* lib/target-supports.exp (check_weak_available): Return true for AIX.
[official-gcc.git] / gcc / ChangeLog
blob5b1adf0676d827e777cf1c82ee9517d69c36d7d3
1 2013-09-03  David Edelsohn  <dje.gcc@gmail.com>
3         * config/rs6000/rs6000.h (ASM_OUTPUT_DEF_FROM_DECLS): Emit lglobl for
4         function descriptor.
6 2013-09-03  Richard Biener  <rguenther@suse.de>
8         * tree-affine.c (add_elt_to_tree): Fix association issue,
9         avoid useless converts and make sure to always return a
10         properly typed result.
12 2013-09-03  Richard Biener  <rguenther@suse.de>
14         PR middle-end/57656
15         * fold-const.c (negate_expr_p): Fix division case.
16         (negate_expr): Likewise.
18 2013-09-03  Richard Biener  <rguenther@suse.de>
20         PR lto/58285
21         * tree-streamer-out.c: Include tm.h.
22         * Makefile.in (tree-streamer-out.o): Depend on $(TM_H).
24 2013-09-03  Jan Hubicka  <jh@suse.cz>
26         * tree-profile.c (tree_profiling): Cleanup CFG when done.
28 2013-09-03  Alan Modra  <amodra@gmail.com>
30         * config.gcc (powerpc*-*-linux*): Add support for little-endian
31         multilibs to big-endian target and vice versa.
32         * config/rs6000/t-linux64: Use := assignment on all vars.
33         (MULTILIB_EXTRA_OPTS): Remove fPIC.
34         (MULTILIB_OSDIRNAMES): Specify using mapping from multilib_options.
35         * config/rs6000/t-linux64le: New file.
36         * config/rs6000/t-linux64bele: New file.
37         * config/rs6000/t-linux64lebe: New file.
39 2013-09-02  Jan Hubicka  <jh@suse.cz>
41         * ipa-inline-transform.c (inline_transform): Do not
42         optimize_inline_calls when not optimizing.
44 2013-09-02  Jan Hubicka  <jh@suse.cz>
46         * lto-symtab.c (lto_symtab_merge_symbols): Add comments; merge
47         duplicated nodes for assembler names.
48         * symtab.c (symtab_unregister_node): Do not attempt to unlink
49         hard registers from assembler name hash.
51 2013-09-02  Jan Hubicka  <jh@suse.cz>
53         * ipa-split.c (execute_split_functions): Split externally visible
54         functions called once.
56 2013-09-02  Martin Jambor  <mjambor@suse.cz>
58         PR ipa/58106
59         * ipa-prop.c (ipa_edge_duplication_hook): Always put new rdesc to the
60         linked list.  When finding the correct duplicate, also consider also
61         the caller in additon to its inlined_to node.
63 2013-09-02  James Greenhalgh  <james.greenhalgh@arm.com>
65         * config/aarch64/aarch64-simd-builtins.def
66         (dup_lane_scalar): Remove.
67         * config/aarch64/aarch64-simd.md
68         (aarch64_simd_dup): Add 'w->w' alternative.
69         (aarch64_dup_lane<mode>): Allow for VALL.
70         (aarch64_dup_lane_scalar<mode>): Remove.
71         (aarch64_dup_lane_<vswap_width_name><mode>): New.
72         (aarch64_get_lane_signed<mode>): Add w->w altenative.
73         (aarch64_get_lane_unsigned<mode>): Likewise.
74         (aarch64_get_lane<mode>): Likewise.
75         * config/aarch64/aarch64.c (aarch64_evpc_dup): New.
76         (aarch64_expand_vec_perm_const_1): Use aarch64_evpc_dup.
77         * config/aarch64/iterators.md (VSWAP_WIDTH): New.
78         (VCON): Change container of V2SF.
79         (vswap_width_name): Likewise.
80         * config/aarch64/arm_neon.h
81         (__aarch64_vdup_lane_any): New.
82         (__aarch64_vdup<q>_lane<q>_<fpsu><8,16,32,64>): Likewise.
83         (vdup<q>_n_<psuf><8,16,32,64>): Convert to C implementation.
84         (vdup<q>_lane<q>_<fpsu><8,16,32,64>): Likewise.
86 2013-09-02  Eric Botcazou  <ebotcazou@adacore.com>
88         PR middle-end/56382
89         * expr.c (emit_move_complex): Do not move complex FP values as parts if
90         the source or the destination is a single hard register.
92 2013-09-02  Richard Biener  <rguenther@suse.de>
94         PR middle-end/57511
95         * tree-scalar-evolution.c (instantiate_scev_name): Allow
96         non-linear SCEVs.
98 2013-09-02  Richard Biener  <rguenther@suse.de>
100         * tree-affine.c (add_elt_to_tree): Avoid converting all pointer
101         arithmetic to sizetype.
103 2013-09-02  Bin Cheng  <bin.cheng@arm.com>
105         * tree-ssa-loop-ivopts.c (set_autoinc_for_original_candidates):
106         Find auto-increment use both before and after candidate.
108 2013-09-02  Marek Polacek  <polacek@redhat.com>
110         * Makefile.in (ubsan.o): Add $(TM_P_H) dependency.
112 2013-09-01  Jan Hubicka  <jh@suse.cz>
114         * Makefile.in: Add ipa-profile.o
115         (ipa.o, ipa-devrit.o, ipa-inline-analysis.o): Adjust dependencies.
116         * cgraph.c (struct cgraph_propagate_frequency_data,
117         cgraph_propagate_frequency_1, cgraph_propagate_frequency): Move to
118         ipa-profile.c; replace cgraph_ by ipa_ prefix.
119         * cgraph.h (cgraph_propagate_frequency): Remove.
120         * ipa-inline-analysis.c: Include ipa-utils.h; drop duplicated cfgloop.h.
121         (inline_update_callee_summaries): Update.
122         * ipa-profile.c: New file.
123         * ipa-utils.h (ipa_propagate_frequency): Declare.
124         * ipa.c: Do not include pointer-set.h, hash-table.h, lto-streamer.h,
125         data-streamer.h, value-prof.h
126         (symtab_remove_unreachable_nodes): Update profile.
127         (struct histogram_entry, histogram, histogram_pool, histogram_hash,
128         account_time_size, cmp_counts, dump_histogram,
129         ipa_profile_generate_summary, ipa_profile_write_summary,
130         ipa_profile_read_summary, ipa_profile, gate_ipa_profile,
131         pass_data_ipa_profile, pass_ipa_profile, make_pass_ipa_profile):
132         Move to ipa-profile.c
134 2013-09-01  John David Anglin  <danglin@gcc.gnu.org>
136         * config/pa/pa.md: Allow "const 0" operand 1 in "scc" insns.
138 2013-09-01  Jan Hubicka  <jh@suse.cz>
140         * common.opt (fdevirtualize-speculatively): New function.
141         * invoke.texi (fdevirtualize-speculatively): Document.
142         * ipa-devirt.c: Include ipa-inline.h
143         (likely_target_p): New function.
144         (ipa_devirt): New function.
145         (gate_ipa_devirt): New function.
146         (pass_data_ipa_devirt): New static var.
147         (pass_ipa_devirt): Likewise.
148         (make_pass_ipa_devirt): New function.
149         * opts.c (default_options): Add OPT_fdevirtualize_speculatively.
150         (common_handle_option): Disable devirtualization when
151         value range profiling is available.
152         * passes.def (pass_ipa_devirt): Add.
153         * timever.def (TV_IPA_DEVIRT): New timevar.
154         * tree-pass.h (make_pass_ipa_devirt): 
156 2013-09-01  Iain Sandoe  <iain@codesourcery.com>
158         * config/darwin.h (LINK_COMMAND_SPEC_A): Revise sanitizer specs to
159         include sanitize(undefined).
161 2013-08-31  Diego Novillo  <dnovillo@google.com>
163         * Makefile.in (TREE_CORE_H): Define.
164         (TREE_H): Use.
165         (GTFILES): Add tree-core.h.
166         * builtins.c (built_in_class_names): Use BUILT_IN_LAST to
167         size the array.
168         * tree-core.h: New file.
169         Move all data structures, enum, typedefs, global
170         declarations and constants from ...
171         * tree.h: ... here.
173 2013-08-31  Jan Hubicka  <jh@suse.cz>
175         * bulitins.c (expand_builtin): Do not early exit for gcov
176         instrumented functions.
178 2013-08-31  Marek Polacek  <polacek@redhat.com>
180         * ubsan.c: Include tm_p.h.
182 2013-08-31  Jan Hubicka  <jh@suse.cz>
184         * gimple-streamer-in.c (input_gimple_stmt): Silence parameter unused
185         warning.
187         * cgraph.c (cgraph_get_body): Update call of lto_input_function_body.
188         * gimple-streamer-in.c (input_gimple_stmt): Move sanity check to ...
189         * tree-cfg.c (verify_gimple_label): ... here.
190         * ipa-utils.c: Include lto-streamer.h, ipa-inline.h
191         (ipa_merge_profiles): New function.
192         * lto-streamer-in.c (lto_read_body): Take node instead of fn_decl.
193         (lto_input_function_body): Likewise.
194         * ipa-utils.h (ipa_merge_profiles): Declare.
195         * lto-streamer.h (lto_input_function_body): Update prototype.
196         (emit_label_in_global_context_p): Remove.
197         * lto-symtab.c: Include ipa-utils.h
198         (lto_cgraph_replace_node): Use ipa_merge_profiles.
200 2013-08-31  Jan Hubicka  <jh@suse.cz>
202         * cgraph.c (cgraph_speculative_call_info): Fix ref lookup
204 2013-08-31  Jan Hubicka  <jh@suse.cz>
206         * basic-block.h (apply_scale): Make scale parmeter gcov_type.
208 2013-08-31  Uros Bizjak  <ubizjak@gmail.com>
210         * config/alpha/alpha.c (alpha_emit_conditional_move): Update
211         "cmp" RTX before signed_comparison_operator check to account
212         for "code" changes.
214 2013-08-30  Jan Hubicka  <jh@suse.cz>
216         * ipa-prop.c (ipa_set_jf_known_type): Check that we add only records.
217         (detect_type_change_1): Rename to ...
218         (detect_type_change): ... this one; early return on non-polymorphic
219         types.
220         (detect_type_change_ssa): Add comp_type parameter; update
221         use of detect_type_change.
222         (compute_complex_assign_jump_func): Add param_type parameter;
223         update use of detect_type_change_ssa.
224         (compute_complex_ancestor_jump_func): Likewise.
225         (ipa_get_callee_param_type): New function.
226         (ipa_compute_jump_functions_for_edge): Compute parameter type;
227         update calls to the jump function computation functions.
229 2013-08-30  Teresa Johnson  <tejohnson@google.com>
230             Steven Bosscher  <steven@gcc.gnu.org>
232         * cfgrtl.c (fixup_new_cold_bb): New routine.
233         (commit_edge_insertions): Invoke fixup_partitions.
234         (find_partition_fixes): New routine.
235         (fixup_partitions): Ditto.
236         (verify_hot_cold_block_grouping): Update comments.
237         (rtl_verify_edges): Invoke find_partition_fixes.
238         (rtl_verify_bb_pointers): Update comments.
239         (rtl_verify_bb_layout): Ditto.
240         * basic-block.h (probably_never_executed_edge_p): Declare.
241         (fixup_partitions): Ditto.
242         * cfgcleanup.c (try_optimize_cfg): Invoke fixup_partitions.
243         * bb-reorder.c (sanitize_hot_paths): New function.
244         (find_rarely_executed_basic_blocks_and_crossing_edges): Invoke
245         sanitize_hot_paths.
246         * predict.c (probably_never_executed_edge_p): New routine.
247         * cfg.c (check_bb_profile): Add partition insanity warnings.
249 2013-08-30  Meador Inge  <meadori@codesourcery.com>
251         * tree-vrp.c (check_array_ref): Bail out on zero-length arrays.
253 2013-08-30  Marek Polacek  <polacek@redhat.com>
255         * Makefile.in (ubsan.o): Add.
256         (c-family/c-ubsan.o): Add.
257         (builtins.o): Add ubsan.h dependency.
258         * ubsan.h: New file.
259         * ubsan.c: New file.
260         * common.opt: Add -fsanitize=undefined option.
261         (flag_sanitize): Add variable.
262         (fsanitize=): Add option.  Add Driver.
263         (fsanitize=thread): Remove option.
264         (fsanitize=address): Likewise.
265         (static-libubsan): New option.
266         * doc/invoke.texi: Document the new flag and -static-libubsan.
267         * sanitizer.def (DEF_SANITIZER_BUILTIN): Define.
268         (BUILT_IN_UBSAN_HANDLE_BUILTIN_UNREACHABLE): Define.
269         * builtin-attrs.def (ATTR_COLD): Define.
270         (ATTR_COLD_NOTHROW_LEAF_LIST): Define.
271         * builtins.def (BUILT_IN_UBSAN_HANDLE_DIVREM_OVERFLOW,
272         BUILT_IN_UBSAN_HANDLE_SHIFT_OUT_OF_BOUNDS): Define.
273         * flag-types.h (sanitize_code): New enum.
274         * opts.c (common_handle_option): Parse command line arguments
275         of -fsanitize=.  Add -fsanitize=unreachable option.
276         * varasm.c (get_variable_section): Adjust.
277         (assemble_noswitch_variable): Likewise.
278         (assemble_variable): Likewise.
279         (output_constant_def_contents): Likewise.
280         (categorize_decl_for_section): Likewise.
281         (place_block_symbol): Likewise.
282         (output_object_block): Likewise.
283         * builtins.def: Likewise.
284         * toplev.c (compile_file): Likewise.
285         (process_options): Likewise.
286         * cppbuiltin.c: Likewise.
287         * tsan.c (tsan_pass): Likewise.
288         (tsan_gate): Likewise.
289         (tsan_gate_O0): Likewise.
290         * cfgexpand.c (partition_stack_vars): Likewise.
291         (expand_stack_vars): Likewise.
292         (defer_stack_allocation): Likewise.
293         (expand_used_vars): Likewise.
294         * cfgcleanup.c (old_insns_match_p): Likewise.
295         * asan.c (asan_finish_file): Likewise.
296         (asan_instrument): Likewise.
297         (gate_asan): Likewise.
298         (initialize_sanitizer_builtins): Build BT_FN_VOID_PTR_PTR_PTR.
299         (ATTR_COLD_NOTHROW_LEAF_LIST): Define.
300         (asan_global_struct): Use pointer_sized_int_node instead
301         calling build_nonstandard_integer_type.
302         (initialize_sanitizer_builtins): Likewise.
303         (asan_finish_file): Likewise.
304         * gcc.c: Document %{%:function(args):X}.
305         (static_spec_functions): Add sanitize.
306         (handle_spec_function): Add retval_nonnull argument and if non-NULL,
307         store funcval != NULL there.
308         (do_spec_1): Adjust handle_spec_function caller.
309         (handle_braces): Allow %:function(args) as condition.
310         (sanitize_spec_function): New function.
311         (ADD_STATIC_LIBUBSAN_LIBS): Define.
312         (LIBUBSAN_SPEC): Likewise.
313         (LIBUBSAN_EARLY_SPEC): Likewise.
314         (SANITIZER_SPEC): Handle libubsan.
315         (SANITIZER_EARLY_SPEC): Likewise.
316         * config/darwin.h (LINK_COMMAND_SPEC_A): Use %:sanitize(address)
317         instead of fsanitize=address.
318         * config/arm/linux-eabi.h (ASAN_CC1_SPEC): Use %:sanitize(address)
319         instead of fsanitize=address*.
320         * builtins.c: Include ubsan.h.
321         (fold_builtin_0): Instrument __builtin_unreachable.
322         * config/rs6000/rs6000.h (FRAME_GROWS_DOWNWARD): Use flag_sanitize
323         instead of flag_asan.
324         * tree.h (enum tree_index): Add TI_POINTER_SIZED_TYPE.
325         (pointer_sized_int_node): Define.
326         * tree.c (build_common_tree_nodes): Initialize pointer_sized_int_node.
328 2013-08-30  Mike Stump  <mikestump@comcast.net>
330         * doc/install.texi (Prerequisites): Note regression in Tcl 8.6
331         with RE patterns.
333 2013-08-29  Jan Hubicka  <jh@suse.cz>
335         * cgraph.c (cgraph_function_body_availability): Handle weakref
336         correctly.
337         * passes.def: Remove pass_fixup_cfg.
338         * ipa-inline.c (ipa_inline): When not optimizing, do not inline;
339         track when we need to remove functions.
340         (gate_ipa_inline): Execute inlining always; add comment why.
341         (pass_data_ipa_inline): Remove TODO_remove_functions.
342         * ipa-inline-analysis.c (inline_generate_summary): When not optimizing
343         do not produce summaries.
344         * symtab.c (change_decl_assembler_name): Handle renaming of weakrefs.
345         (symtab_nonoverwritable_alias): Assert we are not called on weakref.
346         * varpool.c (cgraph_variable_initializer_availability): Fix weakrefs,
347         constant pool and vtable.
349 2013-08-30  Tejas Belagod  <tejas.belagod@arm.com>
351         * config/aarch64/arm_neon.h (__AARCH64_UINT64_C, __AARCH64_INT64_C):
352         New arm_neon.h's internal macros to specify 64-bit constants.
353         Avoid using stdint.h's macros.
355 2013-08-30  Joern Rennecke  <joern.rennecke@embecosm.com>
357         * recog.c (verify_changes): Verify that changes[i].old is non-zero
358         before applying REG_P.
