a char_span.
5245 (struct char_span): Move to input.h.
5246 (struct correction): Update for fields in char_span becoming
5248 (struct source_line): Update for location_get_source_line
5249 returning a char_span.
5250 (layout::print_line): Likewise.
5251 * edit-context.c (edited_file::print_content): Likewise.
5252 (edited_file::print_diff_hunk): Likewise.
5253 (edited_file::print_run_of_changed_lines): Likewise.
5254 (edited_file::get_num_lines): Likewise.
5255 (edited_line::edited_line): Likewise.
5256 * final.c (asm_show_source): Likewise.
5257 * input.c (location_get_source_line): Convert return type
5258 from const char * to char_span, losing the final "line_len"
5260 (dump_location_info): Update for the above.
5261 (get_substring_ranges_for_loc): Likewise. Use a char_span
5262 when handling the literal within the line.
5263 (test_reading_source_line): Update for location_get_source_line
5264 returning a char_span.
5265 * input.h (class char_span): Move here from
5266 diagnostic-show-locus.c, converting from a struct to a class.
5267 Make data members private.
5268 (char_span::operator bool): New.
5269 (char_span::length): New.
5270 (char_span::get_buffer): New.
5271 (char_span::operator[]): New.
5272 (char_span::subspan): Make const.
5273 (char_span::xstrdup): New.
5274 (location_get_source_line): Convert return type from const char *
5275 to char_span, losing the final "line_size" param.
5277 2018-04-30 Jan Hubicka <jh@suse.cz>
5279 * lto-wrapper.c (ltrans_priorities): New static var.
5280 (cmp_priority): New.
5281 (run_gcc): Read priorities and if doing parallel build order
5282 the Makefile by them.
5284 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5286 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5288 2018-04-30 Richard Biener <rguenther@suse.de>
5290 * tree-cfg.c (verify_address): Remove base argument, add
5291 flag whether to check TREE_ADDRESSABLE and do that.
5292 (verify_expr): Remove.
5293 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5294 (verify_gimple_assign_single): Likewise.
5295 (verify_gimple_switch): Likewise.
5296 (verify_expr_location_1): Dereference tp once. Add (disabled)
5297 piece from verify_expr.
5298 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5300 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5302 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5304 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5306 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5307 (small_data_pattern): Likewise.
5308 (arc_rewrite_small_data): Likewise.
5309 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5310 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5311 (get_symbol_alignment): New function.
5312 (legitimate_small_data_address_p): Likewise.
5313 (legitimate_scaled_address): Update, call
5314 legitimate_small_data_address_p.
5315 (output_sdata): New static variable.
5316 (arc_print_operand): Update how we handle small data operands.
5317 (arc_print_operand_address): Likewise.
5318 (arc_legitimate_address_p): Update, use
5319 legitimate_small_data_address_p.
5320 (arc_rewrite_small_data_p): Remove.
5321 (arc_rewrite_small_data_1): Likewise.
5322 (arc_rewrite_small_data): Likewise.
5323 (small_data_pattern): Likewise.
5324 (compact_sda_memory_operand): Update to use
5325 legitimate_small_data_address_p and get_symbol_alignment.
5326 (prepare_move_operands): Don't rewite sdata pattern.
5327 (prepare_extend_operands): Remove.
5328 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5330 (zero_extendqisi2): Likewise.
5331 (zero_extendhisi2): Likewise.
5332 (extendqihi2): Likewise.
5333 (extendqisi2): Likewise.
5334 (extendhisi2): Likewise.
5338 * config/arc/constraints.md (Usd): Change it to memory constraint.
5340 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5342 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5343 as source of std instructions.
5344 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5345 allow 6-bit constants as source for store instructions.
5346 (movdi_insn): Update instruction pattern to allow 6-bit constants
5347 as source for store instructions.
5349 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5351 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5353 2018-04-30 Nathan Sidwell <nathan@acm.org>
5354 Sandra Loosemore <sandra@codesourcery.com>
5356 * dumpfile.c (dump_open): Allow '-' for stdout.
5357 * doc/invoke.texi (Developer Options): Document dump filename
5358 determination early. Document stdin/stdout selection.
5360 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5362 Microblaze Target: PIC data text relative
5364 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5365 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5367 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5368 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5369 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5370 New addressing mode for data-text relative position indepenedent code.
5371 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5372 'ADDRESS_SYMBOLIC_TXT_REL'.
5373 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5374 (microblaze_legitimate_pic_operand): Exclude function calls from
5375 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5376 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5378 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5379 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5380 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5381 for 'address + offset'.
5382 (microblaze_expand_prologue): Add new function prologue call for
5384 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5385 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5386 table in case of TARGET_PIC_DATA_TEXT_REL.
5387 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5388 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5389 Add new macros 'UNSPEC_TEXT',
5390 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5391 + exclude function calls from 'UNSPEC_PLT' in case of data text
5393 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5394 new target hook for generating address diff vector tables in case of
5396 * doc/tm.texi : Regenerate.
5397 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5398 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5399 of addr diff vector generation.
5400 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5401 target hook definition.
5402 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5403 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5404 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5405 option for data text relative.
5407 2018-04-30 Richard Biener <rguenther@suse.de>
5409 * tree-chrec.h (evolution_function_is_constant_p): Remove
5411 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5413 2018-04-30 Richard Biener <rguenther@suse.de>
5416 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5418 2018-04-30 Richard Biener <rguenther@suse.de>
5420 PR tree-optimization/28364
5421 PR tree-optimization/85275
5422 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5423 copying first exit test.
5425 2018-04-28 Mark Wielaard <mark@klomp.org>
5427 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5429 (dwarf_AT): Handle DW_AT_addr_base.
5430 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5432 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5435 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5436 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5437 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5438 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5440 2018-04-28 Richard Biener <rguenther@suse.de>
5442 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5443 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5444 to reflect use. Only add interesting stmts.
5446 2018-04-27 Martin Jambor <mjambor@suse.cz>
5449 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5450 the jump function allows for passing through aggregate values.
5452 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5454 * input.h (in_system_header_at): Convert from macro to inline
5456 (from_macro_expansion_at): Likewise.
5457 (from_macro_definition_at): Likewise.
5459 2018-04-27 Jeff Law <law@redhat.com>
5461 * config.gcc: Mark tile* targets as deprecated/obsolete.
5463 2018-04-27 Richard Biener <rguenther@suse.de>
5465 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5468 2018-04-27 Richard Biener <rguenther@suse.de>
5470 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5472 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5474 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5475 with Yd constraint. Set "preferred_for_speed" attribute from
5476 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5478 (*movdi_internal): Ditto.
5479 (movti_interunit splitters): Remove
5480 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5481 (movdi_interunit splitters): Ditto.
5482 * config/i386/constraints.md (Ye): Remove.
5483 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5485 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5488 * config/aarch64/constraints.md (Usg): Limit to 31.
5491 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5493 PR tree-optimization/85529
5494 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5495 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5496 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5497 zero extension or masking of the MSB bit.
5498 (optimize_range_tests): Add FIRST_BB argument, pass it through
5499 to optimize_range_tests_var_bound.
5500 (maybe_optimize_range_tests, reassociate_bb): Adjust
5501 optimize_range_tests callers.
5503 2018-04-26 Richard Biener <rguenther@suse.de>
5504 Jakub Jelinek <jakub@redhat.com>
5506 * cgraph.h (symbol_table): Just declare debug method here.
5507 * symtab.c (symbol_table::debug): Define.
5509 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5511 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5513 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5515 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5516 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5517 (*movdi_internal): Substitute Yi and Yj constraint with x
5518 and Ym and Yn constraint with y constraint. Update "isa"
5519 attribute and set "preferred_for_speed" attribute from
5520 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5521 (*movsi_internal): Ditto.
5522 (*movdf_internal): Ditto.
5523 (*movsf_internal): Ditto.
5524 (*zero_extendsidi2): Ditto.
5525 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5526 (sse2_loadld): Ditto.
5527 (*vec_extract<ssevecmodelower>_0): Ditto.
5528 (*vec_extractv4si_0_zext_sse4): Ditto.
5529 (vec_concatv2di): Ditto.
5530 (*vec_dup<mode>): Ditto.
5531 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5532 * config/i386/constraints.md (Yi): Remove.
5537 2018-04-26 Nathan Sidwell <nathan@acm.org>
5539 * dumpfile.c (dump_open): New.
5540 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5541 (dump_finish): Detect stdio/stderr by value not name.
5543 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5545 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5547 2018-04-26 Tom de Vries <tom@codesourcery.com>
5550 * config/nvptx/nvptx.c (verify_neutering_jumps)
5551 (verify_neutering_labels): New function
5552 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5554 2018-04-26 Tom de Vries <tom@codesourcery.com>
5557 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5558 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5561 2018-04-26 Richard Biener <rguenther@suse.de>
5562 Tom de Vries <tom@codesourcery.com>
5565 * lto-streamer-out.c (output_function): Fixup loops if required to match
5566 discovery done in the reader.
5568 2018-04-26 Richard Biener <rguenther@suse.de>
5570 PR tree-optimization/85116
5571 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5572 have a loop exit from the single latch predecessor. Remove
5573 case of header with just condition.
5574 (ch_base::copy_headers): Exclude infinite loops from any
5576 (pass_ch::execute): Record exits.
5578 2018-04-26 Richard Biener <rguenther@suse.de>
5580 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5581 prologue cost vector and pass it to vect_get_load_cost.
5582 (vect_get_peeling_costs_all_drs): Likewise.
5583 (vect_peeling_hash_get_lowest_cost): Likewise.
5584 (vect_enhance_data_refs_alignment): Likewise.
5586 2018-04-26 Richard Biener <rguenther@suse.de>
5589 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5590 checking of integer<->pointer conversions.
5591 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5592 sign-/zero-extending pointer types.
5593 (expand_omp_for_static_chunk): Likewise.
5595 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5596 Jean Lee <xiaoyur347@gmail.com>
5598 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5599 (TARGET_ASAN_SHADOW_OFFSET): Define.
5600 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5601 true for -fsanitize=address.
5603 2018-04-25 Mark Wielaard <mark@klomp.org>
5605 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5608 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5610 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5611 than "alu", remove explicit "memory" and "imm_disp" attributes.
5612 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5615 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5616 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5617 gen_lowpart_no_emit.
5619 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5622 * config/i386/i386.c (ix86_expand_builtin): Change memory
5623 operand to XI, extend p0 to Pmode.
5624 * config/i386/i386.md: Change unspec volatile and operand
5625 1 mode to XI, change operand 0 mode to P.
5627 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5629 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5630 GET_MODE_MASK before any checking.
5631 (nds32_can_use_bset_p): Likewise.
5632 (nds32_can_use_btgl_p): Likewise.
5634 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5636 * config/nds32/nds32-doubleword.md: New define_split pattern for
5637 illegal register number.
5639 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5641 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5643 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5645 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5647 2018-04-25 Richard Biener <rguenther@suse.de>
5649 * lto-streamer.h (LTO_major_version): Bump to 8.
5651 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5653 * BASE-VER: Set to 9.0.0.
5655 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5657 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5658 in __abskf2 and __powikf2.
5660 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5663 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5664 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5665 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5666 Use the above on operand 2. Reindent.
5667 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5669 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5672 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5674 * config/i386/i386.opt (mcet): Removed.
5675 * doc/install.texi: Remove -mcet documentation.
5676 * doc/invoke.texi: Likewise.
5678 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5681 * doc/install.texi: Remove -mcet from bootstrap-cet.
5683 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5686 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5687 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5691 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5692 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5693 containing a CONST_VECTOR.
5695 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5697 * doc/install.texi: Update newlib dependency for nvptx.
5699 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5702 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5703 instead of INTVAL when shifting x left.
5705 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5707 PR tree-optimization/85478
5708 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5709 vect_grouped_store_supported for single element vectors.
5711 2018-04-24 Richard Biener <rguenther@suse.de>
5714 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5715 load cost increase to the case of non-constant step.
5717 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5720 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5721 destination if any_malformed_asm.
5723 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5726 * expr.c (store_field): In the bitfield case, if the value comes from
5727 a function call and is returned in registers by means of a PARALLEL,
5728 do not change the mode of the temporary unless BLKmode and VOIDmode.
5730 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5732 PR rtl-optimization/85423
5733 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5734 dependencies to debug insns when the previous insn is non-debug.
5736 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5738 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5739 enums into a single definition.
5740 (fls): Fix predicates and printing.
5743 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5745 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5746 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5747 and short u6 immediate.
5748 (check_if_valid_sleep_operand): Remove.
5749 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5751 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5753 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5754 flag_always_save_lp condition.
5755 * config/nds32/nds32.opt (malways-save-lp): New option.
5757 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5759 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5760 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5761 * config/nds32/nds32.h
5762 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5763 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5765 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5767 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5768 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5770 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5771 Chung-Ju Wu <jasonwucj@gmail.com>
5773 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5775 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5776 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5778 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5780 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5782 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5784 * config/nds32/nds32-protos.h (nds32_data_alignment,
5785 nds32_local_alignment): Declare.
5786 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5787 nds32_local_alignment): New functions.
5788 (TARGET_CONSTANT_ALIGNMENT): Define.
5789 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5791 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5793 * config/nds32/nds32.c
5794 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5795 (TARGET_MODES_TIEABLE_P): Likewise.
5797 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5799 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5802 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5803 Chung-Ju Wu <jasonwucj@gmail.com>
5805 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5806 for unaligned access.
5807 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5809 * config/nds32/nds32-intrinsic.md: Likewise.
5810 * config/nds32/nds32_intrinsic.h: Likewise.
5811 * config/nds32/nds32.h (nds32_builtins): Likewise.
5812 * config/nds32/nds32.opt (munaligned-access): New option.
5813 * config/nds32/nds32.c (nds32_asm_file_start): Display
5814 flag_unaligned_access status.
5816 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5818 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5819 -mno-relax is present.
5820 * config/riscv/linux.h (LINK_SPEC): Ditto.
5822 2018-04-20 Martin Sebor <msebor@redhat.com>
5825 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5827 (gimple_fold_builtin_stxcpy_chk): Same.
5828 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5830 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5833 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5834 __powikf2 when long double is IEEE 128-bit.
5836 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5838 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5839 step to make sure stack always aligned.
5841 2018-04-20 Carl Love <cel@us.ibm.com>
5844 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5845 size check for arg0.
5847 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5848 Tom de Vries <tom@codesourcery.com>
5851 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5852 Emit insns for calls too.
5853 (nvptx_find_par): Always look for worker-level predecessor insn.
5854 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5856 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5857 (nvptx_process_pars): Propagate frames for calls.
5859 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5862 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5864 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5865 (ix86_handle_option): Don't handle OPT_mibt.
5866 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5868 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5870 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5871 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5872 (ix86_target_macros): Define __CET__ with flag_cf_protection
5873 for -fcf-protection.
5874 * config/i386/i386.c (isa2_opts): Remove -mibt.
5875 * config/i386/i386.h (TARGET_IBT): Removed.
5876 (TARGET_IBT_P): Likewise.
5877 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5878 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5879 * config/i386/i386.opt (mcet): Update help message.
5882 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5883 -mcet as an alias for -mshstk.
5885 2018-04-20 Richard Biener <rguenther@suse.de>
5888 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5889 complexity by forcing a single use of the multiply operand.
5891 2018-04-20 Martin Jambor <mjambor@suse.cz>
5894 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5895 recursion dependency to only apply to non-clones.
5897 2018-04-20 Martin Jambor <mjambor@suse.cz>
5900 * ipa-cp.c (create_specialized_node): Check that clones of
5901 self-recursive edges exist during IPA-CP.
5903 2018-04-19 Toon Moene <toon@moene.org>
5905 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5908 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5910 PR tree-optimization/85467
5911 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5912 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5913 VECTOR_CST element to type.
5915 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5918 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5919 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5920 (builtin_longjmp): Likewise.
5921 (save_stack_nonlocal): New pattern.
5922 (restore_stack_nonlocal): Likewise.
5924 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5927 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5928 Replace ASM_OUTPUT_LABEL with fprintf.
5930 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5933 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5934 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5935 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5936 define __IBT__ and __SHSTK__ for -fcf-protection.
5937 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5939 (ix86_trampoline_init): Likewise.
5940 (x86_output_mi_thunk): Likewise.
5941 (ix86_notrack_prefixed_insn_p): Likewise.
5942 (ix86_option_override_internal): Don't disallow -fcf-protection.
5943 * config/i386/i386.md (rdssp<mode>): Also enable for
5945 (incssp<mode>): Likewise.
5946 (nop_endbr): Likewise.
5947 * config/i386/i386.opt (mcet): Change help message to built-in
5951 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5952 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5953 enable CET built-in functions.
5955 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5957 * common/config/i386/i386-common.c
5958 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5959 OPTION_MASK_ISA_MOVDIRI_UNSET,
5960 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5961 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5962 * config.gcc (movdirintrin.h): New header.
5963 * config/i386/cpuid.h (bit_MOVDIRI,
5964 bit_MOVDIR64B): New bits.
5965 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5967 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5968 (VOID, PVOID, PCVOID)): New function types.
5969 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5970 __builtin_ia32_directstoreu_u64,
5971 __builtin_ia32_movdir64b): New builtins.
5972 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5973 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5975 (ix86_valid_target_attribute_inner_p): Ditto.
5976 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5977 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5978 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5979 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5980 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5981 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5982 (movdiri<mode>, movdir64b_<mode>): New.
5983 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5984 * config/i386/immintrin.h: Include movdirintrin.h.
5985 * config/i386/movdirintrin.h: New file.
5986 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5988 2018-04-19 Richard Biener <rguenther@suse.de>
5991 * cfg.c (clear_bb_flags): When loop state says we have
5992 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5994 2018-04-19 Richard Biener <rguenther@suse.de>
5996 PR tree-optimization/84737
5997 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5998 copying restrict info.
5999 (vect_setup_realignment): Use it.
6000 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6001 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6002 the first DR to all generated stores.
6003 (vectorizable_load): Likewise for loads.
6005 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6007 PR tree-optimization/85446
6008 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6009 the integral and pointer types to have the same precision.
6011 * doc/install.texi: Document --disable-cet being the default and
6014 2018-04-18 Martin Liska <mliska@suse.cz>
6016 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6019 2018-04-18 Martin Liska <mliska@suse.cz>
6022 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6025 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6026 arguments if they are comparable.
6028 2018-04-18 Martin Liska <mliska@suse.cz>
6031 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6034 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6037 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6040 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6041 ENDBR after calling __morestack.
6043 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6046 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6047 by using gcc_base_ver to generate a gcc_driver_version, and use
6048 it when generating GCC_DRIVER_NAME.
6049 * configure: Regenerate.
6051 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6054 * config.gcc: Obsolete powerpc*-*-*spe*.
6056 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6059 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6060 (stabstr_D): Change type of unum from unsigned int to
6061 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6064 2018-04-17 Jim Wilson <jimw@sifive.com>
6067 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6068 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6069 Set arg_pointer_offset after using pretend_args_size.
6071 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6073 PR rtl-optimization/85431
6074 * dse.c (record_store): Ignore zero width stores.
6077 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6078 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6079 __builtin_stack_restore rather than after it.
6080 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6081 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6082 argument instead of virtual_dynamic_stack_rtx.
6084 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6086 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6088 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6089 Add note to error message to explain internal mapping of overloaded
6090 built-in function name to non-overloaded built-in function name.
6091 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6094 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6097 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6098 where the inputs overlap with the output.
6100 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6102 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6103 (=v, v) alternative and explicit "memory" attribute.
6104 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6105 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6107 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6108 "sselog1" type instead of "sselog".
6109 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6110 "sselog". Remove explicit "memory" attribute.
6111 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6112 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6114 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6115 "sselog1" type instead of "sselog". Remove explicit "memory"
6117 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6118 use "sselog1" type instead of "sselog". Remove explicit "memory"
6120 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6121 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6123 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6124 "sselog1" type instead of "sselog". Remove explicit "memory"
6126 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6127 use "sselog1" type instead of "sselog". Remove explicit "memory"
6131 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6134 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6137 2018-04-17 Martin Jambor <mjambor@suse.cz>
6140 * ipa-cp.c (create_specialized_node): Call
6141 expand_all_artificial_thunks if necessary.
6143 2018-04-17 Martin Liska <mliska@suse.cz>
6146 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6147 in message, remote space in between '_G' and '('.
6149 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6152 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6153 avx512f_vmcmp<mode>3<round_saeonly_name>,
6154 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6155 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6156 avx512f_rndscale<mode><round_saeonly_name>,
6157 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6158 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6159 Use %<iptr>2 instead of %2 for -masm=intel.
6160 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6161 avx512f_vcvttss2usi<round_saeonly_name>,
6162 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6164 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6165 avx512f_vcvttsd2usi<round_saeonly_name>,
6166 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6167 Use %q1 instead of %1 for -masm=intel.
6168 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6169 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6170 of %3 for -masm=intel.
6171 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6173 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6175 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6177 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6178 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6180 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6182 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6183 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6184 %g1 and one with %0 and %1.
6185 (avx512er_vmrcp28<mode><round_saeonly_name>,
6186 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6188 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6189 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6190 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6191 of %0 and %{%4%} for -masm=intel.
6192 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6193 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6194 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6195 order of %0 and %{%5%}%{z%} for -masm=intel.
6197 2018-04-17 Jan Hubicka <jh@suse.cz>
6200 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6202 2018-04-17 Martin Liska <mliska@suse.cz>
6205 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6206 for target_clone error message. Make default implementation
6207 clone to be a local declaration.
6208 (separate_attrs): Add new argument and check for an empty
6210 (expand_target_clones): Handle it.
6211 (ipa_target_clone): Make redirection just for target_clones
6214 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6215 Tom de Vries <tom@codesourcery.com>
6218 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6219 tiled basic blocks without omp continue statements.
6221 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6224 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6225 vec_extract expression as having side effects to make sure it gets
6228 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6231 * config/i386/i386.c (get_builtin_code_for_version): Check
6234 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6237 * gcc/config.gcc: Support "skylake".
6238 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6240 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6241 (processor_target_table): Add "skylake".
6242 (ix86_option_override_internal): Add "skylake".
6243 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6244 PROCESSOR_CANNONLAKE.
6245 (get_builtin_code_for_version): Fix priority for
6246 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6247 PROCESSOR_SKYLAKE-AVX512.
6248 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6249 (processor_type): Add PROCESSOR_SKYLAKE.
6251 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6252 Jason Merrill <jason@redhat.com>
6255 * convert.c (convert_to_integer_1): Use direct recursion for
6256 enumeral types and types with a precision less than the number
6257 of bits in their mode.
6259 2018-04-16 Julia Koval <julia.koval@intel.com>
6262 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6263 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6265 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6268 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6269 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6270 and -mno-direct-move.
6272 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6275 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6276 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6277 Ensure negative shifts result in {0}.
6279 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6281 PR rtl-optimization/79916
6282 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6283 regs (if any) to define how to gnerate SD moves when LRA is in
6286 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6288 PR rtl-optimization/85393
6289 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6290 * except.c (expand_dw2_landing_pad_for_region): Make static.
6291 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6292 a label and unconditional jump to old_bb, rather than
6293 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6296 PR rtl-optimization/85376
6297 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6298 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6299 instead of a specific value.
6301 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6302 Bin Cheng <bin.cheng@arm.com>
6304 PR tree-optimization/82965
6305 PR tree-optimization/83991
6306 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6307 by_profile_only parameter.
6308 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6309 information if the loop was predicted to iterate too many times.
6310 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6312 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6315 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6318 2018-04-13 Martin Liska <mliska@suse.cz>
6319 Jakub Jelinek <jakub@redhat.com>
6322 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6323 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6324 * builtins.c (expand_builtin_memory_copy_args): Use
6325 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6326 handle dest_addr == pc_rtx.
6328 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6331 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6332 asked to not generate direct moves.
6333 (fix_trunc<mode>si2_stfiwx): Similar.
6334 (fix_trunc<mode>si2_internal): Similar.
6336 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6339 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6340 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6341 lookup if dest in some wider mode is known to be const0_rtx and
6342 if so, record permanent equivalence for it to be ZERO_EXTEND of
6343 the narrower mode destination.
6345 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6347 * lto-streamer-out.c (output_function): Revert 259346.
6348 * omp-expand.c (expand_oacc_for): Likewise.
6350 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6352 PR rtl-optimization/85354
6353 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6354 * sel-sched.c (sel_global_init): ... here.
6356 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6359 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6360 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6361 mode for PE-COFF targets.
6362 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6363 (i386_pe_asm_lto_end): Likewise.
6364 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6365 (TARGET_ASM_LTO_END): Likewise.
6366 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6367 (i386_pe_asm_lto_start): New function.
6368 (i386_pe_asm_lto_end): Likewise.
6370 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6371 Richard Biener <rguenther@suse.de>
6374 * lto-streamer-out.c (output_function): Fix CFG loop state before
6376 * omp-expand.c (expand_oacc_for): Handle calls to internal
6377 functions like regular functions.
6379 2018-04-12 Richard Biener <rguenther@suse.de>
6382 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6383 for the early LTO debug to properly generate references to it
6384 during DIE emission. Do not re-use that for the skeleton for
6386 (dwarf2out_early_finish): Likewise.
6388 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6391 * config/i386/sse.md
6392 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6393 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6394 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6395 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6396 and output is a reg, avoid creating invalid lowpart subreg, but
6397 instead split into a 512-bit move. Don't split if not AVX512VL,
6398 input is xmm16+ reg and output is a mem.
6399 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6400 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6401 xmm16+ reg and output is a mem.
6403 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6405 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6406 also for flag_dwarf2_cfi_asm.
6408 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6410 PR rtl-optimization/85342
6411 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6412 a bool scalar var inside of the loop instead. Don't try to update
6413 recog_data.operand after failed apply_change_group.
6415 2018-04-12 Tom de Vries <tom@codesourcery.com>
6418 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6419 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6420 array with flexible array member as array without given dimension.
6421 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6422 argument for undefined param to true.
6424 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6427 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6428 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6429 from PowerPC section.
6430 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6431 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6432 help text that is too long.
6433 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6434 help text that is too long.
6435 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6436 help text that is too long.
6438 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6440 * config/alpha/alpha.md (stack_probe_internal): Rename
6441 from "probe_stack". Update all callers.
6443 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6445 PR rtl-optimization/84566
6446 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6447 sched_macro_fuse_insns.
6449 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6452 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6453 (compute_block_dependences): ... from here.
6455 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6457 PR tree-optimization/85331
6458 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6459 from int to HOST_WIDE_INT.
6461 2018-04-11 Martin Jambor <mjambor@suse.cz>
6464 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6465 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6466 not the same as the source val.
6467 (cgraph_edge_brings_value_p): New parameter.
6468 (gather_edges_for_value): Pass destination value to
6469 cgraph_edge_brings_value_p.
6470 (perhaps_add_new_callers): Likewise.
6471 (get_info_about_necessary_edges): Likewise and exclude values brought
6472 only by self-recursive edges.
6473 (create_specialized_node): Redirect only clones of self-calling edges.
6474 (+self_recursive_pass_through_p): New function.
6475 (find_more_scalar_values_for_callers_subset): Use it.
6476 (find_aggregate_values_for_callers_subset): Likewise.
6477 (known_aggs_to_agg_replacement_list): Removed.
6478 (decide_whether_version_node): Re-calculate known constants for all
6479 remaining context clones.
6481 2018-04-11 Richard Biener <rguenther@suse.de>
6484 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6485 from early DWARF output.
6486 (dwarf2out_early_finish): Output line info unconditionally into
6487 early DWARF and add reference to it.
6489 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6492 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6493 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6494 other than V2DFmode using iptr mode attribute.
6495 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6497 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6499 PR rtl-optimization/84659
6500 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6502 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6505 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6507 (output_loc_list): Pass address of a dummy size variable even in the
6508 locview handling loop.
6509 (index_location_lists): Add comment on why skip_loc_list_entry can't
6512 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6515 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6518 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6521 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6522 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6523 and -mstring-compare-inline-limit.
6525 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6528 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6529 for stack clash protection in a register whenever we need it to be in
6532 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6534 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6535 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6537 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6540 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6542 (mlong-double-): Ditto.
6543 * config/rs6000/sysv4.opt (msdata=): Ditto.
6544 (mtls-size=): Ditto.
6546 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6548 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6549 erroneous entries for
6550 "vector int vec_ldl (int, long int *)", and
6551 "vector unsigned int vec_ldl (int, unsigned long int *)".
6552 Add comments and entries for
6553 "vector bool char vec_ldl (int, bool char *)",
6554 "vector bool short vec_ldl (int, bool short *)",
6555 "vector bool int vec_ldl (int, bool int *)",
6556 "vector bool long long vec_ldl (int, bool long long *)",
6557 "vector pixel vec_ldl (int, pixel *)",
6558 "vector long long vec_ldl (int, long long *)",
6559 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6560 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6561 type tree bool_long_long_type_node and correct definition of
6562 bool_V2DI_type_node to make reference to this new type tree.
6563 (rs6000_mangle_type): Replace erroneous reference to
6564 bool_long_type_node with bool_long_long_type_node.
6565 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6566 comments to emphasize sign distinctions for char and int types and
6567 replace RS6000_BTI_bool_long constant with
6568 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6569 use of RS6000_BTI_pixel.
6570 (bool_long_type_node): Remove this macro definition.
6571 (bool_long_long_type_node): New macro definition
6573 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6575 PR rtl-optimization/85300
6576 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6577 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6578 simplify_unary_operation fails.
6580 2018-04-10 Martin Liska <mliska@suse.cz>
6582 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6583 cgraph_edge and ipa_ref.
6585 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6589 * config/i386/sse.md
6590 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6591 computation of the VEC_MERGE selector from mask.
6592 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6593 Fix decoding of the VEC_MERGE selector into mask.
6595 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6597 PR tree-optimization/85286
6598 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6600 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6602 * final.c (final_1): Set insn_last_address as well as
6603 insn_current_address.
6605 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6608 * explow.c (emit_stack_probe): Call validize_mem on memory location
6609 before passing it to gen_probe_stack. Create address operand and
6610 legitimize it for the probe_stack_address case.
6612 2018-04-09 Jan Hubicka <jh@suse.cz>
6615 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6616 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6617 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6619 (free_lang_data): Rebuild type inheritance graph.
6621 2018-04-09 Martin Sebor <msebor@redhat.com>
6623 * invoke.texi (-finline-small-functions): Mention other optimization
6625 (-findirect-inlining, -fpartial-inlining): Same.
6626 (-finline-functions-called-once): Same.
6627 (-freorder-blocks-and-partition): Same.
6629 2018-04-09 Jan Hubicka <jh@suse.cz>
6632 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6633 jumps; choose last target that matches the criteria (i.e.
6634 no partition changes for non-crossing jumps).
6635 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6636 support for redirecting crossing jumps to non-crossing.
6638 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6640 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6641 also for naked functions.
6643 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6645 * config/arc/arc.md (add_shift): New pattern.
6646 (add_shift2): Likewise.
6647 (sub_shift): Likewise.
6648 (sub_shift_cmp0_noout): Likewise.
6649 (compare_si_ashiftsi): Likewise.
6650 (xbfu_cmp0_noout): New combine pattern.
6651 (xbfu_cmp0"): Likewise.
6652 (movsi_set_cc_insn): Place the predicable variant first.
6653 (commutative_binary_cmp0_noout): Remove clobber.
6654 (commutative_binary_cmp0): New pattern.
6655 (noncommutative_binary_cmp0): Likewise.
6656 (noncommutative_binary_cmp0_noout): Likewise.
6657 (noncommutative_binary_comparison_result_used): Removed.
6658 (rsub_cmp0): New pattern.
6659 (rsub_cmp0_noout): Likewise.
6660 (extzvsi): Changed, keep only meaningful variants.
6661 (SQH, SEZ): New iterators.
6662 (SQH_postfix): New mode attribute.
6663 (SEZ_prefix): New code attribute.
6664 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6665 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6666 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6668 (noncommutative_operator): Check the availability of barrel
6671 2018-04-09 Richard Biener <rguenther@suse.de>
6673 PR tree-optimization/85284
6674 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6675 Only use the niter constraining form of simple_iv when the exit
6678 2018-04-09 Tom de Vries <tom@codesourcery.com>
6681 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6682 (define_expand "*memory_barrier"): New define_expand.
6683 (define_insn "memory_barrier"): New insn.
6685 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6687 PR rtl-optimization/80463
6688 PR rtl-optimization/83972
6689 PR rtl-optimization/83480
6691 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6692 correct producer for the insn.
6693 (tidy_control_flow): Fixup seqnos in case of debug insns.
6695 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6697 PR rtl-optimization/83913
6699 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6700 different sched-times when merging exprs.
6702 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6704 PR rtl-optimization/83962
6706 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6707 tidy_fallthru_edge and tidy_control_flow.
6709 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6711 PR rtl-optimization/83530
6713 * sel-sched.c (force_next_insn): New global variable.
6714 (remove_insn_for_debug): When force_next_insn is true, also leave only
6715 next insn in the ready list.
6716 (sel_sched_region): When the region wasn't scheduled, make another pass
6717 over it with force_next_insn set to 1.
6719 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6721 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6723 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6724 for interrupt control.
6725 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6726 functions for interrupt control.
6727 * config/nds32/nds32-intrinsic.md: Likewise.
6728 * config/nds32/nds32_intrinsic.h: Likewise.
6729 * config/nds32/nds32.h (nds32_builtins): Likewise.
6731 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6733 * config/nds32/nds32.c (nds32_init_machine_status,
6734 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6735 strict_aligned_p field.
6736 (nds32_expand_to_rtl_hook): New function.
6737 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6738 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6740 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6741 Chung-Ju Wu <jasonwucj@gmail.com>
6743 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6744 * config/nds32/nds32-n7.md: New file.
6745 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6746 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6748 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6749 * config/nds32/nds32.md (pipeline_model): Add n7.
6750 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6751 * config/nds32/pipelines.md: Include n7 settings.
6753 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6754 Chung-Ju Wu <jasonwucj@gmail.com>
6756 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6757 * config/nds32/nds32-e8.md: New file.
6758 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6759 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6761 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6762 * config/nds32/nds32.md (pipeline_model): Add e8.
6763 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6764 * config/nds32/pipelines.md: Include e8 settings.
6766 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6767 Chung-Ju Wu <jasonwucj@gmail.com>
6769 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6770 * config/nds32/nds32-n8.md: New file.
6771 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6772 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6774 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6775 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6776 * config/nds32/nds32.md (pipeline_model): Add n8.
6777 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6778 * config/nds32/pipelines.md: Include n8 settings.
6780 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6781 Chung-Ju Wu <jasonwucj@gmail.com>
6783 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6784 * config/nds32/nds32-n9-2r1w.md: New file.
6785 * config/nds32/nds32-n9-3r2w.md: New file.
6786 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6787 nds32_register_ports): New or modify for cpu n9.
6788 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6790 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6791 * config/nds32/nds32-utils.c: New file.
6792 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6793 TARGET_MUL_SLOW): Define.
6794 * config/nds32/nds32.md (pipeline_model): New attribute.
6795 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6796 New options that support cpu n9.
6797 * config/nds32/pipelines.md: Include n9 settings.
6798 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6800 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6802 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6803 information if necessary.
6804 (output_cond_branch_compare_zero): Likewise.
6805 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6806 (nds32_target_alignment): Refine for alignment.
6807 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6808 (FUNCTION_BOUNDARY): Modify.
6809 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6811 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6813 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6815 * config/nds32/constants.md (unspec_volatile_element): Add values for
6816 TLB operation and data prefetch.
6817 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6818 functions for TLB operation and data prefetch.
6819 * config/nds32/nds32-intrinsic.md: Likewise.
6820 * config/nds32/nds32_intrinsic.h: Likewise.
6821 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6822 (nds32_print_operand): Likewise.
6823 * config/nds32/nds32.h (nds32_builtins): Likewise.
6825 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6826 Andrew Pinski <pinsika@gcc.gnu.org>
6829 * match.pd: Use constant_boolean_node of correct type instead of
6830 boolean_true_node or boolean_false_node for simplifying
6831 pointer comparisons to zero.
6833 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6835 PR tree-optimization/80021
6836 * tree.c (verify_type_variant): Make error call in verify_variant_match
6837 translatable and remove final full stop.
6839 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6841 * config/nds32/constants.md (unspec_volatile_element): Add
6842 UNSPEC_VOLATILE_EH_RETURN.
6843 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6844 nds32_output_stack_pop): Support dwarf exception handling process.
6845 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6846 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6847 exception handling process.
6848 (nds32_compute_stack_frame): Likewise.
6849 (nds32_return_addr_rtx): Likewise.
6850 (nds32_initial_elimination_offset): Likewise.
6851 (nds32_expand_prologue): Likewise.
6852 (nds32_expand_epilogue): Likewise.
6853 (nds32_dynamic_chain_address): New function.
6854 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6856 (DYNAMIC_CHAIN_ADDRESS): Define.
6857 (EH_RETURN_DATA_REGNO): Define.
6858 (EH_RETURN_STACKADJ_RTX): Define.
6859 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6860 patterns for dwarf exception handling.
6862 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6864 * config/nds32/nds32.h: Clean up obsolete macros.
6866 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6868 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6869 Add enum values for particular instructions.
6870 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6871 particular intrinsic functions.
6872 * config/nds32/nds32-intrinsic.md: Likewise.
6873 * config/nds32/nds32_intrinsic.h: Likewise.
6874 * config/nds32/nds32.h (nds32_builtins): Likewise.
6875 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6876 (btst, ave): New patterns for particular instructions.
6878 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6880 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6881 Add enum values for atomic load/store and memory sync.
6882 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6884 * config/nds32/nds32-intrinsic.md: Likewise.
6885 * config/nds32/nds32_intrinsic.h: Likewise.
6886 * config/nds32/nds32.h (nds32_builtins): Likewise.
6888 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6890 PR tree-optimization/85257
6891 * fold-const.c (native_encode_vector): If not all elts could fit
6892 and off is -1, return 0 rather than offset.
6893 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6894 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6895 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6896 adjust buffer in native_interpret_expr call.
6898 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6900 * config/nds32/constants.md (unspec_volatile_element): Add cache
6901 control enum values.
6902 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6903 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6904 * config/nds32/nds32.c (nds32_cctl_names): New.
6905 (nds32_print_operand): Handle cache control register names.
6906 * config/nds32/nds32.h (nds32_builtins): New enum values.
6907 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6909 * config/nds32/nds32.md (type): Add mmu.
6910 * config/nds32/pipelines.md (simple_insn): Add mmu.
6912 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6914 * config/nds32/nds32.md (type): Remove call.