360 2013-08-30  Jakub Jelinek  <jakub@redhat.com>
362         PR tree-optimization/58277
363         * tree-ssa-strlen.c (strlen_enter_block): If do_invalidate gave up
364         after seeing too many stmts with vdef in between dombb and current
365         bb, invalidate everything.
367 2013-08-30  Richard Biener  <rguenther@suse.de>
369         * fold-const.c (fold_single_bit_test): Fix overflow test.
371 2013-08-30  Eric Botcazou  <ebotcazou@adacore.com>
373         * function.c (assign_parm_setup_reg): For a parameter passed by pointer
374         and which can live in a register, always retrieve the value on entry.
375         * var-tracking.c (add_stores): Treat the copy on entry for a parameter
376         passed by invisible reference specially.
377         (emit_notes_in_bb) <MO_VAL_USE>: Emit notes before the instruction.
378         (vt_add_function_parameter): Correctly deal with a parameter passed by
379         invisible reference.
381 2013-08-30  Jan Hubicka  <jh@suse.cz>
383         * tree.c (set_call_expr_flags): Fix handling of TM_PURE.
385 2013-08-30  Richard Biener  <rguenther@suse.de>
387         PR tree-optimization/58228
388         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Do not
389         allow invariant loads in nested loop vectorization.
391 2013-08-30  Richard Biener  <rguenther@suse.de>
393         PR tree-optimization/58223
394         * tree-loop-distribution.c (has_anti_dependence): Rename to ...
395         (has_anti_or_output_dependence): ... this and adjust to also
396         look for output dependences.
397         (mark_nodes_having_upstream_mem_writes): Adjust.
398         (rdg_flag_uses): Likewise.
400 2013-08-30  Richard Biener  <rguenther@suse.de>
402         PR tree-optimization/58010
403         * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
404         assert that we have a loop-closed PHI.
406 2013-08-29  Jan Hubicka  <jh@suse.cz>
408         * lto-symtab.c (lto_cgraph_replace_node): Free decl_in_state.
409         * cgraph.c (cgraph_release_function_body): Free decl_in_state.
410         * lto-section-in.c (lto_free_function_in_decl_state): New function.
411         (lto_free_function_in_decl_state_for_node): New function.
413 2013-08-29  Xinliang David Li  <davidxl@google.com>
415         * loop-unroll.c (report_unroll_peel): Minor message change.
416         * tree-vect-loop-manip.c (vect_do_peeling_for_alignment):
417         Emit alignment peeling message with default -fopt-info.
418         (vect_loop_versioning): Emit loop version info message.
419         * tree-vectorizer.c (vectorize_loops): Minor message change.
420         (execute_vect_slp): Ditto.
422 2013-08-29  Eric Botcazou  <ebotcazou@adacore.com>
424         * cgraphclones.c (cgraph_create_virtual_clone): Compute the DECL_NAME
425         of the clone from the DECL_NAME of the original function.
427 2013-08-29  Oleg Endo  <olegendo@gcc.gnu.org>
429         * passes.c (register_pass): Add overload.
430         * tree-pass.h (register_pass): Forward declare it.  Add comment.
432 2013-08-29  Jan Hubicka  <jh@suse.cz>
434         * lto-streamer-out.c (hash_tree): Stream DECL_FINAL_P,
435         DECL_CXX_CONSTRUCTOR_P, DECL_CXX_DESTRUCTOR_P and TYPE_FINAL_P.
436         * lto-streamer-in.c (unpack_ts_decl_with_vis_value_fields): Stream
437         DECL_FINAL_P, DECL_CXX_CONSTRUCTOR_P and DECL_CXX_DESTRUCTOR_P.
438         (unpack_ts_type_common_value_fields): Stream TYPE_FINAL_P.
439         * tree-streamer-out.c (pack_ts_decl_with_vis_value_fields):
440         Add DECL_FINAL_P, DECL_CXX_CONSTRUCTOR_P and DECL_CXX_DESTRUCTOR_P.
441         (pack_ts_type_common_value_fields): Add TYPE_FINAL_P.
443 2013-08-29  Teresa Johnson  <tejohnson@google.com>
445         * dumpfile.c (dump_loc): Output column number.
446         * dumpfile.h (OPTGROUP_OTHER): Add and enable under OPTGROUP_ALL.
447         * doc/invoke.texi: Document optall -fopt-info flag.
448         * profile.c (read_profile_edge_counts): Use new dump framework.
449         (compute_branch_probabilities): Ditto.
450         * passes.c (pass_manager::register_one_dump_file): Use OPTGROUP_OTHER
451         when pass not in any opt group.
452         * pass_manager.h (pass_manager::get_pass_profile): New method.
453         * value-prof.c (check_counter): Use new dump framework.
454         (check_ic_target): Ditto.
455         * coverage.c (get_coverage_counts): Ditto.
456         (coverage_init): Setup new dump framework.
458 2013-08-29  Richard Biener  <rguenther@suse.de>
460         PR tree-optimization/58246
461         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Properly
462         handle the dominance check inside a basic-block.
464 2013-08-29  Richard Biener  <rguenther@suse.de>
466         PR middle-end/57287
467         * tree-ssa-copy.c (may_propagate_copy): Allow propagating
468         of default defs that appear in abnormal PHI nodes.
470 2013-08-29  Richard Biener  <rguenther@suse.de>
472         PR tree-optimization/57685
473         * tree-vrp.c (register_edge_assert_for_1): Recurse only for
474         single-use operands to avoid exponential complexity.
476 2013-08-28  Dehao Chen  <dehao@google.com>
478         * ipa-inline.c (edge_badness): Fix integer underflow.
480 2013-08-28  Uros Bizjak  <ubizjak@gmail.com>
482         * gtm-builtins.def (_ITM_free): Declare leaf.
484 2013-08-28  Jakub Jelinek  <jakub@redhat.com>
486         PR target/58067
487         * config/i386/i386.md (*tls_global_dynamic_64_largepic): New insn.
488         (*tls_local_dynamic_base_64_largepic): Likewise.
489         (tls_global_dynamic_64_<mode>, tls_local_dynamic_base_64_<mode>):
490         Remove predicate from call operand.
491         * config/i386/i386.c (ix86_tls_get_addr): For -mcmodel=large -fpic
492         return sum of pic_offset_table_rtx and UNSPEC_PLTOFF of the symbol.
494 2013-08-28  Jeff Law  <law@redhat.com>
496         * tree-ssa-threadedge.c (thread_around_empty_block): Remove
497         checks for the number of predecessors and successors allowed.
498         * tree-ssa-threadupdate.c (mark_threaded_blocks): Ignore requests
499         which require copying a joiner block if there is a request which
500         is a subpath that requires no joiner block copying.
502 2013-08-28  Jan Hubicka  <jh@suse.cz>
504         * lto-streamer-out.c (DFS_write_tree_body): Drop
505         BINFO_INHERITANCE_CHAIN, BINFO_SUBVTT_INDEX and BINFO_VPTR_INDEX.
506         (hash_tree): Do not hash DECL_DEFER_OUTPUT, BINFO_INHERITANCE_CHAIN,
507         BINFO_SUBVTT_INDEX, BINFO_VPTR_INDEX, DECL_IN_TEXT_SECTION.
508         * tree-streamer-in.c (unpack_ts_decl_common_value_fields):
509         Do not read DECL_ERROR_ISSUED.
510         (unpack_ts_decl_with_vis_value_fields): Do not read
511         DECL_DEFER_OUTPUT.
512         (lto_input_ts_binfo_tree_pointers): Do not read
513         BINFO_INHERITANCE_CHAIN, BINFO_SUBVTT_INDEX, BINFO_VPTR_INDEX
514         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Do not
515         write DECL_ERROR_ISSUED..
516         (pack_ts_decl_with_vis_value_fields): Do not write
517         DECL_DEFER_OUTPUT.
518         (write_ts_binfo_tree_pointers): Do not read BINFO_INHERITANCE_CHAIN,
519         BINFO_SUBVTT_INDEX, BINFO_VPTR_INDEX.
520         * print-tree.c (print_node): Do not print DECL_ERROR_ISSUED.
521         * tree.h (tree_decl_common): Update comment.
522         (DECL_ERROR_ISSUED): Remove.
524 2013-08-28  Jakub Jelinek  <jakub@redhat.com>
526         PR middle-end/58257
527         * omp-low.c (copy_var_decl): Copy over TREE_NO_WARNING flag.
529 2013-08-28  Jan Hubicka  <jh@suse.cz>
531         * builtins.def (free): Declare leaf.
533 2013-08-27  David Malcolm  <dmalcolm@redhat.com>
535         * gdbhooks.py: New.
536         * configure.ac (gdbinit.in): Add import of gcc/gdbhooks.py.
537         * configure: Regenerate.
539 2013-08-27  Martin Jambor  <mjambor@suse.cz>
541         * ipa-prop.h (ipa_pass_through_data): New field type_preserved.
542         (ipa_ancestor_jf_data): Likewise.
543         (ipa_get_jf_pass_through_agg_preserved): Fix comment typo.
544         (ipa_get_jf_pass_through_type_preserved): New function.
545         (ipa_get_jf_ancestor_agg_preserved): Fix comment typo.
546         (ipa_get_jf_ancestor_type_preserved): New function.
547         * ipa-cp.c (ipa_get_jf_pass_through_result): Honor type_preserved flag.
548         (ipa_get_jf_ancestor_result): Likewise.
549         (propagate_vals_accross_pass_through): Use
550         ipa_get_jf_pass_through_result to do all the value mappings.
551         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Dump the
552         type_preserved flag.
553         (ipa_set_jf_cst_copy): New function.
554         (ipa_set_jf_simple_pass_through): Set the type_preserved flag.
555         (ipa_set_jf_arith_pass_through): Likewise.
556         (ipa_set_ancestor_jf): Likewise.
557         (compute_complex_assign_jump_func): Set type_preserved instead of
558         punting.
559         (ipa_compute_jump_functions_for_edge): Likewise.
560         (combine_known_type_and_ancestor_jfs): Honor type_preserved.
561         (update_jump_functions_after_inlining): Update type_preserved.
562         Explicitely create jump functions when combining one with pass_through.
563         (ipa_write_jump_function): Stream the type_preserved flags.
564         (ipa_read_jump_function): Likewise.
566 2013-08-27  Jakub Jelinek  <jakub@redhat.com>
567             Aldy Hernandez  <aldyh@redhat.com>
569         * Makefile.in (omp-low.o): Depend on $(TARGET_H).
570         * cfgloop.h (struct loop): Add safelen, force_vect, simduid.
571         * function.h (struct function): Add has_force_vect_loops and
572         has_simduid_loops.
573         * gimple-pretty-print.c (dump_gimple_omp_for): Handle GF_OMP_FOR_KIND*.
574         * gimple.c (gimple_build_omp_critical): Add KIND argument and
575         handle it.
576         * gimple.def: Update CLAUSES comments.
577         * gimple.h (enum gf_mask): Add GF_OMP_FOR_KIND_{FOR,SIMD}.
578         (gimple_build_omp_for): Add argument to prototype.
579         (gimple_omp_for_kind): New.
580         (gimple_omp_for_set_kind): New.
581         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_LINEAR to
582         GOVD_DATA_SHARE_CLASS.
583         (enum omp_region_type): Add ORT_SIMD.
584         (gimple_add_tmp_var): Handle ORT_SIMD.
585         (gimplify_var_or_parm_decl): Same.
586         (is_gimple_stmt): Same.
587         (omp_firstprivatize_variable): Same.
588         (omp_add_variable): Only use splay_tree_insert if lookup failed.
589         (omp_notice_variable): Handle ORT_SIMD.
590         (omp_is_private): Add SIMD argument and handle it as well as ORT_SIMD.
591         (omp_check_private): Handle ORT_SIMD.
592         (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_LINEAR and
593         OMP_CLAUSE_SAFELEN.
594         (gimplify_adjust_omp_clauses_1): Handle GOVD_LINEAR.
595         Handle OMP_CLAUSE_LASTPRIVATE.
596         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_LINEAR and
597         OMP_CLAUSE_SAFELEN.
598         (gimplify_omp_for): Handle OMP_SIMD and OMP_CLAUSE_LINEAR.
599         (gimplify_expr): Handle OMP_SIMD.
600         * internal-fn.c (expand_GOMP_SIMD_LANE): New.
601         (expand_GOMP_SIMD_VF): New.
602         (expand_GOMP_SIMD_LAST_LANE): New.
603         * internal-fn.def (GOMP_SIMD_LANE): New.
604         (GOMP_SIMD_VF): New.
605         (GOMP_SIMD_LAST_LANE): New.
606         * omp-low.c: Include target.h.
607         (extract_omp_for_data): Handle OMP_SIMD, OMP_CLAUSE_LINEAR,
608         OMP_CLAUSE_SAFELEN.
609         (check_omp_nesting_restrictions): Same.
610         (omp_max_vf): New.
611         (lower_rec_simd_input_clauses): New.
612         (lower_rec_input_clauses): Handle OMP_SIMD, GF_OMP_FOR_KIND_SIMD,
613         OMP_CLAUSE_LINEAR.
614         (lower_lastprivate_clauses): Handle OMP_CLAUSE_LINEAR,
615         GF_OMP_FOR_KIND_SIMD, OMP_SIMD.
616         (expand_omp_build_assign): New.
617         (expand_omp_for_init_counts): New.
618         (expand_omp_for_init_vars): New.
619         (extract_omp_for_update_vars): New.
620         (expand_omp_for_generic): Use expand_omp_for_{init,update}_vars
621         and rewrite accordingly.
622         (expand_omp_simd): New.
623         (expand_omp_for): Use expand_omp_simd.
624         (lower_omp_for_lastprivate): Unshare vinit when appropriate.
625         (lower_omp_for): Do not lower the body.
626         * tree-data-ref (get_references_in_stmt): Allow IFN_GOMP_SIMD_LANE
627         in their own loops.
628         * tree-flow.h (find_omp_clause): Remove prototype.
629         * tree-if-conv.c (main_tree_if_conversion): Run if doing if conversion,
630         forcing vectorization of the loop, or if flag_tree_vectorize.
631         (gate_tree_if_conversion): Similarly.
632         * tree-inline.c (remap_gimple_stmt): Pass for kind argument to
633         gimple_build_omp_for.
634         (copy_cfg_body): set has_force_vect_loops and has_simduid_loops.
635         * tree-parloops (create_parallel_loop): Pass kind argument to
636         gimple_build_omp_for.
637         * tree-pretty-print.c (dump_omp_clause): Add cases for
638         OMP_CLAUSE_UNIFORM, OMP_CLAUSE_LINEAR, OMP_CLAUSE_SAFELEN,
639         OMP_CLAUSE__SIMDUID_.
640         (dump_generic_node): Handle OMP_SIMD.
641         * tree-ssa-ccp.c (likely_value): Handle IFN_GOMP_SIMD*.
642         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Do not
643         unroll OMP_SIMD loops here.
644         * tree-ssa-loop.c (gate_tree_vectorize): Run if has_force_vect_loops.
645         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Handle
646         loop->safelen.
647         (vect_analyze_data_refs): Handle simd loops.
648         * tree-vect-loop.c (vectorizable_live_operation): Handle
649         IFN_GOMP_SIMD*.
650         * tree-vect-stmts.c (vectorizable_call): Handle IFN_GOMP_SIMD_LANE.
651         (vectorizable_store): Handle STMT_VINFO_SIMD_LANE_ACCESS_P.
652         (vectorizable_load): Same.
653         * tree-vectorizer.c: Include hash-table.h and tree-ssa-propagate.h.
654         (struct simduid_to_vf): New.
655         (simduid_to_vf::hash): New.
656         (simduid_to-vf::equal): New.
657         (struct simd_array_to_simduid): New.
658         (simd_array_to_simduid::hash): New.
659         (simd_array_to_simduid::equal): New.
660         (adjust_simduid_builtins): New.
661         (struct note_simd_array_uses_struct): New.
662         (note_simd_array_uses_cb): New.
663         (note_simd_array_uses): New.
664         (vectorize_loops): Handle simd hints and adjust simd builtins
665         accordingly.
666         * tree-vectorizer.h (struct _stmt_vec_info): Add
667         simd_lane_access_p field.
668         (STMT_VINFO_SIMD_LANE_ACCESS_P): New macro.
669         * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_LINEAR,
670         OMP_CLAUSE_SAFELEN, OMP_CLAUSE__SIMDUID_, OMP_CLAUSE_UNIFORM.
671         (omp_clause_code_name): Same.
672         (walk_tree_1): Handle OMP_CLAUSE_UNIFORM, OMP_CLAUSE_SAFELEN,
673         OMP_CLAUSE__SIMDUID_, OMP_CLAUSE_LINEAR.
674         * tree.def (OMP_SIMD): New entry.
675         * tree.h (enum omp_clause_code): Add entries for OMP_CLAUSE_LINEAR,
676         OMP_CLAUSE_UNIFORM, OMP_CLAUSE_SAFELEN, OMP_CLAUSE__SIMDUID_.
677         (OMP_CLAUSE_DECL): Adjust range for new clauses.
678         (OMP_CLAUSE_LINEAR_NO_COPYIN): New.
679         (OMP_CLAUSE_LINEAR_NO_COPYOUT): New.
680         (OMP_CLAUSE_LINEAR_STEP): New.
681         (OMP_CLAUSE_SAFELEN_EXPR): New.
682         (OMP_CLAUSE__SIMDUID__DECL): New.
683         (find_omp_clause): New prototype.
685 2013-08-27  H.J. Lu  <hongjiu.lu@intel.com>
687         * config/i386/driver-i386.c (host_detect_local_cpu): Update
688         Haswell processor detection.
690 2013-08-27  Christian Widmer  <shadow@umbrox.de>
692         PR target/57927
693         * config/i386/driver-i386.c (host_detect_local_cpu): Add detection
694         of Ivy Bridge and Haswell processors.  Assume core-avx2 for unknown
695         AVX2 capable processors.
697 2013-08-27  Tejas Belagod  <tejas.belagod@arm.com>
699         * config/aarch64/arm_neon.h: Replace all inline asm implementations
700         of vget_low_* with implementations in terms of other intrinsics.
702 2013-08-27  Marc Glisse  <marc.glisse@inria.fr>
704         PR middle-end/57219
705         * doc/extend.texi (__builtin_isinf_sign): Restrict the return
706         values to -1, 0 and 1.
708 2013-08-27  Vidya Praveen  <vidyapraveen@arm.com>
710         * config/aarch64/aarch64.md (unspec): Add UNSPEC_SISD_SSHL,
711         UNSPEC_SISD_USHL, UNSPEC_USHL_2S, UNSPEC_SSHL_2S, UNSPEC_SISD_NEG.
712         (<optab><mode>3_insn): Remove.
713         (aarch64_ashl_sisd_or_int_<mode>3): New Pattern.
714         (aarch64_lshr_sisd_or_int_<mode>3): Likewise.
715         (aarch64_ashr_sisd_or_int_<mode>3): Likewise.
716         (define_split for aarch64_lshr_sisd_or_int_di3): Likewise.
717         (define_split for aarch64_lshr_sisd_or_int_si3): Likewise.
718         (define_split for aarch64_ashr_sisd_or_int_di3): Likewise.
719         (define_split for aarch64_ashr_sisd_or_int_si3): Likewise.
720         (aarch64_sisd_ushl, aarch64_sisd_sshl): Likewise.
721         (aarch64_ushl_2s, aarch64_sshl_2s, aarch64_sisd_neg_qi): Likewise.
722         (ror<mode>3_insn): Likewise.
723         * config/aarch64/predicates.md (aarch64_simd_register): New.
725 2013-08-27  Richard Biener  <rguenther@suse.de>
727         PR tree-optimization/57521
728         * tree-if-conv.c (if_convertible_bb_p): Verify that at least
729         one edge is non-critical.
730         (find_phi_replacement_condition): Make sure to use a non-critical
731         edge.  Cleanup and remove old bug workarounds.
732         (bb_postdominates_preds): Remove.
733         (if_convertible_loop_p_1): Do not compute post-dominators.
734         (combine_blocks): Do not free post-dominators.
735         (main_tree_if_conversion): Likewise.
736         (pass_data_if_conversion): Add TODO_verify_ssa.
738 2013-08-27  DJ Delorie  <dj@redhat.com>
740         * config/i386/djgpp.h (ASM_DECLARE_FUNCTION_NAME): New.
742 2013-08-27  Yufeng Zhang  <yufeng.zhang@arm.com>
744         * function.c (assign_parm_find_data_types): Set passed_mode and
745         nominal_mode to the TYPE_MODE of nominal_type for the built
746         pointer type in case of the struct-pass-by-reference.
748 2013-08-26  Joern Rennecke  <joern.rennecke@embecosm.com>
750         * config/avr/avr-stdint.h (INT16_TYPE): Change default to "int".
751         (UINT16_TYPE): Change default to "unsigned int".
753         * config/avr/avr.opt (mfract-convert-truncate): New option.
754         * config/avr/avr.c (avr_out_fract): Unless TARGET_FRACT_CONV_TRUNC
755         is set, round negative fractional integers according to n1169
756         when converting to integer types.
758 2013-08-26  Jan Hubicka  <jh@suse.cz>
760         * cgraph.c (cgraph_propagate_frequency): Do not assume that virtual
761         methods can not be called indirectly when their address is not taken.
763 2013-08-26  Jan Hubicka  <jh@suse.cz>
765         * gimple-fold.c (gimple_get_virt_method_for_binfo): Use
766         ctor_for_folding.
768 2013-08-26  Jan Hubicka  <jh@suse.cz>
770         * ipa.c (comdat_can_be_unshared_p_1): C++ constructors and destructors
771         can be unshared.
773 2013-08-26  Joern Rennecke  <joern.rennecke@embecosm.com>
775         * reload.c (find_valid_class): Allow classes that do not include
776         FIRST_PSEUDO_REGISTER - 1.
778 2013-08-26  Jan Hubicka  <jh@suse.cz>
780         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix formatting;
781         fix edge count/frequency when speculation failed; fix type check
782         for the direct call.
784 2013-08-26  Jan Hubicka  <jh@suse.cz>
786         * ipa-prop.c (ipa_print_node_params): Do not ICE during WPA.
788 2013-08-26  Jan Hubicka  <jh@suse.cz>
790         * ipa-inline-transform.c (inline_transform): Be ready for basic block
791         to be changed by edge redirection.
793 2013-08-26  Jan Hubicka  <jh@suse.cz>
795         * cgraph.c (cgraph_speculative_call_info): Fix parameter order and
796         formating; add sanity check.
797         (cgraph_resolve_speculation): Add FIXME about scaling profiles.
798         (cgraph_redirect_edge_call_stmt_to_callee): Fix ICE in debug dump.
799         * ipa-inline.c (heap_edge_removal_hook): Reset node growth cache.
800         (resolve_noninline_speculation): Update callee keys, too.
802 2013-08-26  Jan Hubicka  <jh@suse.cz>
804         * tree.h (tree_decl_with_vis): Add cxx_constructor, cxx_destructor.
805         (DECL_CXX_CONSTRUCTOR_P, DECL_CXX_DESTRUCTOR_P): New macros.
807 2013-08-26  Joern Rennecke  <joern.rennecke@embecosm.com>
809         * config/i386/i386.c (x86_64_elf_select_section): Put ATTRIBUTE_UNUSED
810         into proper place.
812 2013-08-26  Uros Bizjak  <ubizjak@gmail.com>
814         * config/i386/i386.c (ix86_debug_options): Remove prototype.
815         (x86_64_elf_select_section): Ditto.
816         (ix86_handle_tm_regparm_attribute): Remove ATTRIBUTE_UNUSED on used
817         arguments.
818         (ix86_pass_by_reference): Ditto.
819         (output_set_got): Ditto.
820         (ix86_unary_operator_ok): Ditto.
821         (ix86_expand_builtin): Ditto.
823 2013-08-23  Jan Hubicka  <jh@suse.cz>
825         * cgraph.c (cgraph_turn_edge_to_speculative): Fix debug output.
827 2013-08-23  Jan Hubicka  <jh@suse.cz>
829         * tree.h (TYPE_FINAL_P, DECL_FINAL_P): New macros.
830         (tree_decl_with_vis): Add FINAL field.
832 2013-08-23  Jeff Law  <law@redhat.com>
834         * tree-ssa-pre.c (do_regular_insertion): Include the expression in
835         the debugging dump when the expression is fully redundant.
837 2013-08-23  Gabriel Dos Reis  <gdr@integrable-solutions.net>
839         * diagnostic.c (diagnostic_set_caret_max_width): Use pp_buffer.
840         * gimple-pretty-print.c (gimple_dump_bb_buff): Likewise.
841         * pretty-print.c (pp_formatted_text_data): Likewise.
842         (pp_write_text_to_stream): Likewise.
843         (pp_write_text_as_dot_label_to_stream): Likewise.
844         (pp_append_r): Likewise.
845         (pp_format): Likewise.
846         (pp_flush): Likewise.
847         (pp_clear_output_area): Likewise.
848         (pp_append_text): Likewise.
849         (pp_formatted_text): Likewise.
850         (pp_remaining_character_count_for_line): Likewise.
851         (pp_newline): Likewise.
852         (pp_character): Likewise.
853         (output_buffer::~output_buffer): Define.
854         (pretty_printer::~pretty_printer): Destruct output buffer.
855         * pretty-print.h (output_buffer::~output_buffer): Declare.
856         (pretty_printer::~pretty_printer): Declare virtual.
858 2013-08-24  Marc Glisse  <marc.glisse@inria.fr>
860         PR other/57324
861         * hwint.h (HOST_WIDE_INT_UC, HOST_WIDE_INT_1U, HOST_WIDE_INT_M1,
862         HOST_WIDE_INT_M1U): New macros.
863         * fold-const.c (sign_bit_p, build_range_check, fold_unary_loc,
864         fold_binary_loc, fold_ternary_loc): Use the new macros. Use an
865         unsigned -1 for lshift.
866         * cse.c (cse_insn): Likewise.
867         * double-int.c (rshift_double, lshift_double): Likewise.
868         * builtins.c (fold_builtin_bitop): Likewise.
869         * combine.c (force_to_mode): Likewise.
870         * tree.c (integer_pow2p, tree_log2, tree_floor_log2): Likewise.
871         * simplify-rtx.c (simplify_const_unary_operation,
872         simplify_const_binary_operation): Likewise.
873         * tree-stdarg.c (va_list_counter_bump, va_list_ptr_read,
874         check_va_list_escapes): Likewise.
875         * rtlanal.c (nonzero_bits1): Likewise.
876         * expmed.c (expand_smod_pow2): Likewise.
877         * tree-ssa-structalias.c (UNKNOWN_OFFSET): Use HOST_WIDE_INT_MIN.
879 2013-08-23  Jan Hubicka  <jh@suse.cz>
881         * cgraph.c (cgraph_turn_edge_to_speculative): Mark target node
882         as having address taken.
884 2013-08-23  Jan Hubicka  <jh@suse.cz>
886         * ipa-utils.h (method_class_type): Declare.
887         * ipa-devirt.c (method_class_type): Export.
889         * cgraphunit.c (analyze_functions): Do basic devirtualization;
890         do not walk base classes of anonymous types.
892 2013-08-23  Kaz Kojima  <kkojima@gcc.gnu.org>
894         PR rtl-optimization/58220
895         PR regression/58221
896         * final.c (reemit_insn_block_notes): Use NEXT_INSN to
897         handle SEQUENCE insns properly.
899 2013-08-23  Gabriel Dos Reis  <gdr@integrable-solutions.net>
901         * pretty-print.h (pp_newline_and_flush): Declare.  Remove macro
902         definition.
903         (pp_newline_and_indent): Likewise.
904         (pp_separate_with): Likewise.
905         * pretty-print.c (pp_newline_and_flush): Define.
906         (pp_newline_and_indent): Likewise.
907         (pp_separate_with): Likewise.
909 2013-08-23  Jakub Jelinek  <jakub@redhat.com>
911         PR target/58218
912         * config/i386/x86-64.h (TARGET_SECTION_TYPE_FLAGS): Define.
913         * config/i386/i386.c (x86_64_elf_section_type_flags): New function.
915 2013-08-23  Kirill Yukhin  <kirill.yukhin@intel.com>
917         * gcc/config/i386/predicates.md (ext_sse_reg_operand): New.
918         * gcc/config/i386/i386.md (*movti_internal): Use
919         predicate to determine if EVEX is needed.
920         (*movsi_internal): Ditto.
921         (*movdf_internal): Ditto.
922         (*movsf_internal): Ditto.
923         * gcc/config/i386/mmx.md (*mov<mode>_internal): Ditto.
925 2013-08-23  Jakub Jelinek  <jakub@redhat.com>
927         PR tree-optimization/58209
928         * tree-tailcall.c (process_assignment): Handle POINTER_PLUS_EXPR.
929         (find_tail_calls): Give up for pointer result types if m is non-NULL.
930         (adjust_return_value_with_ops): For PLUS_EXPR and pointer result type
931         emit POINTER_PLUS_EXPR.
932         (create_tailcall_accumulator): For pointer result type accumulate in
933         sizetype type.
935 2013-08-22  Paolo Carlini  <paolo.carlini@oracle.com>
937         * configure.ac: Add backslashes missing from the last change.
938         * configure: Regenerate.
940 2013-08-22  Jan Hubicka  <jh@susue.cz>
942         * ipa.c (function_and_variable_visibility): First remember function
943         was global and then make it local.
945 2013-08-22  Julian Brown  <julian@codesourcery.com>
947         * configure.ac: Add aarch64 to list of arches which use "nop" in
948         debug_line test.
949         * configure: Regenerate.
951 2013-08-22  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
953         * config/s390/linux.h (TARGET_LIBC_HAS_FUNCTION): Define as
954         gnu_libc_has_function.
955         * config/s390/tpf.h: Likewise.
957 2013-08-22  Jan Hubicka  <jh@susue.cz>
959         * timevar.c (validate_phases): Add cast.
961 2013-08-22  Jan Hubicka  <jh@susue.cz>
963         * timevar.c (validate_phases): Use size_t for memory.
964         * timevar.h (struct timevar_time_def): Use size_t for ggc_mem.
966 2013-08-22  Gabriel Dos Reis  <gdr@integrable-solutions.net>
968         * pretty-print.h (output_buffer::output_buffer): Declare.
969         (pretty_printer::pretty_printer): Likewise.
970         (pp_construct): Remove.
971         * pretty-print.c (output_buffer::output_buffer): Define.
972         (pretty_printer::pretty_printer): Rename from pp_construct.  Simplify.
973         * gimple-pretty-print.c (print_gimple_stmt): Do not call pp_construct.
974         (print_gimple_expr): Likewise.
975         (print_gimple_seq): Likewise.
976         (gimple_dump_bb): Likewise.
977         * sched-vis.c (dump_value_slim): Likewise.
978         (dump_insn_slim): Likewise.
979         (dump_rtl_slim): Likewise.
980         (str_pattern_slim): Likewise.
981         * tree-mudflap.c (mf_varname_tree): Likewise.
982         * graph.c (print_graph_cfg): Likewise.
983         (start_graph_dump): Likewise.
984         * tree-pretty-print.c (maybe_init_pretty_print): Likewise.  Use
985         placement-new.
986         * diagnostic.c (diagnostic_initialize): Simplify early diagnostic
987         pretty printer initialization.
988         * coretypes.h (diagnostic_context): Remove superflous type alias
989         declaration.
990         (pretty_printer): Likewise.  Declare directly as a class.
991         (pretty_print_info): Remove declaration as class.
992         * asan.c (asan_emit_stack_protection): Remove call to pp_construct
993         and pp_clear_output_area.
994         (asan_add_global): Likewise.
996 2013-08-22  Jan Hubicka  <jh@suse.cz>
998         * cgraphunit.c (analyze_functions) Use update_type_inheritance_graph.
999         * ipa-utils.h (update_type_inheritance_graph): Declare.
1000         (possible_polymorphic_call_target_p): Declare.
1001         (possible_polymorphic_call_target_p): New.
1002         * ipa-devirt.c: Update toplevel comments.
1003         (cached_polymorphic_call_targets): Move up.
1004         (odr_type_d): Move ID down.
1005         (polymorphic_type_binfo_p): Update comment.
1006         (odr_hasher::remove): Likewise;
1007         (get_odr_type): Set anonymous_namespace.
1008         (dump_odr_type): Dump it.
1009         (dump_type_inheritance_graph): Do not ICE when there are no ODR types.
1010         (maybe_record_node): Record node in cached_polymorphic_call_targets.
1011         (record_binfo): Add comment.
1012         (free_polymorphic_call_targets_hash): Do not ICE when cache is not
1013         built.
1014         (devirt_node_removal_hook): Do not iCE when cache is freed.
1015         (possible_polymorphic_call_target_p): New predicate.
1016         (update_type_inheritance_graph): New function.
1018 2013-08-22  Alexander Ivchenko  <alexander.ivchenko@intel.com>
1019             Maxim Kuznetsov  <maxim.kuznetsov@intel.com>
1020             Sergey Lega  <sergey.s.lega@intel.com>
1021             Anna Tikhonova  <anna.tikhonova@intel.com>
1022             Ilya Tocar  <ilya.tocar@intel.com>
1023             Andrey Turetskiy  <andrey.turetskiy@intel.com>
1024             Ilya Verbin  <ilya.verbin@intel.com>
1025             Kirill Yukhin  <kirill.yukhin@intel.com>
1026             Michael Zolotukhin  <michael.v.zolotukhin@intel.com>
1028         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_SET): New.
1029         (OPTION_MASK_ISA_AVX512CD_SET): Ditto.
1030         (OPTION_MASK_ISA_AVX512PF_SET): Ditto.
1031         (OPTION_MASK_ISA_AVX512ER_SET): Ditto.
1032         (OPTION_MASK_ISA_AVX2_UNSET): Update.
1033         (OPTION_MASK_ISA_AVX512F_UNSET): New.
1034         (OPTION_MASK_ISA_AVX512CD_UNSET): Ditto.
1035         (OPTION_MASK_ISA_AVX512PF_UNSET): Ditto.
1036         (OPTION_MASK_ISA_AVX512ER_UNSET): Ditto.
1037         (ix86_handle_option): Handle OPT_mavx512f, OPT_mavx512cd,
1038         OPT_mavx512pf, OPT_mavx512er cases.
1039         * config/i386/constraints.md (v): New constraint.
1040         (Yi, Yj): Replace SSE_REGS with ALL_SSE_REGS.
1041         * config/i386/cpuid.h (bit_AVX512F, bit_AVX512PF, bit_AVX512ER)
1042         (bit_AVX512CD): New.
1043         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1044         AVX512F, AVX512ER, AVX512PF, AVX512CD features.
1045         * config/i386/i386-c.c (ix86_target_macros_internal):
1046         Conditionally define __AVX512F__, __AVX512ER__, __AVX512CD__,
1047         __AVX512PF__.
1048         * config/i386/i386-modes.def (VECTOR_MODES (INT, 128))
1049         (VECTOR_MODES (FLOAT, 128), INT_MODE (XI, 64)): New modes.
1050         * config/i386/i386.c (regclass_map, dbx_register_map)
1051         (dbx64_register_map, svr4_dbx_register_map): Add new SSE registers.