6915 * config/nds32/pipelines.md (simple_insn): Likewise.
6917 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6919 * config/nds32/constants.md (unspec_volatile_element): Add
6920 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6921 UNSPEC_VOLATILE_FMFCFG.
6922 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6923 description for fmfcfg and fmfcsr.
6924 (bdesc_1arg): Add fmtcsr.
6925 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6926 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6927 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6928 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6929 unspec_fmfcfg): New patterns.
6930 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6931 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6932 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6933 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6934 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6935 __nds32__fmfcfg): Define.
6937 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6939 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6940 intrinsic register names.
6941 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6942 intrinsic register enum values and macros.
6944 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6946 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6947 for load/store addressing form.
6948 (nds32_print_operand_address): Likewise.
6950 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6953 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6954 based on LABEL_REF. Remove useless assertion.
6955 (pic_address_needs_scratch): Fix formatting.
6956 (sparc_legitimize_pic_address): Minor tweaks.
6957 (sparc_delegitimize_address): Adjust assertion accordingly.
6958 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6959 into symbolic_operand.
6960 (movsi_high_pic_label_ref): Likewise.
6961 (movsi_lo_sum_pic_label_ref): Likewise.
6962 (movdi_pic_label_ref): Likewise.
6963 (movdi_high_pic_label_ref): Likewise.
6964 (movdi_lo_sum_pic_label_ref): Likewise.
6966 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6968 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6969 custom LIB_SPEC setup.
6971 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6972 Kito Cheng <kito.cheng@gmail.com>
6974 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6975 * config/riscv/freebsd.h: New.
6977 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6979 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6980 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6983 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6984 Kito Cheng <kito.cheng@gmail.com>
6986 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6987 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6988 * config/nds32/nds32-protos.h (nds32_output_call,
6989 nds32_output_return): Declare.
6990 * config/nds32/nds32.md: Refine all the call and return patterns.
6992 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6995 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6996 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6998 PR rtl-optimization/84872
6999 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7000 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7003 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7005 * expr.c (copy_blkmode_to_reg): Revert 254862.
7006 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7008 2018-04-06 Richard Biener <rguenther@suse.de>
7011 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7012 after seeing a component reference with an adjacent field. Treat
7013 refs to arrays at struct end of external decls similar to
7014 refs to unconstrained commons.
7016 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7019 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7020 look through SAVE_EXPRs with non-side-effects argument. Adjust
7022 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7025 2018-04-06 Richard Biener <rguenther@suse.de>
7028 * alias.c (find_base_term): New wrapper around find_base_term
7029 unwinding CSELIB_VAL_PTR changes.
7030 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7033 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7035 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7037 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7038 constant definitions.
7039 ("nop"): lr 0,0 -> nopr r0
7040 ("nop_lr0", "nop_lr1"): New insn definitions.
7042 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7044 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7045 NDS32_V3PUSH_AVAILABLE_P macro.
7047 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7048 Chung-Ju Wu <jasonwucj@gmail.com>
7050 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7051 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7052 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7053 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7054 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7055 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7056 * config/nds32/constraints.md: New constraints and checking for hard
7057 float configuration.
7058 * config/nds32/iterators.md: New mode iterator and attribute for hard
7059 float configuration.
7060 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7062 * config/nds32/nds32-fpu.md: New file.
7063 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7064 deal with hard float code generation.
7065 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7067 (abi_type, float_reg_number): New enum type.
7068 * config/nds32/nds32-predicates.c: New predicates for hard float.
7069 * config/nds32/nds32-protos.h: Declare functions for hard float.
7070 * config/nds32/nds32.c: Implementation for hard float configuration.
7071 * config/nds32/nds32.h: Definitions for hard float configuration.
7072 * config/nds32/nds32.md: Include hard float machine description and
7073 modify patterns for hard float configuration.
7074 * config/nds32/nds32.opt: New options for hard float configuration.
7075 * config/nds32/predicates.md: New predicates for hard float
7078 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7080 * common/config/nds32/nds32-common.c
7081 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7083 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7086 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7087 CONSTRUCTOR_ELT (ctor, ...)->value.
7089 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7092 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7094 2018-04-05 Tom de Vries <tom@codesourcery.com>
7097 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7100 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7101 Kito Cheng <kito.cheng@gmail.com>
7103 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7104 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7105 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7106 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7108 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7109 Kito Cheng <kito.cheng@gmail.com>
7111 * config/nds32/constraints.md (Ufe): New memory constraint.
7112 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7113 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7114 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7116 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7117 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7119 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7121 * config/nds32/nds32.md: Use optimize_size in the condition for
7122 alu-shift instructions.
7124 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7126 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7128 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7130 * config/nds32/nds32.md (negsi2): Refine pattern.
7132 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7133 Chung-Ju Wu <jasonwucj@gmail.com>
7135 * config/nds32/iterators.md (shift_rotate): New code iterator.
7136 (shift): New code attribute.
7137 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7138 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7139 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7140 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7141 bit-wise operations.
7142 (andsi3, *andsi3): Ditto.
7143 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7144 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7145 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7146 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7147 nds32_ior_operand, nds32_xor_operand): New predicates.
7149 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7151 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7152 (addsi3, subsi3): ... this.
7154 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7156 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7158 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7160 * config/nds32/nds32.md: Adjust indention.
7162 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7164 * config/nds32/nds32.md (feature): New attribute.
7166 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7168 * config/nds32/nds32.md (subtype): New attribute.
7170 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7173 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7174 expansion to perform a bitwise AND of the argument followed by a
7175 boolean negation of the result.
7177 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7179 PR rtl-optimization/84878
7180 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7181 the basic block. Assert the use reference is not artificial and that
7182 it has an associated insn.
7184 2018-04-04 Michael Matz <matz@suse.de>
7186 * builtins.c (compute_objsize): Pass correct operand
7187 to array_at_struct_end_p.
7189 2018-04-04 Richard Biener <rguenther@suse.de>
7192 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7193 from contexts for DINFO_LEVEL_TERSE and below.
7195 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7197 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7198 resiter_operand condition.
7199 * config/nds32/nds32.md (*move<mode>): Ditto.
7201 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7202 Monk Chiang <sh.chiang04@gmail.com>
7204 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7206 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7208 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7210 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7211 Kito Cheng <kito.cheng@gmail.com>
7213 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7214 nds32_cond_code_str, output_cond_branch,
7215 output_cond_branch_compare_zero, nds32_expand_cbranch,
7216 nds32_expand_cstore, nds32_expand_movcc,
7217 nds32_output_cbranchsi4_equality_zero,
7218 nds32_output_cbranchsi4_equality_reg,
7219 nds32_output_cbranchsi4_equality_reg_or_const_int,
7220 nds32_output_cbranchsi4_greater_less_zero: New functions.
7221 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7222 nds32_expand_cstore, nds32_expand_movcc,
7223 nds32_output_cbranchsi4_equality_zero,
7224 nds32_output_cbranchsi4_equality_reg,
7225 nds32_output_cbranchsi4_equality_reg_or_const_int,
7226 nds32_output_cbranchsi4_greater_less_zero): Declare.
7227 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7228 nds32_rimm11s_operand): New predicates.
7229 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7230 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7233 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7235 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7236 * config/nds32/nds32.md: Ditto.
7237 * config/nds32/pipelines.md: Ditto.
7239 2018-04-04 Richard Biener <rguenther@suse.de>
7241 PR tree-optimization/85168
7242 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7243 propagating abnormals.
7245 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7247 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7249 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7250 Kito Cheng <kito.cheng@gmail.com>
7252 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7253 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7254 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7255 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7256 * config/nds32/nds32.md (sibcall_internal): New.
7257 (sibcall_register): Remove.
7258 (sibcall_immediate): Remove.
7259 (sibcall_value_internal): New.
7260 (sibcall_value_register): Remove.
7261 (sibcall_value_immediate): Remove.
7262 * config/nds32/predicates.md (nds32_general_register_operand): New.
7263 (nds32_call_address_operand): New.
7265 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7267 PR rtl-optimization/85167
7268 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7269 bb_defs if *split_p, instead preinitialize it to NULL.
7271 PR tree-optimization/85156
7272 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7273 evaluating the argument multiple times.
7275 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7277 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7279 (_mm_cvtpd_ps): Likewise.
7280 (_mm_cvttpd_epi32): Likewise.
7281 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7282 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7283 vector, pixel, and bool following altivec.h include.
7285 2018-04-03 Martin Sebor <msebor@redhat.com>
7287 * doc/extend.texi (Common Function Attributes): Clarify.
7288 (const attribute): Likewise.
7289 (pure attribute): Likewise.
7291 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7294 * config/i386/i386.c (ix86_expand_vector_set): Use
7295 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7297 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7299 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7300 instructions when changing rounding bits to preserve precision bits
7301 in the x87 control word.
7303 2018-04-03 Martin Liska <mliska@suse.cz>
7305 PR tree-optimization/82491
7306 * rtl.h (strip_offset_and_add): Replace += suboffset with
7307 poly_uint64 () + suboffset.
7309 2018-03-29 Martin Liska <mliska@suse.cz>
7310 Martin Jambor <mjambor@suse.cz>
7313 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7314 param_type is not an integral or pointer type.
7316 2018-04-03 Richard Biener <rguenther@suse.de>
7318 * sese.h (recompute_all_dominators): Remove.
7320 2018-04-02 Martin Sebor <msebor@redhat.com>
7322 * doc/invoke.texi (-Wrestrict): Fix typos.
7324 2018-04-02 Jim Wilson <jimw@sifive.com>
7326 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7327 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7328 (<optab>di3, <optab>si3_extend): Likewise.
7329 (<optab>si3_mask, <optab>si3_mask_1): New.
7330 (<optab>di3_mask, <optab>di3_mask_1): New.
7331 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7332 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7333 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7335 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7337 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7340 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7342 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7343 (nds32_canonicalize_comparison): New function.
7345 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7346 Kito Cheng <kito.cheng@gmail.com>
7347 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7349 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7350 * config/nds32/constants.md (unspec_volatile_element): Add
7351 UNSPEC_VOLATILE_RELAX_GROUP.
7352 * config/nds32/nds32-relax-opt.c: New file.
7353 * config/nds32/nds32-predicates.c
7354 (nds32_symbol_load_store_p): New function.
7355 * config/nds32/nds32-protos.h
7356 (nds32_symbol_load_store_p): Declare function.
7357 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7358 * config/nds32/nds32.c
7359 (nds32_register_pass): New function to register pass.
7360 (nds32_register_passes): New function to register passes.
7361 * config/nds32/nds32.md (relax_group): New pattern.
7362 * config/nds32/nds32.opt (mrelax-hint): New option.
7363 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7365 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7367 * config/nds32/t-nds32: Modify files dependency.
7369 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7371 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7372 (PROFILE_HOOK): Define its implementation.
7374 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7376 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7377 type and 32-bit size.
7379 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7382 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7383 (V_128_256): New mode iterator.
7384 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7385 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7386 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7388 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7389 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7391 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7394 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7395 NaN inputs correctly.
7397 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7400 * config/rs6000/vsx.md (??r): New mode attribute.
7401 (*vsx_mov<mode>_64bit): Use it.
7402 (*vsx_mov<mode>_32bit): Likewise.
7404 2018-03-30 Martin Sebor <msebor@redhat.com>
7406 PR tree-optimization/84818
7407 * builtins.c (check_access): Use warning_n.
7409 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7412 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7414 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7417 2018-03-30 Julia Koval <julia.koval@intel.com>
7420 * x86-tune.def (movx, partial_reg_dependency): Enable for
7423 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7426 * lra-constraints.c (process_alt_operands): Move setting
7427 this_alternative_matches below.
7429 2018-03-29 Martin Liska <mliska@suse.cz>
7432 * doc/invoke.texi: Document how LTO works with debug info.
7433 Describe auto-load support of binutils. Mention 'x86-64'
7434 as valid option value of -march option.
7436 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7438 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7441 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7442 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7443 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7446 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7449 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7450 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7451 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7452 for RS6000_BTM_POWERPC64.
7453 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7454 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7455 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7460 2018-03-28 Carl Love <cel@us.ibm.com>
7463 2017-09-27 Carl Love <cel@us.ibm.com>
7465 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7466 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7467 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7470 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7472 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7473 instead of __vector bool.
7474 (_mm_max_pu8): Likewise.
7475 (_mm_min_pi16): Likewise.
7477 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7480 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7481 (DIVWEUO): Likewise.
7483 (DIVDEUO): Likewise.
7484 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7485 DIVWEUO and DIVDEUO.
7486 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7487 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7488 (div_extend): Likewise.
7489 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7491 (__builtin_divweuo): Likewise.
7492 (__builtin_divdeo): Likewise.
7493 (__builtin_divdeuo): Likewise.
7495 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7498 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7499 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7501 PR tree-optimization/82004
7502 * gimple-match-head.c (optimize_pow_to_exp): New function.
7503 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7504 Don't fold to exp if optimize_pow_to_exp is false.
7506 2018-03-28 Martin Liska <mliska@suse.cz>
7509 * calls.c (initialize_argument_information): Fix trailing space.
7510 * common.opt: Fix typo and provide better explanation for
7511 -fsanitize-coverage option.
7512 * config/i386/i386.opt: Fix typo.
7514 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7515 Martin Liska <mliska@suse.cz>
7518 * gimplify.c (asan_poison_variable): Don't do the check for
7519 gimplify_omp_ctxp here.
7520 (gimplify_decl_expr): Do it here.
7521 (gimplify_target_expr): Likewise.
7523 2018-03-28 Martin Liska <mliska@suse.cz>
7526 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7527 chkp_type_bounds_count if MPX is not enabled.
7529 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7531 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7533 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7536 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7537 function to create the function decl for complex long double
7538 multiply and divide for -mabi=ieeelongdouble.
7539 (init_float128_ieee): Call it.
7541 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7544 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7545 -fcf-protection=branch -mibt.
7546 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7548 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7551 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7552 arm_word_relocations.
7554 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7557 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7558 extern array declarations.
7560 2018-03-27 Richard Biener <rguenther@suse.de>
7563 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7564 explicit single_use checks.
7566 2018-03-27 Richard Biener <rguenther@suse.de>
7568 PR tree-optimization/85082
7569 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7572 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7574 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7575 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7576 Turn on fasynchronous-unwind-tables and funwind-tables.
7578 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7581 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7582 (*bmi_blsr_<mode>_ccz): Ditto.
7584 2018-03-26 Tom de Vries <tom@codesourcery.com>
7586 PR tree-optimization/85063
7587 * omp-general.c (offloading_function_p): New function. Factor out
7589 * omp-offload.c (pass_omp_target_link::gate): ... here.
7590 * omp-general.h (offloading_function_p): Declare.
7591 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7592 with attribute omp declare target for offloading functions.
7594 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7596 PR tree-optimization/84005
7597 * tree-data-ref.h (get_base_for_alignment): Declare.
7598 * tree-data-ref.c (get_base_for_alignment_1): New function.
7599 (get_base_for_alignment): Likewise.
7600 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7601 get_base_for_alignment to find a suitable base object, instead
7602 of always using drb->base_address.
7604 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7607 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7608 known size by default.
7610 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7613 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7614 and non BLKmode operands.
7616 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7619 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7620 Clean up attributes.
7622 2018-03-23 Richard Biener <rguenther@suse.de>
7625 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7626 we are going to emit early debug for LTO.
7628 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7631 * function.c (match_asm_constraints_1): Don't optimize if input
7632 doesn't satisfy general_operand predicate for output's mode.
7635 * alias.c (write_dependence_p): Don't require for x_canonicalized
7636 non-VOIDmode if x has VOIDmode.
7639 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7640 just don't try to optimize it rather than assert it never happens.
7642 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7644 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7645 macro expansions for definition of ST_INTERNAL_<mode> and
7646 LD_INTERNAL_<mode> builtins.
7647 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7649 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7651 (altivec_expand_st_builtin): Likewise.
7652 (altivec_expand_builtin): Remove calls to deleted functions.
7653 (rs6000_address_for_altivec): Delete this function.
7654 * config/rs6000/vector.md: Remove expands for
7655 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7657 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7660 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7661 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7662 re-computing once computed.
7663 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7664 (arm_init_machine_status): Initialize
7665 machine->static_chain_stack_bytes.
7667 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7670 * doc/extend.texi: Add four new prototypes for vec_ld.
7671 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7672 definitions for more logical presentation.
7673 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7674 entries for V1TI variants of __builtin_altivec_ld builtin.
7675 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7676 handling of V1TI variant of LVX icode pattern.
7677 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7678 (rs6000_gimple_fold_builtin): Likewise.
7679 (altivec_init_builtins): Add code to define
7680 __builtin_altivec_lvx_v1ti function.
7682 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7685 * function.c (match_asm_constraints_1): Don't do the optimization
7686 if input isn't a REG, SUBREG, MEM or constant.
7688 2018-03-22 Tom de Vries <tom@codesourcery.com>
7690 PR tree-optimization/84956
7691 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7692 bb_has_abnormal_pred.
7694 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7697 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7698 DECL_INITIAL (decl) to decl at the end.
7699 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7702 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7704 * doc/extend.texi (__builtin_tgmath): Document when complex
7705 integer types are treated as _Complex _Float64.
7707 2018-03-21 Tom de Vries <tom@codesourcery.com>
7709 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7711 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7713 PR tree-optimization/84960
7714 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7715 if it is ENTRY block, move them into single succ of ENTRY in that case.
7717 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7719 PR tree-optimization/84811
7720 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7721 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7722 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7723 (known_subrange_p): Update accordingly. Cast each value involved
7724 in the size comparison, rather than casting the result of the
7727 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7729 PR tree-optimization/84982
7730 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7731 by flipping the least significant bit rather than all bits from
7732 bitpos to bitpos + bitsize - 1.
7734 2018-03-21 Nathan Sidwell <nathan@acm.org>
7736 * doc/extend.texi (Deprecated Features): Remove mention of
7737 long-deleted deprecations.
7739 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7742 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7743 * configure: Regenerate.
7745 2018-03-21 Tom de Vries <tom@codesourcery.com>
7747 PR tree-optimization/83126
7748 * tree-parloops.c (num_phis): New function.
7749 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7751 2018-03-21 Nathan Sidwell <nathan@acm.org>
7753 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7754 mention anon-struct/union members and trailing attributes.
7756 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7758 PR tree-optimization/84969
7759 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7760 builtin memset partitions if they set different rhs values.
7762 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7764 PR rtl-optimization/84989
7765 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7766 VEC_DUPLICATE with scalar result mode.
7768 2018-03-21 Martin Liska <mliska@suse.cz>
7771 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7772 not intended return statement.
7774 2018-03-21 Martin Liska <mliska@suse.cz>
7777 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7778 (chkp_find_bound_slots_1): Limit number of iterations.
7780 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7783 * Minor grammar fixes for x86 options.
7785 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7788 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7789 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7791 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7794 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7795 (altivec_lvx_<mode>_1op): Likewise.
7796 (altivec_stvx_<mode>_2op): Likewise.
7797 (altivec_stvx_<mode>_1op): Likewise.
7798 (altivec_lvx_<VM2:mode>): New define_expand.
7799 (altivec_stvx_<VM2:mode>): Likewise.
7800 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7801 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7802 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7803 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7804 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7805 (rs6000_gen_lvx): Likewise.
7806 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7807 (altivec_expand_stv_builtin): Likewise.
7808 (altivec_expand_builtin): Likewise.
7809 * config/rs6000/vector.md: Likewise.
7811 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7814 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7817 2018-03-20 Richard Biener <rguenther@suse.de>
7820 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7821 sign-conversions as zero, fall back to standard scalar_stmt
7824 2018-03-20 Martin Liska <mliska@suse.cz>
7827 * predict.c (rebuild_frequencies): Handle case when we have
7828 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7830 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7833 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7834 flag_section_anchors.
7835 * varasm.c (use_blocks_for_decl_p): Remove hack for
7836 dw2_force_const_mem.
7839 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7841 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7842 be created, use lowpart_subreg of operands[0] rather than operands[0]
7844 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7845 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7846 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7847 and n constraint instead of aarch64_shift_imm_di and Usd.
7848 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7849 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7851 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7854 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7855 to favor GPR over NEON registers.
7856 (<shift>di3_neon): Likewise.
7858 2018-03-20 Tom de Vries <tom@codesourcery.com>
7861 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7862 (nvptx_process_pars): Emit bar.sync asap and alap.
7864 2018-03-20 Tom de Vries <tom@codesourcery.com>
7867 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7868 seen_label if seen_label is already set.
7870 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7873 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7874 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7875 Use 1U instead of 1. Formatting fixes.
7878 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7879 instead of TREE_TYPE (s1) for the return value.
7881 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7883 PR tree-optimization/84946
7884 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7885 bitsize + bitsize in poly_uint64 rather than poly_int64.
7888 * dwarf2asm.c: Include fold-const.c.
7889 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7890 of decl rather than decl itself.
7892 PR rtl-optimization/84643
7893 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7895 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7898 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7899 calling assemble_variable.
7901 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7904 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7905 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7907 2018-03-19 Jim Wilson <jimw@sifive.com>
7910 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7911 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7912 (riscv_first_stack_step): Likewise.
7913 (riscv_option_override): Use STACK_BOUNDARY instead of
7915 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7917 (BIGGEST_ALIGNMENT): Set to 128.
7918 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7919 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7922 2018-03-19 Richard Biener <rguenther@suse.de>
7924 PR tree-optimization/84933
7925 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7926 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7928 2018-03-19 Richard Biener <rguenther@suse.de>
7930 PR tree-optimization/84859
7931 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7932 (cond_if_else_store_replacement): Perform sinking operation on
7933 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7934 Generalize what a BB with a single eligible store is.
7936 2018-03-19 Richard Biener <rguenther@suse.de>
7938 PR tree-optimization/84929
7939 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7940 chrec_is_positive against non-chrec arg.
7942 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7945 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7947 2018-03-18 Martin Liska <mliska@suse.cz>
7949 PR rtl-optimization/84635
7950 * regrename.c (build_def_use): Use matches_mode only when
7953 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7955 PR tree-optimization/84913
7956 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7957 vectorize chains of COND_EXPRs.
7959 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7961 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7963 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7965 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7967 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7969 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7971 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7972 Kito Cheng <kito.cheng@gmail.com>
7974 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7975 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7976 (nds32_adjust_reg_alloc_order): New function.
7977 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7979 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7981 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7982 nds32_print_operand, nds32_print_operand_address): Use
7983 HOST_WIDE_INT_PRINT_DEC instead.
7985 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7987 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7989 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7992 * config/i386/i386.c (initial_ix86_tune_features,
7993 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7995 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7996 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7997 rather than 1u << ix86_tune. Formatting fix.
7998 (ix86_option_override_internal): Change ix86_arch_mask from
7999 unsigned int to unsigned HOST_WIDE_INT, initialize to
8000 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8001 (ix86_function_specific_restore): Likewise.
8003 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8006 * postreload.c (reload_combine_recognize_pattern): Perform
8007 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8008 truncate_int_for_mode the result for the destination's mode.
8011 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8012 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8014 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8017 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8020 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8021 regno_allocno_class_array and sorted_pseudos.
8022 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8023 insns where regno is used.
8025 2018-03-16 Martin Liska <mliska@suse.cz>
8028 * multiple_target.c (create_dispatcher_calls): Redirect
8029 reference in the symbol table.
8031 2018-03-16 Martin Liska <mliska@suse.cz>
8034 * multiple_target.c (create_dispatcher_calls): Redirect also
8037 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8041 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8042 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8043 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8045 2018-03-16 Julia Koval <julia.koval@intel.com>
8047 * doc/invoke.texi (Skylake Server): Add CLWB.
8048 Cannonlake): Remove CLWB.
8050 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8052 PR tree-optimization/84841
8053 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8055 (FLOAT_ONE_CONST_TYPE): Define.
8056 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8057 (sort_by_operand_rank): Put entries with higher constant_type last
8058 rather than first to match comments.
8060 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8062 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8065 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8068 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8069 check and use address of target always.
8071 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8074 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8075 (indirect_thunk_bnd_needed): Likewise.
8076 (indirect_thunks_used): Likewise.
8077 (indirect_thunks_bnd_used): Likewise.
8078 (indirect_return_needed): New.
8079 (indirect_return_bnd_needed): Likewise.
8080 (output_indirect_thunk_function): Add a bool argument for
8082 (output_indirect_thunk_function): Don't generate alias for
8083 function return thunk.
8084 (ix86_code_end): Call output_indirect_thunk_function to generate
8085 function return thunks.
8086 (ix86_output_function_return): Set indirect_return_bnd_needed
8087 and indirect_return_needed instead of indirect_thunk_bnd_needed
8088 and indirect_thunk_needed.
8090 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8092 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8093 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8094 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8096 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8097 Paul Hua <paul.hua.gm@gmail.com>
8100 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8102 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8104 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8105 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8108 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8111 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8112 instead of GET_MODE_SIZE when comparing Units.
8114 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8117 * varasm.c (hash_section): Return an unchangeble hash value
8118 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8119 Return !aarch64_can_use_per_function_literal_pools_p ().
8121 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8124 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8125 rather than address of cmode as last argument to prepare_cmp_insn.
8127 2018-03-15 Julia Koval <julia.koval@intel.com>
8129 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8130 F_AVX512VNNI, F_AVX512BITALG): New.
8132 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8135 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8136 insn for floating-point loads and stores.
8138 2018-03-14 Carl Love <cel@us.ibm.com>
8140 * config/rs6000/rs6000-c.c: Add macro definitions for
8141 ALTIVEC_BUILTIN_VEC_PERMXOR.
8142 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8143 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8144 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8145 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8147 * config/doc/extend.texi: Add prototypes for vec_permxor.
8149 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8152 * diagnostic-show-locus.c (class layout_point): Convert m_line
8153 from int to linenum_type.
8154 (line_span::comparator): Use linenum "compare" function when
8155 comparing line numbers.
8156 (test_line_span): New function.
8157 (layout_range::contains_point): Convert param "row" from int to
8159 (layout_range::intersects_line_p): Likewise.
8160 (layout::will_show_line_p): Likewise.
8161 (layout::print_source_line): Likewise.
8162 (layout::should_print_annotation_line_p): Likewise.
8163 (layout::print_annotation_line): Likewise.
8164 (layout::print_leading_fixits): Likewise.
8165 (layout::annotation_line_showed_range_p): Likewise.
8166 (struct line_corrections): Likewise for field m_row.
8167 (line_corrections::line_corrections): Likewise for param "row".
8168 (layout::print_trailing_fixits): Likewise.
8169 (layout::get_state_at_point): Likewise.
8170 (layout::get_x_bound_for_row): Likewise.
8171 (layout::print_line): Likewise.
8172 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8173 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8174 * input.c (selftest::test_linenum_comparisons): New function.
8175 (selftest::input_c_tests): Call it.
8176 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8177 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8178 * selftest.h (ASSERT_GT): New macro.
8179 (ASSERT_GT_AT): New macro.
8180 (ASSERT_LT): New macro.
8181 (ASSERT_LT_AT): New macro.
8183 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8185 PR rtl-optimization/84780
8186 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8188 2018-03-14 Martin Liska <mliska@suse.cz>
8190 * tree.c (record_node_allocation_statistics): Use
8191 get_stats_node_kind.
8192 (get_stats_node_kind): New function extracted from
8193 record_node_allocation_statistics.
8194 (free_node): Use get_stats_node_kind.
8196 2018-03-14 Richard Biener <rguenther@suse.de>
8198 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8199 that the value-set of ANTIC_IN doesn't grow.
8202 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8204 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8205 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8207 2018-03-14 Julia Koval <julia.koval@intel.com>
8209 * config.gcc (icelake-client, icelake-server): New.
8211 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8212 (initial_ix86_arch_features): Ditto.
8213 (PTA_SKYLAKE): Add SGX.
8214 (PTA_ICELAKE): Remove.
8215 (PTA_ICELAKE_CLIENT): New.
8216 (PTA_ICELAKE_SERVER): New.
8217 (ix86_option_override_internal): Split up icelake on icelake client and
8219 (get_builtin_code_for_version): Ditto.
8220 (fold_builtin_cpu): Ditto.
8221 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8222 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8223 * config/i386/i386.h (processor_type): Ditto.
8224 * doc/invoke.texi: Ditto.
8226 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8229 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8230 INTEGER_CST offset, add it together with bitpos / 8 and
8231 sign extend based on POINTER_SIZE.
8235 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8238 * config/i386/constraints.md (Yc): New register constraint.
8239 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8240 Use Yc constraint for alternative 2 of operand 0. Remove
8241 preferred_for_speed attribute.
8243 2018-03-14 Richard Biener <rguenther@suse.de>
8245 PR tree-optimization/84830
8246 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8247 with the old one to avoid oscillations.
8249 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8252 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8254 (assign_by_spills): Return a flag of reload assignment failure.
8255 Do not process the reload assignment failures. Do not spill other
8256 reload pseudos if they has the same reg class. Update n if
8258 (lra_assign): Add a return arg. Set up from the result of
8259 assign_by_spills call.
8260 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8261 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8262 usage_insns if it is not NULL.
8263 (spill_hard_reg_in_range): New function.
8264 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8265 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8266 function prototypes.
8267 (lra_assign): Change prototype.
8268 * lra.c (lra): Add code to deal with fails by splitting hard reg
8271 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8273 * config/riscv/riscv.opt (mrelax): New option.
8274 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8275 "norelax" when riscv_mrelax is disabled.
8276 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8278 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8281 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8282 reassociation for int modes.
8284 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8286 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8287 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8289 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8290 * config/aarch64/aarch64-sve.md
8291 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8292 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8293 (*extend<mode><Vwide>2): Rename to...
8294 (aarch64_sve_extend<mode><Vwide>2): ...this.
8295 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8296 renaming the old pattern to...
8297 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8299 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8300 define_expand, renaming the old pattern to...
8301 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8302 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8303 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8304 account when deciding which SVE instruction the optab should use.
8305 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8307 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8309 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8310 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8311 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8312 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8313 (tlsdesc_small_<mode>): Turn a define_expand and use
8314 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8315 (tlsdesc_small_advsimd_<mode>): ...this.
8316 (tlsdesc_small_sve_<mode>): New pattern.
8318 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8320 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8321 (UNSPEC_UMUL_HIGHPART): New constants.
8322 (MUL_HIGHPART): New int iteraor.
8323 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8324 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8326 (*<su>mul<mode>3_highpart): New define_insn.
8328 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8331 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8334 2018-03-13 Martin Liska <mliska@suse.cz>
8337 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8339 (sem_item_optimizer::~sem_item_optimizer): Release it.
8340 (sem_item_optimizer::merge_classes): Register variable aliases.
8341 (sem_item_optimizer::fixup_pt_set): New function.
8342 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8343 * ipa-icf.h: Declare new variables and functions.
8345 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8348 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8349 integer_pow2p@2 and test integer_pow2p in condition.
8350 (A < 0 ? C : 0): Similarly for @1.
8353 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8354 characters starting at p contain '\0' character, don't look beyond
8358 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8359 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8362 * reg-stack.c (change_stack): Change update_end var from int to
8363 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8364 also call set_block_for_insn on the newly added insns and rescan.
8367 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8368 on the last operand.
8371 * tree.c (stabilize_reference_1): Return save_expr (e) for
8372 STATEMENT_LIST even if it doesn't have side-effects.
8374 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8376 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8378 2018-03-12 Renlin Li <renlin.li@arm.com>
8380 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8381 aarch64_output_scalar_simd_mov_immediate.
8383 2018-03-12 Martin Sebor <msebor@redhat.com>
8385 PR tree-optimization/83456
8386 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8387 for perfectly overlapping calls to memcpy.
8388 (gimple_fold_builtin_memory_chk): Same.
8389 (gimple_fold_builtin_strcpy): Handle no-warning.
8390 (gimple_fold_builtin_stxcpy_chk): Same.
8391 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8393 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8395 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8396 parameter. Use it for SFmode.
8397 (rs6000_function_arg_advance_1): Adjust.
8398 (rs6000_function_arg): Adjust.
8399 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8401 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8403 PR rtl-optimization/84169
8404 PR rtl-optimization/84780
8405 * combine.c (can_combine_p): Check for a 2-insn combination whether
8406 the destination register is used between the two insns, too.
8408 2018-03-12 Richard Biener <rguenther@suse.de>
8410 PR tree-optimization/84803
8411 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8412 for refs DR analysis didn't process.
8414 2018-03-12 Richard Biener <rguenther@suse.de>
8416 PR tree-optimization/84777
8417 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8418 force-vectorize loops ignore whether we are optimizing for size.
8420 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8422 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8423 (TARGET_MD_ASM_ADJUST): Define.
8425 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8426 Kito Cheng <kito.cheng@gmail.com>
8427 Chung-Ju Wu <jasonwucj@gmail.com>
8429 * config/nds32/nds32.c (nds32_compute_stack_frame,
8430 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8431 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8432 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8433 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8434 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8435 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8436 * config/nds32/nds32.md (prologue, epilogue): Use macro
8437 NDS32_V3PUSH_AVAILABLE_P to do checking.
8439 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8442 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8443 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8444 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8445 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8446 addition of most attributes on !orig_type_die or the attribute not
8447 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8449 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8450 Chung-Ju Wu <jasonwucj@gmail.com>
8452 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8454 * config/nds32/nds32.opt (mvh): New option.
8456 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8457 Chung-Ju Wu <jasonwucj@gmail.com>
8459 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8461 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8462 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8465 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8466 Chung-Ju Wu <jasonwucj@gmail.com>
8468 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8470 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8471 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8473 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8474 Kito Cheng <kito.cheng@gmail.com>
8475 Chung-Ju Wu <jasonwucj@gmail.com>
8477 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8478 UNSPEC_FFMISM and UNSPEC_FLMISM.
8479 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8480 for ffb, ffmism and flmism.
8481 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8482 (unspec_ffmism): Ditto.
8483 (unspec_flmism): Ditto.
8484 (nds32_expand_builtin_impl): Check if string extension is available.
8485 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8486 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8488 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8491 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8494 * lra-assigns.c (assign_by_spills): Return a flag of reload
8495 assignment failure. Do not process the reload assignment
8496 failures. Do not spill other reload pseudos if they has the same
8498 (lra_assign): Add a return arg. Set up from the result of
8499 assign_by_spills call.
8500 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8501 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8502 usage_insns if it is not NULL.
8503 (spill_hard_reg_in_range): New function.
8504 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8505 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8506 function prototypes.
8507 (lra_assign): Change prototype.
8508 * lra.c (lra): Add code to deal with fails by splitting hard reg
8511 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8514 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8516 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8519 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8520 (dw_val_node): Add val_symbolic_view.
8521 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8522 (symview_upper_bound): New.
8523 (new_line_info_table): Initialize symviews_since_reset.
8524 (dwarf2out_source_line): Count symviews_since_reset and set
8525 symview_upper_bound.
8526 (dw_val_equal_p): Handle symview.
8527 (add_AT_symview): New.
8528 (print_dw_val): Handle symview.
8529 (attr_checksum, attr_checksum_ordered): Likewise.
8530 (same_dw_val_p, size_of_die): Likewise.
8531 (value_format, output_die): Likewise.
8532 (add_high_low_attributes): Use add_AT_symview for entry_view.
8533 (dwarf2out_finish): Reset symview_upper_bound, clear
8536 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8539 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8540 Add strict argument and use it.
8541 (rs6000_split_multireg_move): Update for new strict argument.
8542 (mem_operand_gpr): Disallow all non-offsettable addresses.
8543 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8545 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8548 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8549 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8550 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8553 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8554 decl, use remap_type if we want to use the type.
8556 2018-03-09 Martin Sebor <msebor@redhat.com>
8558 PR tree-optimization/84526
8559 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8561 (builtin_access::generic_overlap): Be prepared to handle non-array
8564 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8566 PR rtl-optimization/84682
8567 * lra-constraints.c (process_address_1): Check is_address flag
8568 for address constraints.
8569 (process_alt_operands): Likewise.
8570 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8571 preprocess_constraints.
8572 * recog.h (preprocess_constraints): Add oploc parameter.
8574 * recog.c (preprocess_constraints): Test address_operand for
8575 CT_ADDRESS constraints.
8577 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8580 * lra-assigns.c (assign_by_spills): Return a flag of reload
8581 assignment failure. Do not process the reload assignment
8582 failures. Do not spill other reload pseudos if they has the same
8584 (lra_assign): Add a return arg. Set up from the result of
8585 assign_by_spills call.
8586 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8587 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8588 usage_insns if it is not NULL.
8589 (spill_hard_reg_in_range): New function.
8590 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8591 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8592 function prototypes.
8593 (lra_assign): Change prototype.
8594 * lra.c (lra): Add code to deal with fails by splitting hard reg
8597 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8600 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8601 Accept complain bool parameter. Only emit errors if it is true.
8602 (arm_parse_cpu_option_name): Likewise.
8603 (arm_target_thumb_only): Adjust callers of the above.
8604 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8605 prototype to take a default true bool parameter.
8606 (arm_parse_arch_option_name): Likewise.
8608 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8609 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8613 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8614 * configure: Regenerate.
8615 * configure.ac ("linker --version-script option"): New.
8616 ("linker soname option"): New.
8618 2018-03-09 Richard Biener <rguenther@suse.de>
8620 PR tree-optimization/84775
8621 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8622 immediate uses of predicate stmts and mark them modified.
8625 PR tree-optimization/84178
8626 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8628 (version_loop_for_if_conversion): Delay update_ssa call.
8629 (tree_if_conversion): Delay update_ssa until after predicate
8632 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8635 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8636 when the function accesses prior frames.
8638 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8641 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8642 gen_llsym, otherwise call maybe_gen_llsym.
8645 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8646 has ',' character inside of it.
8648 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8651 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8652 as clobbering CC_REGNUM.
8654 2018-03-08 Richard Biener <rguenther@suse.de>
8657 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8658 (follow_copies_to_constant): Do not follow SSA names registered
8661 2018-03-08 Richard Biener <rguenther@suse.de>
8663 PR tree-optimization/84178
8664 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8666 (version_loop_for_if_conversion): Delay update_ssa call.
8667 (tree_if_conversion): Delay update_ssa until after predicate
8670 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8672 PR tree-optimization/84178
8673 * tree-if-conv.c (release_bb_predicate): Remove the
8674 the assertion that the stmts have NULL use_ops.
8675 Discard the statements, asserting that they haven't
8676 yet been added to a BB.
8678 2018-03-08 Richard Biener <rguenther@suse.de>
8680 PR tree-optimization/84746
8681 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8682 (phi_translate): Pass in destination ANTIC_OUT set.
8683 (phi_translate_1): Likewise. For a simplified result lookup
8684 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8685 (phi_translate_set): Adjust.