1052         (gate_insert_vzeroupper): Disable vzeroupper for TARGET_AVX512F.
1053         (ix86_target_string): Define -mavx512f, -mavx512er, -mavx512cd,
1054         -mavx512pf options.
1055         (ix86_option_override_internal): Define PTA_AVX512F, PTA_AVX512ER,
1056         PTA_AVX512PF, PTA_AVX512CD.  Handle -mavx512f, -mavx512er, -mavx512cd,
1057         -mavx512pf options.  Fix formatting.
1058         (ix86_conditional_register_usage): Squash EXT_REX_SSE_REGs for 32-bit
1059         targets.  Squash EVEX_SSE_REGS if AVX512F is disabled.
1060         (ix86_valid_target_attribute_inner_p): Handle -mavx512f, -mavx512er,
1061         -mavx512cd, -mavx512pf options.
1062         (standard_sse_constant_opcode): Add vpternlogd for 512-bit modes.
1063         (print_reg, ix86_print_operand): Handle 'g' to output 512-bit operands.
1064         (ix86_preferred_output_reload_class): Replace SSE_REGS with
1065         ALL_SSE_REGS.
1066         (ix86_hard_regno_mode_ok): Support 512-bit registers.
1067         (ix86_set_reg_reg_cost): Ditto.
1068         (x86_order_regs_for_local_alloc): Ditto.
1069         (MAX_VECT_LEN): Extend to 64-byte.
1070         (ix86_spill_class): Replace SSE_REGS with ALL_SSE_REGS.
1071         * config/i386/i386.h (TARGET_AVX512F, TARGET_AVX512PF)
1072         (TARGET_AVX512ER, TARGET_AVX512CD): New.
1073         (BIGGEST_ALIGNMENT): Extend to 512-bits.
1074         (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS): Add new registers.
1075         (CALL_USED_REGISTERS, REG_ALLOC_ORDER): Likewise.
1076         (VALID_AVX512F_SCALAR_MODE, VALID_AVX512F_REG_MODE): New.
1077         (SSE_REG_MODE_P): Support new modes.
1078         (FIRST_MMX_REG, FIRST_REX_INT_REG, FIRST_REX_SSE_REG): Add comments.
1079         (FIRST_EXT_REX_SSE_REG, LAST_EXT_REX_SSE_REG): New.
1080         (reg_class, REG_CLASS_NAMES): Add EVEX_SSE_REGS, ALL_SSE_REGS.
1081         (SSE_CLASS_P, MAYBE_SSE_CLASS_P): Replace SSE_REGS with ALL_SSE_REGS.
1082         (REG_CLASS_CONTENTS): Add new registers.
1083         (SSE_REGNO_P, SSE_REGNO, HARD_REGNO_RENAME_OK): Support new registers.
1084         (EXT_REX_SSE_REGNO_P): New.
1085         (HI_REGISTER_NAMES): Add new registers.
1086         * config/i386/i386.md: Define constants for new registers.
1087         (mode): Add new 512-bit modes.
1088         (prefix): Support evex prefix.
1089         (isa): Support avx512f, noavx512f, fma_avx512f.
1090         (ssemodesuffix): Add new 512-bit modes.
1091         (movxi): New.
1092         (*movxi_internal_avx512f): Ditto.
1093         (*movdi_internal): Replace constraint "x" with the new constraint "v".
1094         Support MODE_XI.
1095         (*movsi_internal): Likewise.
1096         (*movdf_internal): Likewise.
1097         (*movsf_internal): Likewise.
1098         (*fop_<mode>_comm_sse): Replace constraint "x" with new constraint "v".
1099         (<code><mode>3): Likewise.
1100         * config/i386/i386.opt (mavx512f, mavx512pf, mavx512er, mavx512cd):
1101         New.
1102         * config/i386/mmx.md (*mov<mode>_internal): Replace constraint "x"
1103         with the new constraint "v".
1104         * config/i386/sse.md (*mov<mode>_internal): Support new registers and
1105         modes.
1106         (<sse>_loadu<ssemodesuffix><avxsizesuffix>): Replace constraint "x"
1107         with the new constraint "v".
1108         (<sse2>_loaddqu<avxsizesuffix>): Likewise.
1109         (<sse2>_storedqu<avxsizesuffix>): Likewise.
1110         (*<plusminus_insn><mode>3): Likewise.
1111         (<sse>_vm<plusminus_insn><mode>3): Likewise.
1112         (*mul<mode>3): Likewise.
1113         (<sse>_vmmul<mode>3): Likewise.
1114         (<sse>_div<mode>3): Likewise.
1115         (<sse>_vmdiv<mode>3): Likewise.
1116         (<sse>_sqrt<mode>2): Likewise.
1117         (<sse>_vmsqrt<mode>2): Likewise.
1118         (*<code><mode>3_finite): Likewise.
1119         (*<code><mode>3) <smaxmin>: Likewise.
1120         (<sse>_vm<code><mode>3): Likewise.
1121         (*<code><mode>3) <any_logic>: Likewise.
1122         (*fma_fmadd_<mode>): Likewise.
1123         (*fma_fmsub_<mode>): Likewise.
1124         (*fma_fnmadd_<mode>): Likewise.
1125         (*fma_fnmsub_<mode>): Likewise.
1126         (*fma_fmaddsub_<mode>): Likewise.
1127         (*fma_fmsubadd_<mode>): Likewise.
1128         (*fmai_fmadd_<mode>): Likewise.
1129         (*fmai_fmsub_<mode>): Likewise.
1130         (*fmai_fnmadd_<mode>): Likewise.
1131         (*fmai_fnmsub_<mode>): Likewise.
1132         (sse_cvtsi2ss): Likewise.
1133         (sse_cvtsi2ssq): Likewise.
1134         (sse_cvtss2si): Likewise.
1135         (sse_cvtss2si_2): Likewise.
1136         (sse_cvtss2siq): Likewise.
1137         (sse_cvtss2siq_2): Likewise.
1138         (sse_cvttss2si): Likewise.
1139         (sse_cvtss2siq_2): Likewise.
1140         (float<sseintvecmodelower><mode>2): Likewise.
1141         (sse2_cvtsd2si_2): Likewise.
1142         (sse2_cvtsd2siq_2): Likewise.
1143         (*<plusminus_insn><mode>3): Likewise.
1144         (*<sse2_avx2>_<plusminus_insn><mode>3): Likewise.
1145         (*<sse4_1_avx2>_mul<mode>3): Likewise.
1146         (ashr<mode>3): Likewise.
1147         (<shift_insn><mode>3): Likewise.
1148         (avx2_<code><mode>3): Likewise.
1149         (*avx2_<code><mode>3): Likewise.
1150         (*andnot<mode>3): Likewise.
1151         (*<code><mode>3) <any_logic>: Likewise.
1152         (abs<mode>2): Likewise.
1153         (avx2_permvar<mode>): Likewise.
1154         (avx2_perm<mode>_1): Likewise.
1155         (*avx_vpermilp<mode>): Likewise.
1156         (avx_vpermilvar<mode>3): Likewise.
1157         (avx2_ashrv<mode>): Likewise.
1158         (avx2_<shift_insn>v<mode>): Likewise.
1159         * doc/invoke.texi: Document -mavx512f, -mavx512pf, -mavx512er,
1160         -mavx512cd.
1161         * doc/rtl.texi: Document XImode.
1163 2013-08-21  Jeff Law  <law@redhat.com>
1165         * tree-flow.h (register_jump_thread): Pass vector of edges
1166         instead of each important edge.
1167         * tree-ssa-threadedge.c (thread_across_edge): Build the jump
1168         thread path into a vector and pass that to register_jump_thread.
1169         * tree-ssa-threadupdate.c (register_jump_thread): Conver the
1170         passed in edge vector to the current 3-edge form.
1172         Revert:
1173         2013-08-20  Alexey Makhalov  <makhaloff@gmail.com>
1175         * dce.c (fini_dce): Call df_analyze again just in case
1176         delete_unmarked_insns removed anything.
1178 2013-08-21  Joern Rennecke  <joern.rennecke@embecosm.com>
1180         * reload.h (struct reg_equivs): Rename to ..
1181         (struct reg_equivs_s): .. this.
1183 2013-08-20  Martin Liska  <marxin.liska@gmail.com>
1185         * ipa.c (ipa_profile_read_summary): Fix buffer overflow.
1187 2013-08-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1189         * config/sol2-10.h (TARGET_LIBC_HAS_FUNCTION): Don't nest comment.
1191 2013-08-21  Jeff Law  <law@redhat.com>
1193         * tree-vrp.c (simplify_stmt_for_jump_threading): Try to
1194         simplify assignments too.  If the RHS collapses to a singleton
1195         range, then return the value for the range.
1197 2013-08-21  Kirill Yukhin  <kirill.yukhin@intel.com>
1199         * config/i386/sse.md (V16): Rename to...
1200         (VMOVE): this.
1201         (mov<mode>): Update iterator name.
1202         (*mov<mode>_internal): Ditto.
1203         (push<mode>1): Ditto.
1204         (movmisalign<mode>): Ditto.
1206 2013-08-20  Jan Hubicka  <jh@suse.cz>
1208         PR bootstrap/58186
1209         * cgraph.c (cgraph_add_edge_to_call_site_hash): Overwrite hash
1210         entry for direct edges.
1211         (cgraph_turn_edge_to_speculative): Fix setting of can_throw_external.
1213 2013-08-20  David Malcolm  <dmalcolm@redhat.com>
1215         Revert my last two changes, r201865 and r201864:
1217         Revert r201865:
1218         2013-08-20  David Malcolm  <dmalcolm@redhat.com>
1220         Make opt_pass and gcc::pass_manager be GC-managed, so that pass
1221         instances can own GC refs.
1223         * Makefile.in (GTFILES): Add pass_manager.h and tree-pass.h.
1224         * context.c (gcc::context::gt_ggc_mx): Traverse passes_.
1225         (gcc::context::gt_pch_nx): Likewise.
1226         (gcc::context::gt_pch_nx):  Likewise.
1227         * ggc.h (gt_ggc_mx <T>): New.
1228         (gt_pch_nx_with_op <T>): New.
1229         (gt_pch_nx <T>): New.
1230         * passes.c (opt_pass::gt_ggc_mx): New.
1231         (opt_pass::gt_pch_nx): New.
1232         (opt_pass::gt_pch_nx_with_op): New.
1233         (pass_manager::gt_ggc_mx): New.
1234         (pass_manager::gt_pch_nx): New.
1235         (pass_manager::gt_pch_nx_with_op): New.
1236         (pass_manager::operator new): Use
1237         ggc_internal_cleared_alloc_stat rather than xcalloc.
1238         * pass_manager.h (class pass_manager): Add GTY((user)) marking.
1239         (pass_manager::gt_ggc_mx): New.
1240         (pass_manager::gt_pch_nx): New.
1241         (pass_manager::gt_pch_nx_with_op): New.
1242         * tree-pass.h (class opt_pass): Add GTY((user)) marking.
1243         (opt_pass::operator new): New.
1244         (opt_pass::gt_ggc_mx): New.
1245         (opt_pass::gt_pch_nx): New.
1246         (opt_pass::gt_pch_nx_with_op): New.
1248         Revert r201864:
1249         2013-08-20  David Malcolm  <dmalcolm@redhat.com>
1251         * Makefile.in (GTFILES): Add context.h.
1252         * context.c (gcc::context::operator new): New.
1253         (gcc::context::gt_ggc_mx): New.
1254         (gcc::context::gt_pch_nx): New.
1255         (gcc::context::gt_pch_nx): New.
1256         * context.h (gcc::context): Add GTY((user)) marking.
1257         (gcc::context::operator new): New.
1258         (gcc::context::gt_ggc_mx): New.
1259         (gcc::context::gt_pch_nx): New.
1260         (gcc::context::gt_pch_nx): New.
1261         (g): Add GTY marking.
1262         (gt_ggc_mx (gcc::context *)): New.
1263         (gt_pch_nx (gcc::context *)): New.
1264         (gt_pch_nx (gcc::context *ctxt, gt_pointer_operator op,
1265         void *cookie)): New.
1266         * gengtype.c (open_base_files) <ifiles>: Add context.h.
1268 2013-08-20  Alexey Makhalov  <makhaloff@gmail.com>
1270         * dce.c (fini_dce): Call df_analyze again just in case
1271         delete_unmarked_insns removed anything.
1273 2013-08-20  Teresa Johnson  <tejohnson@google.com>
1275         PR rtl-optimizations/57451
1276         * final.c (reemit_insn_block_notes): Prevent lexical blocks
1277         from crossing split section boundaries.
1279 2013-08-20  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
1281         * config/arm/linux-elf.h (MULTILIB_DEFAULTS): Remove definition.
1282         * config/arm/t-linux-eabi (MULTILIB_OPTIONS): Document association
1283         with MULTLIB_DEFAULTS.
1285 2013-08-20  Nick Clifton  <nickc@redhat.com>
1287         * target.def (narrow_volatile_bitfield): Note that the default
1288         value is false, not !TARGET_STRICT_ALIGN.
1289         * doc/tm.texi: Regenerate.
1291 2013-08-20  Pavel Chupin  <pavel.v.chupin@intel.com>
1293         Fix LIB_SPEC for systems without libpthread.
1295         * config/gnu-user.h: Introduce GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC.
1296         * config/arm/linux-eabi.h: Use GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC
1297         for Android.
1298         * config/i386/linux-common.h: Likewise.
1299         * config/mips/linux-common.h: Likewise.
1301 2013-08-20  Zhouyi Zhou  <yizhouzhou@ict.ac.cn>
1303         * tree-ssa-ccp.c (get_default_value): Remove redundant condition
1304         checks.
1306 2013-08-20  David Malcolm  <dmalcolm@redhat.com>
1308         Make opt_pass and gcc::pass_manager be GC-managed, so that pass
1309         instances can own GC refs.
1311         * Makefile.in (GTFILES): Add pass_manager.h and tree-pass.h.
1312         * context.c (gcc::context::gt_ggc_mx): Traverse passes_.
1313         (gcc::context::gt_pch_nx): Likewise.
1314         (gcc::context::gt_pch_nx):  Likewise.
1315         * ggc.h (gt_ggc_mx <T>): New.
1316         (gt_pch_nx_with_op <T>): New.
1317         (gt_pch_nx <T>): New.
1318         * passes.c (opt_pass::gt_ggc_mx): New.
1319         (opt_pass::gt_pch_nx): New.
1320         (opt_pass::gt_pch_nx_with_op): New.
1321         (pass_manager::gt_ggc_mx): New.
1322         (pass_manager::gt_pch_nx): New.
1323         (pass_manager::gt_pch_nx_with_op): New.
1324         (pass_manager::operator new): Use
1325         ggc_internal_cleared_alloc_stat rather than xcalloc.
1326         * pass_manager.h (class pass_manager): Add GTY((user)) marking.
1327         (pass_manager::gt_ggc_mx): New.
1328         (pass_manager::gt_pch_nx): New.
1329         (pass_manager::gt_pch_nx_with_op): New.
1330         * tree-pass.h (class opt_pass): Add GTY((user)) marking.
1331         (opt_pass::operator new): New.
1332         (opt_pass::gt_ggc_mx): New.
1333         (opt_pass::gt_pch_nx): New.
1334         (opt_pass::gt_pch_nx_with_op): New.
1336 2013-08-20  David Malcolm  <dmalcolm@redhat.com>
1338         * Makefile.in (GTFILES): Add context.h.
1339         * context.c (gcc::context::operator new): New.
1340         (gcc::context::gt_ggc_mx): New.
1341         (gcc::context::gt_pch_nx): New.
1342         (gcc::context::gt_pch_nx): New.
1343         * context.h (gcc::context): Add GTY((user)) marking.
1344         (gcc::context::operator new): New.
1345         (gcc::context::gt_ggc_mx): New.
1346         (gcc::context::gt_pch_nx): New.
1347         (gcc::context::gt_pch_nx): New.
1348         (g): Add GTY marking.
1349         (gt_ggc_mx (gcc::context *)): New.
1350         (gt_pch_nx (gcc::context *)): New.
1351         (gt_pch_nx (gcc::context *ctxt, gt_pointer_operator op,
1352         void *cookie)): New.
1353         * gengtype.c (open_base_files) <ifiles>: Add context.h.
1355 2013-08-20  Alan Modra  <amodra@gmail.com>
1357         PR target/57865
1358         * config/rs6000/rs6000.c (rs6000_emit_prologue): Correct ool_adjust.
1359         (rs6000_emit_epilogue): Likewise.
1361 2013-08-19  Dehao Chen  <dehao@google.com>
1363         * value-prof.c (gimple_ic): Fix the bug of adding EH edge.
1365 2013-08-19  Peter Bergner  <bergner@vnet.ibm.com>
1366             Jakub Jelinek  <jakub@redhat.com>
1368         * builtins.def (BUILT_IN_FABSD32): New DFP ABS builtin.
1369         (BUILT_IN_FABSD64): Likewise.
1370         (BUILT_IN_FABSD128): Likewise.
1371         * builtins.c (expand_builtin): Add support for new DFP ABS builtins.
1372         (fold_builtin_1): Likewise.
1373         * config/rs6000/dfp.md (*negtd2_fpr): Handle non-overlapping
1374         destination and source operands.
1375         (*abstd2_fpr): Likewise.
1376         (*nabstd2_fpr): Likewise.
1378 2013-08-19  Richard Sandiford  <rdsandiford@googlemail.com>
1380         * config/mips/mips.c (mips_adjust_insn_length): Add checks for
1381         JUMP_P and INSN_P.
1383 2013-08-19  Aldy Hernandez  <aldyh@redhat.com>
1385         * doc/invoke.texi (-fcilkplus): Clarify that implementation is
1386         incomplete.
1388 2013-08-19  Alexander Ivchenko  <alexander.ivchenko@intel.com>
1390         * target.def (TARGET_LIBC_HAS_FUNCTION): New target hook.
1391         * builtins.c (default_libc_has_function): New.
1392         (gnu_libc_has_function): Ditto.
1393         (no_c99_libc_has_function): Ditto.
1394         (expand_builtin_cexpi): Using new target hook TARGET_LIBC_HAS_FUNCTION
1395         instead of TARGET_HAS_SINCOS and TARGET_C99_FUNCTIONS.
1396         (fold_builtin_sincos): Likewise.
1397         (fold_builtin_cexp): Likewise.
1398         * builtins.def (DEF_C94_BUILTIN): Likewise.
1399         (DEF_C99_BUILTIN): Likewise.
1400         (DEF_C99_C90RES_BUILTIN): Likewise.
1401         (DEF_C99_COMPL_BUILTIN): New define. Change all complex c99 builtin
1402         definitions to using this define.
1403         * config/darwin-protos.h (darwin_libc_has_function): New.
1404         * config/darwin.c (darwin_libc_has_function): Ditto.
1405         * config/alpha/linux.h: Remove TARGET_C99_FUNCTIONS and
1406         TARGET_HAS_SINCOS. Redefine TARGET_LIBC_HAS_FUNCTION.
1407         * config/darwin.h: Ditto.
1408         * config/elfos.h: Ditto.
1409         * config/freebsd.h: Ditto.
1410         * config/i386/cygming.h: Ditto.
1411         * config/i386/djgpp.h: Ditto.
1412         * config/i386/i386-interix.h: Ditto.
1413         * config/microblaze/microblaze.h: Ditto.
1414         * config/mmix/mmix.h: Ditto.
1415         * config/gnu-user.h: Ditto.
1416         * config/ia64/hpux.h: Ditto.
1417         * config/pa/pa-hpux.h: Ditto.
1418         * config/pdp11/pdp11.h: Ditto.
1419         * config/picochip/picochip.h: Ditto.
1420         * config/linux.h: Ditto.
1421         * config/netbsd.h: Ditto.
1422         * config/openbsd.h: Ditto.
1423         * config/rs6000/aix43.h: Ditto.
1424         * config/rs6000/aix51.h: Ditto.
1425         * config/rs6000/aix52.h: Ditto.
1426         * config/rs6000/aix53.h: Ditto.
1427         * config/rs6000/aix61.h: Ditto.
1428         * config/rs6000/darwin.h: Ditto.
1429         * config/rs6000/linux.h: Ditto.
1430         * config/rs6000/linux64.h: Ditto.
1431         * config/s390/tpf.h: Ditto.
1432         * config/sol2-10.h: Ditto.
1433         * config/sol2.h: Ditto.
1434         * config/vms/vms.h: Ditto.
1435         * config/vxworks.h: Ditto.
1436         * config/linux-android.c (linux_android_libc_has_function):
1437         New linux-specific implementation of TARGET_LIBC_HAS_FUNCTION.
1438         * config/linux-protos.h (linux_android_libc_has_function):
1439         New declaration.
1440         * config/i386/i386.c (ix86_libc_has_function): New.
1441         * config/i386/i386-protos.h
1442         (ix86_libc_has_function): New declaration.
1443         * config/i386/i386.md
1444         ("isinfxf2"): Change condition for TARGET_LIBC_HAS_FUNCTION.
1445         ("isinf<mode>2): Likewise.
1446         * convert.c (convert_to_integer): Using new target hook
1447         TARGET_LIBC_HAS_FUNCTION istead of TARGET_HAS_SINCOS and
1448         TARGET_C99_FUNCTIONS.
1449         * fortran/f95-lang.c (gfc_init_builtin_functions): Ditto.
1450         * tree-ssa-math-opts.c (execute_cse_sincos): Ditto.
1451         * coretypes.h (function_class): New enum for different
1452         classes of functions.
1453         * defaults.h: Remove TARGET_C99_FUNCTIONS and TARGET_HAS_SINCOS.
1454         * doc/tm.texi.in (TARGET_C99_FUNCTIONS): Remove documentation.
1455         (TARGET_HAS_SINCOS): Likewise.
1456         (TARGET_LIBC_HAS_FUNCTION): New.
1457         * doc/tm.texi: Regenerated.
1458         * targhooks.h (default_libc_has_function): New declaration.
1459         (no_c99_libc_has_function): Ditto.
1460         (gnu_libc_has_function): Ditto.
1461         * system.h: Add the poisoning of TARGET_C99_FUNCTIONS
1462         and TARGET_HAS_SINCOS.
1464 2013-08-18  Jan Hubicka  <jh@suse.cz>
1466         * Makeifle-in (ipa-devirt.o): New.
1467         (GTFILES): Add ipa-utils.h and ipa-devirt.c
1468         * cgraphunit.c (decide_is_symbol_needed): Do not care about virtuals.
1469         (analyze_functions): Look into possible targets of polymorphic call.
1470         * dumpfile.c (dump_files): Add type-inheritance dump.
1471         * dumpfile.h (TDI_inheritance): New.
1472         * ipa-devirt.c: New file.
1473         * ipa-utils.h (odr_type_d): Forward declare.
1474         (odr_type): New type.
1475         (build_type_inheritance_graph): Declare.
1476         (possible_polymorphic_call_targets): Declare and introduce inline
1477         variant when only edge is pased.
1478         (dump_possible_polymorphic_call_targets): Likewise.
1479         * timevar.def (TV_IPA_INHERITANCE, TV_IPA_VIRTUAL_CALL): New.
1480         * tree.c (type_in_anonymous_namespace_p): Break out from ...
1481         (types_same_for_odr): ... here.
1482         * tree.h (type_in_anonymous_namespace_p): Declare.
1484 2013-08-18  Jakub Jelinek  <jakub@redhat.com>
1486         PR tree-optimization/58006
1487         * tree-parloops.c (take_address_of): Don't ICE if get_name
1488         returns NULL.
1489         (eliminate_local_variables_stmt): Remove clobber stmts.
1491 2013-08-18  Eric Botcazou  <ebotcazou@adacore.com>
1493         * cgraphunit.c (handle_alias_pairs): Reset the alias flag after the
1494         error message is issued for an alias to undefined symbol.
1496 2013-08-18  Jan Hubicka  <jh@suse.cz>
1498         * cgraph.c (cgraph_create_indirect_edge): Discover
1499         polymorphic calls and record basic info into indirect_info.
1500         * gimple-fold.c (gimple_fold_call): When doing BINFO based
1501         devirtualization, ignore objc function calls.
1502         * ipa-cp.c (initialize_node_lattices): Be ready for polymorphic
1503         call with no parm index info.
1504         * ipa-prop.c (ipa_analyze_call_uses): Likewise.
1505         * tree.c (virtual_method_call_p): New function.
1506         * tree.h (virtual_method_call_p): Declare.
1508 2013-08-16  Jan Hubicka  <jh@suse.cz>
1510         PR middle-end/58179
1511         * tree.c (obj_type_ref_class): Do not ICE on non-method calls.
1513 2013-08-16  David Edelsohn  <dje.gcc@gmail.com>
1515         * config/rs6000/rs6000.md (rs6000_get_timebase_ppc32): Add length
1516         attribute.
1518 2013-08-16  David Malcolm  <dmalcolm@redhat.com>
1520         * gengtype.c (type_for_name): Add special-case support for
1521         locating types within the "gcc::" namespace.
1522         (open_base_files): Emit a "using namespace gcc" directive.
1524 2013-08-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
1526         PR target/58160
1527         * config/rs6000/predicates.md (fusion_gpr_mem_load): Allow the
1528         memory rtx to contain ZERO_EXTEND and SIGN_EXTEND.
1530         * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): Pass operands
1531         array instead of each individual operand as a separate argument.
1532         (emit_fusion_gpr_load): Likewise.
1533         (expand_fusion_gpr_load): Add new function declaration.
1535         * config/rs6000/rs6000.c (fusion_gpr_load_p): Change the calling
1536         signature to have the operands passed as an array, instead of as
1537         separate arguments.  Allow ZERO_EXTEND to be in the memory
1538         address, and also SIGN_EXTEND if -mpower8-fusion-sign.  Do not
1539         depend on the register live/dead flags when peepholes are run.
1540         (expand_fusion_gpr_load): New function to be called from the
1541         peephole2 pass, to change the register that addis sets to be the
1542         target register.
1543         (emit_fusion_gpr_load): Change the calling signature to have the
1544         operands passed as an array, instead of as separate arguments.
1545         Allow ZERO_EXTEND to be in the memory address, and also
1546         SIGN_EXTEND if -mpower8-fusion-sign.
1548         * config/rs6000/rs6000.md (UNSPEC_FUSION_GPR): Delete unused
1549         unspec enumeration.
1550         (power8 fusion peephole/peephole2): Rework the fusion peepholes to
1551         adjust the register addis loads up in the peephole2 pass.  Do not
1552         depend on the register live/dead state when the peephole pass is done.
1554 2013-08-16  David Malcolm  <dmalcolm@redhat.com>
1556         * gengtype.c (create_user_defined_type): Ensure that the kind
1557         is set to TYPE_USER_STRUCT, fixing a bug seen when an incomplete
1558         declaration is seen before the GTY((user)) marking.
1560 2013-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1562         PR target/58105
1563         * config/i386/i386.c (make_resolver_func): Set DECL_UNINLINABLE.
1565 2013-08-16  Jan Hubicka  <jh@suse.cz>
1567         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Add new
1568         arugment expected_type.
1569         (gimple_fold_call): Use it.
1570         * gimple.h (gimple_extract_devirt_binfo_from_cst): Update prototype.
1571         * ipa-cp.c (ipa_get_indirect_edge_target_1): Update.
1572         * ipa-prop.c (ipa_analyze_virtual_call_uses): Use obj_type_ref_class.
1573         (try_make_edge_direct_virtual_call): Likewise.
1574         * tree.c (obj_type_ref_class): New.
1575         * tree.h (obj_type_ref_class): Use it.
1577 2013-08-16  Gabriel Dos Reis  <gdr@integrable-solutions.net>
1579         * sched-vis.c (rtl_slim_pp_initialized): Remove.
1580         (rtl_slim_pp): Likewise.
1581         (init_rtl_slim_pretty_print): Likewise.
1582         (dump_value_slim):  Don't call it.  Use local pretty printer.
1583         (dump_insn_slim): Likewise.
1584         (dump_rtl_slim): Likewise.
1585         (str_pattern_slim): Likewise.
1586         * tree-mudflap.c (mf_varname_tree): Use local pretty printer.
1587         Simplify.
1589 2013-08-16  Jakub Jelinek  <jakub@redhat.com>
1591         PR tree-optimization/58164
1592         * gimple.c (walk_stmt_load_store_addr_ops): For visit_addr
1593         walk gimple_goto_dest of GIMPLE_GOTO.
1595         PR tree-optimization/58165
1596         * tree-call-cdce.c (shrink_wrap_one_built_in_call): If
1597         bi_call must be the last stmt in a bb, don't split_block, instead
1598         use fallthru edge from it and give up if there is none.
1599         Release conds vector when returning early.
1601 2013-08-14  Xinliang David Li  <davidxl@google.com>
1603         * config/i386/i386.c (ix86_option_override_internal):
1604         Remove unused variable and field.
1606 2013-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1608         PR target/57949
1609         * doc/invoke.texi: Add documentation of mcompat-align-parm option.
1610         * config/rs6000/rs6000.opt: Add mcompat-align-parm option.
1611         * config/rs6000/rs6000.c (rs6000_function_arg_boundary): For AIX
1612         and Linux, correct BLKmode alignment when 128-bit alignment is
1613         required and compatibility flag is not set.
1614         (rs6000_gimplify_va_arg): For AIX and Linux, honor specified alignment
1615         for zero-size arguments when compatibility flag is not set.
1617 2013-08-14  Jakub Jelinek  <jakub@redhat.com>
1619         PR tree-optimization/58145
1620         * tree-sra.c (build_ref_for_offset): If prev_base has
1621         TREE_THIS_VOLATILE or TREE_SIDE_EFFECTS, propagate it to MEM_REF.
1623 2013-08-14  Xinliang David Li  <davidxl@google.com>
1625         * config/i386/i386.c (ix86_option_override_internal):
1626         Fix uninitialized variable error.
1628 2013-08-14  Xinliang David Li  <davidxl@google.com>
1630         * config/i386/i386.opt: Define two new options.
1631         * config/i386/x86-tune.def: Add arch selector field in macros.
1632         * config/i386/i386.h: Adjust macro definition.
1633         * config/i386/i386.c (ix86_option_override_internal):
1634         Refactor the code.
1635         (parse_mtune_ctrl_str): New function.
1636         (set_ix86_tune_features): New function.
1637         (ix86_function_specific_restore): Call the new helper function.
1639 2013-08-14  Andrey Belevantsev  <abel@ispras.ru>
1641         PR rtl-optimization/57662
1642         * sel-sched.c (code_motion_process_successors): When the current insn
1643         is removed after the recursive traversal, break from the loop.
1644         Add comments and debug printouts.
1646 2013-08-14  Jakub Jelinek  <jakub@redhat.com>
1647             Alexandre Oliva  <aoliva@redhat.com>
1649         PR target/58067
1650         * config/i386/i386.c (ix86_delegitimize_address): For CM_MEDIUM_PIC
1651         and CM_LARGE_PIC ix86_cmodel fall thru into the -m32 code, handle
1652         there also UNSPEC_PLTOFF.
1654 2013-08-14  Marek Polacek  <polacek@redhat.com>
1656         * ipa-inline-analysis.c (add_clause): Avoid shifting integer
1657         NUM_CONDITIONS bit positions.
1659 2013-08-13  Cary Coutant  <ccoutant@google.com>
1661         * dwarf2out.c (CHECKSUM_BLOCK): New macro.
1662         (attr_checksum): Hash vector contents instead of pointer.
1663         (attr_checksum_ordered): Likewise.
1665 2013-08-13  Uros Bizjak  <ubizjak@gmail.com>
1667         * config/i386/sse.md (*sse2_maskmovdqu): Emit addr32 prefix
1668         when Pmode != word_mode.  Add length_address attribute.
1669         (sse3_monitor_<mode>): Merge from sse3_monitor and
1670         sse3_monitor64_<mode> insn patterns.  Emit addr32 prefix when
1671         Pmode != word_mode.  Update insn length attribute.
1672         * config/i386/i386.c (ix86_option_override_internal): Update
1673         ix86_gen_monitor selection for merged sse3_monitor insn.
1675 2013-08-13  Julian Brown  <julian@codesourcery.com>
1677         * config/rs6000/rs6000.c (rs6000_legitimize_reload_address): Don't
1678         perform invalid legitimization on greater-than-word-size modes for
1679         TARGET_E500_DOUBLE.
1681 2013-08-13  Vladimir Makarov  <vmakarov@redhat.com>
1683         * ira.c (setup_class_translate_array): Use aclass instead of cl
1684         for classes not fully covered by allocno classes.
1686 2013-08-13  Jakub Jelinek  <jakub@redhat.com>
1688         PR tree-optimization/57661
1689         * tree-inline.h (struct copy_body_data): Add blocks_to_copy field.
1690         * tree-inline.c (tree_function_versioning): Initialize it.
1691         (remap_gimple_stmt): Return GIMPLE_NOP for MEM_REF lhs clobber stmts
1692         if id->blocks_to_copy and MEM_REF's SSA_NAME is defined in a block
1693         that is not being copied.
1695         PR sanitizer/56417
1696         * asan.c (instrument_strlen_call): Fix typo in comment.
1697         Use char * type even for the lhs of POINTER_PLUS_EXPR.
1699 2013-08-13  Steve Ellcey  <sellcey@mips.com>
1701         * config/mips/mips.md (prefetch): Use lw instead of ld on
1702         loongson in 32bit mode.
1704 2013-08-13  Nick Clifton  <nickc@redhat.com>
1706         * config.gcc: (avr-linux): Allow for tmake_file not being empty.
1708 2013-08-13  Jan Hubicka  <jh@suse.cz>
1710         * cgraph.c (cgraph_turn_edge_to_speculative): Return newly
1711         introduced edge; fix typo in sanity check.
1712         (cgraph_resolve_speculation): Export; improve diagnostic.
1713         (cgraph_redirect_edge_call_stmt_to_callee): Better diagnostic; cancel
1714         speculation at type mismatch.
1715         * cgraph.h (cgraph_turn_edge_to_speculative): Update.
1716         (cgraph_resolve_speculation): Declare.
1717         (symtab_can_be_discarded): New function.
1718         * value-prof.c (gimple_ic_transform): Remove actual transform code.
1719         * ipa-inline-transform.c (speculation_removed): New global var.
1720         (clone_inlined_nodes): See if speculation can be removed.
1721         (inline_call): If speculations was removed, we growths may not match.
1722         * ipa-inline.c (can_inline_edge_p): Add DISREGARD_LIMITS parameter.
1723         (speculation_useful_p): New function.
1724         (resolve_noninline_speculation): New function.
1725         (inline_small_functions): Resolve useless speculations.
1726         * ipa-inline.h (speculation_useful_p): Declare
1727         * ipa.c (can_replace_by_local_alias): Simplify.
1728         (ipa_profile): Produce speculative calls in non-lto, too;
1729         add simple cost model; produce local aliases.