8686 (do_pre_regular_insertion): Likewise.
8687 (do_pre_partial_partial_insertion): Likewise.
8689 2018-03-08 Martin Liska <mliska@suse.cz>
8691 PR gcov-profile/84735
8692 * doc/gcov.texi: Document usage of profile files.
8693 * gcov-io.h: Document changes in the format.
8695 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8699 * dwarf2out.c (struct dw_line_info_table): Update comments for
8701 (FORCE_RESET_NEXT_VIEW): New.
8702 (FORCE_RESETTING_VIEW_P): New.
8703 (RESETTING_VIEW_P): Check for -1 too.
8704 (ZERO_VIEW_P): Likewise.
8705 (new_line_info_table): Force-reset next view.
8706 (dwarf2out_begin_function): Likewise.
8707 (dwarf2out_source_line): Simplify zero_view_p initialization.
8708 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8709 view directly. Omit view when omitting .loc at line 0.
8711 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8713 PR tree-optimization/84740
8714 * tree-switch-conversion.c (process_switch): Call build_constructors
8715 only if info.phi_count is non-zero.
8717 PR tree-optimization/84739
8718 * tree-tailcall.c (find_tail_calls): Check call arguments against
8719 DECL_ARGUMENTS (current_function_decl) rather than
8720 DECL_ARGUMENTS (func) when checking for tail recursion.
8722 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8724 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8725 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8726 Volker Reichelt's entry and add entries for people that perform
8727 GCC fuzzy testing and report numerous bugs.
8729 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8732 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8733 readonly data in sdata, if that is disabled.
8734 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8735 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8736 -mreadonly-in-sdata option.
8738 2018-03-07 Martin Sebor <msebor@redhat.com>
8740 PR tree-optimization/84468
8741 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8742 basic block when looking for nul assignment.
8744 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8747 * except.h (output_function_exception_table): Adjust prototype.
8748 * except.c (output_function_exception_table): Remove FNNAME parameter
8749 and add SECTION parameter. Ouput one part of the table at a time.
8750 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8751 the first part of the exception table and emit unwind directives.
8752 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8753 (i386_pe_seh_cold_init): Likewise.
8754 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8755 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8756 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8757 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8758 * config/i386/winnt.c: Include except.h.
8759 (struct seh_frame_state): Add reg_offset, after_prologue and
8760 in_cold_section fields.
8761 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8762 (i386_pe_seh_cold_init): New function.
8763 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8764 to seh->in_cold_section.
8765 (seh_emit_push): Record the offset of the push.
8766 (seh_emit_save): Record the offet of the save.
8767 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8768 Test seh->after_prologue to disregard the epilogue.
8769 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8770 (i386_pe_end_cold_function): New function.
8772 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8775 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8776 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8779 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8781 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8782 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8785 * multiple_target.c: Include tree-inline.h and intl.h.
8786 (expand_target_clones): Diagnose and fail if node->definition and
8787 !tree_versionable_function_p (node->decl).
8789 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8791 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8793 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8794 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8795 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8797 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8800 * combine.c (try_combine): Use reg_or_subregno instead of handling
8801 just paradoxical SUBREGs and REGs.
8803 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8805 * config/arc/arc.c (arc_finalize_pic): Remove function.
8806 (arc_must_save_register): We use single base PIC register, remove
8807 checks to save/restore the PIC register.
8808 (arc_expand_prologue): Likewise.
8809 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8811 (arc_verify_short): Likewise.
8812 (arc_attr_type): Likewise.
8813 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8814 (walk_stores): Likewise.
8815 (arc_address_cost): Make it static.
8816 (arc_verify_short): Likewise.
8817 (branch_dest): Likewise.
8818 (arc_attr_type): Likewise.
8819 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8820 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8821 (arc_final_prescan_insn): Remove inserting the nops due to
8822 hardware hazards. It is done in reorg step.
8823 (insn_length_variant_t): Remove.
8824 (insn_length_parameters_t): Likewise.
8825 (arc_insn_length_parameters): Likewise.
8826 (arc_get_insn_variants): Likewise.
8827 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8829 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8832 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8835 PR tree-optimization/84687
8836 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8838 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8840 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8842 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8843 Rename to ppc_speculation_barrier.
8844 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8845 __builtin_ppc_speculation_barrier.
8847 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8850 * combine.c (combine_simplify_rtx): Don't try to simplify if
8851 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8854 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8856 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8857 to 32 bytes when compiling for POWER9.
8859 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8862 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8863 regparm >= 3 with no arg reg available also for calls with
8864 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8867 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8869 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8871 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8874 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8876 2018-03-05 Richard Biener <rguenther@suse.de>
8878 PR tree-optimization/84486
8879 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8880 When inserting a __builtin_assume_aligned call set the LHS
8881 SSA name alignment info accordingly.
8883 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8885 PR tree-optimization/84114
8886 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8887 Avoid reassociation of FLOAT_MODE addition.
8889 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8891 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8892 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8893 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8894 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8895 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8896 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8897 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8899 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8900 __builtin_ia32_wbinvd): New builtins.
8901 (SPECIAL_ARGS2): New.
8902 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8903 (SPECIAL_ARGS2): New.
8904 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8905 (ix86_valid_target_attribute_inner_p): Ditto.
8906 (ix86_init_mmx_sse_builtins): Add special_args2.
8907 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8908 TARGET_WBNOINVD_P): New.
8909 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8910 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8911 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8912 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8913 * config/i386/pconfigintrin.h: New file.
8914 * config/i386/wbnoinvdintrin.h: Ditto.
8915 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8917 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8919 2018-03-05 Richard Biener <rguenther@suse.de>
8921 PR tree-optimization/84670
8922 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8924 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8925 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8926 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8927 doesn't grow if all successors (recursively) were visited at least
8930 2018-03-05 Richard Biener <rguenther@suse.de>
8932 PR tree-optimization/84650
8933 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8934 if executed in the loop pipeline.
8936 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8938 * doc/configfiles.texi (Configuration Files): Move info about
8939 conditionalizing $target-protos.h to...
8940 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8941 differs from $target-protos.h.
8943 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8944 Chung-Ju Wu <jasonwucj@gmail.com>
8946 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8947 * config/nds32/nds32-multiple.md (setmemsi): Define.
8948 * config/nds32/nds32-memory-manipulation.c
8949 (nds32_gen_dup_4_byte_to_word_value): New.
8950 (emit_setmem_word_loop): New.
8951 (emit_setmem_byte_loop): New.
8952 (nds32_expand_setmem_loop): New.
8953 (nds32_expand_setmem_loop_v3m): New.
8954 (nds32_expand_setmem_unroll): New.
8955 (nds32_expand_setmem): New.
8957 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8958 Chung-Ju Wu <jasonwucj@gmail.com>
8960 * config/nds32/nds32-memory-manipulation.c
8961 (nds32_emit_load_store): New.
8962 (nds32_emit_post_inc_load_store): New.
8963 (nds32_emit_mem_move): New.
8964 (nds32_emit_mem_move_block): New.
8965 (nds32_expand_movmemsi_loop_unknown_size): New.
8966 (nds32_expand_movmemsi_loop_known_size): New.
8967 (nds32_expand_movmemsi_loop): New.
8968 (nds32_expand_movmemsi_unroll): New.
8969 (nds32_expand_movmemqi): Rename ...
8970 (nds32_expand_movmemsi): ... to this.
8971 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8972 (movmemsi): ... to this.
8973 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8974 (nds32_expand_movmemsi): ... to this.
8976 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8977 Monk Chiang <sh.chiang04@gmail.com>
8978 Chung-Ju Wu <jasonwucj@gmail.com>
8980 * config/nds32/nds32-protos.h
8981 (nds32_expand_load_multiple): New arguments.
8982 (nds32_expand_store_multiple): Ditto.
8983 (nds32_valid_multiple_load_store): Rename ...
8984 (nds32_valid_multiple_load_store_p): ... to this.
8985 * config/nds32/nds32-memory-manipulation.c
8986 (nds32_expand_load_multiple): Refine implementation.
8987 (nds32_expand_store_multiple): Ditto.
8988 * config/nds32/nds32-multiple.md
8989 (load_multiple): Update nds32_expand_load_multiple interface.
8990 (store_multiple): Update nds32_expand_store_multiple interface.
8991 * config/nds32/nds32-predicates.c
8992 (nds32_valid_multiple_load_store): Rename ...
8993 (nds32_valid_multiple_load_store_p): ... to this and refine
8995 * config/nds32/predicates.md
8996 (nds32_load_multiple_and_update_address_operation): New predicate.
8997 (nds32_store_multiple_and_update_address_operation): New predicate.
8999 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9000 Chung-Ju Wu <jasonwucj@gmail.com>
9002 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9003 (combo): New attribute.
9004 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9006 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9008 * config/nds32/nds32.opt: Change -mcmodel= default value.
9010 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9011 Monk Chiang <sh.chiang04@gmail.com>
9012 Chung-Ju Wu <jasonwucj@gmail.com>
9014 * config/nds32/constants.md (unspec_element): New enum.
9015 * config/nds32/constraints.md (Umw): New constraint.
9016 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9017 * config/nds32/nds32-intrinsic.md: Likewise.
9018 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9019 (nds32_valid_smw_lwm_base_p): New.
9020 (nds32_output_smw_single_word): New.
9021 (nds32_output_lmw_single_word): New.
9022 (nds32_expand_unaligned_load): New.
9023 (nds32_expand_unaligned_store): New.
9024 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9025 (nds32_output_smw_single_word): Declare.
9026 (nds32_output_lmw_single_word): Declare.
9027 (nds32_expand_unaligned_load): Declare.
9028 (nds32_expand_unaligned_store): Declare.
9029 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9030 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9031 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9032 NDS32_BUILTIN_UASTORE_DW.
9033 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9036 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9037 Kito Cheng <kito.cheng@gmail.com>
9038 Chung-Ju Wu <jasonwucj@gmail.com>
9040 * config/nds32/nds32-intrinsic.c
9041 (nds32_expand_builtin_null_ftype_reg): Delete.
9042 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9043 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9044 (nds32_read_argument): New.
9045 (nds32_legitimize_target): Ditto.
9046 (nds32_legitimize_argument): Ditto.
9047 (nds32_check_constant_argument): Ditto.
9048 (nds32_expand_unop_builtin): Ditto.
9049 (nds32_expand_unopimm_builtin): Ditto.
9050 (nds32_expand_binop_builtin): Ditto.
9051 (nds32_builtin_decl_impl): Ditto.
9052 (builtin_description): Ditto.
9053 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9054 (nds32_init_builtins_impl): Ditto.
9055 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9056 (nds32_builtin_decl): New.
9057 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9058 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9060 2018-03-02 Jeff Law <law@redhat.com>
9062 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9063 (redundant_insn, fill_simple_delay_slots): Likewise.
9064 (fill_slots_from_thread): Likewise.
9065 * resource.c (mark_referenced_resources): Likewise.
9066 (mark_set_resources, find_dead_or_set_registers): Likewise.
9068 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9070 * substring-locations.h (format_warning_va): Formatting fix for
9072 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9074 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9075 * substring-locations.c: Include intl.h.
9076 (format_warning_va): Turned into small wrapper around
9077 format_warning_n_va, renamed to ...
9078 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9079 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9081 (format_warning_at_substring_n): New function.
9082 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9083 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9084 format_warning_at_substring with just a shorter name instead of
9085 const function pointer.
9086 (fmtwarn_n): New function.
9087 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9088 appropriate, get rid of all the fmtstr temporaries, move conditionals
9089 with G_() wrapped string literals directly into fmtwarn arguments,
9090 cast dir.len to (int), formatting fixes.
9092 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9094 * doc/invoke.texi: Remove "Cilk Plus" references.
9096 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9097 Richard Biener <rguenther@suse.de>
9100 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9101 for error or warning attributes if CALL_FROM_THUNK_P is set.
9104 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9107 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9108 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9111 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9112 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9114 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9115 instead of -1U in last predictors element's probability member.
9117 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9120 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9121 arguments if they are comparable.
9123 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9125 PR tree-optimization/84634
9126 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9127 masks and masked_loop_p with a single loop_masks, making sure it's
9128 null for bb vectorization.
9130 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9132 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9133 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9134 loop->force_vectorize to check whether there is no alias.
9136 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9139 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9141 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9143 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9144 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9145 instead of a loop around prev_real_insn.
9146 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9150 * config/i386/i386.c (ix86_print_operand): Use conditional
9151 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9154 2018-03-02 Richard Biener <rguenther@suse.de>
9156 PR tree-optimization/84427
9157 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9158 (bitmap_set_subtract_values): Rewrite to handle multiple
9161 (prune_clobbered_mems): Likewise.
9162 (phi_translate): Take edge instead of pred/phiblock.
9163 (phi_translate_1): Likewise.
9164 (phi_translate_set): Likewise. Insert all translated
9165 exprs for a value into the set, keeping possibly multiple
9166 expressions per value.
9167 (compute_antic_aux): Adjust for phi_translate changes.
9168 When intersecting union the expressions and prune those
9169 not in the final value set, keeping possibly multiple
9170 expressions per value. Do not use value-insertion
9171 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9172 all expressions. Add verification that the value-sets
9173 only shrink during iteration.
9174 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9175 (do_pre_regular_insertion): Likewise.
9176 (do_pre_partial_partial_insertion): Likewise.
9178 2018-03-02 Richard Biener <rguenther@suse.de>
9181 * config/darwin.c (saved_debug_info_level): New static global.
9182 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9183 (darwin_asm_lto_end): Restore debug info generation settings.
9185 2018-03-01 Martin Liska <mliska@suse.cz>
9188 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9191 2018-03-01 Richard Biener <rguenther@suse.de>
9194 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9195 type annotation with LTO when debug was disabled at compile-time.
9197 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9199 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9201 (mips_final_postscan_insn): Likewise.
9203 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9205 PR rtl-optimization/84528
9206 * alias.c (init_alias_target): Add commentary.
9207 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9208 a unique base value if the frame pointer is not eliminated
9209 to the stack pointer.
9211 2018-03-01 Tom de Vries <tom@codesourcery.com>
9213 PR rtl-optimization/83327
9214 * lra-int.h (hard_regs_spilled_into): Declare.
9215 * lra.c (hard_regs_spilled_into): Define.
9216 (init_reg_info): Init hard_regs_spilled_into.
9217 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9218 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9219 (process_bb_lives): Handle hard_regs_spilled_into.
9220 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9221 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9223 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9225 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9226 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9227 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9228 * config/rs6000/aix72.h: New file.
9230 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9232 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9233 instead of warning_at with conditional singular and plural messages
9237 * stor-layout.c (update_alignment_for_field): For
9238 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9239 && !DECL_PACKED (field), do the alignment update, just use
9240 only desired_align instead of MAX (type_align, desired_align)
9242 (place_field): Don't do known_align < desired_align handling
9243 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9244 is non-NULL, instead do it after rli->prev_field handling and
9245 only if not within a bitfield word. For DECL_PACKED (field)
9246 use type_align of BITS_PER_UNIT.
9248 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9250 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9251 superfluous parentheses and trailing spaces.
9253 2018-02-28 Richard Biener <rguenther@suse.de>
9255 PR tree-optimization/84584
9256 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9257 SCoPs with fake exit edge.
9259 2018-02-28 Martin Liska <mliska@suse.cz>
9262 * timevar.c (timer::print): Fix format to properly print 100%
9265 2018-02-28 Richard Biener <rguenther@suse.de>
9268 * genmatch.c (capture_info::walk_match): Do not mark
9269 captured expressions without operands as expr_p given
9270 they act more like predicates and should be subject to
9271 "lost tail" side-effect preserving.
9273 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9275 PR rtl-optimization/81611
9276 * auto-inc-dec.c (attempt_change): Move dead note from
9277 mem_insn if it's the next use of regno
9278 (find_address): Take address use of reg holding
9279 non-incremented value. Add parm to limit search to the named
9281 (merge_in_block): Attempt to use a mem insn that is the next
9282 use of the original regno.
9284 2018-02-27 Martin Sebor <msebor@redhat.com>
9287 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9288 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9290 2018-02-27 Martin Sebor <msebor@redhat.com>
9292 PR translation/84207
9293 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9294 n argument to unsigned HOST_WIDE_INT.
9295 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9296 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9297 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9298 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9300 2018-02-27 Richard Biener <rguenther@suse.de>
9302 PR tree-optimization/84512
9303 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9304 Do not use the estimate returned from record_stmt_cost for
9305 the scalar iteration cost but sum properly using add_stmt_cost.
9307 2018-02-27 Richard Biener <rguenther@suse.de>
9309 PR tree-optimization/84466
9310 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9311 Adjust last change to less strictly validate use operands.
9313 2018-02-27 Martin Liska <mliska@suse.cz>
9315 PR gcov-profile/84548
9316 * gcov.c (process_file): Allow partial overlap and consider it
9317 also as group functions.
9318 (output_lines): Properly calculate range of lines for a group.
9320 2018-02-27 Martin Liska <mliska@suse.cz>
9322 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9323 'ggc' suffixes. Change first column width.
9324 (timer::print): Fix formatting of the column.
9326 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9328 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9329 preserve inline entry blocks for the sake of debug inline
9330 entry point markers alone.
9331 (remove_unused_locals): Suggest in comments a better place to
9332 force the preservation of inline entry blocks that are
9333 otherwise unused, but do not preserve them.
9335 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9337 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9339 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9342 * config/i386/constraints.md (Bs): Replace
9343 ix86_indirect_branch_register with
9344 TARGET_INDIRECT_BRANCH_REGISTER.
9346 * config/i386/i386.md (indirect_jump): Likewise.
9347 (tablejump): Likewise.
9348 (*sibcall_memory): Likewise.
9349 (*sibcall_value_memory): Likewise.
9350 Peepholes of indirect call and jump via memory: Likewise.
9351 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9352 (*sibcall_value_GOT_32): Likewise.
9353 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9354 (GOT_memory_operand): Likewise.
9355 (call_insn_operand): Likewise.
9356 (sibcall_insn_operand): Likewise.
9357 (GOT32_symbol_operand): Likewise.
9358 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9360 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9362 PR rtl-optimization/83496
9363 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9364 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9365 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9366 redundant insn, if any.
9367 (relax_delay_slots): Likewise.
9368 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9370 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9372 PR tree-optimization/83965
9373 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9374 that grouped statements are part of a reduction chain. Return
9375 true if the statement is not marked as a reduction itself but
9377 (vect_recog_dot_prod_pattern): Don't check whether the statement
9378 is part of a group here.
9379 (vect_recog_sad_pattern): Likewise.
9380 (vect_recog_widen_sum_pattern): Likewise.
9382 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9385 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9387 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9390 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9392 (ix86_output_indirect_function_return): New prototype.
9393 (ix86_split_simple_return_pop_internal): Likewise.
9394 * config/i386/i386.c (indirect_return_via_cx): New.
9395 (indirect_return_via_cx_bnd): Likewise.
9396 (indirect_thunk_name): Handle return va CX_REG.
9397 (output_indirect_thunk_function): Create alias for
9398 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9399 (ix86_output_indirect_jmp): Remove the bool argument.
9400 (ix86_output_indirect_function_return): New function.
9401 (ix86_split_simple_return_pop_internal): Likewise.
9402 * config/i386/i386.md (*indirect_jump): Don't pass false
9403 to ix86_output_indirect_jmp.
9404 (*tablejump_1): Likewise.
9405 (simple_return_pop_internal): Change it to define_insn_and_split.
9406 Call ix86_split_simple_return_pop_internal to split it for
9408 (simple_return_indirect_internal): Call
9409 ix86_output_indirect_function_return instead of
9410 ix86_output_indirect_jmp.
9412 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9415 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9416 memset and value initialization afterwards.
9418 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9420 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9422 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9425 * common/config/aarch64/aarch64-common.c
9426 (aarch_option_optimization_table[]): Switch
9427 off fomit-frame-pointer
9429 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9430 Chung-Ju Wu <jasonwucj@gmail.com>
9432 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9434 (store_multiple): Ditto.
9436 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9438 * config.gcc: Add --with-cpu support for nds32 target.
9439 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9440 * config/nds32/nds32.opt: Add -mcpu= option.
9442 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9444 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9445 isel=yes): Warn for these deprecated options.
9447 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9449 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9450 ISA_2_5_MASKS_EMBEDDED.
9452 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9454 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9455 p->max as pointers rather than using iterative_hash_expr.
9457 2018-02-23 Carl Love <cel@us.ibm.com>
9459 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9460 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9462 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9463 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9464 P8V_BUILTIN_VEC_VUNSIGNED2.
9466 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9469 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9470 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9472 * lra-constraints.c (curr_insn_transform): Set up
9473 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9475 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9476 * lra-eliminations.c (spill_pseudos): Ditto.
9477 (process_insn_for_elimination): Ditto.
9478 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9479 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9480 LRA_NON_CLOBBERED_ALT.
9482 2018-02-22 Martin Sebor <msebor@redhat.com>
9484 PR tree-optimization/84480
9485 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9486 to maybe_diag_stxncpy_trunc. Call it.
9487 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9488 from gimple_fold_builtin_strcpy. Print inlining stack.
9489 (handle_builtin_stxncpy): Print inlining stack.
9490 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9492 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9495 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9496 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9497 and -fcheck-pointer-bounds are used together.
9498 (indirect_thunk_prefix): New enum.
9499 (indirect_thunk_need_prefix): New function.
9500 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9501 "_nt" instead of "_bnd" for NOTRACK prefix.
9502 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9503 (output_indirect_thunk_function): Likewise.
9505 (ix86_code_end): Update output_indirect_thunk_function calls.
9506 (ix86_output_indirect_branch_via_reg): Replace
9507 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9508 (ix86_output_indirect_branch_via_push): Likewise.
9509 (ix86_output_function_return): Likewise.
9510 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9511 incompatible with -fcf-protection=branch and
9512 -fcheck-pointer-bounds.
9514 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9517 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9518 Change gcc_assert call to output_operand_lossage.
9520 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9522 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9524 2018-02-22 DJ Delorie <dj@redhat.com>
9525 Sebastian Perta <sebastian.perta@renesas.com>
9526 Oleg Endo <olegendo@gcc.gnu.org>
9528 * config/rx/rx.c (rx_rtx_costs): New function.
9529 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9531 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9533 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9535 2018-02-22 Martin Liska <mliska@suse.cz>
9538 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9539 Add "native" as a possible value.
9541 2018-02-22 Martin Liska <mliska@suse.cz>
9544 * config/i386/i386.c (ix86_option_override_internal):
9545 Add "native" as a possible value for -march and -mtune.
9547 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9550 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9551 to all type variants.
9553 PR tree-optimization/84503
9554 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9555 width as info->bitpos + info->bitsize - start.
9556 (merged_store_group::merge_overlapping): Simplify width computation.
9557 (check_no_overlap): New function.
9558 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9559 start + width and last_order of the group, fail if check_no_overlap
9561 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9562 to group if check_no_overlap fails.
9564 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9566 * config/rs6000/altivec.md: Delete contraint arguments to
9567 define_expand, define_split, and define_peephole2, and in
9568 define_insn_and_split if always unused.
9569 * config/rs6000/darwin.md: Ditto.
9570 * config/rs6000/dfp.md: Ditto.
9571 * config/rs6000/rs6000.md: Ditto.
9572 * config/rs6000/sync.md: Ditto.
9573 * config/rs6000/vector.md: Ditto.
9574 * config/rs6000/vsx.md: Ditto.
9576 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9578 * config/rs6000/altivec.md: Write output control strings as braced
9579 blocks instead of double-quoted strings.
9580 * config/rs6000/darwin.md: Ditto.
9581 * config/rs6000/rs6000.md: Ditto.
9582 * config/rs6000/vector.md: Ditto.
9583 * config/rs6000/vsx.md: Ditto.
9585 2018-02-21 Jason Merrill <jason@redhat.com>
9587 PR c++/84314 - ICE with templates and fastcall attribute.
9588 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9590 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9592 * ipa-cp.c (determine_versionability): Fix comment typos.
9594 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9597 * ipa-cp.c (determine_versionability): Do not version functions caling
9600 2018-02-21 Martin Liska <mliska@suse.cz>
9603 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9604 Add "native" as a possible value.
9605 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9606 the macro when native cpu detection is available.
9608 2018-02-21 Martin Liska <mliska@suse.cz>
9611 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9612 Add "native" as a possible value.
9613 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9614 when native cpu detection is available.
9616 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9617 Martin Sebor <msebor@redhat.com>
9619 PR tree-optimization/84478
9620 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9622 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9623 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9624 support which is conservatively correct, for 2 only stay conservative
9625 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9626 argument to the 2 argument get_range_strlen, adjust 6 arg
9627 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9629 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9630 (gimple_fold_builtin_strlen): Pass true as last argument to
9633 2018-02-20 Martin Sebor <msebor@redhat.com>
9636 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9637 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9638 (builtin_memref::builtin_memref): Factor out parts into
9639 set_base_and_offset and call it.
9641 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9644 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9645 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9646 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9647 search at the associated MODE_INT.
9649 2018-02-20 Jeff Law <law@redhat.com>
9652 PR tree-optimization/81592
9654 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9655 for range data rather than using global data.
9656 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9657 range data rather than using global data.
9658 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9659 pass it to children as needed.
9660 (struct directive::fmtresult): Similarly.
9661 (struct directive::set_width): Similarly.
9662 (struct directive::set_precision): Similarly.
9663 (format_integer, format_directive, parse_directive): Similarly.
9664 (format_none): Accept unnamed vr_values parameter.
9665 (format_percent, format_floating, format_character): Similarly.
9666 (format_string, format_plain): Similarly.
9667 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9668 the EVRP range analyzer for range data rather than using global data.
9669 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9670 gimple-ssa-evrp-analyze.h
9671 (class sprintf_dom_walker): Add after_dom_children member function.
9672 Add evrp_range_analyzer member.
9673 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9674 range analyzer as needed.
9675 (sprintf_dom_walker::after_dom_children): New member function.
9676 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9678 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9679 (evrp_range_analyzer::pop_to_marker): Likewise.
9681 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9683 PR tree-optimization/84419
9684 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9685 with the required type if its current type is compatible but
9688 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9691 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9692 after vectorization.
9694 2018-02-20 Martin Liska <mliska@suse.cz>
9697 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9698 possible values if we don't have a hint.
9700 2018-02-20 Martin Liska <mliska@suse.cz>
9704 * final.c (shorten_branches): Build align_tab array with one
9706 * opts.c (finish_options): Add alignment option limit check.
9707 (MAX_CODE_ALIGN): Likewise.
9708 (MAX_CODE_ALIGN_VALUE): Likewise.
9709 * doc/invoke.texi: Document maximum allowed option value for
9710 all -falign-* options.
9712 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9715 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9716 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9717 * var-tracking.c (emit_note_insn_var_location): Remove all references
9718 to NOTE_INSN_CALL_ARG_LOCATION.
9719 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9720 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9721 Use copy_rtx_if_shared.
9722 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9723 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9724 (dwarf2out_var_location): Remove handling of
9725 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9727 * final.c (final_scan_insn): Remove all references to
9728 NOTE_INSN_CALL_ARG_LOCATION.
9729 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9730 before dumping final insns.
9731 * except.c (emit_note_eh_region_end): Remove all references to
9732 NOTE_INSN_CALL_ARG_LOCATION.
9733 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9734 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9735 * config/arc/arc.c (hwloop_optimize): Likewise.
9736 * config/arm/arm.c (create_fix_barrier): Likewise.
9737 * config/s390/s390.c (s390_chunkify_start): Likewise.
9738 * config/sh/sh.c (find_barrier): Likewise.
9739 * config/i386/i386.c (rest_of_insert_endbranch,
9740 ix86_seh_fixup_eh_fallthru): Likewise.
9741 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9742 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9743 * config/frv/frv.c (frv_function_prologue): Likewise.
9744 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9746 (note_outside_basic_block_p): Remove all references to
9747 NOTE_INSN_CALL_ARG_LOCATION.
9748 * gengtype.c (adjust_field_rtx_def): Likewise.
9749 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9751 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9752 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9755 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9758 PR tree-optimization/84452
9759 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9760 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9763 2018-02-19 Martin Liska <mliska@suse.cz>
9766 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9768 2018-02-19 Martin Liska <mliska@suse.cz>
9769 Richard Sandiford <richard.sandiford@linaro.org>
9771 PR tree-optimization/82491
9772 * gimple-fold.c (get_base_constructor): Make earlier bail out
9775 2018-02-19 Carl Love <cel@us.ibm.com>
9777 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9778 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9780 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9781 P8V_BUILTIN_VEC_NEG.
9783 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9785 * config/rl78/rl78.md (movdf): New define expand.
9787 2018-02-19 Martin Liska <mliska@suse.cz>
9790 * doc/invoke.texi: Fix typo.
9791 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9793 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9795 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9796 handle rs6000_single_float and rs6000_double_float specially for
9799 2018-02-16 Jeff Law <law@redhat.com>
9801 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9804 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9807 * ipa-inline.c (inline_small_functions): Fix a typo.
9809 2018-02-16 Nathan Sidwell <nathan@acm.org>
9811 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9813 2018-02-16 Carl Love <cel@us.ibm.com>
9815 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9816 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9817 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9818 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9819 expansion to P8V_BUILTIN_VEC_FLOAT2.
9821 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9823 PR rtl-optimization/70023
9824 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9825 src_regno into account.
9827 2018-02-16 Carl Love <cel@us.ibm.com>
9829 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9830 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9831 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9832 * config/rs6000/rs6000.c: Remove case statements for
9833 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9834 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9835 and P9V_BUILTIN_VEC_VINSERT4B.
9836 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9837 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9838 * config/rs6000/vsx.md:
9839 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9842 2018-02-16 Carl Love <cel@us.ibm.com>
9844 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9846 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9848 * config/rs6000/rs6000-c.c: Add the definitions for
9849 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9850 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9851 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9852 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9853 definition for insert4b and define insn *insert3b_internal.
9854 * doc/extend.texi: Add documentation for vec_extract4b.
9856 2018-02-16 Nathan Sidwell <nathan@acm.org>
9858 * doc/extend.texi (Backwards Compatibility): Mention friend
9859 injection. Note for-scope is deprecated.
9860 * doc/invoke.texi (-ffriend-injection): Deprecate.
9862 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9864 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9865 that moved to I2, also allow destinations that are a paradoxical
9866 subreg (instead of a normal reg).
9868 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9871 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9874 2018-02-16 Richard Biener <rguenther@suse.de>
9876 PR tree-optimization/84037
9877 PR tree-optimization/84016
9879 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9880 Adjust vec_construct for the fact we need additional higher latency
9881 128bit inserts for AVX256 and AVX512 vector builds.
9882 (ix86_add_stmt_cost): Scale vector construction cost for
9885 2018-02-16 Richard Biener <rguenther@suse.de>
9887 PR tree-optimization/84417
9888 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9889 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9890 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9892 2018-02-16 Martin Liska <mliska@suse.cz>
9895 * internal-fn.def (ASAN_CHECK): Set proper flags.
9896 (ASAN_MARK): Likewise.
9898 2018-02-16 Julia Koval <julia.koval@intel.com>
9900 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9901 from PTA_CANNONLAKE.
9903 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9906 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9907 Use ++iter rather than iter++ for std::list iterators.
9908 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9909 defer deleting them until all nodes in the forest are processed. Do
9910 free even leaf nodes. Change to_process into auto_vec.
9913 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9914 * vec.h (vec_default_construct): Use memset instead of placement new
9915 if BROKEN_VALUE_INITIALIZATION is defined.
9916 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9917 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9920 PR rtl-optimization/83723
9921 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9922 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9923 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9925 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9927 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9929 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9931 PR rtl-optimization/81443
9932 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9933 from inner REGs to paradoxical SUBREGs.
9935 2018-02-16 Richard Biener <rguenther@suse.de>
9937 PR tree-optimization/84399
9938 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9939 For operands we can analyze at their definition make sure we can
9940 analyze them at each use as well.
9942 2018-02-16 Richard Biener <rguenther@suse.de>
9944 PR tree-optimization/84190
9945 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9946 volatile accesses if the decl isn't volatile.
9948 2018-02-15 Jason Merrill <jason@redhat.com>
9950 PR c++/84314 - ICE with templates and fastcall attribute.
9951 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9952 TYPE_CANONICAL on an existing type.
9954 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9956 PR tree-optimization/84383
9957 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9958 dstoff nor call operand_equal_p if dstbase is NULL.
9960 PR tree-optimization/84334
9961 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9962 also a CONSTANT_CLASS_P, punt.
9964 2018-02-14 Jim Wilson <jimw@sifive.com>
9966 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9967 first SMALL_OPERAND check. New local min_second_step. Move assert
9968 to where locals are set. Add TARGET_RVC support.
9969 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9971 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9973 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9975 2018-02-14 Martin Sebor <msebor@redhat.com>
9977 PR tree-optimization/83698
9978 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9979 arrays constrain the offset range to their bounds.
9980 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9981 (builtin_access::overlap): Avoid setting the size of overlap if it's
9983 (maybe_diag_overlap): Also consider arrays when deciding what values
9984 of offsets to include in diagnostics.
9986 2018-02-14 Martin Sebor <msebor@redhat.com>
9989 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9990 that correspond to the kind of a declaration.
9992 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9995 * config/pa/pa.md: Load address of PIC label using the linkage table
9996 if the label is nonlocal.
9998 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10000 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10001 warning message if user requests -maltivec=be.
10002 * doc/invoke.texi: Document deprecation of -maltivec=be.
10004 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10007 * config/rs6000/rs6000-c.c: Update definitions for
10008 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10009 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10011 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10014 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10015 add _get_ssp intrinsics. Remove argument from
10016 __builtin_ia32_rdssp[d|q].
10017 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10018 * config/i386/i386-builtin.def: Remove argument from
10019 __builtin_ia32_rdssp[d|q].
10020 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10021 ix86_expand_special_args_builtin for _rdssp[d|q].
10022 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10023 Clear register before usage.
10024 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10025 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10027 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10029 PR tree-optimization/84357
10030 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10031 operand 1 of an ARRAY_REF too.
10033 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10036 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10037 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10039 (set_of_reg): New struct.
10040 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10041 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10042 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10044 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10045 Split into bitclr, bitset, bitinvert patterns if appropriate.
10046 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10047 use rx_fuse_in_memory_bitop.
10048 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10049 to named insn, correct maximum insn length.
10051 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10054 * machmode.def: Define a complex mode for PARTIAL_INT.
10055 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10057 * doc/rtl.texi: Document CSPImode.
10058 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10060 (msp430_hard_regno_nregs_with_padding): Likewise.
10062 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10065 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10067 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10069 PR rtl-optimization/84169
10070 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10071 we generated a parallel as new i3 and we split that to new i2 and i3
10072 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10073 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10074 those to i2, not i1. Partially rewrite this scan code.
10076 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10079 * stor-layout.c (place_field): For variable length fields, adjust
10080 offset_align afterwards not just based on the field's alignment,
10081 but also on the size.
10083 PR middle-end/84309
10084 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10085 of exps and logs in the use_exp2 case.
10087 2018-02-13 Jeff Law <law@redhat.com>
10089 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10090 entry for "vector".
10092 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10095 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10099 * common.opt (gas-loc-support, gas-locview-support): New.
10100 (ginline-points, ginternal-reset-location-views): New.
10101 * doc/invoke.texi: Document them. Use @itemx where intended.
10102 (gvariable-location-views): Adjust.
10103 * target.def (reset_location_view): New.
10104 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10105 (TARGET_RESET_LOCATION_VIEW): New.
10106 * doc/tm.texi: Rebuilt.
10107 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10108 (dwarf2out_default_as_locview_support): New.
10109 (output_asm_line_debug_info): Use option variables.
10110 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10111 (output_loc_list): Likewise.
10112 (add_high_low_attributes): Check option variables.
10113 Don't output entry view attribute in strict mode.
10114 (gen_inlined_subroutine_die): Check option variables.
10115 (dwarf2out_inline_entry): Likewise.
10116 (init_sections_and_labels): Likewise.
10117 (dwarf2out_early_finish): Likewise.
10118 (maybe_reset_location_view): New, from...
10119 (dwarf2out_var_location): ... here. Call it.
10120 * debug.h (dwarf2out_default_as_loc_support): Declare.
10121 (dwarf2out_default_as_locview_support): Declare.
10122 * hooks.c (hook_int_rtx_insn_0): New.
10123 * hooks.h (hook_int_rtx_insn_0): Declare.
10124 * toplev.c (process_options): Take -gas-loc-support and
10125 -gas-locview-support from dwarf2out. Enable
10126 -gvariable-location-views by default only with locview
10127 assembler support. Enable -ginternal-reset-location-views by
10128 default only if the target defines the corresponding hook.
10129 Enable -ginline-points by default if location views are
10130 enabled; force it disabled if statement frontiers are
10132 * tree-inline.c (expand_call_inline): Check option variables.
10133 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10135 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10137 PR tree-optimization/84321
10138 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10139 handling. Also check whether the anti-range contains any values
10140 that satisfy the mask; switch to a VR_RANGE if not.
10142 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10145 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10147 2018-02-13 Martin Jambor <mjambor@suse.cz>
10150 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10151 of call statements, also set location of a load to a temporary.
10153 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10155 * config/rl78/rl78.c (add_vector_labels): New function.
10156 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10157 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10158 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10159 which checks that no arguments are passed.
10160 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10161 * doc/extend.texi: Documentation for the new attribute.
10163 2018-02-13 Andreas Schwab <schwab@suse.de>
10165 * config/riscv/linux.h (CPP_SPEC): Define.
10167 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10170 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10171 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10172 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10173 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10174 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10175 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10176 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10177 temporarily for AES and PCLMUL builtins.
10179 PR tree-optimization/84339
10180 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10181 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10184 PR middle-end/84309
10185 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10186 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10187 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10189 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10191 * omp-simd-clone.h: New file.
10192 * omp-simd-clone.c: Include omp-simd-clone.h.
10193 (expand_simd_clones): No longer static.
10194 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10195 cgraph.h and omp-simd-clone.h.
10196 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10197 (vect_recog_widen_shift_pattern): Formatting fix.
10198 (vect_pattern_recog_1): Don't check optab for calls.
10201 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10202 operands[2] into a REG before using gen_lowpart on it.
10204 2018-02-12 Jeff Law <law@redhat.com>
10207 * config/sh/sh.c (find_barrier): Consider a sibling call
10210 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10211 successfully back substituting a reg.
10213 2018-02-12 Richard Biener <rguenther@suse.de>
10215 PR tree-optimization/84037
10216 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10217 parameter, move visited init to caller.