1731 2013-08-13  David Malcolm  <dmalcolm@redhat.com>
1733         * config/i386/t-i386 (i386.o): Rename stray PIPELINE_H to
1734         PASS_MANAGER_H.
1736 2013-08-12  Paolo Carlini  <paolo.carlini@oracle.com>
1738         * config/i386/i386.c (ix86_function_versions): Use error + inform.
1740 2013-08-12  Uros Bizjak  <ubizjak@gmail.com>
1742         * config/i386/i386.md (floatunssi<mode>2 expand): Use MODEF mode
1743         iterator instead of X87MODEF.
1745 2013-08-12  Perez Read  <netfirewall@gmail.com>
1747         PR target/58132
1748         * config/i386/i386.md (*movabs<mode>_1):  Add <ptrsize> PTR before
1749         operand 0 for intel asm alternative.
1750         (*movabs<mode>_2): Ditto for operand 1.
1752 2013-08-12  James Greenhalgh  <james.greenhalgh@arm.com>
1754         * config/aarch64/arm_none.h
1755         (vdup<bhsd>_lane_<su><8,16,32,64>): Fix macro call.
1757 2013-08-12  Nick Clifton  <nickc@redhat.com>
1759         * config.gcc (m32r-linux): Allow for tmake_file not being empty.
1761 2013-08-12  Yuri Rumyantsev  <ysrumyan@gmail.com>
1763         * config/i386/i386.md (floatunssi<mode>2 expand): Add new
1764         expand for QI/HImode operand to produce more effictive code for
1765         unsigned char(short) --> float(double) conversion.
1767 2013-08-12  Alexander Monakov  <amonakov@ispras.ru>
1769         * doc/invoke.texi: Mention that -ftls-model does not force the final
1770         model.
1772 2013-08-12  Marek Polacek  <polacek@redhat.com>
1773             Marc Glisse  <marc.glisse@inria.fr>
1775         PR tree-optimization/57980
1776         * tree-tailcall.c (process_assignment): Call build_minus_one_cst
1777         when creating -1 constant.
1779 2013-08-10  Jan Hubicka  <jh@suse.cz>
1781         Workaround binutils PR14342.
1782         * tree-profile.c (init_ic_make_global_vars): Add LTO path.
1783         (gimple_init_edge_profiler): Likewise.
1784         (gimple_gen_ic_func_profiler): Likewise.
1786 2013-08-09  Jan Hubicka  <jh@suse.cz>
1788         * cgraph.c (cgraph_create_edge_1): Clear speculative flag.
1790 2013-08-09  Xinliang David Li  <davidxl@google.com>
1792         * config/i386/stringop.def: New file.
1793         * config/i386/stringop.opt: New file.
1794         * config/i386/i386-opts.h: Include stringopt.def.
1795         * config/i386/i386.opt: Include stringopt.opt.
1796         * config/i386/i386.c (ix86_option_override_internal):
1797         Override default size based stringop inline strategies with options.
1798         * config/i386/i386.c (ix86_parse_stringop_strategy_string):
1799         New function.
1801 2013-08-09  Jan Hubicka  <jh@suse.cz>
1803         * ipa-ref.c (ipa_clear_stmts_in_references): Clear lto_stmt_uid, too.
1805 2013-08-09  Jan Hubicka  <jh@suse.cz>
1807         * cgraph.c (cgraph_resolve_speculation): Cut frequency to
1808         CGRAPH_FREQ_MAX.
1809         (dump_cgraph_node): Dump profile-id.
1810         * cgraph.h (cgraph_indirect_call_info): Add common_target_id
1811         and common_target_probability.
1812         * lto-cgraph.c (lto_output_edge): Stream common targets.
1813         (lto_output_node): Stream profile ids.
1814         (input_node): Stream profile ids.
1815         (input_edge): Stream common targets.
1816         * lto-streamer-in.c (fixup_call_stmt_edges_1): Fix formatting.
1817         * ipa.c: Include value-prof.h
1818         (ipa_profile_generate_summary): Turn indirect call statement histograms
1819         into common targets.
1820         (ipa_profile): Turn common targets into speculative edges.
1822 2013-08-09  Jan Hubicka  <jh@suse.cz>
1824         * cgraph.h (cgraph_node): Add profile_id.
1825         * value-prof.c (cgraph_node_map): Turn into pointer_map.
1826         (init_node_map): Rewrite to handle hashes increas of incremental IDs.
1827         (del_node_map): Update.
1828         (find_func_by_funcdef_no): Replace by ...
1829         (find_func_by_profile_id): ... this one.
1830         (gimple_ic_transform): Do not remove useful histograms when
1831         speculation is not done; dump info when indirect call removal
1832         can happen at LTO.
1833         * value-prof.h (find_func_by_profile_id, gimple_ic): Declare.
1834         * gcov-io.h (__gcov_indirect_call_profiler): Replace by ...
1835         (__gcov_indirect_call_profiler_v2): .. this one.
1836         * profile.h (init_node_map): Update.
1837         * coverage.c (coverage_compute_profile_id): New function.
1838         * coverage.h (coverage_compute_profile_id): Declare.
1839         * tree-profile.c (init_ic_make_global_vars): Make
1840         __gcov_indirect_call_callee and  __gcov_indirect_call_counters global.
1841         (gimple_init_edge_profiler): Update prototype of
1842         __gcov_indirect_call_profiler.
1843         (gimple_gen_ic_func_profiler): Simplify.
1844         (tree_profiling): Use init_node_map
1846 2013-08-09  Jan Hubicka  <jh@suse.cz>
1848         * cgraphbuild.c (cgraph_rebuild_references): Rebuild only
1849         non-speculative refs.
1850         * cgraph.c (cgraph_update_edge_in_call_site_hash): New function.
1851         (cgraph_add_edge_to_call_site_hash): Deal with speculative calls.
1852         (cgraph_set_call_stmt): Likewise.
1853         (cgraph_create_edge_1): Fix release checking compilatoin;
1854         clear lto_stmt_uid.
1855         (cgraph_free_edge): Free indirect info.
1856         (cgraph_turn_edge_to_speculative): New function.
1857         (cgraph_speculative_call_info): New function.
1858         (cgraph_make_edge_direct): Return direct edge; handle speculation.
1859         (cgraph_redirect_edge_call_stmt_to_callee): Expand speculative edges.
1860         (dump_cgraph_node): Dump speculation.
1861         (verify_edge_count_and_frequency): Accept speculative edges.
1862         (verify_edge_corresponds_to_fndecl): Handle partitioned cgraph.
1863         (verify_cgraph_node): Handle speculation.
1864         * cgraph.h (cgraph_edge): Add SPECULATIVE flag.
1865         (cgraph_set_call_stmt): Update prototype.
1866         (cgraph_make_edge_direct): Update prototype.
1867         (cgraph_speculative_call_info): Declare.
1868         * ipa-cp.c (ipcp_discover_new_direct_edges): Be ready for edge
1869         to change; update call of ipa_find_references.
1870         * ipa-ref.c (ipa_record_reference): Fix return value; clear
1871         lto_stmt_uid and speculative flags.
1872         (ipa_dump_references): Dump speculation.
1873         (ipa_clone_references): Clone speculative flag.
1874         (ipa_clone_referring): Likewise.
1875         (ipa_clone_ref): New function.
1876         (ipa_find_reference): Look into lto_stmt_uids
1877         (ipa_clear_stmts_in_references): Do not clear speculative calls.
1878         * ipa-ref.h (ipa_ref): Add lto_stmt_uid and speculative flags.
1879         (ipa_find_reference): Update declaration.
1880         (ipa_clone_ref): Declare.
1881         * lto-cgraph.c (lto_output_edge): Make lto_stmt_uids start from 0;
1882         stream speculative flag.
1883         (lto_output_ref): Stream statements uids and speculation.
1884         (input_ref): Likewise.
1885         (input_edge): Stream speuclation.
1886         * cgraphclones.c (cgraph_clone_edge): Clone speculation.
1887         (cgraph_set_call_stmt_including_clones): Handle speculation.
1888         * ipa-inline.c (heap_edge_removal_hook): New function.
1889         (inline_small_functions): Register it.
1890         * lto-streamer-in.c (fixup_call_stmt_edges_1): Bounds checking;
1891         also initialize refs.
1892         * ipa-prop.c (ipa_make_edge_direct_to_target): Be ready for
1893         edge to change.
1894         (try_make_edge_direct_simple_call): Likewise.
1895         (try_make_edge_direct_simple_call): Likewise.
1896         (update_indirect_edges_after_inlining): Likewise.
1897         (remove_described_reference): Look proper lto_stmt_uid.
1898         (propagate_controlled_uses): Likewise.
1899         (propagate_controlled_uses): Liekwise.
1900         * tree-inline.c (copy_bb): Copy speculative edges.
1901         (redirect_all_calls): New function.
1902         (copy_cfg_body): Do redirection after loop info is updated.
1903         (delete_unreachable_blocks_update_callgraph): Updadte speculation.
1905 2013-08-09  Jan Hubicka  <jh@suse.cz>
1907         * lto-streamer-out.c (output_function): Renumber PHIs.
1908         * lto-streamer-in.c (input_function): Likewise.
1910 2013-08-09  James Greenhalgh  <james.greenhalgh@arm.com>
1912         * config/aarch64/aarch64-simd-builtins.def (get_lane_signed): Remove.
1913         (get_lane_unsigned): Likewise.
1914         (dup_lane_scalar): Likewise.
1915         (get_lane): enable for VALL.
1916         * config/aarch64/aarch64-simd.md
1917         (aarch64_dup_lane_scalar<mode>): Remove.
1918         (aarch64_get_lane_signed<mode>): Likewise.
1919         (aarch64_get_lane_unsigned<mode>): Likewise.
1920         (aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): New.
1921         (aarch64_get_lane_zero_extendsi<mode>): Likewise.
1922         (aarch64_get_lane<mode>): Enable for all vector modes.
1923         (aarch64_get_lanedi): Remove misleading constraints.
1924         * config/aarch64/arm_neon.h
1925         (__aarch64_vget_lane_any): Define.
1926         (__aarch64_vget<q>_lane_<fpsu><8,16,32,64>): Likewise.
1927         (vget<q>_lane_<fpsu><8,16,32,64>): Use __aarch64_vget_lane macros.
1928         (vdup<bhsd>_lane_<su><8,16,32,64>): Likewise.
1929         * config/aarch64/iterators.md (VDQQH): New.
1930         (VDQQHS): Likewise.
1931         (vwcore): Likewise.
1933 2013-08-09  Eric Botcazou  <ebotcazou@adacore.com>
1935         * configure.ac: Add GAS check for LEON instructions on SPARC.
1936         * configure: Regenerate.
1937         * config.in: Likewise.
1938         * config.gcc (with_cpu): Remove sparc-leon*-* and deal with LEON in the
1939         sparc*-*-* block.
1940         * config/sparc/sparc.opt (LEON, LEON3): New masks.
1941         * config/sparc/sparc.h (ASM_CPU32_DEFAULT_SPEC): Set to AS_LEON_FLAG
1942         for LEON or LEON3.
1943         (ASM_CPU_SPEC): Pass AS_LEON_FLAG if -mcpu=leon or -mcpu=leon3.
1944         (AS_LEON_FLAG): New macro.
1945         * config/sparc/sparc.c (sparc_option_override): Set MASK_LEON for leon
1946         and MASK_LEON3 for leon3 and unset them if HAVE_AS_LEON is not defined.
1947         Deal with LEON and LEON3 for the memory model.
1948         * config/sparc/sync.md (atomic_compare_and_swap<mode>): Enable if LEON3
1949         (atomic_compare_and_swap<mode>_1): Likewise.
1950         (*atomic_compare_and_swap<mode>_1): Likewise.
1952 2013-08-09  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1954         * config/arm/neon.md (vcond): Fix floating-point vector
1955         comparisons against 0.
1957 2013-08-08  Vladimir Makarov  <vmakarov@redhat.com>
1959         * lra-constraints.c (emit_spill_move): Remove assert.
1960         (process_alt_operands): Add more debugging
1961         output.  Increase reject for spilling into memory.  Decrease
1962         reject for reloading scratch.
1963         (split_reg): Use HARD_REGNO_CALLER_SAVE_MODE.
1965 2013-08-08  Steve Ellcey  <sellcey@mips.com>
1967         * config/mips/mti-linux.h (SYSROOT_SUFFIX_SPEC): Add nan2008.
1968         * config/mips/t-mti-elf (MULTILIB_OPTIONS): Make mips16 and
1969         micromips incompatible.  Add nan2008.
1970         (MULTILIB_DIRNAMES): Add nan2008.
1971         (MULTILIB_EXCEPTIONS): Remove mips16/micromips entry.
1972         * config/mips/t-mti-linux (MULTILIB_OPTIONS): Make mips16
1973         and micromips incompatible.  Add nan2008.
1974         (MULTILIB_DIRNAMES): Add nan2008.
1975         (MULTILIB_EXCEPTIONS): Remove mips16/micromips entry.
1977 2013-08-08  Richard Sandiford  <rdsandiford@googlemail.com>
1979         PR rtl-optimization/58079
1980         * combine.c (combine_simplify_rtx): Avoid using SUBST if
1981         simplify_comparison has widened a comparison with an integer.
1983 2013-08-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1985         * config/arm/neon.md (movmisalign<mode>): Disable when we
1986         don't allow unaligned accesses.
1987         (*movmisalign<mode>_neon_store): Likewise.
1988         (*movmisalign<mode>_neon_load): Likewise.
1989         (*movmisalign<mode>_neon_store): Likewise.
1990         (*movmisalign<mode>_neon_load): Likewise.
1992 2013-08-08  Jan Hubicka  <jh@suse.cz>
1994         * cgraphbuild.c (build_cgraph_edges): Do not walk into debugs.
1995         (make_pass_rebuild_cgraph_edges): Also clear references.
1996         * cgraph.c (verify_cgraph_node): Add basic ipa-ref verifier.
1997         * ipa-inline-transform.c (inline_transform): Remove all references
1998         after inlining.
1999         * cgraphunit.c (expand_function): Remove all references after
2000         expansion.
2001         * ipa-ref.c (ipa_ref_has_aliases_p): Fix formatting.
2002         (ipa_find_reference): Rewrite to iterator.
2003         (remove_stmt_references): Likewise.
2004         (ipa_clear_stmts_in_references): New function.
2005         * ipa-ref.h (ipa_clear_stmts_in_references): Declare.
2006         * cgraphclones.c (cgraph_materialize_all_clones): Remove or
2007         clear references.
2008         * ipa-split.c (split_function): Remove references in split function.
2010 2013-08-08  Richard Earnshaw  <rearnsha@arm.com>
2012         PR target/57431
2013         * config/arm/arm/neon.md (neon_vld1_dupdi): New expand pattern.
2014         (neon_vld1_dup<mode> VD iterator): Iterate over VD not VDX.
2016 2013-08-08  Richard Earnshaw  <rearnsha@arm.com>
2018         PR target/56979
2019         * config/arm/arm.c (aapcs_vfp_allocate): Decompose the argument if the
2020         suggested mode for the assignment isn't compatible with the
2021         registers required.
2023 2013-08-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2025         PR target/58065
2026         * config/arm/arm.h (MALLOC_ABI_ALIGNMENT): Define.
2028 2013-08-07  Xinliang David Li  <davidxl@google.com>
2030         * config/i386/i386.opt: New option -mtune-ctrl=.
2031         * config/i386/x86-tune.def: New file.
2032         * config/i386/i386.h: include x86-tune.def.
2033         * config/i386/i386.c (ix86_option_override_internal):
2034         Parsing -mtune-ctrl= option and set tune features.
2036 2013-08-07  Oleg Endo  <olegendo@gcc.gnu.org>
2038         PR other/12081
2039         * config/rs6000/rs6000.c (gen_2arg_fn_t): Remove typedef.
2040         (rs6000_emit_swdiv, rs6000_emit_swrsqrt): Don't cast result of GEN_FCN
2041         to gen_2arg_fn_t.
2043 2013-08-07  Eric Botcazou  <ebotcazou@adacore.com>
2045         * rtl.h (update_alignments): Declare.
2046         * final.c (grow_label_align): New function extracted from...
2047         (shorten_branches): ...here.  Call it.
2048         (update_alignments): New function.
2049         * reorg.c (sibling_labels): New variable.
2050         (get_label_before): Add SIBLING parameter.  If it is non-zero, push
2051         the new label along with it onto the sibling_labels vector.
2052         (fill_simple_delay_slots): Adjust call to get_label_before.
2053         (fill_slots_from_thread): Likewise.
2054         (relax_delay_slots): Likewise.
2055         (make_return_insns): Likewise.
2056         (dbr_schedule): Invoke update_alignment on the sibling_labels vector.
2058 2013-08-07  Eric Botcazou  <ebotcazou@adacore.com>
2060         * diagnostic.c (diagnostic_classify_diagnostic): Accept zero index and
2061         document its semantics.
2062         (diagnostic_report_diagnostic): Adjust accordingly.
2064 2013-08-07  David Malcolm  <dmalcolm@redhat.com>
2066         * config/sparc/sparc.c (insert_pass_work_around_errata): Move into...
2067         (sparc_option_override): ...and port to new C++ pass API.
2068         * config/sparc/t-sparc (sparc.o): Add dep on CONTEXT_H
2070 2013-08-07  Peter Bergner  <bergner@vnet.ibm.com>
2072         * config/rs6000/rs6000.c (htm_expand_builtin) <case 0>: Remove.
2074 2013-08-06  Caroline Tice  <cmtice@google.com>
2076         * gcc.c (VTABLE_VERIFICATION_SPEC): New definition.
2077         (LINK_COMMAND_SPEC): Add VTABLE_VERIFICATION_SPEC.
2078         * tree-pass.h: Add pass_vtable_verify.
2079         * varasm.c (assemble_variable): Add code to properly set the comdat
2080         section and name for the .vtable_map_vars section.
2081         (assemble_vtyv_preinit_initializer): New function.
2082         (default_sectin_type_flags):  Make sure .vtable_map_vars section has
2083         LINK_ONCE flag.
2084         * output.h: Add function decl for assemble_vtv_preinit_initializer.
2085         * vtable-verify.c: New file.
2086         * vtable-verify.h: New file.
2087         * flag-types.h (enum vtv_priority): Defintions for flag_vtable_verify
2088         initialiation levels.
2089         * timevar.def (TV_VTABLE_VERIFICATION): New definition.
2090         * passes.def: Insert pass_vtable_verify.
2091         * aclocal.m4: Reorder includes.
2092         * doc/invoke.texi:  Document the -fvtable-verify=, -fvtv-debug, and
2093         -fvtv-counts options.
2094         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add vtv_start*.o,
2095         as appropriate, if -fvtable-verify=... is used.
2096         (GNU_USER_TARGET_ENDFILE_SPEC): Add vtv_end*.o as appropriate, if
2097         -fvtable-verify=... is used.
2098         * Makefile.in (OBJS):  Add vtable-verify.o to list.
2099         (vtable-verify.o): Add new build rule.
2100         (GTFILES): Add vtable-verify.c to list.
2101         * common.opt (fvtable-verify=): New flag.
2102         (vtv_priority): Values for fvtable-verify= flag.
2103         (fvtv-counts): New flag.
2104         (fvtv-debug): New flag.
2105         * tree.h (save_vtable_map_decl): New extern function decl.
2107 2013-08-07  David Malcolm  <dmalcolm@redhat.com>
2109         * config/rl78/rl78.c (rl78_devirt_pass): Convert from a struct to...
2110         (pass_rl78_devirt): ...new subclass of rtl_opt_pass along with...
2111         (pass_data_rl78_devirt): ...new pass_data instance and...
2112         (make_pass_rl78_devirt): ...new function.
2113         (rl78_asm_file_start): Port pass registration to new C++ API.
2115 2013-08-07  David Malcolm  <dmalcolm@redhat.com>
2117         * coretypes.h (rtl_opt_pass): Add.
2118         (gcc::context): Add.
2119         * config/epiphany/epiphany.c (pass_mode_switch_use): New.
2120         (epiphany_init): Port to new C++ pass API.
2121         (epiphany_optimize_mode_switching): Likewise.
2122         * pass_manager.h (pass_manager::get_pass_split_all_insns): New.
2123         (pass_manager::get_pass_mode_switching): New.
2124         (pass_manager::get_pass_peephole2): New.
2125         * mode-switching.c (pass_mode_switching): Add clone method.
2126         * recog.c (pass_peephole2): Add clone method.
2127         (pass_split_all_insns): Add clone method.
2129 2013-08-06  David Malcolm  <dmalcolm@redhat.com>
2131         * config/mips/mips.c (insert_pass_mips_machine_reorg2): Move into...
2132         (mips_option_override): ...here, porting to new C++ API for passes.
2134 2013-08-06  Jan Hubicka  <jh@suse.cz>
2136         * cgraph.c (cgraph_get_body): New function based on lto.c
2137         implementation.
2138         * cgraph.h (cgraph_get_body): Declare.
2139         * cgraphclones.c (cgraph_create_virtual_clone): Commonize WPA and
2140         LTO paths.
2141         * cgraphunit.c (expand_function): Get body prior expanding.
2142         * ipa.c (function_and_variable_visibility): Use gimple_has_body_p test.
2143         * lto-cgraph.c (lto_output_node): Do not stream bodies we don't
2144         really need.
2145         * passes.c (do_per_function_toporder): Get body.
2146         * tree-inline.c (expand_call_inline): Get body prior inlining it.
2147         * tree-ssa-structalias.c (ipa_pta_execute): Get body; skip clones.
2149 2013-08-06  Martin Jambor  <mjambor@suse.cz>
2151         PR fortran/57987
2152         * cgraphunit.c (cgraph_finalize_function): Assert that nested function
2153         is not re-finalized.  Rename second parameter to no_collect.
2155 2013-08-06  Martin Jambor  <mjambor@suse.cz>
2157         PR middle-end/58041
2158         * gimple-ssa-strength-reduction.c (replace_ref): Make sure built
2159         MEM_REF has proper alignment information.
2161 2013-08-05  Oleg Endo  <olegendo@gcc.gnu.org>
2163         PR other/12081
2164         * recog.h (rtx (*insn_gen_fn) (rtx, ...)): Replace typedef with new
2165         class insn_gen_fn.
2166         * expr.c (move_by_pieces_1, store_by_pieces_2): Replace argument
2167         rtx (*) (rtx, ...) with insn_gen_fn.
2168         * genoutput.c (output_insn_data): Cast gen_? function pointers to
2169         insn_gen_fn::stored_funcptr.  Add initializer braces.
2171 2013-08-05  David Malcolm  <dmalcolm@redhat.com>
2173         Rewrite how instances of passes are cloned to remove assumptions
2174         about their sizes (thus allowing pass subclasses to have
2175         additional data fields, albeit non-GC-managed ones at this point).
2177         * passes.c (make_pass_instance): Now that passes have clone
2178         methods, rewrite this function to eliminate XNEW and memcpy
2179         calls that used hardcoded sizes.  Since this function no longer
2180         creates pass instances, rename it to...
2181         (add_pass_instance): ...this.  Document the old way that passes were
2182         numbered and flagged, and rework this function to continue using it.
2183         (next_pass_1): Add an initial_pass argument for use by
2184         add_pass_instance.
2185         (position_pass): When adding multiple instances of a pass, use
2186         the pass's clone method, rather than relying on the XNEW/memcpy
2187         within the former make_pass_instance (now add_pass_instance).
2188         (pass_manager::pass_manager): When invoking next_pass_1, also supply
2189         the initial instance of the current pass within the pass manager.
2191 2013-08-05  David Malcolm  <dmalcolm@redhat.com>
2193         This is the automated part of the conversion of passes from C
2194         structs to C++ classes.
2196         Patch autogenerated by refactor_passes.py from
2197         https://github.com/davidmalcolm/gcc-refactoring-scripts
2198         revision 03fe39476a4c4ea450b49e087cfa817b5f92021e
2200         * asan.c (pass_asan): Convert from a global struct to a subclass of
2201         gimple_opt_pass along with...
2202         (pass_data_asan): ...new pass_data instance and...
2203         (make_pass_asan): ...new function.
2204         (pass_asan_O0): Convert from a global struct to a subclass of
2205         gimple_opt_pass along with...
2206         (pass_data_asan_O0): ...new pass_data instance and...
2207         (make_pass_asan_O0): ...new function.
2208         * auto-inc-dec.c (pass_inc_dec): Convert from a global struct to a
2209         subclass of rtl_opt_pass along with...
2210         (pass_data_inc_dec): ...new pass_data instance and...
2211         (make_pass_inc_dec): ...new function.
2212         * bb-reorder.c (pass_reorder_blocks): Convert from a global struct to
2213         a subclass of rtl_opt_pass along with...
2214         (pass_data_reorder_blocks): ...new pass_data instance and...
2215         (make_pass_reorder_blocks): ...new function.
2216         (pass_duplicate_computed_gotos): Convert from a global struct to a
2217         subclass of rtl_opt_pass along with...
2218         (pass_data_duplicate_computed_gotos): ...new pass_data instance and...
2219         (make_pass_duplicate_computed_gotos): ...new function.
2220         (pass_partition_blocks): Convert from a global struct to a subclass of
2221         rtl_opt_pass along with...
2222         (pass_data_partition_blocks): ...new pass_data instance and...
2223         (make_pass_partition_blocks): ...new function.
2224         * bt-load.c (pass_branch_target_load_optimize1): Convert from a global
2225         struct to a subclass of rtl_opt_pass along with...
2226         (pass_data_branch_target_load_optimize1): ...new pass_data instance
2227         and...
2228         (make_pass_branch_target_load_optimize1): ...new function.
2229         (pass_branch_target_load_optimize2): Convert from a global struct to a
2230         subclass of rtl_opt_pass along with...
2231         (pass_data_branch_target_load_optimize2): ...new pass_data instance
2232         and...
2233         (make_pass_branch_target_load_optimize2): ...new function.
2234         * cfgcleanup.c (pass_jump): Convert from a global struct to a subclass
2235         of rtl_opt_pass along with...
2236         (pass_data_jump): ...new pass_data instance and...
2237         (make_pass_jump): ...new function.
2238         (pass_jump2): Convert from a global struct to a subclass of
2239         rtl_opt_pass along with...
2240         (pass_data_jump2): ...new pass_data instance and...
2241         (make_pass_jump2): ...new function.
2242         * cfgexpand.c (pass_expand): Convert from a global struct to a
2243         subclass of rtl_opt_pass along with...
2244         (pass_data_expand): ...new pass_data instance and...
2245         (make_pass_expand): ...new function.
2246         * cfgrtl.c (pass_free_cfg): Convert from a global struct to a subclass
2247         of rtl_opt_pass along with...
2248         (pass_data_free_cfg): ...new pass_data instance and...
2249         (make_pass_free_cfg): ...new function.
2250         (pass_into_cfg_layout_mode): Convert from a global struct to a
2251         subclass of rtl_opt_pass along with...
2252         (pass_data_into_cfg_layout_mode): ...new pass_data instance and...
2253         (make_pass_into_cfg_layout_mode): ...new function.
2254         (pass_outof_cfg_layout_mode): Convert from a global struct to a
2255         subclass of rtl_opt_pass along with...
2256         (pass_data_outof_cfg_layout_mode): ...new pass_data instance and...
2257         (make_pass_outof_cfg_layout_mode): ...new function.
2258         * cgraphbuild.c (pass_build_cgraph_edges): Convert from a global
2259         struct to a subclass of gimple_opt_pass along with...
2260         (pass_data_build_cgraph_edges): ...new pass_data instance and...
2261         (make_pass_build_cgraph_edges): ...new function.
2262         (pass_rebuild_cgraph_edges): Convert from a global struct to a
2263         subclass of gimple_opt_pass along with...
2264         (pass_data_rebuild_cgraph_edges): ...new pass_data instance and...
2265         (make_pass_rebuild_cgraph_edges): ...new function.
2266         (pass_remove_cgraph_callee_edges): Convert from a global struct to a
2267         subclass of gimple_opt_pass along with...
2268         (pass_data_remove_cgraph_callee_edges): ...new pass_data instance
2269         and...
2270         (make_pass_remove_cgraph_callee_edges): ...new function.
2271         * combine-stack-adj.c (pass_stack_adjustments): Convert from a global
2272         struct to a subclass of rtl_opt_pass along with...
2273         (pass_data_stack_adjustments): ...new pass_data instance and...
2274         (make_pass_stack_adjustments): ...new function.
2275         * combine.c (pass_combine): Convert from a global struct to a subclass
2276         of rtl_opt_pass along with...
2277         (pass_data_combine): ...new pass_data instance and...
2278         (make_pass_combine): ...new function.
2279         * compare-elim.c (pass_compare_elim_after_reload): Convert from a
2280         global struct to a subclass of rtl_opt_pass along with...
2281         (pass_data_compare_elim_after_reload): ...new pass_data instance
2282         and...
2283         (make_pass_compare_elim_after_reload): ...new function.
2284         * cprop.c (pass_rtl_cprop): Convert from a global struct to a subclass
2285         of rtl_opt_pass along with...
2286         (pass_data_rtl_cprop): ...new pass_data instance and...
2287         (make_pass_rtl_cprop): ...new function.
2288         * cse.c (pass_cse): Convert from a global struct to a subclass of
2289         rtl_opt_pass along with...
2290         (pass_data_cse): ...new pass_data instance and...
2291         (make_pass_cse): ...new function.
2292         (pass_cse2): Convert from a global struct to a subclass of
2293         rtl_opt_pass along with...
2294         (pass_data_cse2): ...new pass_data instance and...
2295         (make_pass_cse2): ...new function.
2296         (pass_cse_after_global_opts): Convert from a global struct to a
2297         subclass of rtl_opt_pass along with...
2298         (pass_data_cse_after_global_opts): ...new pass_data instance and...
2299         (make_pass_cse_after_global_opts): ...new function.
2300         * dce.c (pass_ud_rtl_dce): Convert from a global struct to a subclass
2301         of rtl_opt_pass along with...
2302         (pass_data_ud_rtl_dce): ...new pass_data instance and...
2303         (make_pass_ud_rtl_dce): ...new function.
2304         (pass_fast_rtl_dce): Convert from a global struct to a subclass of
2305         rtl_opt_pass along with...
2306         (pass_data_fast_rtl_dce): ...new pass_data instance and...
2307         (make_pass_fast_rtl_dce): ...new function.
2308         * df-core.c (pass_df_initialize_opt): Convert from a global struct to
2309         a subclass of rtl_opt_pass along with...
2310         (pass_data_df_initialize_opt): ...new pass_data instance and...
2311         (make_pass_df_initialize_opt): ...new function.
2312         (pass_df_initialize_no_opt): Convert from a global struct to a
2313         subclass of rtl_opt_pass along with...
2314         (pass_data_df_initialize_no_opt): ...new pass_data instance and...
2315         (make_pass_df_initialize_no_opt): ...new function.
2316         (pass_df_finish): Convert from a global struct to a subclass of
2317         rtl_opt_pass along with...
2318         (pass_data_df_finish): ...new pass_data instance and...
2319         (make_pass_df_finish): ...new function.
2320         * dse.c (pass_rtl_dse1): Convert from a global struct to a subclass of
2321         rtl_opt_pass along with...
2322         (pass_data_rtl_dse1): ...new pass_data instance and...
2323         (make_pass_rtl_dse1): ...new function.
2324         (pass_rtl_dse2): Convert from a global struct to a subclass of
2325         rtl_opt_pass along with...
2326         (pass_data_rtl_dse2): ...new pass_data instance and...
2327         (make_pass_rtl_dse2): ...new function.
2328         * dwarf2cfi.c (pass_dwarf2_frame): Convert from a global struct to a
2329         subclass of rtl_opt_pass along with...
2330         (pass_data_dwarf2_frame): ...new pass_data instance and...
2331         (make_pass_dwarf2_frame): ...new function.
2332         * except.c (pass_set_nothrow_function_flags): Convert from a global
2333         struct to a subclass of rtl_opt_pass along with...
2334         (pass_data_set_nothrow_function_flags): ...new pass_data instance
2335         and...
2336         (make_pass_set_nothrow_function_flags): ...new function.
2337         (pass_convert_to_eh_region_ranges): Convert from a global struct to a
2338         subclass of rtl_opt_pass along with...
2339         (pass_data_convert_to_eh_region_ranges): ...new pass_data instance
2340         and...
2341         (make_pass_convert_to_eh_region_ranges): ...new function.
2342         * final.c (pass_compute_alignments): Convert from a global struct to a
2343         subclass of rtl_opt_pass along with...
2344         (pass_data_compute_alignments): ...new pass_data instance and...
2345         (make_pass_compute_alignments): ...new function.
2346         (pass_final): Convert from a global struct to a subclass of
2347         rtl_opt_pass along with...
2348         (pass_data_final): ...new pass_data instance and...
2349         (make_pass_final): ...new function.
2350         (pass_shorten_branches): Convert from a global struct to a subclass of
2351         rtl_opt_pass along with...
2352         (pass_data_shorten_branches): ...new pass_data instance and...
2353         (make_pass_shorten_branches): ...new function.
2354         (pass_clean_state): Convert from a global struct to a subclass of
2355         rtl_opt_pass along with...
2356         (pass_data_clean_state): ...new pass_data instance and...
2357         (make_pass_clean_state): ...new function.
2358         * function.c (pass_instantiate_virtual_regs): Convert from a global
2359         struct to a subclass of rtl_opt_pass along with...
2360         (pass_data_instantiate_virtual_regs): ...new pass_data instance and...
2361         (make_pass_instantiate_virtual_regs): ...new function.
2362         (pass_leaf_regs): Convert from a global struct to a subclass of
2363         rtl_opt_pass along with...
2364         (pass_data_leaf_regs): ...new pass_data instance and...
2365         (make_pass_leaf_regs): ...new function.
2366         (pass_thread_prologue_and_epilogue): Convert from a global struct to a
2367         subclass of rtl_opt_pass along with...
2368         (pass_data_thread_prologue_and_epilogue): ...new pass_data instance
2369         and...
2370         (make_pass_thread_prologue_and_epilogue): ...new function.
2371         (pass_match_asm_constraints): Convert from a global struct to a
2372         subclass of rtl_opt_pass along with...
2373         (pass_data_match_asm_constraints): ...new pass_data instance and...
2374         (make_pass_match_asm_constraints): ...new function.
2375         * fwprop.c (pass_rtl_fwprop): Convert from a global struct to a
2376         subclass of rtl_opt_pass along with...
2377         (pass_data_rtl_fwprop): ...new pass_data instance and...
2378         (make_pass_rtl_fwprop): ...new function.
2379         (pass_rtl_fwprop_addr): Convert from a global struct to a subclass of
2380         rtl_opt_pass along with...
2381         (pass_data_rtl_fwprop_addr): ...new pass_data instance and...
2382         (make_pass_rtl_fwprop_addr): ...new function.
2383         * gcse.c (pass_rtl_pre): Convert from a global struct to a subclass of
2384         rtl_opt_pass along with...
2385         (pass_data_rtl_pre): ...new pass_data instance and...
2386         (make_pass_rtl_pre): ...new function.
2387         (pass_rtl_hoist): Convert from a global struct to a subclass of
2388         rtl_opt_pass along with...