10218 (vect_slp_analyze_operations): Separate cost from validity
10219 check, initialize visited once for all instances.
10220 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10222 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10224 (vect_model_promotion_demotion_cost): Likewise.
10225 (vectorizable_bswap): Guard cost modeling with !slp_node
10226 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10228 (vectorizable_call): Likewise.
10229 (vectorizable_conversion): Likewise.
10230 (vectorizable_assignment): Likewise.
10231 (vectorizable_shift): Likewise.
10232 (vectorizable_operation): Likewise.
10233 (vectorizable_store): Likewise.
10234 (vectorizable_load): Likewise.
10235 (vectorizable_condition): Likewise.
10236 (vectorizable_comparison): Likewise.
10238 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10241 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10242 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10243 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10245 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10247 PR middle-end/83665
10248 * params.def (inline-min-speedup): Increase from 8 to 15.
10249 (max-inline-insns-auto): Decrease from 40 to 30.
10250 * ipa-split.c (consider_split): Add some buffer for function to
10251 be considered inlining candidate.
10252 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10255 2018-02-12 Richard Biener <rguenther@suse.de>
10257 PR tree-optimization/84037
10258 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10259 matched stmts if we cannot swap the non-matched ones.
10261 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10263 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10264 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10265 _mm_maskz_scalef_round_ss): New intrinsics.
10266 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10267 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10268 __builtin_ia32_scalefss_round): Remove.
10269 (__builtin_ia32_scalefsd_mask_round,
10270 __builtin_ia32_scalefss_mask_round): New intrinsics.
10271 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10272 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10273 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10274 "<round_constraint>")): Changed to ...
10275 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10276 "<round_scalar_constraint>")): ... this.
10277 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10278 %0, %1, %2<round_op3>}"): Changed to ...
10279 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10280 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10281 %2<round_scalar_mask_op3>}"): ... this.
10282 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10284 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10286 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10287 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10288 (_mm_maskz_sqrt_round_ss): New intrinsics.
10289 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10290 (__builtin_ia32_sqrtsd_mask_round)
10291 (__builtin_ia32_sqrtss_mask_round): New builtins.
10292 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10293 (__builtin_ia32_sqrtss_round): Remove.
10294 (__builtin_ia32_sqrtsd_mask_round)
10295 (__builtin_ia32_sqrtss_mask_round): New builtins.
10296 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10297 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10298 ((match_operand:VF_128 1 "vector_operand"
10299 "xBm,<round_constraint>")): Changed to ...
10300 ((match_operand:VF_128 1 "vector_operand"
10301 "xBm,<round_scalar_constraint>")): ... this.
10302 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10303 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10304 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10305 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10306 %<iptr>1<round_scalar_mask_op3>}): ... this.
10307 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10308 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10310 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10313 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10314 Cast vec_cmpeq result to correct type.
10315 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10316 Cast vec_cmpgt result to correct type.
10318 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10320 * final.c (final_scan_insn_1): Renamed from...
10321 (final_scan_insn): ... this. New wrapper, to recover
10322 seen from the outermost call in recursive ones.
10323 * config/sparc/sparc.c (output_return): Drop seen from call.
10324 (output_sibcall): Likewise.
10325 * config/visium/visium.c (output_branch): Likewise.
10327 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10329 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10332 2018-02-10 Alan Modra <amodra@gmail.com>
10335 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10336 Specify LR as an input.
10338 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10341 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10342 remove_member_access_dummy_vars): New functions.
10343 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10344 lower_omp_1, execute_lower_omp): Use them.
10346 PR rtl-optimization/84308
10347 * shrink-wrap.c (spread_components): Release todo vector.
10349 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10351 PR rtl-optimization/57193
10352 * ira-color.c (struct allocno_color_data): Add member
10353 conflict_allocno_hard_prefs.
10354 (update_conflict_allocno_hard_prefs): New.
10355 (bucket_allocno_compare_func): Add a preference based on
10356 conflict_allocno_hard_prefs.
10357 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10358 (color_allocnos): Remove a dead code. Initiate
10359 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10361 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10364 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10365 constraint from =wa to wa. Avoid a subreg on the output operand,
10366 instead use a pseudo and subreg it in a move.
10367 (p9_xxbrd_<mode>): Changed to ...
10368 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10369 (p9_xxbrd_v2df): New expander.
10370 (p9_xxbrw_<mode>): Changed to ...
10371 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10372 (p9_xxbrw_v4sf): New expander.
10374 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10376 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10378 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10381 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10382 multiply in 32-bit mode.
10383 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10384 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10387 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10389 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10390 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10391 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10392 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10394 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10397 * dwarf2out.c (is_trivial_indirect_ref): New function.
10398 (dwarf2out_late_global_decl): Do not generate a location
10399 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10400 and that are not defined in the current unit.
10402 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10404 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10405 instead of a libcall for UNORDERED.
10407 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10410 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10411 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10413 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10416 * config/s390/s390.c (s390_set_current_function): Invoke
10417 s390_indirect_branch_settings also if fndecl didn't change.
10419 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10421 * config/rs6000/rs6000.md (blockage): Set length to zero.
10423 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10425 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10427 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10430 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10431 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10435 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10436 PARALLEL incoming that failed vt_get_decl_and_offset check.
10438 PR middle-end/84237
10439 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10440 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10442 (get_variable_section): For decls in named .bss* sections pass true as
10443 second argument to bss_initializer_p.
10445 2018-02-09 Marek Polacek <polacek@redhat.com>
10446 Jakub Jelinek <jakub@redhat.com>
10449 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10450 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10451 Sync some changes from cxx_fold_indirect_ref.
10453 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10455 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10457 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10458 (BLOCK_INLINE_ENTRY_LABEL): New.
10459 (dwarf2out_var_location): Disregard inline entry markers.
10460 (inline_entry_data): New struct.
10461 (inline_entry_data_hasher): New hashtable type.
10462 (inline_entry_data_hasher::hash): New.
10463 (inline_entry_data_hasher::equal): New.
10464 (inline_entry_data_table): New variable.
10465 (add_high_low_attributes): Add DW_AT_entry_pc and
10466 DW_AT_GNU_entry_view attributes if a pending entry is found
10467 in inline_entry_data_table. Add old entry_pc attribute only
10468 if debug nonbinding markers are disabled.
10469 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10470 markers are enabled.
10471 (block_within_block_p, dwarf2out_inline_entry): New.
10472 (dwarf2out_finish): Check that no entries remained in
10473 inline_entry_data_table.
10474 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10475 (final_scan_insn, notice_source_line): Likewise.
10476 (rest_of_clean_state): Skip inline entry markers.
10477 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10479 * gimple.c (gimple_build_debug_inline_entry): New.
10480 * gimple.h (enum gimple_debug_subcode): Add
10481 GIMPLE_DEBUG_INLINE_ENTRY.
10482 (gimple_build_debug_inline_entry): Declare.
10483 (gimple_debug_inline_entry_p): New.
10484 (gimple_debug_nonbind_marker_p): Adjust.
10485 * insn-notes.def (INLINE_ENTRY): New.
10486 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10487 inline entry marker notes.
10488 (print_insn): Likewise.
10489 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10490 (INSN_DEBUG_MARKER_KIND): Likewise.
10491 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10492 * tree-inline.c (expand_call_inline): Build and insert
10493 debug_inline_entry stmt.
10494 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10495 inline entry blocks early, if nonbind markers are enabled.
10496 (dump_scope_block): Dump fragment info.
10497 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10498 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10499 (gimple_build_debug_inline_entry): New.
10500 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10501 Enable/disable inline entry points too.
10502 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10503 (DEBUG_INSN): Describe inline entry markers.
10505 * common.opt (gvariable-location-views): New.
10506 (gvariable-location-views=incompat5): New.
10507 * config.in: Rebuilt.
10508 * configure: Rebuilt.
10509 * configure.ac: Test assembler for view support.
10510 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10511 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10512 * dwarf2out.c (var_loc_view): New typedef.
10513 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10514 (dwarf2out_locviews_in_attribute): New.
10515 (dwarf2out_locviews_in_loclist): New.
10516 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10517 (enum dw_line_info_opcode): Add LI_adv_address.
10518 (struct dw_line_info_table): Add view.
10519 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10520 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10521 (zero_view_p): New variable.
10522 (ZERO_VIEW_P): New macro.
10523 (output_asm_line_debug_info): New.
10524 (struct var_loc_node): Add view.
10525 (add_AT_view_list, AT_loc_list): New.
10526 (add_var_loc_to_decl): Add view param. Test it against last.
10527 (new_loc_list): Add view params. Record them.
10528 (AT_loc_list_ptr): Handle loc and view lists.
10529 (view_list_to_loc_list_val_node): New.
10530 (print_dw_val): Handle dw_val_class_view_list.
10531 (size_of_die): Likewise.
10532 (value_format): Likewise.
10533 (loc_list_has_views): New.
10534 (gen_llsym): Set vl_symbol too.
10535 (maybe_gen_llsym, skip_loc_list_entry): New.
10536 (dwarf2out_maybe_output_loclist_view_pair): New.
10537 (output_loc_list): Output view list or entries too.
10538 (output_view_list_offset): New.
10539 (output_die): Handle dw_val_class_view_list.
10540 (output_dwarf_version): New.
10541 (output_compilation_unit_header): Use it.
10542 (output_skeleton_debug_sections): Likewise.
10543 (output_rnglists, output_line_info): Likewise.
10544 (output_pubnames, output_aranges): Update version comments.
10545 (output_one_line_info_table): Output view numbers in asm comments.
10546 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10547 Call maybe_gen_llsym.
10548 (loc_list_from_tree_1): Adjust.
10549 (add_AT_location_description): Create view list attribute if
10550 needed, check it's absent otherwise.
10551 (convert_cfa_to_fb_loc_list): Adjust.
10552 (maybe_emit_file): Call output_asm_line_debug_info for test.
10553 (dwarf2out_var_location): Reset views as needed. Precompute
10554 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10555 attribute. Set view.
10556 (new_line_info_table): Reset next view.
10557 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10558 (dwarf2out_source_line): Likewise. Output view resets and labels to
10559 the assembler, or select appropriate line info opcodes.
10560 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10561 (optimize_string_length): Catch it. Adjust.
10562 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10563 dw_val_class_view_list, and remove it if no longer needed.
10564 (hash_loc_list): Hash view numbers.
10565 (loc_list_hasher::equal): Compare them.
10566 (optimize_location_lists): Check whether a view list symbol is
10567 needed, and whether the locview attribute is present, and
10568 whether they match. Remove the locview attribute if no longer
10570 (index_location_lists): Call skip_loc_list_entry for test.
10571 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10572 Use output_dwarf_version.
10573 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10574 (struct dw_val_node): Add val_view_list.
10575 * final.c (SEEN_NEXT_VIEW): New.
10576 (set_next_view_needed): New.
10577 (clear_next_view_needed): New.
10578 (maybe_output_next_view): New.
10579 (final_start_function): Rename to...
10580 (final_start_function_1): ... this. Take pointer to FIRST,
10581 add SEEN parameter. Emit param bindings in the initial view.
10582 (final_start_function): Reintroduce SEEN-less interface.
10583 (final): Rename to...
10584 (final_1): ... this. Take SEEN parameter. Output final pending
10585 next view at the end.
10586 (final): Reintroduce seen-less interface.
10587 (final_scan_insn): Output pending next view before switching
10588 sections or ending a block. Mark the next view as needed when
10589 outputting variable locations. Notify debug backend of section
10590 changes, and of location view changes.
10591 (rest_of_handle_final): Adjust.
10592 * toplev.c (process_options): Autodetect value for debug variable
10593 location views option. Warn on incompat5 without -gdwarf-5.
10594 * doc/invoke.texi (gvariable-location-views): New.
10595 (gvariable-location-views=incompat5): New.
10596 (gno-variable-location-views): New.
10598 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10600 PR tree-optimization/84136
10601 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10602 that the result of find_edge is non-NULL.
10604 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10607 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10608 storing integer register in SImode. Fix cost of 256 and 512
10609 byte aligned SSE register store.
10611 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10613 * config/i386/i386.c (ix86_multiplication_cost): Fix
10614 multiplication cost for TARGET_AVX512DQ.
10616 2018-02-08 Marek Polacek <polacek@redhat.com>
10618 PR tree-optimization/84238
10619 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10622 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10624 PR tree-optimization/84265
10625 * tree-vect-stmts.c (vectorizable_store): Don't treat
10626 VMAT_CONTIGUOUS accesses as grouped.
10627 (vectorizable_load): Likewise.
10629 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10631 PR tree-optimization/81635
10632 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10633 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10634 (test_round_for_mask): New functions.
10635 (wide_int_cc_tests): Call test_round_for_mask.
10636 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10637 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10638 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10639 range returned by get_range_info.
10641 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10644 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10645 * symtab.c: Include builtins.h
10646 (symtab_node::output_to_lto_symbol_table_p): Move here
10647 from lto-streamer-out.c:output_symbol_p.
10648 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10649 (output_symbol_p): Move all logic to symtab.c
10650 (produce_symtab): Update.
10652 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10654 * config/s390/s390-opts.h (enum indirect_branch): Define.
10655 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10656 (s390_indirect_branch_via_thunk)
10657 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10658 (enum s390_indirect_branch_type): Define.
10659 * config/s390/s390.c (struct s390_frame_layout, struct
10660 machine_function): Remove.
10661 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10662 (indirect_branch_table_label_no, indirect_branch_table_name):
10664 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10665 (enum s390_indirect_branch_option): Define.
10666 (s390_return_addr_from_memory): New function.
10667 (s390_handle_string_attribute): New function.
10668 (s390_attribute_table): Add new attribute handler.
10669 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10670 (s390_indirect_branch_via_thunk): New function.
10671 (s390_indirect_branch_via_inline_thunk): New function.
10672 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10673 sibling call optimization for non z10 compiles.
10674 (s390_emit_call): Force indirect branch target to be a single
10675 register. Add r1 clobber for non-z10 compiles.
10676 (s390_emit_epilogue): Emit return jump via return_use expander.
10677 (s390_reorg): Handle JUMP_INSNs as execute targets.
10678 (s390_option_override_internal): Perform validity checks for the
10679 new command line options.
10680 (s390_indirect_branch_attrvalue): New function.
10681 (s390_indirect_branch_settings): New function.
10682 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10683 (s390_output_indirect_thunk_function): New function.
10684 (s390_code_end): Implement target hook.
10685 (s390_case_values_threshold): Implement target hook.
10686 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10688 * config/s390/s390.h (struct s390_frame_layout)
10689 (struct machine_function): Move here from s390.c.
10690 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10691 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10692 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10693 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10694 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10695 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10696 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10697 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10698 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10699 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10700 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10701 (mnemonic attribute): Add values which aren't recognized
10703 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10704 pattern for branch conversion. Fix mnemonic attribute.
10705 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10706 indirect branch via thunk if requested.
10707 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10708 ("*indirect_jump"): Disable for branch conversion using out of
10710 ("indirect_jump_via_thunk<mode>_z10")
10711 ("indirect_jump_via_thunk<mode>")
10712 ("indirect_jump_via_inlinethunk<mode>_z10")
10713 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10714 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10715 ("casesi_jump_via_inlinethunk<mode>_z10")
10716 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10717 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10718 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10719 ("*indirect2_jump"): Disable for branch conversion.
10720 ("casesi_jump"): Turn into expander and expand patterns for branch
10722 ("return_use"): New expander.
10723 ("*return"): Emit return via thunk and rename it to ...
10724 ("*return<mode>"): ... this one.
10725 * config/s390/s390.opt: Add new options and and enum for the
10728 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10730 * lra-constraints.c (match_reload): Unconditionally use
10731 gen_lowpart_SUBREG, rather than selecting between that
10732 and equivalent gen_rtx_SUBREG code.
10734 2018-02-08 Richard Biener <rguenther@suse.de>
10736 PR tree-optimization/84233
10737 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10738 changed flag instead of boguously re-using phi_inserted.
10740 2018-02-08 Martin Jambor <mjambor@suse.cz>
10742 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10743 static local variables.
10745 2018-02-08 Richard Biener <rguenther@suse.de>
10747 PR tree-optimization/84278
10748 * tree-vect-stmts.c (vectorizable_store): When looking for
10749 smaller vector types to perform grouped strided loads/stores
10750 make sure the mode is supported by the target.
10751 (vectorizable_load): Likewise.
10753 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10755 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10756 Increase LDP/STP opportunities by adding adjacent callee-saves.
10758 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10760 PR rtl-optimization/84068
10761 PR rtl-optimization/83459
10762 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10764 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10766 PR tree-optimization/84224
10767 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10768 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10769 non-zero arguments.
10771 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10774 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10775 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10776 count, remove one USE.
10778 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10780 * doc/install.texi (Configuration): Document the
10781 --with-long-double-format={ibm,ieee} PowerPC configuration
10785 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10786 Convert from define_expand to be define_insn_and_split. Rework
10787 float/double/_Float128 conversions to QI/HI/SImode to work with
10788 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10789 conversions to QI/HImode types did a store and then a load to
10790 truncate the value. For conversions to VSX registers, don't split
10791 the insn, instead emit the code directly. Use the code iterator
10792 any_fix to combine signed and unsigned conversions.
10793 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10794 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10795 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10796 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10797 (fix_<mode>di2_hw): Likewise.
10798 (fixuns_<mode>di2_hw): Likewise.
10799 (fix_<mode>si2_hw): Likewise.
10800 (fixuns_<mode>si2_hw): Likewise.
10801 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10802 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10803 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10804 fix<uns>_trunc<SFDF:mode>si2_p8.
10805 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10806 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10807 (fix<uns>_<mode>_mem): Likewise.
10808 (fctiw<u>z_<mode>_mem): Likewise.
10809 (fix<uns>_<mode>_mem): Likewise.
10810 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10811 the register allocator from doing a direct move to the GPRs to do
10812 a store, and instead use the ISA 3.0 store byte/half-word from
10813 vector register instruction. For IEEE 128-bit floating point,
10814 also optimize stores of 32-bit ints.
10815 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10817 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10819 * genextract.c (push_pathstr_operand): New function to support
10821 (walk_rtx): Call push_pathstr_operand.
10822 (print_path): Support [a-zA-Z].
10824 2018-02-07 Richard Biener <rguenther@suse.de>
10826 PR tree-optimization/84037
10827 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10828 (cse_and_gimplify_to_preheader): Declare.
10829 (vect_get_place_in_interleaving_chain): Likewise.
10830 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10832 (_loop_vec_info::~_loop_vec_info): Delete it.
10833 (cse_and_gimplify_to_preheader): New function.
10834 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10835 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10836 (vectorizable_load): Likewise. For grouped stores always base
10837 the IV on the first element.
10838 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10839 condition before gimplifying.
10841 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10843 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10844 *DIV_EXPR and *MOD_EXPR.
10846 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10849 * config/i386/i386.c (ix86_option_override_internal): Mask out
10850 the CF_SET bit when checking -fcf-protection.
10852 2018-02-07 Tom de Vries <tom@codesourcery.com>
10855 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10858 2018-02-07 Richard Biener <rguenther@suse.de>
10860 PR tree-optimization/84204
10861 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10864 PR tree-optimization/84205
10865 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10866 special-case isl_ast_op_zdiv_r.
10868 PR tree-optimization/84223
10869 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10870 Only add conditions from within the region.
10871 (gather_bbs::after_dom_children): Adjust.
10873 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10876 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10877 * config/avr/avr.md: Only post-reload split REG-REG moves if
10878 either register is GENERAL_REG_P.
10880 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10882 PR tree-optimization/84235
10883 * tree-ssa-scopedtables.c
10884 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10885 if the subtraction is performed in floating point type where NaNs are
10886 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10887 build 1. Formatting fix.
10889 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10892 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10893 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10894 and skip it regardless of bb boundaries. Use CALL_P macro,
10895 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10896 unnecessarily, formatting fix.
10898 2018-02-06 Michael Collison <michael.collison@arm.com>
10900 * config/arm/thumb2.md:
10901 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10902 (*thumb_mov_notscc): Ditto.
10904 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10907 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10908 unsigned_fix, not "s".
10910 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10912 * configure.ac (gcc_fn_eh_frame_ro): New function.
10913 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10914 correct .eh_frame permissions.
10915 * configure: Regenerate.
10917 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10919 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10920 irrelevant options.
10922 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10924 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10925 Display warning message for -mno-speculate-indirect-jumps.
10927 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10929 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10931 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10933 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10935 PR tree-optimization/84225
10936 * tree-eh.c (find_trapping_overflow): Only call
10937 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10939 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10942 * config/i386/i386.c: Reimplement the check of possible options
10943 -mibt/-mshstk conbination. Change error messages.
10944 * doc/invoke.texi: Fix a typo: remove extra '='.
10946 2018-02-06 Marek Polacek <polacek@redhat.com>
10948 PR tree-optimization/84228
10949 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10951 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10954 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10955 emitted arch directives.
10956 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10957 __ARM_FEATURE_COPROC before changing architectures.
10959 2018-02-06 Richard Biener <rguenther@suse.de>
10961 * config/i386/i386.c (print_reg): Fix typo.
10962 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10964 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10966 * configure: Regenerate.
10968 2018-02-05 Martin Sebor <msebor@redhat.com>
10970 PR tree-optimization/83369
10971 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10974 2018-02-05 Martin Liska <mliska@suse.cz>
10976 * doc/invoke.texi: Cherry-pick upstream r323995.
10978 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10980 * ira.c (ira_init_register_move_cost): Adjust comment.
10982 2018-02-05 Martin Liska <mliska@suse.cz>
10984 PR gcov-profile/84137
10985 * doc/gcov.texi: Fix typo in documentation.
10987 2018-02-05 Martin Liska <mliska@suse.cz>
10989 PR gcov-profile/83879
10990 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10991 using dlopen functionality.
10993 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10995 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10996 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10997 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10998 _mm_maskz_range_round_ss): New intrinsics.
10999 (__builtin_ia32_rangesd128_round)
11000 (__builtin_ia32_rangess128_round): Remove.
11001 (__builtin_ia32_rangesd128_mask_round,
11002 __builtin_ia32_rangess128_mask_round): New builtins.
11003 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11004 __builtin_ia32_rangess128_round): Remove.
11005 (__builtin_ia32_rangesd128_mask_round,
11006 __builtin_ia32_rangess128_mask_round): New builtins.
11007 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11008 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11009 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11010 "<round_saeonly_constraint>")): Changed to ...
11011 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11012 "<round_saeonly_scalar_constraint>")): ... this.
11013 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11014 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11015 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11016 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11017 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11019 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11021 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11023 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11024 Remove all values except native, 8540 and 8548.
11026 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11028 * config/i386/i386.c (ix86_output_function_return): Pass
11029 INVALID_REGNUM, instead of -1, as invalid register number to
11030 indirect_thunk_name and output_indirect_thunk.
11032 2018-02-02 Julia Koval <julia.koval@intel.com>
11034 * config.gcc: Add -march=icelake.
11035 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11036 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11037 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11038 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11039 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11040 (processor_target_table): Add icelake.
11041 (ix86_option_override_internal): Handle new PTAs.
11042 (get_builtin_code_for_version): Handle icelake.
11043 (M_INTEL_COREI7_ICELAKE): New.
11044 (fold_builtin_cpu): Handle icelake.
11045 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11046 * doc/invoke.texi: Add -march=icelake.
11048 2018-02-02 Julia Koval <julia.koval@intel.com>
11050 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11051 to wide_int_bitmask.
11052 * wide-int-bitmask.h: New.
11054 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11057 * config/i386/i386.md: Replace Pmode with word_mode in
11058 builtin_setjmp_setup and builtin_longjmp to support x32.
11060 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11064 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11066 (rs6000_supported_cpu_names): New static variable.
11067 (linux_cpu_translation_table): Likewise.
11068 (elf_platform) <cpu>: Define new static variable and use it.
11069 Translate kernel AT_PLATFORM name to canonical name if needed.
11070 Error if platform name is unknown.
11072 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11075 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11077 2018-02-01 Jeff Law <law@redhat.com>
11080 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11081 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11082 the scratch if RELEASE_VIA_POP is false.
11083 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11084 If we have to save a temporary register, decrement SIZE appropriately.
11085 Pass new arguments to release_scratch_register_on_entry.
11086 (ix86_adjust_stack_and_probe): Likewise.
11087 (ix86_emit_probe_stack_range): Pass new arguments to
11088 release_scratch_register_on_entry.
11090 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11092 PR rtl-optimization/84157
11093 * combine.c (change_zero_ext): Use REG_P predicate in
11094 front of HARD_REGISTER_P predicate.
11096 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11098 * config/avr/avr.c (avr_option_override): Move disabling of
11099 -fdelete-null-pointer-checks to...
11100 * common/config/avr/avr-common.c (avr_option_optimization_table):
11103 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11105 PR tree-optimization/81635
11106 * tree-data-ref.c (split_constant_offset_1): For types that
11107 wrap on overflow, try to use range info to prove that wrapping
11110 2018-02-01 Renlin Li <renlin.li@arm.com>
11113 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11114 TAILCALL_ADDR_REGS.
11115 (aarch64_register_move_cost): Likewise.
11116 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11117 TAILCALL_ADDR_REGS.
11118 (REG_CLASS_NAMES): Likewise.
11119 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11120 TAILCALL_ADDR_REGS. Remove IP registers.
11121 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11123 2018-02-01 Richard Biener <rguenther@suse.de>
11125 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11126 for specifying RPO order and allow NULL for that.
11127 * domwalk.c (dom_walker::dom_walker): Likewise.
11128 (dom_walker::walk): Handle NULL RPO order.
11129 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11131 (rewrite_update_dom_walker): Likewise.
11132 (mark_def_dom_walker): Likewise.
11134 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11136 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11137 (aarch64_maybe_expand_sve_subreg_move): Declare.
11138 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11139 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11141 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11142 that are semantically a reverse operation.
11143 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11144 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11145 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11147 (aarch64_can_change_mode_class): For big-endian, forbid changes
11148 between two SVE modes if they have different element sizes.
11150 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11152 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11153 the TImode handling for big-endian targets.
11155 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11157 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11158 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11160 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11161 Remove BSWAP handing for big-endian targets and use the form of
11162 LD1RQ appropariate for the mode.
11164 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11166 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11167 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11168 duplicated element.
11170 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11173 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11174 check for operands that need to go through aarch64_sve_reload_be.
11176 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11178 PR tree-optimization/81661
11179 PR tree-optimization/84117
11180 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11181 * tree-eh.c: Include gimplify.h.
11182 (find_trapping_overflow, replace_trapping_overflow,
11183 rewrite_to_non_trapping_overflow): New functions.
11184 * tree-vect-loop.c: Include tree-eh.h.
11185 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11186 * tree-data-ref.c: Include tree-eh.h.
11187 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11189 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11191 PR rtl-optimization/84123
11192 * combine.c (change_zero_ext): Check if hard register satisfies
11193 can_change_dest_mode before calling gen_lowpart_SUBREG.
11195 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11198 * ira.c (ira_init_register_move_cost): Remove assert.
11200 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11202 PR rtl-optimization/84071
11203 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11204 * doc/tm.texi: Regenerate.
11206 2018-01-31 Richard Biener <rguenther@suse.de>
11208 PR tree-optimization/84132
11209 * tree-data-ref.c (analyze_miv_subscript): Properly
11210 check whether evolution_function_is_affine_multivariate_p
11211 before calling gcd_of_steps_may_divide_p.
11213 2018-01-31 Julia Koval <julia.koval@intel.com>
11216 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11217 * config/i386/i386.md (rdpid_rex64) New.
11218 (rdpid): Make 32bit only.
11220 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11223 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11224 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11226 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11229 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11231 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11233 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11235 PR rtl-optimization/84071
11236 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11237 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11239 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11241 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11242 (arc_attribute_table): Add 'aux' attribute.
11243 (arc_in_small_data_p): Consider aux like variables.
11244 (arc_is_aux_reg_p): New function.
11245 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11246 (arc_get_aux_arg): New function.
11247 (prepare_move_operands): Handle aux-register access.
11248 (arc_handle_aux_attribute): New function.
11249 * doc/extend.texi (ARC Variable attributes): Add subsection.
11251 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11253 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11254 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11255 (arc_attribute_table): Add 'uncached' attribute.
11256 (arc_print_operand): Print '.di' flag for uncached memory
11258 (arc_in_small_data_p): Do not consider for small data the uncached
11260 (arc_is_uncached_mem_p): New function.
11261 * config/arc/predicates.md (compact_store_memory_operand): Check
11262 for uncached memory accesses.
11263 (nonvol_nonimm_operand): Likewise.
11264 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11266 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11269 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11270 falign-loops=): Add Optimization flag.
11272 2018-01-30 Jeff Law <law@redhat.com>
11275 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11276 INT_REGISTERS_SAVED. Check it prior to calling
11277 get_scratch_register_on_entry.
11278 (ix86_adjust_stack_and_probe): Similarly.
11279 (ix86_emit_probe_stack_range): Similarly.
11280 (ix86_expand_prologue): Corresponding changes.
11282 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11285 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11286 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11288 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11291 * lra-constraints.c (curr_insn_transform): Process AND in the
11294 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11296 PR rtl-optimization/83986
11297 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11298 dependence against last_pending_memory_flush in addition to
11299 pending_jump_insns.
11301 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11303 PR tree-optimization/81611
11304 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11307 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11310 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11313 2018-01-30 Richard Biener <rguenther@suse.de>
11314 Jakub Jelinek <jakub@redhat.com>
11316 PR tree-optimization/84111
11317 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11318 inner loops added during recursion, as they don't have up-to-date
11321 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11324 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11325 (can_inline_edge_by_limits_p): ... here.
11326 (can_early_inline_edge_p, check_callers,
11327 update_caller_keys, update_callee_keys, recursive_inlining,
11328 add_new_edges_to_heap, speculation_useful_p,
11329 inline_small_functions,
11330 inline_small_functions, flatten_function,
11331 inline_to_all_callers_1): Update.
11333 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11335 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11338 2018-01-30 Richard Biener <rguenther@suse.de>
11340 PR tree-optimization/83008
11341 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11342 invariant and constant vector uses in stmts when they need
11343 more than one stmt.
11345 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11348 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11349 * configure: Regenerate.
11351 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11353 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11355 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11356 Use gen_rtx_REG rather than gen_lowpart.
11358 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11360 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11361 rather than 0 when creating partial subregs.
11363 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11365 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11368 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11371 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11372 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11373 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11374 flags. This restores the settings used before the 2017-07-24.
11375 Turning off pre increment/decrement/modify allows IVOPTS to
11376 optimize DF/SF loops where the index is an int.
11378 2018-01-29 Richard Biener <rguenther@suse.de>
11379 Kelvin Nilsen <kelvin@gcc.gnu.org>
11382 * tree-vect-stmts.c (vectorizable_call): Don't call
11383 targetm.vectorize_builtin_md_vectorized_function if callee is
11386 2018-01-22 Carl Love <cel@us.ibm.com>
11388 * doc/extend.tex: Fix typo in second arg in
11389 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11391 2018-01-29 Richard Biener <rguenther@suse.de>
11393 PR tree-optimization/84086
11394 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11395 (flush_ssaname_freelist): When SSA names were released reset
11396 the SCEV hash table.
11398 2018-01-29 Richard Biener <rguenther@suse.de>
11400 PR tree-optimization/84057
11401 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11402 removed paths when removing edges.
11404 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11406 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11407 -mfunction-return=@var{choice}.
11409 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11411 PR diagnostic/84034
11412 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11413 Handle CR like TAB.
11414 (layout::print_source_line): Likewise.
11415 (test_get_line_width_without_trailing_whitespace): Add test cases.
11417 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11419 PR middle-end/84040
11420 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11423 2018-01-26 Jim Wilson <jimw@sifive.com>
11425 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11427 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11430 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11432 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11433 and CMP + SUB-immediate -> SUBS.
11435 2018-01-26 Martin Sebor <msebor@redhat.com>
11437 PR tree-optimization/83896
11438 * tree-ssa-strlen.c (get_string_len): Rename...
11439 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11440 Avoid assuming length is constant.
11441 (handle_char_store): Use HOST_WIDE_INT for string length.
11443 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11446 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11447 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11449 2018-01-26 Richard Biener <rguenther@suse.de>
11451 PR rtl-optimization/84003
11452 * dse.c (record_store): Only record redundant stores when
11453 the earlier store aliases at least all accesses the later one does.
11455 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11457 PR rtl-optimization/83985
11458 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11459 REG_CFA_RESTORE insns.
11460 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11461 REG_CFA_RESTORE insns here.
11464 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11465 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11467 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11469 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11470 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11471 (arc_init): Likewise.
11472 (arc_override_options): Likewise.
11473 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11475 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11477 * config/arc/arc.h (TARGET_DBNZ): Define.
11478 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11479 properly set the tune attribute.
11480 (dbnz): Use TARGET_DBNZ guard.
11481 * config/arc/arc.opt (mtune): Add core3 option.
11483 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11485 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11486 recognize new pic like addresses.
11487 (arc_delegitimize_address): Clean up.
11489 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11491 * config/arc/arc-arches.def: Option mrf16 valid for all
11493 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11494 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11495 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11496 * config/arc/arc-tables.opt: Regenerate.
11497 * config/arc/arc.c (arc_conditional_register_usage): Handle
11498 reduced register file case.
11499 (arc_file_start): Set must have build attributes.
11500 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11501 mrf16 option value.
11502 * config/arc/arc.opt (mrf16): Add new option.
11503 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11504 * config/arc/genmultilib.awk: Handle new mrf16 option.
11505 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11506 * config/arc/t-multilib: Regenerate.
11507 * doc/invoke.texi (ARC Options): Document mrf16 option.
11509 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11511 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11512 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11513 (arc_attribute_table): Add 'secure_call' attribute.
11514 (arc_print_operand): Print secure call operand.
11515 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11517 (arc_is_secure_call_p): New function. * config/arc/arc.md
11518 (call_i): Add support for sjli instruction.
11519 (call_value_i): Likewise.
11520 * config/arc/constraints.md (Csc): New constraint.
11522 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11523 John Eric Martin <John.Martin@emmicro-us.com>
11525 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11526 * config/arc/arc.c (_arc_jli_section): New struct.
11527 (arc_jli_section): New type.
11528 (rc_jli_sections): New static variable.
11529 (arc_handle_jli_attribute): New function.
11530 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11531 (arc_file_end): New function.
11532 (TARGET_ASM_FILE_END): Define.
11533 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11534 (arc_add_jli_section): New function.
11535 (jli_call_scan): Likewise.
11536 (arc_reorg): Call jli_call_scan.
11537 (arc_output_addsi): Remove 'S' from printing asm operand.
11538 (arc_is_jli_call_p): New function.
11539 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11541 (movhi_insn): Likewise.
11542 (movsi_insn): Likewise.
11543 (movsi_set_cc_insn): Likewise.
11544 (loadqi_update): Likewise.
11545 (load_zeroextendqisi_update): Likewise.
11546 (load_signextendqisi_update): Likewise.
11547 (loadhi_update): Likewise.
11548 (load_zeroextendhisi_update): Likewise.
11549 (load_signextendhisi_update): Likewise.
11550 (loadsi_update): Likewise.
11551 (loadsf_update): Likewise.
11552 (movsicc_insn): Likewise.
11553 (bset_insn): Likewise.
11554 (bxor_insn): Likewise.
11555 (bclr_insn): Likewise.
11556 (bmsk_insn): Likewise.
11557 (bicsi3_insn): Likewise.
11558 (cmpsi_cc_c_insn): Likewise.
11559 (movsi_ne): Likewise.
11560 (movsi_cond_exec): Likewise.
11561 (clrsbsi2): Likewise.
11562 (norm_f): Likewise.
11572 (call_i): Remove 'S' asm letter, add jli instruction.
11573 (call_value_i): Likewise.
11574 * config/arc/arc.op (mjli-always): New option.
11575 * config/arc/constraints.md (Cji): New constraint.
11576 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11578 (subsf3_fpx): Likewise.
11579 (mulsf3_fpx): Likewise.
11580 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11582 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11583 function attrbutes.
11584 * doc/invoke.texi (ARC): Document mjli-always option.
11586 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11588 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11589 avoid addition with 0 and use incw and decw where possible.
11591 2018-01-26 Richard Biener <rguenther@suse.de>
11593 PR tree-optimization/81082
11594 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11595 association if it requires casting to unsigned.
11596 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11597 from fold_plusminus_mult_expr to catch important cases late when
11598 range info is available.
11600 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11602 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11603 * configure.ac (hidden_linkonce): New test.
11604 * configure: Regenerate.
11605 * config.in: Regenerate.
11607 2018-01-26 Julia Koval <julia.koval@intel.com>
11609 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11610 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11611 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11612 _mm_mask_bitshuffle_epi64_mask): Fix type.
11613 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11614 USI_FTYPE_V4DI_V4DI_USI): Remove.
11615 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11616 __builtin_ia32_vpshufbitqmb256_mask,
11617 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11618 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11619 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11621 2018-01-26 Alan Modra <amodra@gmail.com>
11624 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11625 UNSPEC_VBPERMQ. Sort other unspecs.
11627 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11629 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11631 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11633 PR middle-end/83055
11634 * predict.c (drop_profile): Do not push/pop cfun; update also
11636 (handle_missing_profiles): Fix logic looking for zero profiles.
11638 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11640 PR middle-end/83977
11641 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11642 on functions with #pragma omp declare simd or functions with simd
11644 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11645 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11646 Remove trailing \n from warning_at calls.
11648 2018-01-25 Tom de Vries <tom@codesourcery.com>
11651 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11652 for neutered workers.
11654 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11657 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11658 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11660 2018-01-24 Jeff Law <law@redhat.com>
11663 * i386.c (get_probe_interval): Move to earlier point.
11664 (ix86_compute_frame_layout): If -fstack-clash-protection and
11665 the frame is larger than the probe interval, then use pushes
11666 to save registers rather than reg->mem moves.
11667 (ix86_expand_prologue): Remove conditional for int_registers_saved
11670 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11673 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11674 min/max for never referenced object.
11676 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11678 PR middle-end/83977
11679 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11681 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11682 attributes from DECL_ATTRIBUTES (decl) without affecting
11683 DECL_ATTRIBUTES (current_function_decl).
11684 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11685 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11687 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11689 PR tree-optimization/83979
11690 * fold-const.c (fold_comparison): Use constant_boolean_node
11691 instead of boolean_{true,false}_node.
11693 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11695 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11698 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11700 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11701 Simplify the clause that sets the length attribute.
11702 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11703 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11704 clause that sets the length attribute.
11705 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11707 2018-01-24 Tom de Vries <tom@codesourcery.com>
11710 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11711 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11712 Add strict parameter.
11713 (prevent_branch_around_nothing): Insert dummy insn between branch to
11714 label and label with no ptx insn inbetween.
11715 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11717 2018-01-24 Tom de Vries <tom@codesourcery.com>
11720 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11721 for neutered threads in warp.
11722 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11724 2018-01-24 Richard Biener <rguenther@suse.de>
11726 PR tree-optimization/83176
11727 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11730 2018-01-24 Richard Biener <rguenther@suse.de>
11732 PR tree-optimization/82819
11733 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11734 code generating pluses that are no-ops in the target precision.
11736 2018-01-24 Richard Biener <rguenther@suse.de>
11738 PR middle-end/84000
11739 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11741 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11743 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11744 to merge probabilities.
11745 * predict.c (probably_never_executed): Also mark as cold functions
11746 with global 0 profile and guessed local profile.
11747 * profile-count.c (profile_probability::combine_with_count): New
11749 * profile-count.h (profile_probability::operator*,
11750 profile_probability::operator*=, profile_probability::operator/,
11751 profile_probability::operator/=): Reduce precision to adjusted
11752 and set value to guessed on contradictory divisions.
11753 (profile_probability::combine_with_freq): Remove.
11754 (profile_probability::combine_wiht_count): Declare.
11755 (profile_count::force_nonzero):: Set to adjusted.
11756 (profile_count::probability_in):: Set quality to adjusted.
11757 * tree-ssa-tail-merge.c (replace_block_by): Use
11758 combine_with_count.
11760 2018-01-23 Andrew Waterman <andrew@sifive.com>
11761 Jim Wilson <jimw@sifive.com>
11763 * config/riscv/riscv.c (riscv_stack_boundary): New.
11764 (riscv_option_override): Set riscv_stack_boundary. Handle
11765 riscv_preferred_stack_boundary_arg.
11766 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11767 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11768 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11769 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11770 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11771 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11773 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11776 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11777 of struct ix86_frame.
11778 (ix86_expand_epilogue): Likewise. Add a local variable for
11779 the reg_save_offset field in struct ix86_frame.
11781 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11783 PR tree-optimization/82604
11784 * tree-loop-distribution.c (enum partition_kind): New enum item
11785 PKIND_PARTIAL_MEMSET.
11786 (partition_builtin_p): Support above new enum item.
11787 (generate_code_for_partition): Ditto.
11788 (compute_access_range): Differentiate cases that equality can be
11789 proven at all loops, the innermost loops or no loops.
11790 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11791 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11792 (finalize_partitions, distribute_loop): Don't fuse partition of
11793 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11794 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11795 parloop is enabled.
11797 2018-01-23 Martin Liska <mliska@suse.cz>
11799 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11800 order to ignore the predictor.
11801 (PRED_POLYMORPHIC_CALL): Likewise.
11802 (PRED_RECURSIVE_CALL): Likewise.
11804 2018-01-23 Martin Liska <mliska@suse.cz>
11806 * tree-profile.c (tree_profiling): Print function header to
11807 aware reader which function we are working on.
11808 * value-prof.c (gimple_find_values_to_profile): Do not print
11809 not interesting value histograms.
11811 2018-01-23 Martin Liska <mliska@suse.cz>
11813 * profile-count.h (enum profile_quality): Add
11814 profile_uninitialized as the first value. Do not number values
11815 as they are zero based.
11816 (profile_count::verify): Update sanity check.
11817 (profile_probability::verify): Likewise.
11819 2018-01-23 Nathan Sidwell <nathan@acm.org>
11821 * doc/invoke.texi (ffor-scope): Deprecate.
11823 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11825 PR tree-optimization/83510
11826 * domwalk.c (set_all_edges_as_executable): New function.
11827 (dom_walker::dom_walker): Convert bool param
11828 "skip_unreachable_blocks" to enum reachability. Move setup of
11829 edge flags to set_all_edges_as_executable and only do it when
11830 reachability is REACHABLE_BLOCKS.
11831 * domwalk.h (enum dom_walker::reachability): New enum.
11832 (dom_walker::dom_walker): Convert bool param
11833 "skip_unreachable_blocks" to enum reachability.
11834 (set_all_edges_as_executable): New decl.
11835 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11836 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11838 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11839 but converting true to REACHABLE_BLOCKS.
11840 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11842 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11843 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11844 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11846 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11847 if check_all_array_refs will be called.
11849 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11851 * tree.c (selftest::test_location_wrappers): Add more test
11854 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11856 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11857 (selftest::test_bit_in_range): Likewise.
11859 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11862 * doc/sourcebuild.texi (vect_float): Say that the selector
11863 only describes the situation when -funsafe-math-optimizations is on.
11864 (vect_float_strict): Document.
11866 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11868 PR tree-optimization/83965
11869 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11870 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11871 instead of checking only for a reduction.
11872 (vect_recog_widen_sum_pattern): Likewise.
11874 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11876 * predict.c (probably_never_executed): Only use precise profile info.
11877 (compute_function_frequency): Skip after inlining hack since we now
11878 have quality checking.
11880 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11882 * profile-count.h (profile_probability::very_unlikely,
11883 profile_probability::unlikely, profile_probability::even): Set
11884 precision to guessed.
11886 2018-01-23 Richard Biener <rguenther@suse.de>
11888 PR tree-optimization/83963
11889 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11890 Properly terminate dominator walk when crossing the exit edge not
11891 when visiting its source block.
11893 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11896 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11897 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11899 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11901 PR tree-optimization/83957
11902 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11903 semicolon after for body surrounded by braces.
11905 PR tree-optimization/83081
11906 * profile-count.h (profile_probability::split): New method.
11907 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11908 Use profile_probability::split.
11909 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11910 when splitting a single conditional jump into 2.
11912 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11914 PR tree-optimization/69452
11915 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11918 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11920 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11921 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11922 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11924 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11926 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11928 * config/rl78/rl78.md (movdi): New define_expand.
11929 * config/rl78/rl78.c (rl78_split_movdi): New function.
11931 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11934 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11936 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11937 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11938 128-bit to produce an UNSPEC move to get the double word with the
11939 signbit and then a shift directly to do signbit.
11940 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11941 implementation with a new version that just does either a direct
11942 move or a regular move. Move memory interface to separate insns.
11943 Move insns so they are next to the expander.
11944 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11945 with signbit move. Split big and little endian case.
11946 (signbit<mode>2_dm_mem_le): Likewise.
11947 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11948 (signbit<mode>2_dm2): Likewise.
11950 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11952 * config/rl78/rl78.md (anddi3): New define_expand.
11954 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11956 * config/rl78/rl78.md (umindi3): New define_expand.
11958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11960 * config/rl78/rl78.md (smindi3): New define_expand.
11962 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11964 * config/rl78/rl78.md (smaxdi3): New define_expand.
11966 2018-01-22 Carl Love <cel@us.ibm.com>
11968 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11969 LVX_V1TI): Add macro expansion.
11970 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11971 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11972 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11973 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11974 Change check to determine if the instruction is a byte reversing
11975 entry. Fix typo in comment.
11976 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11977 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11978 Add def_builtin calls for new builtins.
11979 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11980 Add define_insn expansion.
11982 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11984 * config/rl78/rl78.md (umaxdi3): New define_expand.
11986 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11988 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11989 for non-QImode registers.
11991 2018-01-22 Richard Biener <rguenther@suse.de>
11993 PR tree-optimization/83963
11994 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11995 including the loop exit block.
11996 (scop_detection::merge_sese): Likewise.
11997 (scop_detection::add_scop): Do it here instead.
11999 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12001 * doc/sourcebuild.texi (arm_softfloat): Document.
12003 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12006 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12007 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12008 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12010 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12011 David Edelsohn <dje.gcc@gmail.com>
12014 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12015 Change "crset eq" to "crset 2".
12016 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12017 (*call_indirect_aix<mode>_nospec): Likewise.
12018 (*call_value_indirect_aix<mode>_nospec): Likewise.
12019 (*call_indirect_elfv2<mode>_nospec): Likewise.
12020 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12021 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12022 change assembly output from . to $.
12023 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12024 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12025 (*tablejump<mode>_internal1_nospec): Likewise.
12027 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12030 * config/sh/sh_optimize_sett_clrt.cc:
12031 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12033 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12035 PR tree-optimization/83940
12036 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12037 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12038 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12039 use it to pass back the definition type.
12040 (vect_check_store_rhs): Likewise rhs_dt_out.
12041 (vect_build_gather_load_calls): Add a mask_dt argument and use
12042 it instead of a call to vect_is_simple_use.
12043 (vectorizable_store): Update calls to vect_check_load_store_mask
12044 and vect_check_store_rhs. Use the dt returned by the latter instead
12045 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12046 instead of calls to vect_is_simple_use. Pass the scalar rather
12047 than the vector operand to vect_is_simple_use when handling
12048 second and subsequent copies of an rhs value.
12049 (vectorizable_load): Update calls to vect_check_load_store_mask
12050 and vect_build_gather_load_calls. Use the cached mask_dt and
12051 gs_info.offset_dt instead of calls to vect_is_simple_use.
12053 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12055 PR middle-end/83945
12056 * tree-emutls.c: Include gimplify.h.
12057 (lower_emutls_2): New function.
12058 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12059 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12060 it before further processing.
12063 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12064 UINTVAL (trueop1) instead of INTVAL (op1).
12066 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12070 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12071 INCOMING_FRAME_SP_OFFSET if not defined.
12072 (scan_trace): Add ENTRY argument. If true and
12073 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12074 emit a note to adjust the CFA offset.
12075 (create_cfi_notes): Adjust scan_trace callers.
12076 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12077 INCOMING_FRAME_SP_OFFSET in the CIE.
12078 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12079 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12081 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12082 * doc/tm.texi: Regenerated.
12084 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12086 PR rtl-optimization/83147
12087 * lra-constraints.c (remove_inheritance_pseudos): Use
12088 lra_substitute_pseudo_within_insn.
12090 2018-01-19 Tom de Vries <tom@codesourcery.com>
12091 Cesar Philippidis <cesar@codesourcery.com>
12094 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12096 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12099 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12100 spaces for function labels.
12102 2018-01-19 Martin Liska <mliska@suse.cz>
12104 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12105 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12106 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12107 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12108 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12109 (PRED_CONST_RETURN): Change from 69 to 65.
12110 (PRED_NULL_RETURN): Change from 91 to 71.
12111 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12112 (PRED_LOOP_GUARD): Change from 66 to 73.
12114 2018-01-19 Martin Liska <mliska@suse.cz>
12116 * predict.c (predict_insn_def): Add new assert.
12117 (struct branch_predictor): Change type to signed integer.
12118 (test_prediction_value_range): Amend test to cover
12119 PROB_UNINITIALIZED.
12120 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12121 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12122 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12123 (PRED_LOOP_IV_COMPARE): Likewise.
12124 * predict.h (PROB_UNINITIALIZED): Define new constant.
12126 2018-01-19 Martin Liska <mliska@suse.cz>
12128 * predict.c (dump_prediction): Add new format for
12129 analyze_brprob.py script which is enabled with -details
12131 * profile-count.h (precise_p): New function.
12133 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12135 PR tree-optimization/83922
12136 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12137 there are no statements that need masking.
12138 (vect_active_double_reduction_p): New function.
12139 (vect_analyze_loop_operations): Use it when handling phis that
12140 are not in the loop header.
12142 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12144 PR tree-optimization/83914
12145 * tree-vect-loop.c (vectorizable_induction): Don't convert
12146 init_expr or apply the peeling adjustment for inductions
12147 that are nested within the vectorized loop.
12149 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12151 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12154 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12158 * function.h (gimplify_parameters): Add gimple_seq * argument.
12159 * function.c: Include gimple.h and options.h.
12160 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12161 for the added local temporaries if needed.
12162 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12163 if there are any parameter cleanups, wrap whole body into a
12164 try/finally with the cleanups.
12166 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12169 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12170 Use GET_MODE_CLASS for scalar floating point.
12172 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12176 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12177 Fix call of call_cgraph_insertion_hooks.
12179 2018-01-18 Martin Sebor <msebor@redhat.com>
12181 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12183 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12186 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12189 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12192 * common.opt: (-ffile-prefix-map): New option.
12193 * opts.c (common_handle_option): Defer it.
12194 * opts-global.c (handle_common_deferred_options): Handle it.
12195 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12196 * file-prefix-map.h: New file.
12197 (remap_debug_filename, add_debug_prefix_map): ...here.
12198 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12199 * final.c (debug_prefix_map, add_debug_prefix_map
12200 remap_debug_filename): Move to...
12201 * file-prefix-map.c: New file.
12202 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12203 generalize, get rid of alloca(), use strrchr() instead of strchr().
12204 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12205 Implement in terms of add_prefix_map().
12206 (remap_macro_filename, remap_debug_filename): Implement in term of
12208 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12209 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12210 * dbxout.c: Include file-prefix-map.h.
12211 * varasm.c: Likewise.
12212 * vmsdbgout.c: Likewise.
12213 * xcoffout.c: Likewise.
12214 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12215 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12216 * doc/invoke.texi (-ffile-prefix-map): Document.
12217 (-fdebug-prefix-map): Update description.
12219 2018-01-18 Martin Liska <mliska@suse.cz>
12221 * config/i386/i386.c (indirect_thunk_name): Document that also
12223 (output_indirect_thunk): Document why both instructions
12224 (pause and lfence) are generated.
12226 2018-01-18 Richard Biener <rguenther@suse.de>
12228 PR tree-optimization/83887
12229 * graphite-scop-detection.c
12230 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12231 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12232 (scop_detection::merge_sese): Re-implement with a flood-fill
12233 algorithm that properly finds a SESE region if it exists.
12235 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12238 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12239 pointer_diff optimizations use view_convert instead of convert.
12241 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12243 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12244 Generate different code for -mno-speculate-indirect-jumps.
12245 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12246 (*call_indirect_aix<mode>): Disable for
12247 -mno-speculate-indirect-jumps.
12248 (*call_indirect_aix<mode>_nospec): New define_insn.
12249 (*call_value_indirect_aix<mode>): Disable for
12250 -mno-speculate-indirect-jumps.
12251 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12252 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12253 -mno-speculate-indirect-jumps.
12254 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12256 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12258 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12259 long double type, set the flags for noting the default long double
12260 type, even if we don't pass or return a long double type.
12262 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12265 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12268 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12270 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12271 support for merge[hl].
12272 (fold_mergehl_helper): New helper function.
12273 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12274 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12275 (altivec_vmrglw_direct): Add xxmrglw insn.
12277 2018-01-17 Andrew Waterman <andrew@sifive.com>
12279 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12280 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12282 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12285 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12286 call the lto_location_cache before reading the
12287 DECL_SOURCE_LOCATION of the types.
12289 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12290 Richard Sandiford <richard.sandiford@linaro.org>
12292 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12293 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12294 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12295 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12296 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12298 * config/aarch64/constraints.md (aarch64_movti_operand):
12300 * config/aarch64/predicates.md (Uti): Add new constraint.
12302 2018-01-17 Carl Love <cel@us.ibm.com>
12304 * config/rs6000/vsx.md (define_expand xl_len_r,
12305 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12306 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12308 (define_expand, define_insn): Move the shift left from the
12309 define_insn to the define_expand for lxvl and stxvl instructions.
12310 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12311 and XL_LEN_R definitions to PURE.
12313 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12315 * config/i386/i386.c (indirect_thunk_name): Declare regno
12316 as unsigned int. Compare regno with INVALID_REGNUM.
12317 (output_indirect_thunk): Ditto.
12318 (output_indirect_thunk_function): Ditto.
12319 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12320 in the call to output_indirect_thunk_function.
12322 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12324 PR middle-end/83884
12325 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12326 rather than the size of inner_type to determine the stack slot size
12327 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12329 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12332 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12335 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12337 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12338 endian Linux systems to optionally enable multilibs for selecting
12339 the long double type if the user configured an explicit type.
12340 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12341 have no long double multilibs if not defined.
12342 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12343 warn if the user used -mabi={ieee,ibm}longdouble and we built
12344 multilibs for long double.
12345 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12346 appropriate multilib option.
12347 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12349 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12350 for building long double multilibs.
12351 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12353 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12355 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12358 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12360 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12363 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12366 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12369 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12371 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12372 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12374 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12376 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12377 different rtl trees depending on TARGET_64BIT.
12378 (rs6000_gen_lvx): Likewise.
12380 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12382 * config/visium/visium.md (nop): Tweak comment.
12383 (hazard_nop): Likewise.
12385 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12387 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12388 -mspeculate-indirect-jumps.
12389 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12390 for -mno-speculate-indirect-jumps.
12391 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12392 (*call_value_indirect_elfv2<mode>): Disable for
12393 -mno-speculate-indirect-jumps.
12394 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12395 (indirect_jump): Emit different RTL for
12396 -mno-speculate-indirect-jumps.
12397 (*indirect_jump<mode>): Disable for
12398 -mno-speculate-indirect-jumps.
12399 (*indirect_jump<mode>_nospec): New define_insn.
12400 (tablejump): Emit different RTL for
12401 -mno-speculate-indirect-jumps.
12402 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12403 (tablejumpsi_nospec): New define_expand.
12404 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12405 (tablejumpdi_nospec): New define_expand.
12406 (*tablejump<mode>_internal1): Disable for
12407 -mno-speculate-indirect-jumps.
12408 (*tablejump<mode>_internal1_nospec): New define_insn.
12409 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12412 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12414 * caller-save.c (insert_save): Drop unnecessary parameter. All
12417 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12418 Richard Biener <rguenth@suse.de>
12421 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12422 return early, inline manually is_gimple_sizepos. Make sure if we
12423 call gimplify_expr we don't end up with a gimple constant.
12424 * tree.c (variably_modified_type_p): Don't return true for
12425 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12426 * gimplify.h (is_gimple_sizepos): Remove.
12428 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12430 PR tree-optimization/83857
12431 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12432 vectorizable_live_operation for pure SLP statements.
12433 (vectorizable_live_operation): Handle PHIs.
12435 2018-01-16 Richard Biener <rguenther@suse.de>
12437 PR tree-optimization/83867
12438 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12439 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12441 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12444 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12445 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12446 If off is not INTEGER_CST, issue a may not be aligned warning
12447 rather than isn't aligned. Use isn%'t rather than isn't.
12448 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12450 <case MULT_EXPR>: Improve the case when bottom and one of the
12451 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12452 operand, in that case check if the other operand is multiple of
12453 bottom divided by the INTEGER_CST operand.
12455 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12458 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12459 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12460 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12461 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12462 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12463 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12464 (pa_function_arg_size): New function.
12466 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12468 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12469 in a separate statement.
12471 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12473 PR tree-optimization/83847
12474 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12475 group gathers and scatters.
12477 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12479 PR rtl-optimization/86620
12480 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12482 PR rtl-optimization/83213
12483 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12484 to last if both are JUMP_INSNs.
12486 PR tree-optimization/83843
12487 * gimple-ssa-store-merging.c
12488 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12489 store_immediate_info for bswap/nop orig_stores.
12491 2018-01-15 Andrew Waterman <andrew@sifive.com>
12493 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12495 <UDIV>: Increase cost if !TARGET_DIV.
12497 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12499 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12500 (define_attr "cr_logical_3op"): New.
12501 (cceq_ior_compare): Adjust.
12502 (cceq_ior_compare_complement): Adjust.
12503 (*cceq_rev_compare): Adjust.
12504 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12505 (is_cracked_insn): Adjust.
12506 (insn_must_be_first_in_group): Adjust.
12507 * config/rs6000/40x.md: Adjust.
12508 * config/rs6000/440.md: Adjust.
12509 * config/rs6000/476.md: Adjust.
12510 * config/rs6000/601.md: Adjust.
12511 * config/rs6000/603.md: Adjust.
12512 * config/rs6000/6xx.md: Adjust.
12513 * config/rs6000/7450.md: Adjust.
12514 * config/rs6000/7xx.md: Adjust.
12515 * config/rs6000/8540.md: Adjust.
12516 * config/rs6000/cell.md: Adjust.
12517 * config/rs6000/e300c2c3.md: Adjust.
12518 * config/rs6000/e500mc.md: Adjust.
12519 * config/rs6000/e500mc64.md: Adjust.
12520 * config/rs6000/e5500.md: Adjust.
12521 * config/rs6000/e6500.md: Adjust.
12522 * config/rs6000/mpc.md: Adjust.
12523 * config/rs6000/power4.md: Adjust.
12524 * config/rs6000/power5.md: Adjust.
12525 * config/rs6000/power6.md: Adjust.
12526 * config/rs6000/power7.md: Adjust.
12527 * config/rs6000/power8.md: Adjust.
12528 * config/rs6000/power9.md: Adjust.
12529 * config/rs6000/rs64.md: Adjust.
12530 * config/rs6000/titan.md: Adjust.
12532 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12534 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12535 ix86_indirect_branch_register logic.
12537 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12539 * config/i386/constraints.md (Bs): Update
12540 ix86_indirect_branch_register check. Don't check
12541 ix86_indirect_branch_register with GOT_memory_operand.
12543 * config/i386/predicates.md (GOT_memory_operand): Don't check
12544 ix86_indirect_branch_register here.
12545 (GOT32_symbol_operand): Likewise.
12547 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12549 * config/i386/predicates.md (constant_call_address_operand):
12550 Rewrite ix86_indirect_branch_register logic.
12551 (sibcall_insn_operand): Likewise.
12553 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12555 * config/i386/constraints.md (Bs): Replace
12556 ix86_indirect_branch_thunk_register with
12557 ix86_indirect_branch_register.
12559 * config/i386/i386.md (indirect_jump): Likewise.
12560 (tablejump): Likewise.
12561 (*sibcall_memory): Likewise.
12562 (*sibcall_value_memory): Likewise.
12563 Peepholes of indirect call and jump via memory: Likewise.
12564 * config/i386/i386.opt: Likewise.
12565 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12566 (GOT_memory_operand): Likewise.
12567 (call_insn_operand): Likewise.
12568 (sibcall_insn_operand): Likewise.
12569 (GOT32_symbol_operand): Likewise.
12571 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12573 PR middle-end/83837
12574 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12575 type rather than type addr's type points to.
12576 (expand_omp_atomic_mutex): Likewise.
12577 (expand_omp_atomic): Likewise.
12579 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12582 * config/i386/i386.c (output_indirect_thunk_function): Use
12583 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12584 for __x86_return_thunk.
12586 2018-01-15 Richard Biener <rguenther@suse.de>
12588 PR middle-end/83850
12589 * expmed.c (extract_bit_field_1): Fix typo.
12591 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12594 * config/arm/iterators.md (VF): New mode iterator.
12595 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12596 Remove integer-related logic from pattern.
12597 (neon_vabd<mode>_3): Likewise.
12599 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12601 PR middle-end/82694
12602 * common.opt (fstrict-overflow): No longer an alias.
12603 (fwrapv-pointer): New option.
12604 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12605 also for pointer types based on flag_wrapv_pointer.
12606 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12607 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12608 opts->x_flag_wrapv got set.
12609 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12610 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12611 POINTER_TYPE_OVERFLOW_UNDEFINED.
12612 * match.pd: Likewise in address comparison pattern.
12613 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12615 2018-01-15 Richard Biener <rguenther@suse.de>
12618 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12619 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12620 Reset type names to their identifier if their TYPE_DECL doesn't
12621 have linkage (and thus is used for ODR and devirt).
12622 (save_debug_info_for_decl): Remove.
12623 (save_debug_info_for_type): Likewise.
12624 (add_tree_to_fld_list): Adjust.
12625 * tree-pretty-print.c (dump_generic_node): Make dumping of
12626 type names more robust.
12628 2018-01-15 Richard Biener <rguenther@suse.de>
12630 * BASE-VER: Bump to 8.0.1.
12632 2018-01-14 Martin Sebor <msebor@redhat.com>
12635 * builtins.c (check_access): Avoid warning when the no-warning bit
12638 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12640 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12641 * ira-color (allocno_hard_regs_compare): Likewise.
12643 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12646 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12647 Use .pushsection/.popsection.
12649 2018-01-14 Martin Sebor <msebor@redhat.com>
12652 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12654 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12656 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12657 entry from extra_headers.
12658 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12659 extra_headers, make the list bitwise identical to the i?86-*-* one.
12661 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12663 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12664 -mcmodel=large with -mindirect-branch=thunk,
12665 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12666 -mfunction-return=thunk-extern.
12667 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12668 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12669 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12671 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12673 * config/i386/i386.c (print_reg): Print the name of the full
12674 integer register without '%'.
12675 (ix86_print_operand): Handle 'V'.
12676 * doc/extend.texi: Document 'V' modifier.
12678 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12680 * config/i386/constraints.md (Bs): Disallow memory operand for
12681 -mindirect-branch-register.
12683 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12684 (GOT_memory_operand): Likewise.
12685 (call_insn_operand): Likewise.
12686 (sibcall_insn_operand): Likewise.
12687 (GOT32_symbol_operand): Likewise.
12688 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12689 for -mindirect-branch-register.
12690 (tablejump): Likewise.
12691 (*sibcall_memory): Likewise.
12692 (*sibcall_value_memory): Likewise.
12693 Disallow peepholes of indirect call and jump via memory for
12694 -mindirect-branch-register.
12695 (*call_pop): Replace m with Bw.
12696 (*call_value_pop): Likewise.
12697 (*sibcall_pop_memory): Replace m with Bs.
12698 * config/i386/i386.opt (mindirect-branch-register): New option.
12699 * doc/invoke.texi: Document -mindirect-branch-register option.
12701 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12703 * config/i386/i386-protos.h (ix86_output_function_return): New.
12704 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12705 set function_return_type.
12706 (indirect_thunk_name): Add ret_p to indicate thunk for function
12708 (output_indirect_thunk_function): Pass false to
12709 indirect_thunk_name.
12710 (ix86_output_indirect_branch_via_reg): Likewise.
12711 (ix86_output_indirect_branch_via_push): Likewise.
12712 (output_indirect_thunk_function): Create alias for function
12713 return thunk if regno < 0.
12714 (ix86_output_function_return): New function.
12715 (ix86_handle_fndecl_attribute): Handle function_return.
12716 (ix86_attribute_table): Add function_return.
12717 * config/i386/i386.h (machine_function): Add
12718 function_return_type.
12719 * config/i386/i386.md (simple_return_internal): Use
12720 ix86_output_function_return.
12721 (simple_return_internal_long): Likewise.
12722 * config/i386/i386.opt (mfunction-return=): New option.
12723 (indirect_branch): Mention -mfunction-return=.
12724 * doc/extend.texi: Document function_return function attribute.
12725 * doc/invoke.texi: Document -mfunction-return= option.
12727 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12729 * config/i386/i386-opts.h (indirect_branch): New.
12730 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12731 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12732 with local indirect jump when converting indirect call and jump.
12733 (ix86_set_indirect_branch_type): New.
12734 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12735 (indirectlabelno): New.
12736 (indirect_thunk_needed): Likewise.
12737 (indirect_thunk_bnd_needed): Likewise.
12738 (indirect_thunks_used): Likewise.
12739 (indirect_thunks_bnd_used): Likewise.
12740 (INDIRECT_LABEL): Likewise.
12741 (indirect_thunk_name): Likewise.
12742 (output_indirect_thunk): Likewise.
12743 (output_indirect_thunk_function): Likewise.
12744 (ix86_output_indirect_branch_via_reg): Likewise.
12745 (ix86_output_indirect_branch_via_push): Likewise.
12746 (ix86_output_indirect_branch): Likewise.
12747 (ix86_output_indirect_jmp): Likewise.
12748 (ix86_code_end): Call output_indirect_thunk_function if needed.
12749 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12751 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12752 (ix86_attribute_table): Add indirect_branch.
12753 * config/i386/i386.h (machine_function): Add indirect_branch_type
12754 and has_local_indirect_jump.
12755 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12757 (tablejump): Likewise.
12758 (*indirect_jump): Use ix86_output_indirect_jmp.
12759 (*tablejump_1): Likewise.
12760 (simple_return_indirect_internal): Likewise.
12761 * config/i386/i386.opt (mindirect-branch=): New option.
12762 (indirect_branch): New.
12765 (thunk-inline): Likewise.
12766 (thunk-extern): Likewise.
12767 * doc/extend.texi: Document indirect_branch function attribute.
12768 * doc/invoke.texi: Document -mindirect-branch= option.
12770 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12773 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12775 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12777 * ipa-inline.c (want_inline_small_function_p): Return false if
12778 inlining has already failed with CIF_FINAL_ERROR.
12779 (update_caller_keys): Call want_inline_small_function_p before
12781 (update_callee_keys): Likewise.
12783 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12785 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12787 (rs6000_quadword_masked_address_p): Likewise.
12788 (quad_aligned_load_p): Likewise.
12789 (quad_aligned_store_p): Likewise.
12790 (const_load_sequence_p): Add comment to describe the outer-most loop.
12791 (mimic_memory_attributes_and_flags): New function.
12792 (rs6000_gen_stvx): Likewise.
12793 (replace_swapped_aligned_store): Likewise.
12794 (rs6000_gen_lvx): Likewise.
12795 (replace_swapped_aligned_load): Likewise.
12796 (replace_swapped_load_constant): Capitalize argument name in
12797 comment describing this function.
12798 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12799 and stores that access quad-word aligned addresses and replace
12800 with stvx or lvx instructions when appropriate.
12801 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12802 New function prototype.
12803 (rs6000_quadword_masked_address_p): Likewise.
12804 (rs6000_gen_lvx): Likewise.
12805 (rs6000_gen_stvx): Likewise.
12806 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12807 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12808 when memory address is aligned.
12809 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12810 this split to select lvx instruction when memory address is aligned.
12811 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12812 instruction when memory address is aligned.
12813 (*vsx_le_perm_load_v16qi): Likewise.
12814 (four unnamed splitters): Modify to select the stvx instruction
12815 when memory is aligned.
12817 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12819 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12820 which appears in the queue multiple times.
12822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12823 Alan Hayward <alan.hayward@arm.com>
12824 David Sherwood <david.sherwood@arm.com>
12826 * tree-vectorizer.h (vec_lower_bound): New structure.
12827 (_loop_vec_info): Add check_nonzero and lower_bounds.
12828 (LOOP_VINFO_CHECK_NONZERO): New macro.
12829 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12830 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12831 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12832 fields. Make seg_len the distance travelled, not including the
12834 (dr_direction_indicator): Declare.
12835 (dr_zero_step_indicator): Likewise.
12836 (dr_known_forward_stride_p): Likewise.
12837 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12839 (runtime_alias_check_p): Allow runtime alias checks with
12841 (operator ==): Compare access_size and align.
12842 (prune_runtime_alias_test_list): Rework for new distinction between
12843 the access_size and seg_len.
12844 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12846 (get_segment_min_max): New function.
12847 (create_intersect_range_checks): Use it.
12848 (dr_step_indicator): New function.
12849 (dr_direction_indicator): Likewise.
12850 (dr_zero_step_indicator): Likewise.
12851 (dr_known_forward_stride_p): Likewise.
12852 * tree-loop-distribution.c (data_ref_segment_size): Return
12853 DR_STEP * (niters - 1).
12854 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12856 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12857 (vect_preserves_scalar_order_p): New function, split out from...
12858 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12859 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12860 (vect_vfa_access_size): New function.
12861 (vect_vfa_align): Likewise.
12862 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12863 (dump_lower_bound): New function.
12864 (vect_check_lower_bound): Likewise.
12865 (vect_small_gap_p): Likewise.
12866 (vectorizable_with_step_bound_p): Likewise.
12867 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12868 depencies if the vectorization factor is 1. Convert the checks
12869 for nonzero steps into checks on the bounds of DR_STEP. Try using
12870 a bunds check for variable steps if the minimum required step is
12871 relatively small. Update calls to the dr_with_seg_len
12872 constructor and to vect_compile_time_alias.
12873 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12875 (vect_loop_versioning): Call it.
12876 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12878 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12881 Alan Hayward <alan.hayward@arm.com>
12882 David Sherwood <david.sherwood@arm.com>
12884 * doc/sourcebuild.texi (vect_scatter_store): Document.
12885 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12887 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12889 * genopinit.c (main): Add supports_vec_scatter_store and
12890 supports_vec_scatter_store_cached to target_optabs.
12891 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12892 IFN_MASK_SCATTER_STORE.
12893 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12895 * internal-fn.h (internal_store_fn_p): Declare.
12896 (internal_fn_stored_value_index): Likewise.
12897 * internal-fn.c (scatter_store_direct): New macro.
12898 (expand_scatter_store_optab_fn): New function.
12899 (direct_scatter_store_optab_supported_p): New macro.
12900 (internal_store_fn_p): New function.
12901 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12902 IFN_MASK_SCATTER_STORE.
12903 (internal_fn_mask_index): Likewise.
12904 (internal_fn_stored_value_index): New function.
12905 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12906 for scatter stores.
12907 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12908 * optabs-query.c (supports_vec_scatter_store_p): New function.
12909 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12910 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12911 true for scatter stores.
12912 (vect_gather_scatter_fn_p): Handle scatter stores too.
12913 (vect_check_gather_scatter): Consider using scatter stores if
12914 supports_vec_scatter_store_p.
12915 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12916 scatter stores too.
12917 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12918 internal_fn_stored_value_index.
12919 (check_load_store_masking): Handle scatter stores too.
12920 (vect_get_store_rhs): Make public.
12921 (vectorizable_call): Use internal_store_fn_p.
12922 (vectorizable_store): Handle scatter store internal functions.
12923 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12924 when deciding whether the end of the group has been reached.
12925 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12926 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12927 (mask_scatter_store<mode>): New insns.
12929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12930 Alan Hayward <alan.hayward@arm.com>
12931 David Sherwood <david.sherwood@arm.com>
12933 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12934 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12935 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12937 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12938 Use vect_truncate_gather_scatter_offset if we can't treat the
12939 operation as a normal gather load or scatter store.
12940 (get_group_load_store_type): Take the gather_scatter_info
12941 as argument. Try using a gather load or scatter store for
12942 single-element groups.
12943 (get_load_store_type): Update calls to get_group_load_store_type
12944 and vect_use_strided_gather_scatters_p.
12946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12947 Alan Hayward <alan.hayward@arm.com>
12948 David Sherwood <david.sherwood@arm.com>
12950 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12951 optional tree argument.
12952 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12954 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12955 but continue to use the current value as a fallback.
12956 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12957 to compare the updates.
12958 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12959 (get_load_store_type): Use it when handling a strided access.
12960 (vect_get_strided_load_store_ops): New function.
12961 (vect_get_data_ptr_increment): Likewise.
12962 (vectorizable_load): Handle strided gather loads. Always pass
12963 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12966 Alan Hayward <alan.hayward@arm.com>
12967 David Sherwood <david.sherwood@arm.com>
12969 * doc/md.texi (gather_load@var{m}): Document.
12970 (mask_gather_load@var{m}): Likewise.
12971 * genopinit.c (main): Add supports_vec_gather_load and
12972 supports_vec_gather_load_cached to target_optabs.
12973 * optabs-tree.c (init_tree_optimization_optabs): Use
12974 ggc_cleared_alloc to allocate target_optabs.
12975 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12976 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12978 * internal-fn.h (internal_load_fn_p): Declare.
12979 (internal_gather_scatter_fn_p): Likewise.
12980 (internal_fn_mask_index): Likewise.
12981 (internal_gather_scatter_fn_supported_p): Likewise.
12982 * internal-fn.c (gather_load_direct): New macro.
12983 (expand_gather_load_optab_fn): New function.
12984 (direct_gather_load_optab_supported_p): New macro.
12985 (direct_internal_fn_optab): New function.
12986 (internal_load_fn_p): Likewise.
12987 (internal_gather_scatter_fn_p): Likewise.
12988 (internal_fn_mask_index): Likewise.
12989 (internal_gather_scatter_fn_supported_p): Likewise.
12990 * optabs-query.c (supports_at_least_one_mode_p): New function.
12991 (supports_vec_gather_load_p): Likewise.
12992 * optabs-query.h (supports_vec_gather_load_p): Declare.
12993 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12994 and memory_type field.
12995 (NUM_PATTERNS): Bump to 15.
12996 * tree-vect-data-refs.c: Include internal-fn.h.
12997 (vect_gather_scatter_fn_p): New function.
12998 (vect_describe_gather_scatter_call): Likewise.
12999 (vect_check_gather_scatter): Try using internal functions for
13000 gather loads. Recognize existing calls to a gather load function.
13001 (vect_analyze_data_refs): Consider using gather loads if
13002 supports_vec_gather_load_p.
13003 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13004 (vect_get_gather_scatter_offset_type): Likewise.
13005 (vect_convert_mask_for_vectype): Likewise.
13006 (vect_add_conversion_to_patterm): Likewise.
13007 (vect_try_gather_scatter_pattern): Likewise.
13008 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13009 (vect_vect_recog_func_ptrs): Add it.
13010 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13011 internal_fn_mask_index and internal_gather_scatter_fn_p.
13012 (check_load_store_masking): Take the gather_scatter_info as an
13013 argument and handle gather loads.
13014 (vect_get_gather_scatter_ops): New function.
13015 (vectorizable_call): Check internal_load_fn_p.
13016 (vectorizable_load): Likewise. Handle gather load internal
13018 (vectorizable_store): Update call to check_load_store_masking.
13019 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13020 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13021 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13022 (aarch64_gather_scale_operand_d): New predicates.
13023 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13024 (mask_gather_load<mode>): New insns.
13026 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13027 Alan Hayward <alan.hayward@arm.com>
13028 David Sherwood <david.sherwood@arm.com>
13030 * optabs.def (fold_left_plus_optab): New optab.
13031 * doc/md.texi (fold_left_plus_@var{m}): Document.
13032 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13033 * internal-fn.c (fold_left_direct): Define.
13034 (expand_fold_left_optab_fn): Likewise.
13035 (direct_fold_left_optab_supported_p): Likewise.
13036 * fold-const-call.c (fold_const_fold_left): New function.
13037 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13038 * tree-parloops.c (valid_reduction_p): New function.
13039 (gather_scalar_reductions): Use it.
13040 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13041 (vect_finish_replace_stmt): Declare.
13042 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13043 (needs_fold_left_reduction_p): New function, split out from...
13044 (vect_is_simple_reduction): ...here. Accept reductions that
13045 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13046 (vect_force_simple_reduction): Also store the reduction type in
13047 the assignment's STMT_VINFO_REDUC_TYPE.
13048 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13049 (merge_with_identity): New function.
13050 (vect_expand_fold_left): Likewise.
13051 (vectorize_fold_left_reduction): Likewise.
13052 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13053 scalar phi in place for it. Check for target support and reject
13054 cases that would reassociate the operation. Defer the transform
13055 phase to vectorize_fold_left_reduction.
13056 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13057 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13058 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13062 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13063 call to ifc_temp_var.
13065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13066 Alan Hayward <alan.hayward@arm.com>
13067 David Sherwood <david.sherwood@arm.com>
13069 * target.def (legitimize_address_displacement): Take the original
13070 offset as a poly_int.
13071 * targhooks.h (default_legitimize_address_displacement): Update
13073 * targhooks.c (default_legitimize_address_displacement): Likewise.
13074 * doc/tm.texi: Regenerate.
13075 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13076 as an argument, moving assert of ad->disp == ad->disp_term to...
13077 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13078 Try calling targetm.legitimize_address_displacement before expanding
13079 the address rather than afterwards, and adjust for the new interface.
13080 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13081 Match the new hook interface. Handle SVE addresses.
13082 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13083 new hook interface.
13085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13087 * Makefile.in (OBJS): Add early-remat.o.
13088 * target.def (select_early_remat_modes): New hook.
13089 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13090 * doc/tm.texi: Regenerate.
13091 * targhooks.h (default_select_early_remat_modes): Declare.
13092 * targhooks.c (default_select_early_remat_modes): New function.
13093 * timevar.def (TV_EARLY_REMAT): New timevar.
13094 * passes.def (pass_early_remat): New pass.
13095 * tree-pass.h (make_pass_early_remat): Declare.
13096 * early-remat.c: New file.
13097 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13099 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13102 Alan Hayward <alan.hayward@arm.com>
13103 David Sherwood <david.sherwood@arm.com>
13105 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13106 vfm1 with a bound_epilog parameter.
13107 (vect_do_peeling): Update calls accordingly, and move the prologue
13108 call earlier in the function. Treat the base bound_epilog as 0 for
13109 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13110 this base when peeling for gaps.
13111 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13112 with fully-masked loops.
13113 (vect_estimate_min_profitable_iters): Handle the single peeled
13114 iteration in that case.
13116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13117 Alan Hayward <alan.hayward@arm.com>
13118 David Sherwood <david.sherwood@arm.com>
13120 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13121 single-element interleaving even if the size is not a power of 2.
13122 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13123 accesses for single-element interleaving if the group size is
13126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13127 Alan Hayward <alan.hayward@arm.com>
13128 David Sherwood <david.sherwood@arm.com>
13130 * doc/md.texi (fold_extract_last_@var{m}): Document.
13131 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13132 * optabs.def (fold_extract_last_optab): New optab.
13133 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13134 * internal-fn.c (fold_extract_direct): New macro.
13135 (expand_fold_extract_optab_fn): Likewise.
13136 (direct_fold_extract_optab_supported_p): Likewise.
13137 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13138 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13139 EXTRACT_LAST_REDUCTION.
13140 (get_initial_def_for_reduction): Do not create an initial vector
13141 for EXTRACT_LAST_REDUCTION reductions.
13142 (vectorizable_reduction): Leave the scalar phi in place for
13143 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13144 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13145 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13146 transform phase to vectorizable_condition.
13147 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13149 (vect_finish_stmt_generation): ...here.
13150 (vect_finish_replace_stmt): New function.
13151 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13152 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13154 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13157 Alan Hayward <alan.hayward@arm.com>
13158 David Sherwood <david.sherwood@arm.com>
13160 * doc/md.texi (extract_last_@var{m}): Document.
13161 * optabs.def (extract_last_optab): New optab.
13162 * internal-fn.def (EXTRACT_LAST): New internal function.
13163 * internal-fn.c (cond_unary_direct): New macro.
13164 (expand_cond_unary_optab_fn): Likewise.
13165 (direct_cond_unary_optab_supported_p): Likewise.
13166 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13167 loops using EXTRACT_LAST.
13168 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13169 (extract_last_<mode>): ...this optab.
13170 (vec_extract<mode><Vel>): Update accordingly.
13172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13173 Alan Hayward <alan.hayward@arm.com>
13174 David Sherwood <david.sherwood@arm.com>
13176 * target.def (empty_mask_is_expensive): New hook.
13177 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13178 * doc/tm.texi: Regenerate.
13179 * targhooks.h (default_empty_mask_is_expensive): Declare.
13180 * targhooks.c (default_empty_mask_is_expensive): New function.
13181 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13182 if the target says that empty masks are expensive.
13183 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13185 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13188 Alan Hayward <alan.hayward@arm.com>
13189 David Sherwood <david.sherwood@arm.com>
13191 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13192 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13193 (vect_use_loop_mask_for_alignment_p): New function.
13194 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13195 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13196 niters_skip argument. Make sure that the first niters_skip elements
13197 of the first iteration are inactive.
13198 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13199 Update call to vect_set_loop_masks_directly.
13200 (get_misalign_in_elems): New function, split out from...
13201 (vect_gen_prolog_loop_niters): ...here.
13202 (vect_update_init_of_dr): Take a code argument that specifies whether
13203 the adjustment should be added or subtracted.
13204 (vect_update_init_of_drs): Likewise.
13205 (vect_prepare_for_masked_peels): New function.
13206 (vect_do_peeling): Skip prologue peeling if we're using a mask
13207 instead. Update call to vect_update_inits_of_drs.
13208 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13210 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13211 alignment. Do not include the number of peeled iterations in
13212 the minimum threshold in that case.
13213 (vectorizable_induction): Adjust the start value down by
13214 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13215 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13216 Take the number of skipped iterations into account when calculating
13218 * tree-vect-stmts.c (vect_gen_while_not): New function.
13220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13221 Alan Hayward <alan.hayward@arm.com>
13222 David Sherwood <david.sherwood@arm.com>
13224 * doc/sourcebuild.texi (vect_fully_masked): Document.
13225 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13226 default value to 0.
13227 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13229 (vect_analyze_loop_2): ...here. Don't check the vectorization
13230 factor against the number of loop iterations if the loop is
13233 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13234 Alan Hayward <alan.hayward@arm.com>
13235 David Sherwood <david.sherwood@arm.com>
13237 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13238 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13239 (dump_groups): Update accordingly.
13240 (iv_use::mem_type): New member variable.
13241 (address_p): New function.
13242 (record_use): Add a mem_type argument and initialize the new
13244 (record_group_use): Add a mem_type argument. Use address_p.
13245 Remove obsolete null checks of base_object. Update call to record_use.
13246 (find_interesting_uses_op): Update call to record_group_use.
13247 (find_interesting_uses_cond): Likewise.
13248 (find_interesting_uses_address): Likewise.
13249 (get_mem_type_for_internal_fn): New function.
13250 (find_address_like_use): Likewise.
13251 (find_interesting_uses_stmt): Try find_address_like_use before
13252 calling find_interesting_uses_op.
13253 (addr_offset_valid_p): Use the iv mem_type field as the type
13254 of the addressed memory.
13255 (add_autoinc_candidates): Likewise.
13256 (get_address_cost): Likewise.
13257 (split_small_address_groups_p): Use address_p.
13258 (split_address_groups): Likewise.
13259 (add_iv_candidate_for_use): Likewise.
13260 (autoinc_possible_for_pair): Likewise.
13261 (rewrite_groups): Likewise.
13262 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13263 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13264 (get_alias_ptr_type_for_ptr_address): New function.
13265 (rewrite_use_address): Rewrite address uses in calls that were
13266 identified by find_address_like_use.
13268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13269 Alan Hayward <alan.hayward@arm.com>
13270 David Sherwood <david.sherwood@arm.com>
13272 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13274 * gimple-expr.h (is_gimple_addressable: Likewise.
13275 * gimple-expr.c (is_gimple_address): Likewise.
13276 * internal-fn.c (expand_call_mem_ref): New function.
13277 (expand_mask_load_optab_fn): Use it.
13278 (expand_mask_store_optab_fn): Likewise.
13280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13281 Alan Hayward <alan.hayward@arm.com>
13282 David Sherwood <david.sherwood@arm.com>
13284 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13285 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13286 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13287 (cond_umax@var{mode}): Document.
13288 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13289 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13290 (cond_umin_optab, cond_umax_optab): New optabs.
13291 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13292 (COND_IOR, COND_XOR): New internal functions.
13293 * internal-fn.h (get_conditional_internal_fn): Declare.
13294 * internal-fn.c (cond_binary_direct): New macro.
13295 (expand_cond_binary_optab_fn): Likewise.
13296 (direct_cond_binary_optab_supported_p): Likewise.
13297 (get_conditional_internal_fn): New function.
13298 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13299 Cope with reduction statements that are vectorized as calls rather
13301 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13302 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13303 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13304 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13305 (UNSPEC_COND_EOR): New unspecs.
13306 (optab): Add mappings for them.
13307 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13308 (sve_int_op, sve_fp_op): New int attributes.
13310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13311 Alan Hayward <alan.hayward@arm.com>
13312 David Sherwood <david.sherwood@arm.com>
13314 * optabs.def (while_ult_optab): New optab.
13315 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13316 * internal-fn.def (WHILE_ULT): New internal function.
13317 * internal-fn.h (direct_internal_fn_supported_p): New override
13318 that takes two types as argument.
13319 * internal-fn.c (while_direct): New macro.
13320 (expand_while_optab_fn): New function.
13321 (convert_optab_supported_p): Likewise.
13322 (direct_while_optab_supported_p): New macro.
13323 * wide-int.h (wi::udiv_ceil): New function.
13324 * tree-vectorizer.h (rgroup_masks): New structure.
13325 (vec_loop_masks): New typedef.
13326 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13327 and fully_masked_p.
13328 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13329 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13330 (vect_max_vf): New function.
13331 (slpeel_make_loop_iterate_ntimes): Delete.
13332 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13333 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13334 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13335 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13336 internal-fn.h, stor-layout.h and optabs-query.h.
13337 (vect_set_loop_mask): New function.
13338 (add_preheader_seq): Likewise.
13339 (add_header_seq): Likewise.
13340 (interleave_supported_p): Likewise.
13341 (vect_maybe_permute_loop_masks): Likewise.
13342 (vect_set_loop_masks_directly): Likewise.
13343 (vect_set_loop_condition_masked): Likewise.
13344 (vect_set_loop_condition_unmasked): New function, split out from
13345 slpeel_make_loop_iterate_ntimes.
13346 (slpeel_make_loop_iterate_ntimes): Rename to..
13347 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13348 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13349 (vect_do_peeling): Update call accordingly.
13350 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13352 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13353 mask_compare_type, can_fully_mask_p and fully_masked_p.
13354 (release_vec_loop_masks): New function.
13355 (_loop_vec_info): Use it to free the loop masks.
13356 (can_produce_all_loop_masks_p): New function.
13357 (vect_get_max_nscalars_per_iter): Likewise.
13358 (vect_verify_full_masking): Likewise.
13359 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13360 retries, and free the mask rgroups before retrying. Check loop-wide
13361 reasons for disallowing fully-masked loops. Make the final decision
13362 about whether use a fully-masked loop or not.
13363 (vect_estimate_min_profitable_iters): Do not assume that peeling
13364 for the number of iterations will be needed for fully-masked loops.
13365 (vectorizable_reduction): Disable fully-masked loops.
13366 (vectorizable_live_operation): Likewise.
13367 (vect_halve_mask_nunits): New function.
13368 (vect_double_mask_nunits): Likewise.
13369 (vect_record_loop_mask): Likewise.
13370 (vect_get_loop_mask): Likewise.
13371 (vect_transform_loop): Handle the case in which the final loop
13372 iteration might handle a partial vector. Call vect_set_loop_condition
13373 instead of slpeel_make_loop_iterate_ntimes.
13374 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13375 (check_load_store_masking): New function.
13376 (prepare_load_store_mask): Likewise.
13377 (vectorizable_store): Handle fully-masked loops.
13378 (vectorizable_load): Likewise.
13379 (supportable_widening_operation): Use vect_halve_mask_nunits for
13381 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13382 (vect_gen_while): New function.
13383 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13384 (aarch64_uqdec<mode>): New insn.
13386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13387 Alan Hayward <alan.hayward@arm.com>
13388 David Sherwood <david.sherwood@arm.com>
13390 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13391 (reduc_xor_scal_optab): New optabs.
13392 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13393 (reduc_xor_scal_@var{m}): Document.
13394 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13395 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13396 internal functions.
13397 * fold-const-call.c (fold_const_call): Handle them.
13398 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13399 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13400 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13401 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13402 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13403 (UNSPEC_XORV): New unspecs.
13404 (optab): Add entries for them.
13405 (BITWISEV): New int iterator.
13406 (bit_reduc_op): New int attributes.
13408 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13409 Alan Hayward <alan.hayward@arm.com>
13410 David Sherwood <david.sherwood@arm.com>
13412 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13413 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13414 * optabs.def (vec_shl_insert_optab): New optab.
13415 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13416 (duplicate_and_interleave): Likewise.
13417 * tree-vect-loop.c: Include internal-fn.h.
13418 (neutral_op_for_slp_reduction): New function, split out from
13419 get_initial_defs_for_reduction.
13420 (get_initial_def_for_reduction): Handle option 2 for variable-length
13421 vectors by loading the neutral value into a vector and then shifting
13422 the initial value into element 0.
13423 (get_initial_defs_for_reduction): Replace the code argument with
13424 the neutral value calculated by neutral_op_for_slp_reduction.
13425 Use gimple_build_vector for constant-length vectors.
13426 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13427 but the first group_size elements have a neutral value.
13428 Use duplicate_and_interleave otherwise.
13429 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13430 Update call to get_initial_defs_for_reduction. Handle SLP
13431 reductions for variable-length vectors by creating one vector
13432 result for each scalar result, with the elements associated
13433 with other scalar results stubbed out with the neutral value.
13434 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13435 Require IFN_VEC_SHL_INSERT for double reductions on
13436 variable-length vectors, or SLP reductions that have
13437 a neutral value. Require can_duplicate_and_interleave_p
13438 support for variable-length unchained SLP reductions if there
13439 is no neutral value, such as for MIN/MAX reductions. Also require
13440 the number of vector elements to be a multiple of the number of
13441 SLP statements when doing variable-length unchained SLP reductions.
13442 Update call to vect_create_epilog_for_reduction.
13443 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13444 and remove initial values.
13445 (duplicate_and_interleave): Make public.
13446 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13447 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13449 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13450 Alan Hayward <alan.hayward@arm.com>
13451 David Sherwood <david.sherwood@arm.com>
13453 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13454 (can_duplicate_and_interleave_p): New function.
13455 (vect_get_and_check_slp_defs): Take the vector of statements
13456 rather than just the current one. Remove excess parentheses.
13457 Restriction rejectinon of vect_constant_def and vect_external_def
13458 for variable-length vectors to boolean types, or types for which
13459 can_duplicate_and_interleave_p is false.
13460 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13461 (duplicate_and_interleave): New function.
13462 (vect_get_constant_vectors): Use gimple_build_vector for
13463 constant-length vectors and suitable variable-length constant
13464 vectors. Use duplicate_and_interleave for other variable-length
13465 vectors. Don't defer the update when inserting new statements.
13467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13468 Alan Hayward <alan.hayward@arm.com>
13469 David Sherwood <david.sherwood@arm.com>
13471 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13472 min_profitable_iters doesn't go negative.
13474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13475 Alan Hayward <alan.hayward@arm.com>
13476 David Sherwood <david.sherwood@arm.com>
13478 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13479 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13480 * optabs.def (vec_mask_load_lanes_optab): New optab.
13481 (vec_mask_store_lanes_optab): Likewise.
13482 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13483 (MASK_STORE_LANES): Likewise.
13484 * internal-fn.c (mask_load_lanes_direct): New macro.
13485 (mask_store_lanes_direct): Likewise.
13486 (expand_mask_load_optab_fn): Handle masked operations.
13487 (expand_mask_load_lanes_optab_fn): New macro.
13488 (expand_mask_store_optab_fn): Handle masked operations.
13489 (expand_mask_store_lanes_optab_fn): New macro.
13490 (direct_mask_load_lanes_optab_supported_p): Likewise.
13491 (direct_mask_store_lanes_optab_supported_p): Likewise.
13492 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13494 (vect_load_lanes_supported): Likewise.
13495 * tree-vect-data-refs.c (strip_conversion): New function.
13496 (can_group_stmts_p): Likewise.
13497 (vect_analyze_data_ref_accesses): Use it instead of checking
13498 for a pair of assignments.
13499 (vect_store_lanes_supported): Take a masked_p parameter.
13500 (vect_load_lanes_supported): Likewise.
13501 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13502 vect_store_lanes_supported and vect_load_lanes_supported.
13503 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13504 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13505 parameter. Don't allow gaps for masked accesses.
13506 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13507 and vect_load_lanes_supported.
13508 (get_load_store_type): Take a masked_p parameter and update
13509 call to get_group_load_store_type.
13510 (vectorizable_store): Update call to get_load_store_type.
13511 Handle IFN_MASK_STORE_LANES.
13512 (vectorizable_load): Update call to get_load_store_type.
13513 Handle IFN_MASK_LOAD_LANES.
13515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13516 Alan Hayward <alan.hayward@arm.com>
13517 David Sherwood <david.sherwood@arm.com>
13519 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13521 * config/aarch64/aarch64-protos.h
13522 (aarch64_sve_struct_memory_operand_p): Declare.
13523 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13524 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13525 (VPRED, vpred): Handle SVE structure modes.
13526 * config/aarch64/constraints.md (Utx): New constraint.
13527 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13528 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13529 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13530 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13531 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13532 structure modes. Split into pieces after RA.
13533 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13534 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13536 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13537 SVE structure modes.
13538 (aarch64_classify_address): Likewise.
13539 (sizetochar): Move earlier in file.
13540 (aarch64_print_operand): Handle SVE register lists.
13541 (aarch64_array_mode): New function.
13542 (aarch64_sve_struct_memory_operand_p): Likewise.
13543 (TARGET_ARRAY_MODE): Redefine.
13545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13546 Alan Hayward <alan.hayward@arm.com>
13547 David Sherwood <david.sherwood@arm.com>
13549 * target.def (array_mode): New target hook.
13550 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13551 * doc/tm.texi: Regenerate.
13552 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13553 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13554 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13555 targetm.array_mode.
13556 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13560 Alan Hayward <alan.hayward@arm.com>
13561 David Sherwood <david.sherwood@arm.com>
13563 * fold-const.c (fold_binary_loc): Check the argument types
13564 rather than the result type when testing for a vector operation.
13566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13568 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13569 * doc/tm.texi: Regenerate.
13571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13572 Alan Hayward <alan.hayward@arm.com>
13573 David Sherwood <david.sherwood@arm.com>
13575 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13576 (sve): Document new AArch64 extension.
13577 * doc/md.texi (w): Extend the description of the AArch64
13578 constraint to include SVE vectors.
13579 (Upl, Upa): Document new AArch64 predicate constraints.
13580 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13582 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13583 (msve-vector-bits=): New option.
13584 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13585 SVE when these are disabled.
13586 (sve): New extension.
13587 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13588 modes. Adjust their number of units based on aarch64_sve_vg.
13589 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13590 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13591 aarch64_addr_query_type.
13592 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13593 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13594 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13595 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13596 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13597 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13598 (aarch64_simd_imm_zero_p): Delete.
13599 (aarch64_check_zero_based_sve_index_immediate): Declare.
13600 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13601 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13602 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13603 (aarch64_sve_float_mul_immediate_p): Likewise.
13604 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13605 rather than an rtx.
13606 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13607 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13608 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13609 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13610 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13611 (aarch64_regmode_natural_size): Likewise.
13612 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13613 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13615 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13616 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13617 for VG and the SVE predicate registers.
13618 (V_ALIASES): Add a "z"-prefixed alias.
13619 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13620 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13621 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13622 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13623 (REG_CLASS_NAMES): Add entries for them.
13624 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13625 and the predicate registers.
13626 (aarch64_sve_vg): Declare.
13627 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13628 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13629 (REGMODE_NATURAL_SIZE): Define.
13630 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13632 * config/aarch64/aarch64.c: Include cfgrtl.h.
13633 (simd_immediate_info): Add a constructor for series vectors,
13634 and an associated step field.
13635 (aarch64_sve_vg): New variable.
13636 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13637 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13638 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13639 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13640 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13641 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13642 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13643 (aarch64_get_mask_mode): New functions.
13644 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13645 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13646 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13647 predicate modes and predicate registers. Explicitly restrict
13648 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13649 to store a vector mode if it is recognized by
13650 aarch64_classify_vector_mode.
13651 (aarch64_regmode_natural_size): New function.
13652 (aarch64_hard_regno_caller_save_mode): Return the original mode
13654 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13655 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13656 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13657 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13659 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13660 does not overlap dest if the function is frame-related. Handle
13662 (aarch64_split_add_offset): New function.
13663 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13664 them aarch64_add_offset.
13665 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13666 and update call to aarch64_sub_sp.
13667 (aarch64_add_cfa_expression): New function.
13668 (aarch64_expand_prologue): Pass extra temporary registers to the
13669 functions above. Handle the case in which we need to emit new
13670 DW_CFA_expressions for registers that were originally saved
13671 relative to the stack pointer, but now have to be expressed
13672 relative to the frame pointer.
13673 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13675 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13676 IP0 and IP1 values for SVE frames.
13677 (aarch64_expand_vec_series): New function.
13678 (aarch64_expand_sve_widened_duplicate): Likewise.
13679 (aarch64_expand_sve_const_vector): Likewise.
13680 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13681 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13682 into the register, rather than emitting a SET directly.
13683 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13684 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13685 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13686 (offset_9bit_signed_scaled_p): New functions.
13687 (aarch64_replicate_bitmask_imm): New function.
13688 (aarch64_bitmask_imm): Use it.
13689 (aarch64_cannot_force_const_mem): Reject expressions involving
13690 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13691 (aarch64_classify_index): Handle SVE indices, by requiring
13692 a plain register index with a scale that matches the element size.
13693 (aarch64_classify_address): Handle SVE addresses. Assert that
13694 the mode of the address is VOIDmode or an integer mode.
13695 Update call to aarch64_classify_symbol.
13696 (aarch64_classify_symbolic_expression): Update call to
13697 aarch64_classify_symbol.
13698 (aarch64_const_vec_all_in_range_p): New function.
13699 (aarch64_print_vector_float_operand): Likewise.
13700 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13701 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13702 and the FP immediates 1.0 and 0.5.
13703 (aarch64_print_address_internal): Handle SVE addresses.
13704 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13705 (aarch64_regno_regclass): Handle predicate registers.
13706 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13708 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13709 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13710 (aarch64_convert_sve_vector_bits): New function.
13711 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13712 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13713 rather than an rtx.
13714 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13715 Handle SVE vector and predicate modes. Accept VL-based constants
13716 that need only one temporary register, and VL offsets that require
13717 no temporary registers.
13718 (aarch64_conditional_register_usage): Mark the predicate registers
13719 as fixed if SVE isn't available.
13720 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13721 Return true for SVE vector and predicate modes.
13722 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13723 rather than an unsigned int. Handle SVE modes.
13724 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13726 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13728 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13729 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13730 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13731 (aarch64_sve_float_mul_immediate_p): New functions.
13732 (aarch64_sve_valid_immediate): New function.
13733 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13734 Explicitly reject structure modes. Check for INDEX constants.
13735 Handle PTRUE and PFALSE constants.
13736 (aarch64_check_zero_based_sve_index_immediate): New function.
13737 (aarch64_simd_imm_zero_p): Delete.
13738 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13739 vector modes. Accept constants in the range of CNT[BHWD].
13740 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13741 ask for an Advanced SIMD mode.
13742 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13743 (aarch64_simd_vector_alignment): Handle SVE predicates.
13744 (aarch64_vectorize_preferred_vector_alignment): New function.
13745 (aarch64_simd_vector_alignment_reachable): Use it instead of
13747 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13748 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13750 (MAX_VECT_LEN): Delete.
13751 (expand_vec_perm_d): Add a vec_flags field.
13752 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13753 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13754 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13756 (aarch64_evpc_rev): Rename to...
13757 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13758 (aarch64_evpc_rev_global): New function.
13759 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13760 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13762 (aarch64_evpc_sve_tbl): New function.
13763 (aarch64_expand_vec_perm_const_1): Update after rename of
13764 aarch64_evpc_rev. Handle SVE permutes too, trying
13765 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13766 than aarch64_evpc_tbl.
13767 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13768 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13769 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13770 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13771 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13772 (aarch64_expand_sve_vcond): New functions.
13773 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13774 of aarch64_vector_mode_p.
13775 (aarch64_dwarf_poly_indeterminate_value): New function.
13776 (aarch64_compute_pressure_classes): Likewise.
13777 (aarch64_can_change_mode_class): Likewise.
13778 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13779 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13780 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13781 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13782 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13783 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13784 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13785 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13787 (Dn, Dl, Dr): Accept const as well as const_vector.
13788 (Dz): Likewise. Compare against CONST0_RTX.
13789 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13790 of "vector" where appropriate.
13791 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13792 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13793 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13794 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13795 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13796 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13797 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13798 (v_int_equiv): Extend to SVE modes.
13799 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13801 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13802 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13803 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13804 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13805 (SVE_COND_FP_CMP): New int iterators.
13806 (perm_hilo): Handle the new unpack unspecs.
13807 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13809 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13810 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13811 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13812 (aarch64_equality_operator, aarch64_constant_vector_operand)
13813 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13814 (aarch64_sve_nonimmediate_operand): Likewise.
13815 (aarch64_sve_general_operand): Likewise.
13816 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13817 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13818 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13819 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13820 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13821 (aarch64_sve_float_arith_immediate): Likewise.
13822 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13823 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13824 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13825 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13826 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13827 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13828 (aarch64_sve_float_arith_operand): Likewise.
13829 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13830 (aarch64_sve_float_mul_operand): Likewise.
13831 (aarch64_sve_vec_perm_operand): Likewise.
13832 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13833 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13834 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13835 as well as const_vector.
13836 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13837 in file. Use CONST0_RTX and CONSTM1_RTX.
13838 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13839 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13840 Use aarch64_simd_imm_zero.
13841 * config/aarch64/aarch64-sve.md: New file.
13842 * config/aarch64/aarch64.md: Include it.
13843 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13844 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13845 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13846 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13847 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13848 (sve): New attribute.
13849 (enabled): Disable instructions with the sve attribute unless
13851 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13852 aarch64_expand_mov_immediate.
13853 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13854 CNT[BHSD] immediates.
13855 (movti): Split CONST_POLY_INT moves into two halves.
13856 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13857 Split additions that need a temporary here if the destination
13858 is the stack pointer.
13859 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13860 (*add<mode>3_poly_1): New instruction.
13861 (set_clobber_cc): New expander.
13863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13865 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13866 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13867 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13868 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13869 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13870 Change innermode from fixed_mode_size to machine_mode.
13871 (simplify_subreg): Update call accordingly. Handle a constant-sized
13872 subreg of a variable-length CONST_VECTOR.
13874 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13875 Alan Hayward <alan.hayward@arm.com>
13876 David Sherwood <david.sherwood@arm.com>
13878 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13879 (add_offset_to_base): New function, split out from...
13880 (create_mem_ref): ...here. When handling a scale other than 1,
13881 check first whether the address is valid without the offset.
13882 Add it into the base if so, leaving the index and scale as-is.
13884 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13887 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13888 fold_for_warn before checking if arg2 is INTEGER_CST.
13890 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13892 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13893 (store_multiple_operation): Delete.
13894 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13895 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13896 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13897 guarded by TARGET_STRING.
13898 (rs6000_output_load_multiple): Delete.
13899 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13900 OPTION_MASK_STRING / TARGET_STRING handling.
13901 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13902 (const rs6000_opt_masks) <"string">: Change mask to 0.
13903 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13904 (MASK_STRING): Delete.
13905 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13907 (load_multiple): Delete.
13914 (store_multiple): Delete.
13921 (movmemsi_8reg): Delete.
13922 (corresponding unnamed define_insn): Delete.
13923 (movmemsi_6reg): Delete.
13924 (corresponding unnamed define_insn): Delete.
13925 (movmemsi_4reg): Delete.
13926 (corresponding unnamed define_insn): Delete.
13927 (movmemsi_2reg): Delete.
13928 (corresponding unnamed define_insn): Delete.
13929 (movmemsi_1reg): Delete.
13930 (corresponding unnamed define_insn): Delete.
13931 * config/rs6000/rs6000.opt (mno-string): New.
13932 (mstring): Replace by deprecation warning stub.
13933 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13935 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13937 * regrename.c (regrename_do_replace): If replacing the same
13938 reg multiple times, try to reuse last created gen_raw_REG.
13941 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13942 main to workaround a bug in GDB.
13944 2018-01-12 Tom de Vries <tom@codesourcery.com>
13947 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13949 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13951 PR rtl-optimization/80481
13952 * ira-color.c (get_cap_member): New function.
13953 (allocnos_conflict_by_live_ranges_p): Use it.
13954 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13955 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13957 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13960 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13961 (*saddl_se_1): Ditto.
13962 (*ssubsi_1): Ditto.
13963 (*ssubl_se_1): Ditto.
13965 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13967 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13968 rather than wi::to_widest for DR_INITs.
13969 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13970 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13971 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13973 (vect_analyze_group_access_1): Note that here.
13975 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13977 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13978 polynomial type sizes.
13980 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13982 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13983 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13984 (gimple_add_tmp_var): Likewise.
13986 2018-01-12 Martin Liska <mliska@suse.cz>
13988 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13989 (gimple_alloc_sizes): Likewise.
13990 (dump_gimple_statistics): Use PRIu64 in printf format.
13991 * gimple.h: Change uint64_t to int.
13993 2018-01-12 Martin Liska <mliska@suse.cz>
13995 * tree-core.h: Use uint64_t instead of int.
13996 * tree.c (tree_node_counts): Likewise.
13997 (tree_node_sizes): Likewise.
13998 (dump_tree_statistics): Use PRIu64 in printf format.
14000 2018-01-12 Martin Liska <mliska@suse.cz>
14002 * Makefile.in: As qsort_chk is implemented in vec.c, add
14003 vec.o to linkage of gencfn-macros.
14004 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14005 passing the info to record_node_allocation_statistics.
14006 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14008 * ggc-common.c (struct ggc_usage): Add operator== and use
14009 it in operator< and compare function.
14010 * mem-stats.h (struct mem_usage): Likewise.
14011 * vec.c (struct vec_usage): Remove operator< and compare
14012 function. Can be simply inherited.
14014 2018-01-12 Martin Jambor <mjambor@suse.cz>
14017 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14018 * tree-ssa-math-opts.c: Include domwalk.h.
14019 (convert_mult_to_fma_1): New function.
14020 (fma_transformation_info): New type.
14021 (fma_deferring_state): Likewise.
14022 (cancel_fma_deferring): New function.
14023 (result_of_phi): Likewise.
14024 (last_fma_candidate_feeds_initial_phi): Likewise.
14025 (convert_mult_to_fma): Added deferring logic, split actual
14026 transformation to convert_mult_to_fma_1.
14027 (math_opts_dom_walker): New type.
14028 (math_opts_dom_walker::after_dom_children): New method, body moved
14029 here from pass_optimize_widening_mul::execute, added deferring logic
14031 (pass_optimize_widening_mul::execute): Moved most of code to
14032 math_opts_dom_walker::after_dom_children.
14033 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14034 * config/i386/i386.c (ix86_option_override_internal): Added
14035 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14037 2018-01-12 Richard Biener <rguenther@suse.de>
14040 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14041 inline instance vars.
14043 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14046 * config/rx/rx.c (rx_is_restricted_memory_address):
14047 Handle SUBREG case.
14049 2018-01-12 Richard Biener <rguenther@suse.de>
14051 PR tree-optimization/80846
14052 * target.def (split_reduction): New target hook.
14053 * targhooks.c (default_split_reduction): New function.
14054 * targhooks.h (default_split_reduction): Declare.
14055 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14056 target requests first reduce vectors by combining low and high
14058 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14059 (get_vectype_for_scalar_type_and_size): Export.
14060 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14061 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14062 * doc/tm.texi: Regenerate.
14063 * config/i386/i386.c (ix86_split_reduction): Implement
14064 TARGET_VECTORIZE_SPLIT_REDUCTION.
14066 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14069 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14070 in PIC mode except for TARGET_VXWORKS_RTP.
14071 * config/sparc/sparc.c: Include cfgrtl.h.
14072 (TARGET_INIT_PIC_REG): Define.
14073 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14074 (sparc_pic_register_p): New predicate.
14075 (sparc_legitimate_address_p): Use it.
14076 (sparc_legitimize_pic_address): Likewise.
14077 (sparc_delegitimize_address): Likewise.
14078 (sparc_mode_dependent_address_p): Likewise.
14079 (gen_load_pcrel_sym): Remove 4th parameter.
14080 (load_got_register): Adjust call to above. Remove obsolete stuff.
14081 (sparc_expand_prologue): Do not call load_got_register here.
14082 (sparc_flat_expand_prologue): Likewise.
14083 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14084 (sparc_use_pseudo_pic_reg): New function.
14085 (sparc_init_pic_reg): Likewise.
14086 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14087 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14089 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14091 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14092 Add item for branch_cost.
14094 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14096 PR rtl-optimization/83565
14097 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14098 not extend the result to a larger mode for rotate operations.
14099 (num_sign_bit_copies1): Likewise.
14101 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14104 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14106 Use values-Xc.o for -pedantic.
14107 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14109 2018-01-12 Martin Liska <mliska@suse.cz>
14112 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14114 (possible_polymorphic_call_targets): Use it.
14115 (ipa_devirt): Likewise.
14117 2018-01-12 Martin Liska <mliska@suse.cz>
14119 * profile-count.h (enum profile_quality): Use 0 as invalid
14120 enum value of profile_quality.
14122 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14124 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14125 -mext-string options.
14127 2018-01-12 Richard Biener <rguenther@suse.de>
14129 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14130 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14131 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14133 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14135 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14137 * configure.ac (--with-long-double-format): Add support for the
14138 configuration option to change the default long double format on
14140 * config.gcc (powerpc*-linux*-*): Likewise.
14141 * configure: Regenerate.
14142 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14143 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14144 used without modification.
14146 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14148 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14149 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14150 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14151 MISC_BUILTIN_SPEC_BARRIER.
14152 (rs6000_init_builtins): Likewise.
14153 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14155 (speculation_barrier): New define_insn.
14156 * doc/extend.texi: Document __builtin_speculation_barrier.
14158 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14161 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14162 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14163 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14165 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14166 integral modes instead of "ss" and "sd".
14167 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14168 vectors with 32-bit and 64-bit elements.
14169 (vecdupssescalarmodesuffix): New mode attribute.
14170 (vec_dup<mode>): Use it.
14172 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14175 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14176 frame if argument is passed on stack.
14178 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14181 * ree.c (combine_reaching_defs): Optimize also
14182 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14183 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14185 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14187 PR middle-end/83189
14188 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14190 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14192 PR middle-end/83718
14193 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14194 after they are computed.
14196 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14198 PR tree-optimization/83695
14199 * gimple-loop-linterchange.cc
14200 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14201 reset cached scev information after interchange.
14202 (pass_linterchange::execute): Remove call to scev_reset_htab.
14204 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14206 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14207 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14208 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14209 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14210 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14211 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14212 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14213 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14214 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14215 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14216 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14217 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14218 (V_lane_reg): Likewise.
14219 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14221 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14222 (vfmal_lane_low<mode>_intrinsic,
14223 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14224 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14225 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14226 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14227 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14228 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14230 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14232 * config/arm/arm-cpus.in (fp16fml): New feature.
14233 (ALL_SIMD): Add fp16fml.
14234 (armv8.2-a): Add fp16fml as an option.
14235 (armv8.3-a): Likewise.
14236 (armv8.4-a): Add fp16fml as part of fp16.
14237 * config/arm/arm.h (TARGET_FP16FML): Define.
14238 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14240 * config/arm/arm-modes.def (V2HF): Define.
14241 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14242 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14243 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14244 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14245 vfmsl_low, vfmsl_high): New set of builtins.
14246 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14247 (vfml_op): New code attribute.
14248 (VFMLHALVES): New int iterator.
14249 (VFML, VFMLSEL): New mode attributes.
14250 (V_reg): Define mapping for V2HF.
14251 (V_hi, V_lo): New mode attributes.
14252 (VF_constraint): Likewise.
14253 (vfml_half, vfml_half_selector): New int attributes.
14254 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14256 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14257 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14259 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14260 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14261 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14262 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14264 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14265 Document new effective target and option set.
14267 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14269 * config/arm/arm-cpus.in (armv8_4): New feature.
14270 (ARMv8_4a): New fgroup.
14271 (armv8.4-a): New arch.
14272 * config/arm/arm-tables.opt: Regenerate.
14273 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14274 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14275 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14276 Add matching rules for -march=armv8.4-a and extensions.
14277 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14279 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14282 * config/rx/rx.md (BW): New mode attribute.
14283 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14285 2018-01-11 Richard Biener <rguenther@suse.de>
14287 PR tree-optimization/83435
14288 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14289 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14290 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14292 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14293 Alan Hayward <alan.hayward@arm.com>
14294 David Sherwood <david.sherwood@arm.com>
14296 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14298 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14299 (aarch64_print_address_internal): Use it to check for a zero offset.
14301 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14302 Alan Hayward <alan.hayward@arm.com>
14303 David Sherwood <david.sherwood@arm.com>
14305 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14306 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14307 Return a poly_int64 rather than a HOST_WIDE_INT.
14308 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14309 rather than a HOST_WIDE_INT.
14310 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14311 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14312 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14313 final_offset from HOST_WIDE_INT to poly_int64.
14314 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14315 to_constant when getting the number of units in an Advanced SIMD
14317 (aarch64_builtin_vectorized_function): Check for a constant number
14319 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14321 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14322 attribute instead of GET_MODE_NUNITS.
14323 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14324 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14325 GET_MODE_SIZE for fixed-size registers.
14326 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14327 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14328 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14329 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14330 (aarch64_print_operand, aarch64_print_address_internal)
14331 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14332 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14333 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14334 Handle polynomial GET_MODE_SIZE.
14335 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14336 wider than SImode without modification.
14337 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14338 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14339 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14340 passing and returning SVE modes.
14341 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14342 rather than GEN_INT.
14343 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14344 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14345 (aarch64_allocate_and_probe_stack_space): Likewise.
14346 (aarch64_layout_frame): Cope with polynomial offsets.
14347 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14348 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14349 polynomial offsets.
14350 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14351 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14352 poly_int64 rather than a HOST_WIDE_INT.
14353 (aarch64_get_separate_components, aarch64_process_components)
14354 (aarch64_expand_prologue, aarch64_expand_epilogue)
14355 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14356 (aarch64_anchor_offset): New function, split out from...
14357 (aarch64_legitimize_address): ...here.
14358 (aarch64_builtin_vectorization_cost): Handle polynomial
14359 TYPE_VECTOR_SUBPARTS.
14360 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14362 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14363 number of elements from the PARALLEL rather than the mode.
14364 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14365 rather than GET_MODE_BITSIZE.
14366 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14367 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14368 (aarch64_expand_vec_perm_const_1): Handle polynomial
14369 d->perm.length () and d->perm elements.
14370 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14371 Apply to_constant to d->perm elements.
14372 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14373 polynomial CONST_VECTOR_NUNITS.
14374 (aarch64_move_pointer): Take amount as a poly_int64 rather
14376 (aarch64_progress_pointer): Avoid temporary variable.
14377 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14378 the mode attribute instead of GET_MODE.
14380 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14381 Alan Hayward <alan.hayward@arm.com>
14382 David Sherwood <david.sherwood@arm.com>
14384 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14385 x exists before using it.
14386 (aarch64_add_constant_internal): Rename to...
14387 (aarch64_add_offset_1): ...this. Replace regnum with separate
14388 src and dest rtxes. Handle the case in which they're different,
14389 including when the offset is zero. Replace scratchreg with an rtx.
14390 Use 2 additions if there is no spare register into which we can
14391 move a 16-bit constant.
14392 (aarch64_add_constant): Delete.
14393 (aarch64_add_offset): Replace reg with separate src and dest
14394 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14395 Use aarch64_add_offset_1.
14396 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14397 an rtx rather than an int. Take the delta as a poly_int64
14398 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14399 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14400 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14401 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14402 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14403 and aarch64_add_sp.
14404 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14405 aarch64_add_constant.
14407 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14409 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14410 Use scalar_float_mode.
14412 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14414 * config/aarch64/aarch64-simd.md
14415 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14416 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14417 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14418 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14419 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14420 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14421 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14422 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14423 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14424 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14426 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14429 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14430 targ_options->x_arm_arch_string is non NULL.
14432 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14434 * config/aarch64/aarch64.h
14435 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14437 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14440 * expmed.c (emit_store_flag_force): Swap if const op0
14441 and change VOIDmode to mode of op0.
14443 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14445 PR rtl-optimization/83761
14446 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14447 than bytes to mode_for_size.
14449 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14451 PR middle-end/83189
14452 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14453 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14456 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14458 PR middle-end/83575
14459 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14460 when in layout mode.
14461 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14462 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14465 2018-01-10 Michael Collison <michael.collison@arm.com>
14467 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14468 * config/aarch64/aarch64-option-extension.def: Add
14469 AARCH64_OPT_EXTENSION of 'fp16fml'.
14470 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14471 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14472 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14473 * config/aarch64/constraints.md (Ui7): New constraint.
14474 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14475 (VFMLA_SEL_W): Ditto.
14478 (VFMLA16_LOW): New int iterator.
14479 (VFMLA16_HIGH): Ditto.
14480 (UNSPEC_FMLAL): New unspec.
14481 (UNSPEC_FMLSL): Ditto.
14482 (UNSPEC_FMLAL2): Ditto.
14483 (UNSPEC_FMLSL2): Ditto.
14484 (f16mac): New code attribute.
14485 * config/aarch64/aarch64-simd-builtins.def
14486 (aarch64_fmlal_lowv2sf): Ditto.
14487 (aarch64_fmlsl_lowv2sf): Ditto.
14488 (aarch64_fmlalq_lowv4sf): Ditto.
14489 (aarch64_fmlslq_lowv4sf): Ditto.
14490 (aarch64_fmlal_highv2sf): Ditto.
14491 (aarch64_fmlsl_highv2sf): Ditto.
14492 (aarch64_fmlalq_highv4sf): Ditto.
14493 (aarch64_fmlslq_highv4sf): Ditto.
14494 (aarch64_fmlal_lane_lowv2sf): Ditto.
14495 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14496 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14497 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14498 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14499 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14500 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14501 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14502 (aarch64_fmlal_lane_highv2sf): Ditto.
14503 (aarch64_fmlsl_lane_highv2sf): Ditto.
14504 (aarch64_fmlal_laneq_highv2sf): Ditto.
14505 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14506 (aarch64_fmlalq_lane_highv4sf): Ditto.
14507 (aarch64_fmlsl_lane_highv4sf): Ditto.
14508 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14509 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14510 * config/aarch64/aarch64-simd.md:
14511 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14512 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14513 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14514 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14515 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14516 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14517 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14518 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14519 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14520 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14521 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14522 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14523 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14524 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14525 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14526 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14527 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14528 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14529 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14530 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14531 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14532 (vfmlsl_low_u32): Ditto.
14533 (vfmlalq_low_u32): Ditto.
14534 (vfmlslq_low_u32): Ditto.
14535 (vfmlal_high_u32): Ditto.
14536 (vfmlsl_high_u32): Ditto.
14537 (vfmlalq_high_u32): Ditto.
14538 (vfmlslq_high_u32): Ditto.
14539 (vfmlal_lane_low_u32): Ditto.
14540 (vfmlsl_lane_low_u32): Ditto.
14541 (vfmlal_laneq_low_u32): Ditto.
14542 (vfmlsl_laneq_low_u32): Ditto.
14543 (vfmlalq_lane_low_u32): Ditto.
14544 (vfmlslq_lane_low_u32): Ditto.
14545 (vfmlalq_laneq_low_u32): Ditto.
14546 (vfmlslq_laneq_low_u32): Ditto.
14547 (vfmlal_lane_high_u32): Ditto.
14548 (vfmlsl_lane_high_u32): Ditto.
14549 (vfmlal_laneq_high_u32): Ditto.
14550 (vfmlsl_laneq_high_u32): Ditto.
14551 (vfmlalq_lane_high_u32): Ditto.
14552 (vfmlslq_lane_high_u32): Ditto.
14553 (vfmlalq_laneq_high_u32): Ditto.
14554 (vfmlslq_laneq_high_u32): Ditto.
14555 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14556 (AARCH64_FL_FOR_ARCH8_4): New.
14557 (AARCH64_ISA_F16FML): New ISA flag.
14558 (TARGET_F16FML): New feature flag for fp16fml.
14559 (doc/invoke.texi): Document new fp16fml option.
14561 2018-01-10 Michael Collison <michael.collison@arm.com>
14563 * config/aarch64/aarch64-builtins.c:
14564 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14565 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14566 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14567 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14568 (AARCH64_ISA_SHA3): New ISA flag.
14569 (TARGET_SHA3): New feature flag for sha3.
14570 * config/aarch64/iterators.md (sha512_op): New int attribute.
14571 (CRYPTO_SHA512): New int iterator.
14572 (UNSPEC_SHA512H): New unspec.
14573 (UNSPEC_SHA512H2): Ditto.
14574 (UNSPEC_SHA512SU0): Ditto.
14575 (UNSPEC_SHA512SU1): Ditto.
14576 * config/aarch64/aarch64-simd-builtins.def
14577 (aarch64_crypto_sha512hqv2di): New builtin.
14578 (aarch64_crypto_sha512h2qv2di): Ditto.
14579 (aarch64_crypto_sha512su0qv2di): Ditto.
14580 (aarch64_crypto_sha512su1qv2di): Ditto.
14581 (aarch64_eor3qv8hi): Ditto.
14582 (aarch64_rax1qv2di): Ditto.
14583 (aarch64_xarqv2di): Ditto.
14584 (aarch64_bcaxqv8hi): Ditto.
14585 * config/aarch64/aarch64-simd.md:
14586 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14587 (aarch64_crypto_sha512su0qv2di): Ditto.
14588 (aarch64_crypto_sha512su1qv2di): Ditto.
14589 (aarch64_eor3qv8hi): Ditto.
14590 (aarch64_rax1qv2di): Ditto.
14591 (aarch64_xarqv2di): Ditto.
14592 (aarch64_bcaxqv8hi): Ditto.
14593 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14594 (vsha512h2q_u64): Ditto.
14595 (vsha512su0q_u64): Ditto.
14596 (vsha512su1q_u64): Ditto.
14597 (veor3q_u16): Ditto.
14598 (vrax1q_u64): Ditto.
14599 (vxarq_u64): Ditto.
14600 (vbcaxq_u16): Ditto.
14601 * config/arm/types.md (crypto_sha512): New type attribute.
14602 (crypto_sha3): Ditto.
14603 (doc/invoke.texi): Document new sha3 option.
14605 2018-01-10 Michael Collison <michael.collison@arm.com>
14607 * config/aarch64/aarch64-builtins.c:
14608 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14609 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14610 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14611 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14612 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14613 (AARCH64_ISA_SM4): New ISA flag.
14614 (TARGET_SM4): New feature flag for sm4.
14615 * config/aarch64/aarch64-simd-builtins.def
14616 (aarch64_sm3ss1qv4si): Ditto.
14617 (aarch64_sm3tt1aq4si): Ditto.
14618 (aarch64_sm3tt1bq4si): Ditto.
14619 (aarch64_sm3tt2aq4si): Ditto.
14620 (aarch64_sm3tt2bq4si): Ditto.
14621 (aarch64_sm3partw1qv4si): Ditto.
14622 (aarch64_sm3partw2qv4si): Ditto.
14623 (aarch64_sm4eqv4si): Ditto.
14624 (aarch64_sm4ekeyqv4si): Ditto.
14625 * config/aarch64/aarch64-simd.md:
14626 (aarch64_sm3ss1qv4si): Ditto.
14627 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14628 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14629 (aarch64_sm4eqv4si): Ditto.
14630 (aarch64_sm4ekeyqv4si): Ditto.
14631 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14632 (sm3part_op): Ditto.
14633 (CRYPTO_SM3TT): Ditto.
14634 (CRYPTO_SM3PART): Ditto.
14635 (UNSPEC_SM3SS1): New unspec.
14636 (UNSPEC_SM3TT1A): Ditto.
14637 (UNSPEC_SM3TT1B): Ditto.
14638 (UNSPEC_SM3TT2A): Ditto.
14639 (UNSPEC_SM3TT2B): Ditto.
14640 (UNSPEC_SM3PARTW1): Ditto.
14641 (UNSPEC_SM3PARTW2): Ditto.
14642 (UNSPEC_SM4E): Ditto.
14643 (UNSPEC_SM4EKEY): Ditto.
14644 * config/aarch64/constraints.md (Ui2): New constraint.
14645 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14646 * config/arm/types.md (crypto_sm3): New type attribute.
14647 (crypto_sm4): Ditto.
14648 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14649 (vsm3tt1aq_u32): Ditto.
14650 (vsm3tt1bq_u32): Ditto.
14651 (vsm3tt2aq_u32): Ditto.
14652 (vsm3tt2bq_u32): Ditto.
14653 (vsm3partw1q_u32): Ditto.
14654 (vsm3partw2q_u32): Ditto.
14655 (vsm4eq_u32): Ditto.
14656 (vsm4ekeyq_u32): Ditto.
14657 (doc/invoke.texi): Document new sm4 option.
14659 2018-01-10 Michael Collison <michael.collison@arm.com>
14661 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14662 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14663 (AARCH64_FL_FOR_ARCH8_4): New.
14664 (AARCH64_FL_V8_4): New flag.
14665 (doc/invoke.texi): Document new armv8.4-a option.
14667 2018-01-10 Michael Collison <michael.collison@arm.com>
14669 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14670 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14671 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14672 * config/aarch64/aarch64-option-extension.def: Add
14673 AARCH64_OPT_EXTENSION of 'sha2'.
14674 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14675 (crypto): Disable sha2 and aes if crypto disabled.
14676 (crypto): Enable aes and sha2 if enabled.
14677 (simd): Disable sha2 and aes if simd disabled.
14678 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14680 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14681 (TARGET_SHA2): New feature flag for sha2.
14682 (TARGET_AES): New feature flag for aes.
14683 * config/aarch64/aarch64-simd.md:
14684 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14685 conditional on TARGET_AES.
14686 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14687 (aarch64_crypto_sha1hsi): Make pattern conditional
14689 (aarch64_crypto_sha1hv4si): Ditto.
14690 (aarch64_be_crypto_sha1hv4si): Ditto.
14691 (aarch64_crypto_sha1su1v4si): Ditto.
14692 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14693 (aarch64_crypto_sha1su0v4si): Ditto.
14694 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14695 (aarch64_crypto_sha256su0v4si): Ditto.
14696 (aarch64_crypto_sha256su1v4si): Ditto.
14697 (doc/invoke.texi): Document new aes and sha2 options.
14699 2018-01-10 Martin Sebor <msebor@redhat.com>
14701 PR tree-optimization/83781
14702 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14705 2018-01-11 Martin Sebor <msebor@gmail.com>
14706 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14708 PR tree-optimization/83501
14709 PR tree-optimization/81703
14711 * tree-ssa-strlen.c (get_string_cst): Rename...
14712 (get_string_len): ...to this. Handle global constants.
14713 (handle_char_store): Adjust.
14715 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14716 Jim Wilson <jimw@sifive.com>
14718 * config/riscv/riscv-protos.h (riscv_output_return): New.
14719 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14720 (riscv_attribute_table, riscv_output_return),
14721 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14722 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14723 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14725 (riscv_expand_prologue): Add early return for naked function.
14726 (riscv_expand_epilogue): Likewise.
14727 (riscv_function_ok_for_sibcall): Return false for naked function.
14728 (riscv_set_current_function): New.
14729 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14730 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14731 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14732 * doc/extend.texi (RISC-V Function Attributes): New.
14734 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14736 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14737 check for 128-bit long double before checking TCmode.
14738 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14739 128-bit long doubles before checking TFmode or TCmode.
14740 (FLOAT128_IBM_P): Likewise.
14742 2018-01-10 Martin Sebor <msebor@redhat.com>
14744 PR tree-optimization/83671
14745 * builtins.c (c_strlen): Unconditionally return zero for the empty
14747 Use -Warray-bounds for warnings.
14748 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14749 for non-constant array indices with COMPONENT_REF, arrays of
14750 arrays, and pointers to arrays.
14751 (gimple_fold_builtin_strlen): Determine and set length range for
14752 non-constant character arrays.
14754 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14756 PR middle-end/81897
14757 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14760 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14762 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14764 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14767 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14768 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14769 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14770 indexed_or_indirect_operand predicate.
14771 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14772 (*vsx_le_perm_load_v8hi): Likewise.
14773 (*vsx_le_perm_load_v16qi): Likewise.
14774 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14775 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14776 (*vsx_le_perm_store_v8hi): Likewise.
14777 (*vsx_le_perm_store_v16qi): Likewise.
14778 (eight unnamed splitters): Likewise.
14780 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14782 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14783 * config/rs6000/emmintrin.h: Likewise.
14784 * config/rs6000/mmintrin.h: Likewise.
14785 * config/rs6000/xmmintrin.h: Likewise.
14787 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14790 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14792 * tree.c (tree_nop_conversion): Return true for location wrapper
14794 (maybe_wrap_with_location): New function.
14795 (selftest::check_strip_nops): New function.
14796 (selftest::test_location_wrappers): New function.
14797 (selftest::tree_c_tests): Call it.
14798 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14799 (maybe_wrap_with_location): New decl.
14800 (EXPR_LOCATION_WRAPPER_P): New macro.
14801 (location_wrapper_p): New inline function.
14802 (tree_strip_any_location_wrapper): New inline function.
14804 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14807 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14808 stack_realign_offset for the largest alignment of stack slot
14810 (ix86_find_max_used_stack_alignment): New function.
14811 (ix86_finalize_stack_frame_flags): Use it. Set
14812 max_used_stack_alignment if we don't realign stack.
14813 * config/i386/i386.h (machine_function): Add
14814 max_used_stack_alignment.
14816 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14818 * config/arm/arm.opt (-mbranch-cost): New option.
14819 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14822 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14825 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14826 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14828 2018-01-10 Richard Biener <rguenther@suse.de>
14831 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14832 early out so it also covers the case where we have a non-NULL
14835 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14837 PR tree-optimization/83753
14838 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14839 for non-strided grouped accesses if the number of elements is 1.
14841 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14844 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14845 * i386.h (TARGET_USE_GATHER): Define.
14846 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14848 2018-01-10 Martin Liska <mliska@suse.cz>
14851 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14852 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14854 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14855 CLEANUP_NO_PARTITIONING is not set.
14857 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14859 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14860 for vectors, as a partial revert of r254296.
14861 * rtl.h (const_vec_p): Delete.
14862 (const_vec_duplicate_p): Don't test for vector CONSTs.
14863 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14864 * expmed.c (make_tree): Likewise.
14867 * common.md (E, F): Use CONSTANT_P instead of checking for
14869 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14870 checking for CONST_VECTOR.
14872 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14874 PR middle-end/83575
14875 * predict.c (force_edge_cold): Handle in more sane way edges
14876 with no prediction.
14878 2018-01-09 Carl Love <cel@us.ibm.com>
14880 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14882 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14883 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14884 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14885 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14886 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14887 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14888 * config/rs6000/rs6000-protos.h: Add extern defition for
14889 rs6000_generate_float2_double_code.
14890 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14892 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14893 (float2_v2df): Add define_expand.
14895 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14898 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14899 op_mode in the force_to_mode call.
14901 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14903 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14904 instead of checking each element individually.
14905 (aarch64_evpc_uzp): Likewise.
14906 (aarch64_evpc_zip): Likewise.
14907 (aarch64_evpc_ext): Likewise.
14908 (aarch64_evpc_rev): Likewise.
14909 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14910 instead of checking each element individually. Return true without
14912 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14913 whether all selected elements come from the same input, instead of
14914 checking each element individually. Remove calls to gen_rtx_REG,
14915 start_sequence and end_sequence and instead assert that no rtl is
14918 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14920 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14921 order of HIGH and CONST checks.
14923 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14925 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14926 if the destination isn't an SSA_NAME.
14928 2018-01-09 Richard Biener <rguenther@suse.de>
14930 PR tree-optimization/83668
14931 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14933 (canonicalize_loop_form): ... here, renamed from ...
14934 (canonicalize_loop_closed_ssa_form): ... this and amended to
14935 swap successor edges for loop exit blocks to make us use
14936 the RPO order we need for initial schedule generation.
14938 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14940 PR tree-optimization/64811
14941 * match.pd: When optimizing comparisons with Inf, avoid
14942 introducing or losing exceptions from comparisons with NaN.
14944 2018-01-09 Martin Liska <mliska@suse.cz>
14947 * asan.c (shadow_mem_size): Add gcc_assert.
14949 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14951 Don't save registers in main().
14954 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14955 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14956 * config/avr/avr.c (avr_set_current_function): Don't error if
14957 naked, OS_task or OS_main are specified at the same time.
14958 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14960 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14962 * common/config/avr/avr-common.c (avr_option_optimization_table):
14963 Switch on -mmain-is-OS_task for optimizing compilations.
14965 2018-01-09 Richard Biener <rguenther@suse.de>
14967 PR tree-optimization/83572
14968 * graphite.c: Include cfganal.h.
14969 (graphite_transform_loops): Connect infinite loops to exit
14970 and remove fake edges at the end.
14972 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14974 * ipa-inline.c (edge_badness): Revert accidental checkin.
14976 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14979 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14980 symbols; not inline clones.
14982 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14985 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14986 hard registers. Formatting fixes.
14988 PR preprocessor/83722
14989 * gcc.c (try_generate_repro): Pass
14990 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14991 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14994 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14995 Kito Cheng <kito.cheng@gmail.com>
14997 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14998 (riscv_leaf_function_p): Delete.
14999 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15001 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15003 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15005 (do_ifelse): New function.
15006 (do_isel): New function.
15007 (do_sub3): New function.
15008 (do_add3): New function.
15009 (do_load_mask_compare): New function.
15010 (do_overlap_load_compare): New function.
15011 (expand_compare_loop): New function.
15012 (expand_block_compare): Call expand_compare_loop() when appropriate.
15013 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15014 option description.
15015 (-mblock-compare-inline-loop-limit): New option.
15017 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15020 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15021 Reverse order of second and third operands in first alternative.
15022 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15023 of first and second elements in UNSPEC_VPERMR vector.
15024 (altivec_expand_vec_perm_le): Likewise.
15026 2018-01-08 Jeff Law <law@redhat.com>
15028 PR rtl-optimizatin/81308
15029 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15030 (process_switch): If group_case_labels makes a change, then set
15032 (pass_convert_switch::execute): If a switch is converted, then
15033 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15035 PR rtl-optimization/81308
15036 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15039 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15041 PR target/83663 - Revert r255946
15042 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15043 generation for cases where splatting a value is not useful.
15044 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15045 across a vec_duplicate and a paradoxical subreg forming a vector
15046 mode to a vec_concat.
15048 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15050 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15051 -march=armv8.3-a variants.
15052 * config/arm/t-multilib: Likewise.
15053 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15055 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15057 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15059 (cceq_ior_compare_complement): Give it a name so I can use it, and
15060 change boolean_or_operator predicate to boolean_operator so it can
15061 be used to generate a crand.
15062 (eqne): New code iterator.
15063 (bd/bd_neg): New code_attrs.
15064 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15065 a single define_insn.
15066 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15067 decrement (bdnzt/bdnzf/bdzt/bdzf).
15068 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15069 with the new names of the branch decrement patterns, and added the
15070 names of the branch decrement conditional patterns.
15072 2018-01-08 Richard Biener <rguenther@suse.de>
15074 PR tree-optimization/83563
15075 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15078 2018-01-08 Richard Biener <rguenther@suse.de>
15080 PR middle-end/83713
15081 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15083 2018-01-08 Richard Biener <rguenther@suse.de>
15085 PR tree-optimization/83685
15086 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15087 references to abnormals.
15089 2018-01-08 Richard Biener <rguenther@suse.de>
15092 * dwarf2out.c (output_indirect_strings): Handle empty
15093 skeleton_debug_str_hash.
15094 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15096 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15098 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15099 (emit_store_direct): Likewise.
15100 (arc_trampoline_adjust_address): Likewise.
15101 (arc_asm_trampoline_template): New function.
15102 (arc_initialize_trampoline): Use asm_trampoline_template.
15103 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15104 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15105 * config/arc/arc.md (flush_icache): Delete pattern.
15107 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15109 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15110 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15113 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15116 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15117 by not USED_FOR_TARGET.
15118 (make_pass_resolve_sw_modes): Likewise.
15120 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15122 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15125 2018-01-08 Richard Biener <rguenther@suse.de>
15127 PR middle-end/83580
15128 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15130 2018-01-08 Richard Biener <rguenther@suse.de>
15132 PR middle-end/83517
15133 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15135 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15137 PR middle-end/81897
15138 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15139 basic blocks with a small number of successors.
15140 (convert_control_dep_chain_into_preds): Improve handling of
15142 (dump_predicates): Split apart into...
15143 (dump_pred_chain): ...here...
15144 (dump_pred_info): ...and here.
15145 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15146 (can_chain_union_be_invalidated_p): Improve check for invalidation
15148 (uninit_uses_cannot_happen): Avoid unnecessary if
15149 convert_control_dep_chain_into_preds yielded nothing.
15151 2018-01-06 Martin Sebor <msebor@redhat.com>
15153 PR tree-optimization/83640
15154 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15155 subtracting negative offset from size.
15156 (builtin_access::overlap): Adjust offset bounds of the access to fall
15157 within the size of the object if possible.
15159 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15161 PR rtl-optimization/83699
15162 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15163 extract_bit_field_as_subreg to cases in which the extracted
15164 value is also a vector.
15166 * lra-constraints.c (process_alt_operands): Test for the equivalence
15167 substitutions when detecting a possible reload cycle.
15169 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15172 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15173 by default if flag_selective_schedling{,2}. Formatting fixes.
15175 PR rtl-optimization/83682
15176 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15177 if it has non-VECTOR_MODE element mode.
15178 (vec_duplicate_p): Likewise.
15180 PR middle-end/83694
15181 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15182 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15184 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15187 * config/i386/i386-builtin.def
15188 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15189 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15190 Require also OPTION_MASK_ISA_AVX512F in addition to
15191 OPTION_MASK_ISA_GFNI.
15192 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15193 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15194 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15195 to OPTION_MASK_ISA_GFNI.
15196 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15197 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15198 OPTION_MASK_ISA_AVX512BW.
15199 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15200 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15201 addition to OPTION_MASK_ISA_GFNI.
15202 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15203 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15204 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15205 to OPTION_MASK_ISA_GFNI.
15206 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15207 a requirement for all ISAs rather than any of them with a few
15209 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15211 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15212 bitmasks to be enabled with 3 exceptions, instead of requiring any
15213 enabled ISA with lots of exceptions.
15214 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15215 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15216 Change avx512bw in isa attribute to avx512f.
15217 * config/i386/sgxintrin.h: Add license boilerplate.
15218 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15219 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15220 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15221 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15223 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15224 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15225 temporarily sse2 rather than sse if not enabled already.
15228 * config/i386/sse.md (VI248_VLBW): Rename to ...
15229 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15230 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15231 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15232 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15233 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15234 mode iterator instead of VI248_VLBW.
15236 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15238 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15239 (record_modified): Skip clobbers; add debug output.
15240 (param_change_prob): Use sreal frequencies.
15242 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15244 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15245 punt for user-aligned variables.
15247 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15249 * tree-chrec.c (chrec_contains_symbols): Return true for
15252 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15255 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15256 of (x|y) == x for BICS pattern.
15258 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15260 PR tree-optimization/83605
15261 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15262 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15265 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15267 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15268 * config/epiphany/rtems.h: New file.
15270 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15271 Uros Bizjak <ubizjak@gmail.com>
15274 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15275 QIreg_operand instead of register_operand predicate.
15276 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15277 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15278 comments instead of -fmitigate[-_]rop.
15280 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15283 * cgraphunit.c (symbol_table::compile): Switch to text_section
15284 before calling assembly_start debug hook.
15285 * run-rtl-passes.c (run_rtl_passes): Likewise.
15288 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15290 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15291 range_int_cst_p rather than !symbolic_range_p before calling
15292 extract_range_from_multiplicative_op_1.
15294 2018-01-04 Jeff Law <law@redhat.com>
15296 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15297 redundant test in assertion.
15299 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15301 * doc/rtl.texi: Document machine_mode wrapper classes.
15303 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15305 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15306 using tree_to_uhwi.
15308 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15310 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15311 the VEC_PERM_EXPR fold to fail.
15313 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15316 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15317 to switched_sections.
15319 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15322 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15323 test for d.testing.
15325 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15328 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15329 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15331 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15334 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15335 is BLKmode and bitpos not zero or mode change is needed.
15337 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15340 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15343 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15346 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15347 instead of MULT rtx. Update all corresponding splitters.
15348 (*saddl_se): Ditto.
15349 (*ssub<modesuffix>): Ditto.
15350 (*ssubl_se): Ditto.
15351 (*cmp_sadd_di): Update split patterns.
15352 (*cmp_sadd_si): Ditto.
15353 (*cmp_sadd_sidi): Ditto.
15354 (*cmp_ssub_di): Ditto.
15355 (*cmp_ssub_si): Ditto.
15356 (*cmp_ssub_sidi): Ditto.
15357 * config/alpha/predicates.md (const23_operand): New predicate.
15358 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15359 Look for ASHIFT, not MULT inner operand.
15360 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15362 2018-01-04 Martin Liska <mliska@suse.cz>
15364 PR gcov-profile/83669
15365 * gcov.c (output_intermediate_file): Add version to intermediate
15367 * doc/gcov.texi: Document new field 'version' in intermediate
15368 file format. Fix location of '-k' option of gcov command.
15370 2018-01-04 Martin Liska <mliska@suse.cz>
15373 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15375 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15377 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15379 2018-01-03 Martin Sebor <msebor@redhat.com>
15381 PR tree-optimization/83655
15382 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15383 checking calls with invalid arguments.
15385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15387 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15388 (vectorizable_mask_load_store): Delete.
15389 (vectorizable_call): Return false for masked loads and stores.
15390 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15391 instead of gimple_assign_rhs1.
15392 (vectorizable_load): Handle IFN_MASK_LOAD.
15393 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15397 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15399 (vectorizable_mask_load_store): ...here.
15400 (vectorizable_load): ...and here.
15402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15404 * tree-vect-stmts.c (vect_build_all_ones_mask)
15405 (vect_build_zero_merge_argument): New functions, split out from...
15406 (vectorizable_load): ...here.
15408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15410 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15412 (vectorizable_mask_load_store): ...here.
15413 (vectorizable_store): ...and here.
15415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15417 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15419 (vectorizable_mask_load_store): ...here.
15421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15423 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15424 (vect_model_store_cost): Take a vec_load_store_type instead of a
15426 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15427 (vect_model_store_cost): Take a vec_load_store_type instead of a
15429 (vectorizable_mask_load_store): Update accordingly.
15430 (vectorizable_store): Likewise.
15431 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15435 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15436 IFN_MASK_LOAD calls here rather than...
15437 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15440 Alan Hayward <alan.hayward@arm.com>
15441 David Sherwood <david.sherwood@arm.com>
15443 * expmed.c (extract_bit_field_1): For vector extracts,
15444 fall back to extract_bit_field_as_subreg if vec_extract
15447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15448 Alan Hayward <alan.hayward@arm.com>
15449 David Sherwood <david.sherwood@arm.com>
15451 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15452 they are variable or constant sized.
15453 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15454 slots for constant-sized data.
15456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15457 Alan Hayward <alan.hayward@arm.com>
15458 David Sherwood <david.sherwood@arm.com>
15460 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15461 handling COND_EXPRs with boolean comparisons, try to find a better
15462 basis for the mask type than the boolean itself.
15464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15466 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15467 is calculated and how it can be overridden.
15468 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15469 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15471 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15475 Alan Hayward <alan.hayward@arm.com>
15476 David Sherwood <david.sherwood@arm.com>
15478 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15479 Remove the mode argument.
15480 (aarch64_simd_valid_immediate): Remove the mode and inverse
15482 * config/aarch64/iterators.md (bitsize): New iterator.
15483 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15484 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15485 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15486 aarch64_simd_valid_immediate.
15487 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15488 (aarch64_reg_or_bic_imm): Likewise.
15489 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15490 with an insn_type enum and msl with a modifier_type enum.
15491 Replace element_width with a scalar_mode. Change the shift
15492 to unsigned int. Add constructors for scalar_float_mode and
15493 scalar_int_mode elements.
15494 (aarch64_vect_float_const_representable_p): Delete.
15495 (aarch64_can_const_movi_rtx_p)
15496 (aarch64_simd_scalar_immediate_valid_for_move)
15497 (aarch64_simd_make_constant): Update call to
15498 aarch64_simd_valid_immediate.
15499 (aarch64_advsimd_valid_immediate_hs): New function.
15500 (aarch64_advsimd_valid_immediate): Likewise.
15501 (aarch64_simd_valid_immediate): Remove mode and inverse
15502 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15503 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15504 and aarch64_float_const_representable_p on the result.
15505 (aarch64_output_simd_mov_immediate): Remove mode argument.
15506 Update call to aarch64_simd_valid_immediate and use of
15507 simd_immediate_info.
15508 (aarch64_output_scalar_simd_mov_immediate): Update call
15511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15512 Alan Hayward <alan.hayward@arm.com>
15513 David Sherwood <david.sherwood@arm.com>
15515 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15516 (mode_nunits): Likewise CONST_MODE_NUNITS.
15517 * machmode.def (ADJUST_NUNITS): Document.
15518 * genmodes.c (mode_data::need_nunits_adj): New field.
15519 (blank_mode): Update accordingly.
15520 (adj_nunits): New variable.
15521 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15523 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15524 listed in adj_nunits.
15525 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15526 listed in adj_nunits. Don't emit case statements for such modes.
15527 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15528 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15529 nothing if adj_nunits is nonnull.
15530 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15531 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15532 (emit_mode_fbit): Update use of print_maybe_const_decl.
15533 (emit_move_size): Likewise. Treat the array as non-const
15535 (emit_mode_adjustments): Handle adj_nunits.
15537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15539 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15540 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15541 (VECTOR_MODES): Use it.
15542 (make_vector_modes): Take the prefix as an argument.
15544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15545 Alan Hayward <alan.hayward@arm.com>
15546 David Sherwood <david.sherwood@arm.com>
15548 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15549 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15550 for MODE_VECTOR_BOOL.
15551 * machmode.def (VECTOR_BOOL_MODE): Document.
15552 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15553 (make_vector_bool_mode): New function.
15554 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15556 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15557 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15559 * stor-layout.c (int_mode_for_mode): Likewise.
15560 * tree.c (build_vector_type_for_mode): Likewise.
15561 * varasm.c (output_constant_pool_2): Likewise.
15562 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15563 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15564 for MODE_VECTOR_BOOL.
15565 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15566 of mode class checks.
15567 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15568 instead of a list of mode class checks.
15569 (expand_vector_scalar_condition): Likewise.
15570 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15573 Alan Hayward <alan.hayward@arm.com>
15574 David Sherwood <david.sherwood@arm.com>
15576 * machmode.h (mode_size): Change from unsigned short to
15578 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15579 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15580 or if measurement_type is not polynomial.
15581 (fixed_size_mode::includes_p): Check for constant-sized modes.
15582 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15583 return a poly_uint16 rather than an unsigned short.
15584 (emit_mode_size): Change the type of mode_size from unsigned short
15585 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15586 (emit_mode_adjustments): Cope with polynomial vector sizes.
15587 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15589 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15591 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15592 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15593 * caller-save.c (setup_save_areas): Likewise.
15594 (replace_reg_with_saved_mem): Likewise.
15595 * calls.c (emit_library_call_value_1): Likewise.
15596 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15597 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15598 (gen_lowpart_for_combine): Likewise.
15599 * convert.c (convert_to_integer_1): Likewise.
15600 * cse.c (equiv_constant, cse_insn): Likewise.
15601 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15602 (cselib_subst_to_values): Likewise.
15603 * dce.c (word_dce_process_block): Likewise.
15604 * df-problems.c (df_word_lr_mark_ref): Likewise.
15605 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15606 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15607 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15608 (rtl_for_decl_location): Likewise.
15609 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15610 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15611 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15612 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15613 (expand_expr_real_1): Likewise.
15614 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15615 (pad_below): Likewise.
15616 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15617 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15618 * ira.c (get_subreg_tracking_sizes): Likewise.
15619 * ira-build.c (ira_create_allocno_objects): Likewise.
15620 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15621 (ira_sort_regnos_for_alter_reg): Likewise.
15622 * ira-costs.c (record_operand_costs): Likewise.
15623 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15624 (resolve_simple_move): Likewise.
15625 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15626 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15627 (lra_constraints): Likewise.
15628 (CONST_POOL_OK_P): Reject variable-sized modes.
15629 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15630 (add_pseudo_to_slot, lra_spill): Likewise.
15631 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15632 * optabs-query.c (get_best_extraction_insn): Likewise.
15633 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15634 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15635 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15636 * recog.c (offsettable_address_addr_space_p): Likewise.
15637 * regcprop.c (maybe_mode_change): Likewise.
15638 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15639 * regrename.c (build_def_use): Likewise.
15640 * regstat.c (dump_reg_info): Likewise.
15641 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15642 (find_reloads, find_reloads_subreg_address): Likewise.
15643 * reload1.c (eliminate_regs_1): Likewise.
15644 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15645 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15646 (simplify_binary_operation_1, simplify_subreg): Likewise.
15647 * targhooks.c (default_function_arg_padding): Likewise.
15648 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15649 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15650 (verify_gimple_assign_ternary): Likewise.
15651 * tree-inline.c (estimate_move_cost): Likewise.
15652 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15653 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15654 (get_address_cost_ainc): Likewise.
15655 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15656 (vect_supportable_dr_alignment): Likewise.
15657 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15658 (vectorizable_reduction): Likewise.
15659 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15660 (vectorizable_operation, vectorizable_load): Likewise.
15661 * tree.c (build_same_sized_truth_vector_type): Likewise.
15662 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15663 * var-tracking.c (emit_note_insn_var_location): Likewise.
15664 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15665 (ADDR_VEC_ALIGN): Likewise.
15667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15668 Alan Hayward <alan.hayward@arm.com>
15669 David Sherwood <david.sherwood@arm.com>
15671 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15673 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15674 or if measurement_type is polynomial.
15675 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15676 * combine.c (make_extraction): Likewise.
15677 * dse.c (find_shift_sequence): Likewise.
15678 * dwarf2out.c (mem_loc_descriptor): Likewise.
15679 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15680 (extract_bit_field, extract_low_bits): Likewise.
15681 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15682 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15683 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15684 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15685 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15686 * reload.c (find_reloads): Likewise.
15687 * reload1.c (alter_reg): Likewise.
15688 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15689 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15690 * tree-if-conv.c (predicate_mem_writes): Likewise.
15691 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15692 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15693 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15694 * valtrack.c (dead_debug_insert_temp): Likewise.
15695 * varasm.c (mergeable_constant_section): Likewise.
15696 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15699 Alan Hayward <alan.hayward@arm.com>
15700 David Sherwood <david.sherwood@arm.com>
15702 * expr.c (expand_assignment): Cope with polynomial mode sizes
15703 when assigning to a CONCAT.
15705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15706 Alan Hayward <alan.hayward@arm.com>
15707 David Sherwood <david.sherwood@arm.com>
15709 * machmode.h (mode_precision): Change from unsigned short to
15711 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15713 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15714 or if measurement_type is not polynomial.
15715 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15716 in which the mode is already known to be a scalar_int_mode.
15717 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15718 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15720 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15721 for GET_MODE_PRECISION.
15722 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15723 for GET_MODE_PRECISION.
15724 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15726 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15727 (expand_field_assignment, make_extraction): Likewise.
15728 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15729 (get_last_value): Likewise.
15730 * convert.c (convert_to_integer_1): Likewise.
15731 * cse.c (cse_insn): Likewise.
15732 * expr.c (expand_expr_real_1): Likewise.
15733 * lra-constraints.c (simplify_operand_subreg): Likewise.
15734 * optabs-query.c (can_atomic_load_p): Likewise.
15735 * optabs.c (expand_atomic_load): Likewise.
15736 (expand_atomic_store): Likewise.
15737 * ree.c (combine_reaching_defs): Likewise.
15738 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15739 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15740 * tree.h (type_has_mode_precision_p): Likewise.
15741 * ubsan.c (instrument_si_overflow): Likewise.
15743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15744 Alan Hayward <alan.hayward@arm.com>
15745 David Sherwood <david.sherwood@arm.com>
15747 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15748 polynomial numbers of units.
15749 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15750 (valid_vector_subparts_p): New function.
15751 (build_vector_type): Remove temporary shim and take the number
15752 of units as a poly_uint64 rather than an int.
15753 (build_opaque_vector_type): Take the number of units as a
15754 poly_uint64 rather than an int.
15755 * tree.c (build_vector_from_ctor): Handle polynomial
15756 TYPE_VECTOR_SUBPARTS.
15757 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15758 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15759 (build_vector_from_val): If the number of units is variable,
15760 use build_vec_duplicate_cst for constant operands and
15761 VEC_DUPLICATE_EXPR otherwise.
15762 (make_vector_type): Remove temporary is_constant ().
15763 (build_vector_type, build_opaque_vector_type): Take the number of
15764 units as a poly_uint64 rather than an int.
15765 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15767 * cfgexpand.c (expand_debug_expr): Likewise.
15768 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15769 (store_constructor, expand_expr_real_1): Likewise.
15770 (const_scalar_mask_from_tree): Likewise.
15771 * fold-const-call.c (fold_const_reduction): Likewise.
15772 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15773 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15774 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15775 (fold_relational_const): Likewise.
15776 (native_interpret_vector): Likewise. Change the size from an
15777 int to an unsigned int.
15778 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15779 TYPE_VECTOR_SUBPARTS.
15780 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15781 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15782 duplicating a non-constant operand into a variable-length vector.
15783 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15784 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15785 * ipa-icf.c (sem_variable::equals): Likewise.
15786 * match.pd: Likewise.
15787 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15788 * print-tree.c (print_node): Likewise.
15789 * stor-layout.c (layout_type): Likewise.
15790 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15791 * tree-cfg.c (verify_gimple_comparison): Likewise.
15792 (verify_gimple_assign_binary): Likewise.
15793 (verify_gimple_assign_ternary): Likewise.
15794 (verify_gimple_assign_single): Likewise.
15795 * tree-pretty-print.c (dump_generic_node): Likewise.
15796 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15797 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15798 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15799 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15800 (vect_shift_permute_load_chain): Likewise.
15801 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15802 (expand_vector_condition, optimize_vector_constructor): Likewise.
15803 (lower_vec_perm, get_compute_type): Likewise.
15804 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15805 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15806 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15807 (vect_recog_mask_conversion_pattern): Likewise.
15808 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15809 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15810 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15811 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15812 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15813 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15814 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15815 (supportable_widening_operation): Likewise.
15816 (supportable_narrowing_operation): Likewise.
15817 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15819 * varasm.c (output_constant): Likewise.
15821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15822 Alan Hayward <alan.hayward@arm.com>
15823 David Sherwood <david.sherwood@arm.com>
15825 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15826 so that both the length == 3 and length != 3 cases set up their
15827 own permute vectors. Add comments explaining why we know the
15828 number of elements is constant.
15829 (vect_permute_load_chain): Likewise.
15831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15832 Alan Hayward <alan.hayward@arm.com>
15833 David Sherwood <david.sherwood@arm.com>
15835 * machmode.h (mode_nunits): Change from unsigned char to
15837 (ONLY_FIXED_SIZE_MODES): New macro.
15838 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15839 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15840 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15842 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15843 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15844 or if measurement_type is not polynomial.
15845 * genmodes.c (ZERO_COEFFS): New macro.
15846 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15848 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15849 Use ZERO_COEFFS when emitting initializers.
15850 * data-streamer.h (bp_pack_poly_value): New function.
15851 (bp_unpack_poly_value): Likewise.
15852 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15853 for GET_MODE_NUNITS.
15854 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15855 for GET_MODE_NUNITS.
15856 * tree.c (make_vector_type): Remove temporary shim and make
15857 the real function take the number of units as a poly_uint64
15858 rather than an int.
15859 (build_vector_type_for_mode): Handle polynomial nunits.
15860 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15861 * emit-rtl.c (const_vec_series_p_1): Likewise.
15862 (gen_rtx_CONST_VECTOR): Likewise.
15863 * fold-const.c (test_vec_duplicate_folding): Likewise.
15864 * genrecog.c (validate_pattern): Likewise.
15865 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15866 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15867 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15868 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15869 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15870 * rtlanal.c (subreg_get_info): Likewise.
15871 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15872 (vect_grouped_load_supported): Likewise.
15873 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15874 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15875 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15876 (simplify_const_unary_operation, simplify_binary_operation_1)
15877 (simplify_const_binary_operation, simplify_ternary_operation)
15878 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15879 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15880 instead of CONST_VECTOR_NUNITS.
15881 * varasm.c (output_constant_pool_2): Likewise.
15882 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15883 explicit-encoded elements in the XVEC for variable-length vectors.
15885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15887 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15890 Alan Hayward <alan.hayward@arm.com>
15891 David Sherwood <david.sherwood@arm.com>
15893 * coretypes.h (fixed_size_mode): Declare.
15894 (fixed_size_mode_pod): New typedef.
15895 * builtins.h (target_builtins::x_apply_args_mode)
15896 (target_builtins::x_apply_result_mode): Change type to
15897 fixed_size_mode_pod.
15898 * builtins.c (apply_args_size, apply_result_size, result_vector)
15899 (expand_builtin_apply_args_1, expand_builtin_apply)
15900 (expand_builtin_return): Update accordingly.
15902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15904 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15905 * cselib.c (cselib_hash_rtx): Likewise.
15906 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15907 CONST_VECTOR encoding.
15909 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15910 Jeff Law <law@redhat.com>
15913 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15914 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15915 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15916 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15919 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15920 explicitly probe *sp in a noreturn function if there were any callee
15921 register saves or frame pointer is needed.
15923 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15926 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15927 BLKmode for ternary, binary or unary expressions.
15930 * var-tracking.c (delete_vta_debug_insn): New inline function.
15931 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15932 insns from get_insns () to NULL instead of each bb separately.
15933 Use delete_vta_debug_insn. No longer static.
15934 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15935 delete_vta_debug_insns callers.
15936 * rtl.h (delete_vta_debug_insns): Declare.
15937 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15938 instead of variable_tracking_main.
15940 2018-01-03 Martin Sebor <msebor@redhat.com>
15942 PR tree-optimization/83603
15943 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15944 arguments past the endof the argument list in functions declared
15945 without a prototype.
15946 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15947 Avoid checking when arguments are null.
15949 2018-01-03 Martin Sebor <msebor@redhat.com>
15952 * doc/extend.texi (attribute const): Fix a typo.
15953 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15954 issuing -Wsuggest-attribute for void functions.
15956 2018-01-03 Martin Sebor <msebor@redhat.com>
15958 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15959 offset_int::from instead of wide_int::to_shwi.
15960 (maybe_diag_overlap): Remove assertion.
15961 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15962 * gimple-ssa-sprintf.c (format_directive): Same.
15963 (parse_directive): Same.
15964 (sprintf_dom_walker::compute_format_length): Same.
15965 (try_substitute_return_value): Same.
15967 2018-01-03 Jeff Law <law@redhat.com>
15969 PR middle-end/83654
15970 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15971 non-constant residual for zero at runtime and avoid probing in
15972 that case. Reorganize code for trailing problem to mirror handling
15975 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15977 PR tree-optimization/83501
15978 * tree-ssa-strlen.c (get_string_cst): New.
15979 (handle_char_store): Call get_string_cst.
15981 2018-01-03 Martin Liska <mliska@suse.cz>
15983 PR tree-optimization/83593
15984 * tree-ssa-strlen.c: Include tree-cfg.h.
15985 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15986 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15987 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15989 (strlen_dom_walker::before_dom_children): Call
15990 gimple_purge_dead_eh_edges. Dump tranformation with details
15992 (strlen_dom_walker::before_dom_children): Update call by adding
15993 new argument cleanup_eh.
15994 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15996 2018-01-03 Martin Liska <mliska@suse.cz>
15999 * cif-code.def (VARIADIC_THUNK): New enum value.
16000 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16003 2018-01-03 Jan Beulich <jbeulich@suse.com>
16005 * sse.md (mov<mode>_internal): Tighten condition for when to use
16006 vmovdqu<ssescalarsize> for TI and OI modes.
16008 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16010 Update copyright years.
16012 2018-01-03 Martin Liska <mliska@suse.cz>
16015 * ipa-visibility.c (function_and_variable_visibility): Skip
16016 functions with noipa attribure.
16018 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16020 * gcc.c (process_command): Update copyright notice dates.
16021 * gcov-dump.c (print_version): Ditto.
16022 * gcov.c (print_version): Ditto.
16023 * gcov-tool.c (print_version): Ditto.
16024 * gengtype.c (create_file): Ditto.
16025 * doc/cpp.texi: Bump @copying's copyright year.
16026 * doc/cppinternals.texi: Ditto.
16027 * doc/gcc.texi: Ditto.
16028 * doc/gccint.texi: Ditto.
16029 * doc/gcov.texi: Ditto.
16030 * doc/install.texi: Ditto.
16031 * doc/invoke.texi: Ditto.
16033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16035 * vector-builder.h (vector_builder::m_full_nelts): Change from
16036 unsigned int to poly_uint64.
16037 (vector_builder::full_nelts): Update prototype accordingly.
16038 (vector_builder::new_vector): Likewise.
16039 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16040 (vector_builder::operator ==): Likewise.
16041 (vector_builder::finalize): Likewise.
16042 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16043 Take the number of elements as a poly_uint64 rather than an
16045 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16046 from unsigned int to poly_uint64.
16047 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16048 (vec_perm_indices::new_vector): Likewise.
16049 (vec_perm_indices::length): Likewise.
16050 (vec_perm_indices::nelts_per_input): Likewise.
16051 (vec_perm_indices::input_nelts): Likewise.
16052 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16053 number of elements per input as a poly_uint64 rather than an
16054 unsigned int. Use the original encoding for variable-length
16055 vectors, rather than clamping each individual element.
16056 For the second and subsequent elements in each pattern,
16057 clamp the step and base before clamping their sum.
16058 (vec_perm_indices::series_p): Handle polynomial element counts.
16059 (vec_perm_indices::all_in_range_p): Likewise.
16060 (vec_perm_indices_to_tree): Likewise.
16061 (vec_perm_indices_to_rtx): Likewise.
16062 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16063 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16064 (tree_vector_builder::new_binary_operation): Handle polynomial
16065 element counts. Return false if we need to know the number
16066 of elements at compile time.
16067 * fold-const.c (fold_vec_perm): Punt if the number of elements
16068 isn't known at compile time.
16070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16072 * vec-perm-indices.h (vec_perm_builder): Change element type
16073 from HOST_WIDE_INT to poly_int64.
16074 (vec_perm_indices::element_type): Update accordingly.
16075 (vec_perm_indices::clamp): Handle polynomial element_types.
16076 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16077 (vec_perm_indices::all_in_range_p): Likewise.
16078 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16080 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16081 polynomial vec_perm_indices element types.
16082 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16083 * fold-const.c (fold_vec_perm): Likewise.
16084 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16085 * tree-vect-generic.c (lower_vec_perm): Likewise.
16086 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16087 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16088 element type to HOST_WIDE_INT.
16090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16091 Alan Hayward <alan.hayward@arm.com>
16092 David Sherwood <david.sherwood@arm.com>
16094 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16095 rather than an int. Use plus_constant.
16096 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16097 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16100 Alan Hayward <alan.hayward@arm.com>
16101 David Sherwood <david.sherwood@arm.com>
16103 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16104 a HOST_WIDE_INT to a poly_int64.
16106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16107 Alan Hayward <alan.hayward@arm.com>
16108 David Sherwood <david.sherwood@arm.com>
16110 * calls.c (load_register_parameters): Cope with polynomial
16111 mode sizes. Require a constant size for BLKmode parameters
16112 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16113 forces a parameter to be padded at the lsb end in order to
16114 fill a complete number of words, require the parameter size
16115 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
16122 from unsigned int to poly_uint64_pod.
16123 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
16130 (reload_inner_reg_of_subreg, push_reload): Use it.
16132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16133 Alan Hayward <alan.hayward@arm.com>
16134 David Sherwood <david.sherwood@arm.com>
16136 * lra-constraints.c (process_alt_operands): Reject matched
16137 operands whose sizes aren't ordered.
16138 (match_reload): Refer to this check here.
16140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16141 Alan Hayward <alan.hayward@arm.com>
16142 David Sherwood <david.sherwood@arm.com>
16144 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16145 that the mode size is in the set {1, 2, 4, 8, 16}.
16147 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16148 Alan Hayward <alan.hayward@arm.com>
16149 David Sherwood <david.sherwood@arm.com>
16151 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16152 Use plus_constant instead of gen_rtx_PLUS.
16154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16155 Alan Hayward <alan.hayward@arm.com>
16156 David Sherwood <david.sherwood@arm.com>
16158 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16159 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16160 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16161 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16162 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16163 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16164 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16165 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16166 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16167 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16169 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16170 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16171 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16172 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16173 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16174 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16175 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16176 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16177 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16178 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16180 * expr.c (emit_move_resolve_push): Treat the input and result
16181 of PUSH_ROUNDING as a poly_int64.
16182 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16183 (emit_push_insn): Likewise.
16184 * lra-eliminations.c (mark_not_eliminable): Likewise.
16185 * recog.c (push_operand): Likewise.
16186 * reload1.c (elimination_effects): Likewise.
16187 * rtlanal.c (nonzero_bits1): Likewise.
16188 * calls.c (store_one_arg): Likewise. Require the padding to be
16189 known at compile time.
16191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16192 Alan Hayward <alan.hayward@arm.com>
16193 David Sherwood <david.sherwood@arm.com>
16195 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16196 Use plus_constant instead of gen_rtx_PLUS.
16198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16199 Alan Hayward <alan.hayward@arm.com>
16200 David Sherwood <david.sherwood@arm.com>
16202 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16203 rather than an int.
16205 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16206 Alan Hayward <alan.hayward@arm.com>
16207 David Sherwood <david.sherwood@arm.com>
16209 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16210 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16211 via stack temporaries. Treat the mode size as polynomial too.
16213 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16214 Alan Hayward <alan.hayward@arm.com>
16215 David Sherwood <david.sherwood@arm.com>
16217 * expr.c (expand_expr_real_2): When handling conversions involving
16218 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16219 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16220 as a poly_uint64 too.
16222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16223 Alan Hayward <alan.hayward@arm.com>
16224 David Sherwood <david.sherwood@arm.com>
16226 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16229 Alan Hayward <alan.hayward@arm.com>
16230 David Sherwood <david.sherwood@arm.com>
16232 * combine.c (can_change_dest_mode): Handle polynomial
16233 REGMODE_NATURAL_SIZE.
16234 * expmed.c (store_bit_field_1): Likewise.
16235 * expr.c (store_constructor): Likewise.
16236 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16237 and polynomial REGMODE_NATURAL_SIZE.
16238 (gen_lowpart_common): Likewise.
16239 * reginfo.c (record_subregs_of_mode): Likewise.
16240 * rtlanal.c (read_modify_subreg_p): Likewise.
16242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16243 Alan Hayward <alan.hayward@arm.com>
16244 David Sherwood <david.sherwood@arm.com>
16246 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16247 numbers of elements.
16249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16250 Alan Hayward <alan.hayward@arm.com>
16251 David Sherwood <david.sherwood@arm.com>
16253 * match.pd: Cope with polynomial numbers of vector elements.
16255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16256 Alan Hayward <alan.hayward@arm.com>
16257 David Sherwood <david.sherwood@arm.com>
16259 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16260 in a POINTER_PLUS_EXPR.
16262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16263 Alan Hayward <alan.hayward@arm.com>
16264 David Sherwood <david.sherwood@arm.com>
16266 * omp-simd-clone.c (simd_clone_subparts): New function.
16267 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16268 (ipa_simd_modify_function_body): Likewise.
16270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16271 Alan Hayward <alan.hayward@arm.com>
16272 David Sherwood <david.sherwood@arm.com>
16274 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16275 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16276 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16277 (expand_vector_condition, vector_element): Likewise.
16278 (subparts_gt): New function.
16279 (get_compute_type): Use subparts_gt.
16280 (count_type_subparts): Delete.
16281 (expand_vector_operations_1): Use subparts_gt instead of
16282 count_type_subparts.
16284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16285 Alan Hayward <alan.hayward@arm.com>
16286 David Sherwood <david.sherwood@arm.com>
16288 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16289 (vect_compile_time_alias): ...this new function. Do the calculation
16290 on poly_ints rather than trees.
16291 (vect_prune_runtime_alias_test_list): Update call accordingly.
16293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16294 Alan Hayward <alan.hayward@arm.com>
16295 David Sherwood <david.sherwood@arm.com>
16297 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16299 (vect_schedule_slp_instance): Likewise.
16301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16302 Alan Hayward <alan.hayward@arm.com>
16303 David Sherwood <david.sherwood@arm.com>
16305 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16306 constant and extern definitions for variable-length vectors.
16307 (vect_get_constant_vectors): Note that the number of units
16308 is known to be constant.
16310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16311 Alan Hayward <alan.hayward@arm.com>
16312 David Sherwood <david.sherwood@arm.com>
16314 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16315 of units as polynomial. Choose between WIDE and NARROW based
16318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16319 Alan Hayward <alan.hayward@arm.com>
16320 David Sherwood <david.sherwood@arm.com>
16322 * tree-vect-stmts.c (simd_clone_subparts): New function.
16323 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16326 Alan Hayward <alan.hayward@arm.com>
16327 David Sherwood <david.sherwood@arm.com>
16329 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16330 vectors as polynomial. Use build_index_vector for
16331 IFN_GOMP_SIMD_LANE.
16333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16334 Alan Hayward <alan.hayward@arm.com>
16335 David Sherwood <david.sherwood@arm.com>
16337 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16338 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16339 for variable-length vectors.
16340 (vectorizable_mask_load_store): Treat the number of units as
16341 polynomial, asserting that it is constant if the condition has
16342 already been enforced.
16343 (vectorizable_store, vectorizable_load): Likewise.
16345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16346 Alan Hayward <alan.hayward@arm.com>
16347 David Sherwood <david.sherwood@arm.com>
16349 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16350 of units as polynomial. Punt if we can't tell at compile time
16351 which vector contains the final result.
16353 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16354 Alan Hayward <alan.hayward@arm.com>
16355 David Sherwood <david.sherwood@arm.com>
16357 * tree-vect-loop.c (vectorizable_induction): Treat the number
16358 of units as polynomial. Punt on SLP inductions. Use an integer
16359 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16360 cast of such a series for variable-length floating-point
16363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16364 Alan Hayward <alan.hayward@arm.com>
16365 David Sherwood <david.sherwood@arm.com>
16367 * tree.h (build_index_vector): Declare.
16368 * tree.c (build_index_vector): New function.
16369 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16370 of units as polynomial, forcibly converting it to a constant if
16371 vectorizable_reduction has already enforced the condition.
16372 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16373 to create a {1,2,3,...} vector.
16374 (vectorizable_reduction): Treat the number of units as polynomial.
16375 Choose vectype_in based on the largest scalar element size rather
16376 than the smallest number of units. Enforce the restrictions
16379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16380 Alan Hayward <alan.hayward@arm.com>
16381 David Sherwood <david.sherwood@arm.com>
16383 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16384 number of units as polynomial.
16386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16387 Alan Hayward <alan.hayward@arm.com>
16388 David Sherwood <david.sherwood@arm.com>
16390 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16391 * target.def (autovectorize_vector_sizes): Return the vector sizes
16392 by pointer, using vector_sizes rather than a bitmask.
16393 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16394 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16395 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16397 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16398 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16399 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16400 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16401 * omp-general.c (omp_max_vf): Likewise.
16402 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16403 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16404 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16405 * tree-vect-slp.c (vect_slp_bb): Likewise.
16406 * doc/tm.texi: Regenerate.
16407 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16409 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16410 the vector size as a poly_uint64 rather than an unsigned int.
16411 (current_vector_size): Change from an unsigned int to a poly_uint64.
16412 (get_vectype_for_scalar_type): Update accordingly.
16413 * tree.h (build_truth_vector_type): Take the size and number of
16414 units as a poly_uint64 rather than an unsigned int.
16415 (build_vector_type): Add a temporary overload that takes
16416 the number of units as a poly_uint64 rather than an unsigned int.
16417 * tree.c (make_vector_type): Likewise.
16418 (build_truth_vector_type): Take the number of units as a poly_uint64
16419 rather than an unsigned int.
16421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16422 Alan Hayward <alan.hayward@arm.com>
16423 David Sherwood <david.sherwood@arm.com>
16425 * target.def (get_mask_mode): Take the number of units and length
16426 as poly_uint64s rather than unsigned ints.
16427 * targhooks.h (default_get_mask_mode): Update accordingly.
16428 * targhooks.c (default_get_mask_mode): Likewise.
16429 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16430 * doc/tm.texi: Regenerate.
16432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16433 Alan Hayward <alan.hayward@arm.com>
16434 David Sherwood <david.sherwood@arm.com>
16436 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16437 * omp-general.c (omp_max_vf): Likewise.
16438 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16439 (expand_omp_simd): Handle polynomial safelen.
16440 * omp-low.c (omplow_simd_context): Add a default constructor.
16441 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16442 (lower_rec_simd_input_clauses): Update accordingly.
16443 (lower_rec_input_clauses): Likewise.
16445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16446 Alan Hayward <alan.hayward@arm.com>
16447 David Sherwood <david.sherwood@arm.com>
16449 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16450 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16451 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16452 (vect_analyze_slp_cost): Likewise.
16453 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16454 (vect_model_load_cost): Likewise.
16456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16457 Alan Hayward <alan.hayward@arm.com>
16458 David Sherwood <david.sherwood@arm.com>
16460 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16461 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16462 from an unsigned int * to a poly_uint64_pod *.
16463 (calculate_unrolling_factor): New function.
16464 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16467 Alan Hayward <alan.hayward@arm.com>
16468 David Sherwood <david.sherwood@arm.com>
16470 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16471 from an unsigned int to a poly_uint64.
16472 (_loop_vec_info::slp_unrolling_factor): Likewise.
16473 (_loop_vec_info::vectorization_factor): Change from an int
16475 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16476 (vect_get_num_vectors): New function.
16477 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16478 (vect_get_num_copies): Use vect_get_num_vectors.
16479 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16480 to an unsigned int *.
16481 (vect_analyze_data_refs): Change min_vf from an int * to a
16483 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16484 than an unsigned HOST_WIDE_INT.
16485 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16486 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16487 to an unsigned int *.
16488 (vect_analyze_data_ref_dependences): Likewise.
16489 (vect_compute_data_ref_alignment): Handle polynomial vf.
16490 (vect_enhance_data_refs_alignment): Likewise.
16491 (vect_prune_runtime_alias_test_list): Likewise.
16492 (vect_shift_permute_load_chain): Likewise.
16493 (vect_supportable_dr_alignment): Likewise.
16494 (dependence_distance_ge_vf): Take the vectorization factor as a
16495 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16496 (vect_analyze_data_refs): Change min_vf from an int * to a
16498 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16499 vfm1 as a poly_uint64 rather than an int. Make the same change
16500 for the returned bound_scalar.
16501 (vect_gen_vector_loop_niters): Handle polynomial vf.
16502 (vect_do_peeling): Likewise. Update call to
16503 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16504 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16506 * tree-vect-loop.c (vect_determine_vectorization_factor)
16507 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16508 (vect_get_known_peeling_cost): Likewise.
16509 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16510 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16511 (vect_transform_loop): Likewise. Use the lowest possible VF when
16512 updating the upper bounds of the loop.
16513 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16514 rather than an int.
16515 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16516 polynomial unroll factors.
16517 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16518 (vect_make_slp_decision): Likewise.
16519 (vect_supported_load_permutation_p): Likewise, and polynomial
16521 (vect_analyze_slp_cost): Handle polynomial vf.
16522 (vect_slp_analyze_node_operations): Likewise.
16523 (vect_slp_analyze_bb_1): Likewise.
16524 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16525 than an unsigned HOST_WIDE_INT.
16526 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16527 (vectorizable_load): Handle polynomial vf.
16528 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16530 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16533 Alan Hayward <alan.hayward@arm.com>
16534 David Sherwood <david.sherwood@arm.com>
16536 * match.pd: Handle bit operations involving three constants
16537 and try to fold one pair.
16539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16541 * tree-vect-loop-manip.c: Include gimple-fold.h.
16542 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16543 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16544 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16545 Add a path that uses a step of VF instead of 1, but disable it
16547 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16548 and niters_no_overflow parameters. Update calls to
16549 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16550 Create a new SSA name if the latter choses to use a ste other
16551 than zero, and return it via niters_vector_mult_vf_var.
16552 * tree-vect-loop.c (vect_transform_loop): Update calls to
16553 vect_do_peeling, vect_gen_vector_loop_niters and
16554 slpeel_make_loop_iterate_ntimes.
16555 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16556 (vect_gen_vector_loop_niters): Update declarations after above changes.
16558 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16560 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16561 128-bit round to integer instructions.
16562 (ceil<mode>2): Likewise.
16563 (btrunc<mode>2): Likewise.
16564 (round<mode>2): Likewise.
16566 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16568 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16569 unaligned VSX load/store on P8/P9.
16570 (expand_block_clear): Allow the use of unaligned VSX
16571 load/store on P8/P9.
16573 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16575 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16577 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16578 swap associated with both a load and a store.
16580 2018-01-02 Andrew Waterman <andrew@sifive.com>
16582 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16583 * config/riscv/riscv.md (clear_cache): Use it.
16585 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16587 * web.c: Remove out-of-date comment.
16589 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16591 * expr.c (fixup_args_size_notes): Check that any existing
16592 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16593 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16594 (emit_single_push_insn): ...here.
16596 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16598 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16599 (const_vector_encoded_nelts): New function.
16600 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16601 (const_vector_int_elt, const_vector_elt): Declare.
16602 * emit-rtl.c (const_vector_int_elt_1): New function.
16603 (const_vector_elt): Likewise.
16604 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16605 of CONST_VECTOR_ELT.
16607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16609 * expr.c: Include rtx-vector-builder.h.
16610 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16611 directly on the tree encoding.
16612 (const_vector_from_tree): Likewise.
16613 * optabs.c: Include rtx-vector-builder.h.
16614 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16615 sequence of "u" values.
16616 * vec-perm-indices.c: Include rtx-vector-builder.h.
16617 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16618 directly on the vec_perm_indices encoding.
16620 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16622 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16623 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16624 * rtx-vector-builder.h: New file.
16625 * rtx-vector-builder.c: Likewise.
16626 * rtl.h (rtx_def::u2): Add a const_vector field.
16627 (CONST_VECTOR_NPATTERNS): New macro.
16628 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16629 (CONST_VECTOR_DUPLICATE_P): Likewise.
16630 (CONST_VECTOR_STEPPED_P): Likewise.
16631 (CONST_VECTOR_ENCODED_ELT): Likewise.
16632 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16633 (unwrap_const_vec_duplicate): Likewise.
16634 (const_vec_series_p): Check for a non-duplicated vector encoding.
16635 Say that the function only returns true for integer vectors.
16636 * emit-rtl.c: Include rtx-vector-builder.h.
16637 (gen_const_vec_duplicate_1): Delete.
16638 (gen_const_vector): Call gen_const_vec_duplicate instead of
16639 gen_const_vec_duplicate_1.
16640 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16641 (gen_const_vec_duplicate): Use rtx_vector_builder.
16642 (gen_const_vec_series): Likewise.
16643 (gen_rtx_CONST_VECTOR): Likewise.
16644 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16645 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16646 Build a new vector rather than modifying a CONST_VECTOR in-place.
16647 (handle_special_swappables): Update call accordingly.
16648 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16649 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16650 Build a new vector rather than modifying a CONST_VECTOR in-place.
16651 (handle_special_swappables): Update call accordingly.
16653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16655 * simplify-rtx.c (simplify_const_binary_operation): Use
16656 CONST_VECTOR_ELT instead of XVECEXP.
16658 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16660 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16661 the selector elements to be different from the data elements
16662 if the selector is a VECTOR_CST.
16663 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16664 ssizetype for the selector.
16666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16668 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16669 before testing each element individually.
16670 * tree-vect-generic.c (lower_vec_perm): Likewise.
16672 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16674 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16675 * selftest-run-tests.c (selftest::run_tests): Call it.
16676 * vector-builder.h (vector_builder::operator ==): New function.
16677 (vector_builder::operator !=): Likewise.
16678 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16679 (vec_perm_indices::all_from_input_p): New function.
16680 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16681 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16682 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16683 instead of reading the VECTOR_CST directly. Detect whether both
16684 vector inputs are the same before constructing the vec_perm_indices,
16685 and update the number of inputs argument accordingly. Use the
16686 utility functions added above. Only construct sel2 if we need to.
16688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16690 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16691 the broadcast of the low byte.
16692 (expand_mult_highpart): Use an explicit encoding for the permutes.
16693 * optabs-query.c (can_mult_highpart_p): Likewise.
16694 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16695 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16696 (vectorizable_bswap): Likewise.
16697 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16698 explicit encoding for the power-of-2 permutes.
16699 (vect_permute_store_chain): Likewise.
16700 (vect_grouped_load_supported): Likewise.
16701 (vect_permute_load_chain): Likewise.
16703 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16705 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16706 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16707 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16708 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16709 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16710 (vect_gen_perm_mask_any): Likewise.
16712 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16714 * int-vector-builder.h: New file.
16715 * vec-perm-indices.h: Include int-vector-builder.h.
16716 (vec_perm_indices): Redefine as an int_vector_builder.
16717 (auto_vec_perm_indices): Delete.
16718 (vec_perm_builder): Redefine as a stand-alone class.
16719 (vec_perm_indices::vec_perm_indices): New function.
16720 (vec_perm_indices::clamp): Likewise.
16721 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16722 (vec_perm_indices::new_vector): New function.
16723 (vec_perm_indices::new_expanded_vector): Update for new
16724 vec_perm_indices class.
16725 (vec_perm_indices::rotate_inputs): New function.
16726 (vec_perm_indices::all_in_range_p): Operate directly on the
16727 encoded form, without computing elided elements.
16728 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16729 encoding. Update for new vec_perm_indices class.
16730 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16731 the given vec_perm_builder.
16732 (expand_vec_perm_var): Update vec_perm_builder constructor.
16733 (expand_mult_highpart): Use vec_perm_builder instead of
16734 auto_vec_perm_indices.
16735 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16736 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16737 or double series encoding as appropriate.
16738 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16739 vec_perm_indices instead of auto_vec_perm_indices.
16740 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16741 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16742 (vect_permute_store_chain): Likewise.
16743 (vect_grouped_load_supported): Likewise.
16744 (vect_permute_load_chain): Likewise.
16745 (vect_shift_permute_load_chain): Likewise.
16746 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16747 (vect_transform_slp_perm_load): Likewise.
16748 (vect_schedule_slp_instance): Likewise.
16749 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16750 (vectorizable_mask_load_store): Likewise.
16751 (vectorizable_bswap): Likewise.
16752 (vectorizable_store): Likewise.
16753 (vectorizable_load): Likewise.
16754 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16755 vec_perm_indices instead of auto_vec_perm_indices. Use
16756 tree_to_vec_perm_builder to read the vector from a tree.
16757 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16758 vec_perm_builder instead of a vec_perm_indices.
16759 (have_whole_vector_shift): Use vec_perm_builder and
16760 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16761 truncation to calc_vec_perm_mask_for_shift.
16762 (vect_create_epilog_for_reduction): Likewise.
16763 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16764 from auto_vec_perm_indices to vec_perm_indices.
16765 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16766 instead of changing individual elements.
16767 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16768 the vector in d.perm.
16769 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16770 from auto_vec_perm_indices to vec_perm_indices.
16771 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16772 instead of changing individual elements.
16773 (arm_vectorize_vec_perm_const): Use new_vector to install
16774 the vector in d.perm.
16775 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16776 Update vec_perm_builder constructor.
16777 (rs6000_expand_interleave): Likewise.
16778 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16779 (rs6000_expand_interleave): Likewise.
16781 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16783 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16784 to qimode could truncate the indices.
16785 * optabs.c (expand_vec_perm_var): Likewise.
16787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16789 * Makefile.in (OBJS): Add vec-perm-indices.o.
16790 * vec-perm-indices.h: New file.
16791 * vec-perm-indices.c: Likewise.
16792 * target.h (vec_perm_indices): Replace with a forward class
16794 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16795 * optabs.h: Include vec-perm-indices.h.
16796 (expand_vec_perm): Delete.
16797 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16798 (expand_vec_perm_const): Declare.
16799 * target.def (vec_perm_const_ok): Replace with...
16800 (vec_perm_const): ...this new hook.
16801 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16802 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16803 * doc/tm.texi: Regenerate.
16804 * optabs.def (vec_perm_const): Delete.
16805 * doc/md.texi (vec_perm_const): Likewise.
16806 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16807 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16808 expand_vec_perm for constant permutation vectors. Assert that
16809 the mode of variable permutation vectors is the integer equivalent
16810 of the mode that is being permuted.
16811 * optabs-query.h (selector_fits_mode_p): Declare.
16812 * optabs-query.c: Include vec-perm-indices.h.
16813 (selector_fits_mode_p): New function.
16814 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16815 is defined, instead of checking whether the vec_perm_const_optab
16816 exists. Use targetm.vectorize.vec_perm_const instead of
16817 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16818 fit in the vector mode before using a variable permute.
16819 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16820 vec_perm_indices instead of an rtx.
16821 (expand_vec_perm): Replace with...
16822 (expand_vec_perm_const): ...this new function. Take the selector
16823 as a vec_perm_indices rather than an rtx. Also take the mode of
16824 the selector. Update call to shift_amt_for_vec_perm_mask.
16825 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16826 Use vec_perm_indices::new_expanded_vector to expand the original
16827 selector into bytes. Check whether the indices fit in the vector
16828 mode before using a variable permute.
16829 (expand_vec_perm_var): Make global.
16830 (expand_mult_highpart): Use expand_vec_perm_const.
16831 * fold-const.c: Includes vec-perm-indices.h.
16832 * tree-ssa-forwprop.c: Likewise.
16833 * tree-vect-data-refs.c: Likewise.
16834 * tree-vect-generic.c: Likewise.
16835 * tree-vect-loop.c: Likewise.
16836 * tree-vect-slp.c: Likewise.
16837 * tree-vect-stmts.c: Likewise.
16838 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16840 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16841 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16842 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16843 (aarch64_vectorize_vec_perm_const): ...this new function.
16844 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16845 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16846 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16847 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16848 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16849 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16850 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16852 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16853 check for NEON modes.
16854 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16855 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16856 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16857 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16859 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16860 the old VEC_PERM_CONST conditions.
16861 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16862 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16863 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16864 (ia64_vectorize_vec_perm_const_ok): Merge into...
16865 (ia64_vectorize_vec_perm_const): ...this new function.
16866 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16867 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16868 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16869 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16870 * config/mips/mips.c (mips_expand_vec_perm_const)
16871 (mips_vectorize_vec_perm_const_ok): Merge into...
16872 (mips_vectorize_vec_perm_const): ...this new function.
16873 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16874 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16875 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16876 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16877 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16878 (rs6000_expand_vec_perm_const): Delete.
16879 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16881 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16882 (altivec_expand_vec_perm_const_le): Take each operand individually.
16883 Operate on constant selectors rather than rtxes.
16884 (altivec_expand_vec_perm_const): Likewise. Update call to
16885 altivec_expand_vec_perm_const_le.
16886 (rs6000_expand_vec_perm_const): Delete.
16887 (rs6000_vectorize_vec_perm_const_ok): Delete.
16888 (rs6000_vectorize_vec_perm_const): New function.
16889 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16890 an element count and rtx array.
16891 (rs6000_expand_extract_even): Update call accordingly.
16892 (rs6000_expand_interleave): Likewise.
16893 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16894 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16895 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16896 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16897 (rs6000_expand_vec_perm_const): Delete.
16898 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16899 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16900 (altivec_expand_vec_perm_const_le): Take each operand individually.
16901 Operate on constant selectors rather than rtxes.
16902 (altivec_expand_vec_perm_const): Likewise. Update call to
16903 altivec_expand_vec_perm_const_le.
16904 (rs6000_expand_vec_perm_const): Delete.
16905 (rs6000_vectorize_vec_perm_const_ok): Delete.
16906 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16907 reference to the SPE evmerge intructions.
16908 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16909 an element count and rtx array.
16910 (rs6000_expand_extract_even): Update call accordingly.
16911 (rs6000_expand_interleave): Likewise.
16912 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16913 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16915 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16917 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16919 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16920 vector mode and that that mode matches the mode of the data
16922 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16923 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16924 directly using expand_vec_perm_1 when forcing selectors into
16926 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16930 * optabs-query.h (can_vec_perm_p): Delete.
16931 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16932 * optabs-query.c (can_vec_perm_p): Split into...
16933 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16934 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16935 particular selector is valid.
16936 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16937 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16938 (vect_grouped_load_supported): Likewise.
16939 (vect_shift_permute_load_chain): Likewise.
16940 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16941 (vect_transform_slp_perm_load): Likewise.
16942 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16943 (vectorizable_bswap): Likewise.
16944 (vect_gen_perm_mask_checked): Likewise.
16945 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16946 implementations of variable permutation vectors into account
16947 when deciding which selector to use.
16948 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16949 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16950 with a false third argument.
16951 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16952 to test whether the constant selector is valid and can_vec_perm_var_p
16953 to test whether a variable selector is valid.
16955 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16957 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16958 * optabs-query.c (can_vec_perm_p): Likewise.
16959 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16960 instead of vec_perm_indices.
16961 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16962 (vect_gen_perm_mask_checked): Likewise,
16963 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16964 (vect_gen_perm_mask_checked): Likewise,
16966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16968 * optabs-query.h (qimode_for_vec_perm): Declare.
16969 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16970 (qimode_for_vec_perm): ...this new function.
16971 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16973 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16975 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16976 does not have a conditional at the top.
16978 2018-01-02 Richard Biener <rguenther@suse.de>
16980 * ipa-inline.c (big_speedup_p): Fix expression.
16982 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16985 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16988 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16992 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16993 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16994 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16995 cond_taken_branch_cost 3->4.
16997 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16999 PR tree-optimization/83581
17000 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17001 TODO_cleanup_cfg if any changes have been made.
17003 PR middle-end/83608
17004 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17005 convert_modes if target mode has the right side, but different mode
17008 PR middle-end/83609
17009 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17010 last argument when extracting from CONCAT. If either from_real or
17011 from_imag is NULL, use expansion through memory. If result is not
17012 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17013 the parts directly to inner mode, if even that fails, use expansion
17016 PR middle-end/83623
17017 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17018 check for bswap in mode rather than HImode and use that in expand_unop
17021 Copyright (C) 2018 Free Software Foundation, Inc.
17023 Copying and distribution of this file, with or without modification,
17024 are permitted in any medium without royalty provided the copyright
17025 notice and this notice are preserved.