2389         (pass_data_rtl_hoist): ...new pass_data instance and...
2390         (make_pass_rtl_hoist): ...new function.
2391         * gimple-low.c (pass_lower_cf): Convert from a global struct to a
2392         subclass of gimple_opt_pass along with...
2393         (pass_data_lower_cf): ...new pass_data instance and...
2394         (make_pass_lower_cf): ...new function.
2395         * gimple-ssa-strength-reduction.c (pass_strength_reduction): Convert
2396         from a global struct to a subclass of gimple_opt_pass along with...
2397         (pass_data_strength_reduction): ...new pass_data instance and...
2398         (make_pass_strength_reduction): ...new function.
2399         * ifcvt.c (pass_rtl_ifcvt): Convert from a global struct to a subclass
2400         of rtl_opt_pass along with...
2401         (pass_data_rtl_ifcvt): ...new pass_data instance and...
2402         (make_pass_rtl_ifcvt): ...new function.
2403         (pass_if_after_combine): Convert from a global struct to a subclass of
2404         rtl_opt_pass along with...
2405         (pass_data_if_after_combine): ...new pass_data instance and...
2406         (make_pass_if_after_combine): ...new function.
2407         (pass_if_after_reload): Convert from a global struct to a subclass of
2408         rtl_opt_pass along with...
2409         (pass_data_if_after_reload): ...new pass_data instance and...
2410         (make_pass_if_after_reload): ...new function.
2411         * init-regs.c (pass_initialize_regs): Convert from a global struct to
2412         a subclass of rtl_opt_pass along with...
2413         (pass_data_initialize_regs): ...new pass_data instance and...
2414         (make_pass_initialize_regs): ...new function.
2415         * ipa-cp.c (pass_ipa_cp): Convert from a global struct to a subclass
2416         of ipa_opt_pass_d along with...
2417         (pass_data_ipa_cp): ...new pass_data instance and...
2418         (make_pass_ipa_cp): ...new function.
2419         * ipa-inline-analysis.c (pass_inline_parameters): Convert from a
2420         global struct to a subclass of gimple_opt_pass along with...
2421         (pass_data_inline_parameters): ...new pass_data instance and...
2422         (make_pass_inline_parameters): ...new function.
2423         * ipa-inline.c (pass_early_inline): Convert from a global struct to a
2424         subclass of gimple_opt_pass along with...
2425         (pass_data_early_inline): ...new pass_data instance and...
2426         (make_pass_early_inline): ...new function.
2427         (pass_ipa_inline): Convert from a global struct to a subclass of
2428         ipa_opt_pass_d along with...
2429         (pass_data_ipa_inline): ...new pass_data instance and...
2430         (make_pass_ipa_inline): ...new function.
2431         * ipa-pure-const.c (pass_local_pure_const): Convert from a global
2432         struct to a subclass of gimple_opt_pass along with...
2433         (pass_data_local_pure_const): ...new pass_data instance and...
2434         (make_pass_local_pure_const): ...new function.
2435         (pass_ipa_pure_const): Convert from a global struct to a subclass of
2436         ipa_opt_pass_d along with...
2437         (pass_data_ipa_pure_const): ...new pass_data instance and...
2438         (make_pass_ipa_pure_const): ...new function.
2439         * ipa-reference.c (pass_ipa_reference): Convert from a global struct
2440         to a subclass of ipa_opt_pass_d along with...
2441         (pass_data_ipa_reference): ...new pass_data instance and...
2442         (make_pass_ipa_reference): ...new function.
2443         * ipa-split.c (pass_split_functions): Convert from a global struct to
2444         a subclass of gimple_opt_pass along with...
2445         (pass_data_split_functions): ...new pass_data instance and...
2446         (make_pass_split_functions): ...new function.
2447         (pass_feedback_split_functions): Convert from a global struct to a
2448         subclass of gimple_opt_pass along with...
2449         (pass_data_feedback_split_functions): ...new pass_data instance and...
2450         (make_pass_feedback_split_functions): ...new function.
2451         * ipa.c (pass_ipa_function_and_variable_visibility): Convert from a
2452         global struct to a subclass of simple_ipa_opt_pass along with...
2453         (pass_data_ipa_function_and_variable_visibility): ...new pass_data
2454         instance and...
2455         (make_pass_ipa_function_and_variable_visibility): ...new function.
2456         (pass_ipa_free_inline_summary): Convert from a global struct to a
2457         subclass of simple_ipa_opt_pass along with...
2458         (pass_data_ipa_free_inline_summary): ...new pass_data instance and...
2459         (make_pass_ipa_free_inline_summary): ...new function.
2460         (pass_ipa_whole_program_visibility): Convert from a global struct to a
2461         subclass of ipa_opt_pass_d along with...
2462         (pass_data_ipa_whole_program_visibility): ...new pass_data instance
2463         and...
2464         (make_pass_ipa_whole_program_visibility): ...new function.
2465         (pass_ipa_profile): Convert from a global struct to a subclass of
2466         ipa_opt_pass_d along with...
2467         (pass_data_ipa_profile): ...new pass_data instance and...
2468         (make_pass_ipa_profile): ...new function.
2469         (pass_ipa_cdtor_merge): Convert from a global struct to a subclass of
2470         ipa_opt_pass_d along with...
2471         (pass_data_ipa_cdtor_merge): ...new pass_data instance and...
2472         (make_pass_ipa_cdtor_merge): ...new function.
2473         * ira.c (pass_ira): Convert from a global struct to a subclass of
2474         rtl_opt_pass along with...
2475         (pass_data_ira): ...new pass_data instance and...
2476         (make_pass_ira): ...new function.
2477         (pass_reload): Convert from a global struct to a subclass of
2478         rtl_opt_pass along with...
2479         (pass_data_reload): ...new pass_data instance and...
2480         (make_pass_reload): ...new function.
2481         * jump.c (pass_cleanup_barriers): Convert from a global struct to a
2482         subclass of rtl_opt_pass along with...
2483         (pass_data_cleanup_barriers): ...new pass_data instance and...
2484         (make_pass_cleanup_barriers): ...new function.
2485         * loop-init.c (pass_loop2): Convert from a global struct to a subclass
2486         of rtl_opt_pass along with...
2487         (pass_data_loop2): ...new pass_data instance and...
2488         (make_pass_loop2): ...new function.
2489         (pass_rtl_loop_init): Convert from a global struct to a subclass of
2490         rtl_opt_pass along with...
2491         (pass_data_rtl_loop_init): ...new pass_data instance and...
2492         (make_pass_rtl_loop_init): ...new function.
2493         (pass_rtl_loop_done): Convert from a global struct to a subclass of
2494         rtl_opt_pass along with...
2495         (pass_data_rtl_loop_done): ...new pass_data instance and...
2496         (make_pass_rtl_loop_done): ...new function.
2497         (pass_rtl_move_loop_invariants): Convert from a global struct to a
2498         subclass of rtl_opt_pass along with...
2499         (pass_data_rtl_move_loop_invariants): ...new pass_data instance and...
2500         (make_pass_rtl_move_loop_invariants): ...new function.
2501         (pass_rtl_unswitch): Convert from a global struct to a subclass of
2502         rtl_opt_pass along with...
2503         (pass_data_rtl_unswitch): ...new pass_data instance and...
2504         (make_pass_rtl_unswitch): ...new function.
2505         (pass_rtl_unroll_and_peel_loops): Convert from a global struct to a
2506         subclass of rtl_opt_pass along with...
2507         (pass_data_rtl_unroll_and_peel_loops): ...new pass_data instance
2508         and...
2509         (make_pass_rtl_unroll_and_peel_loops): ...new function.
2510         (pass_rtl_doloop): Convert from a global struct to a subclass of
2511         rtl_opt_pass along with...
2512         (pass_data_rtl_doloop): ...new pass_data instance and...
2513         (make_pass_rtl_doloop): ...new function.
2514         * lower-subreg.c (pass_lower_subreg): Convert from a global struct to
2515         a subclass of rtl_opt_pass along with...
2516         (pass_data_lower_subreg): ...new pass_data instance and...
2517         (make_pass_lower_subreg): ...new function.
2518         (pass_lower_subreg2): Convert from a global struct to a subclass of
2519         rtl_opt_pass along with...
2520         (pass_data_lower_subreg2): ...new pass_data instance and...
2521         (make_pass_lower_subreg2): ...new function.
2522         * lto-streamer-out.c (pass_ipa_lto_gimple_out): Convert from a global
2523         struct to a subclass of ipa_opt_pass_d along with...
2524         (pass_data_ipa_lto_gimple_out): ...new pass_data instance and...
2525         (make_pass_ipa_lto_gimple_out): ...new function.
2526         (pass_ipa_lto_finish_out): Convert from a global struct to a subclass
2527         of ipa_opt_pass_d along with...
2528         (pass_data_ipa_lto_finish_out): ...new pass_data instance and...
2529         (make_pass_ipa_lto_finish_out): ...new function.
2530         * mode-switching.c (pass_mode_switching): Convert from a global struct
2531         to a subclass of rtl_opt_pass along with...
2532         (pass_data_mode_switching): ...new pass_data instance and...
2533         (make_pass_mode_switching): ...new function.
2534         * modulo-sched.c (pass_sms): Convert from a global struct to a
2535         subclass of rtl_opt_pass along with...
2536         (pass_data_sms): ...new pass_data instance and...
2537         (make_pass_sms): ...new function.
2538         * omp-low.c (pass_expand_omp): Convert from a global struct to a
2539         subclass of gimple_opt_pass along with...
2540         (pass_data_expand_omp): ...new pass_data instance and...
2541         (make_pass_expand_omp): ...new function.
2542         (pass_lower_omp): Convert from a global struct to a subclass of
2543         gimple_opt_pass along with...
2544         (pass_data_lower_omp): ...new pass_data instance and...
2545         (make_pass_lower_omp): ...new function.
2546         (pass_diagnose_omp_blocks): Convert from a global struct to a subclass
2547         of gimple_opt_pass along with...
2548         (pass_data_diagnose_omp_blocks): ...new pass_data instance and...
2549         (make_pass_diagnose_omp_blocks): ...new function.
2550         * passes.c (pass_early_local_passes): Convert from a global struct to
2551         a subclass of simple_ipa_opt_pass along with...
2552         (pass_data_early_local_passes): ...new pass_data instance and...
2553         (make_pass_early_local_passes): ...new function.
2554         (pass_all_early_optimizations): Convert from a global struct to a
2555         subclass of gimple_opt_pass along with...
2556         (pass_data_all_early_optimizations): ...new pass_data instance and...
2557         (make_pass_all_early_optimizations): ...new function.
2558         (pass_all_optimizations): Convert from a global struct to a subclass
2559         of gimple_opt_pass along with...
2560         (pass_data_all_optimizations): ...new pass_data instance and...
2561         (make_pass_all_optimizations): ...new function.
2562         (pass_all_optimizations_g): Convert from a global struct to a subclass
2563         of gimple_opt_pass along with...
2564         (pass_data_all_optimizations_g): ...new pass_data instance and...
2565         (make_pass_all_optimizations_g): ...new function.
2566         (pass_rest_of_compilation): Convert from a global struct to a subclass
2567         of rtl_opt_pass along with...
2568         (pass_data_rest_of_compilation): ...new pass_data instance and...
2569         (make_pass_rest_of_compilation): ...new function.
2570         (pass_postreload): Convert from a global struct to a subclass of
2571         rtl_opt_pass along with...
2572         (pass_data_postreload): ...new pass_data instance and...
2573         (make_pass_postreload): ...new function.
2574         * postreload-gcse.c (pass_gcse2): Convert from a global struct to a
2575         subclass of rtl_opt_pass along with...
2576         (pass_data_gcse2): ...new pass_data instance and...
2577         (make_pass_gcse2): ...new function.
2578         * postreload.c (pass_postreload_cse): Convert from a global struct to
2579         a subclass of rtl_opt_pass along with...
2580         (pass_data_postreload_cse): ...new pass_data instance and...
2581         (make_pass_postreload_cse): ...new function.
2582         * predict.c (pass_profile): Convert from a global struct to a subclass
2583         of gimple_opt_pass along with...
2584         (pass_data_profile): ...new pass_data instance and...
2585         (make_pass_profile): ...new function.
2586         (pass_strip_predict_hints): Convert from a global struct to a subclass
2587         of gimple_opt_pass along with...
2588         (pass_data_strip_predict_hints): ...new pass_data instance and...
2589         (make_pass_strip_predict_hints): ...new function.
2590         * recog.c (pass_peephole2): Convert from a global struct to a subclass
2591         of rtl_opt_pass along with...
2592         (pass_data_peephole2): ...new pass_data instance and...
2593         (make_pass_peephole2): ...new function.
2594         (pass_split_all_insns): Convert from a global struct to a subclass of
2595         rtl_opt_pass along with...
2596         (pass_data_split_all_insns): ...new pass_data instance and...
2597         (make_pass_split_all_insns): ...new function.
2598         (pass_split_after_reload): Convert from a global struct to a subclass
2599         of rtl_opt_pass along with...
2600         (pass_data_split_after_reload): ...new pass_data instance and...
2601         (make_pass_split_after_reload): ...new function.
2602         (pass_split_before_regstack): Convert from a global struct to a
2603         subclass of rtl_opt_pass along with...
2604         (pass_data_split_before_regstack): ...new pass_data instance and...
2605         (make_pass_split_before_regstack): ...new function.
2606         (pass_split_before_sched2): Convert from a global struct to a subclass
2607         of rtl_opt_pass along with...
2608         (pass_data_split_before_sched2): ...new pass_data instance and...
2609         (make_pass_split_before_sched2): ...new function.
2610         (pass_split_for_shorten_branches): Convert from a global struct to a
2611         subclass of rtl_opt_pass along with...
2612         (pass_data_split_for_shorten_branches): ...new pass_data instance
2613         and...
2614         (make_pass_split_for_shorten_branches): ...new function.
2615         * ree.c (pass_ree): Convert from a global struct to a subclass of
2616         rtl_opt_pass along with...
2617         (pass_data_ree): ...new pass_data instance and...
2618         (make_pass_ree): ...new function.
2619         * reg-stack.c (pass_stack_regs): Convert from a global struct to a
2620         subclass of rtl_opt_pass along with...
2621         (pass_data_stack_regs): ...new pass_data instance and...
2622         (make_pass_stack_regs): ...new function.
2623         (pass_stack_regs_run): Convert from a global struct to a subclass of
2624         rtl_opt_pass along with...
2625         (pass_data_stack_regs_run): ...new pass_data instance and...
2626         (make_pass_stack_regs_run): ...new function.
2627         * regcprop.c (pass_cprop_hardreg): Convert from a global struct to a
2628         subclass of rtl_opt_pass along with...
2629         (pass_data_cprop_hardreg): ...new pass_data instance and...
2630         (make_pass_cprop_hardreg): ...new function.
2631         * reginfo.c (pass_reginfo_init): Convert from a global struct to a
2632         subclass of rtl_opt_pass along with...
2633         (pass_data_reginfo_init): ...new pass_data instance and...
2634         (make_pass_reginfo_init): ...new function.
2635         * regmove.c (pass_regmove): Convert from a global struct to a subclass
2636         of rtl_opt_pass along with...
2637         (pass_data_regmove): ...new pass_data instance and...
2638         (make_pass_regmove): ...new function.
2639         * regrename.c (pass_regrename): Convert from a global struct to a
2640         subclass of rtl_opt_pass along with...
2641         (pass_data_regrename): ...new pass_data instance and...
2642         (make_pass_regrename): ...new function.
2643         * reorg.c (pass_delay_slots): Convert from a global struct to a
2644         subclass of rtl_opt_pass along with...
2645         (pass_data_delay_slots): ...new pass_data instance and...
2646         (make_pass_delay_slots): ...new function.
2647         (pass_machine_reorg): Convert from a global struct to a subclass of
2648         rtl_opt_pass along with...
2649         (pass_data_machine_reorg): ...new pass_data instance and...
2650         (make_pass_machine_reorg): ...new function.
2651         * sched-rgn.c (pass_sched): Convert from a global struct to a subclass
2652         of rtl_opt_pass along with...
2653         (pass_data_sched): ...new pass_data instance and...
2654         (make_pass_sched): ...new function.
2655         (pass_sched2): Convert from a global struct to a subclass of
2656         rtl_opt_pass along with...
2657         (pass_data_sched2): ...new pass_data instance and...
2658         (make_pass_sched2): ...new function.
2659         * stack-ptr-mod.c (pass_stack_ptr_mod): Convert from a global struct
2660         to a subclass of rtl_opt_pass along with...
2661         (pass_data_stack_ptr_mod): ...new pass_data instance and...
2662         (make_pass_stack_ptr_mod): ...new function.
2663         * store-motion.c (pass_rtl_store_motion): Convert from a global struct
2664         to a subclass of rtl_opt_pass along with...
2665         (pass_data_rtl_store_motion): ...new pass_data instance and...
2666         (make_pass_rtl_store_motion): ...new function.
2667         * tracer.c (pass_tracer): Convert from a global struct to a subclass
2668         of gimple_opt_pass along with...
2669         (pass_data_tracer): ...new pass_data instance and...
2670         (make_pass_tracer): ...new function.
2671         * trans-mem.c (pass_diagnose_tm_blocks): Convert from a global struct
2672         to a subclass of gimple_opt_pass along with...
2673         (pass_data_diagnose_tm_blocks): ...new pass_data instance and...
2674         (make_pass_diagnose_tm_blocks): ...new function.
2675         (pass_lower_tm): Convert from a global struct to a subclass of
2676         gimple_opt_pass along with...
2677         (pass_data_lower_tm): ...new pass_data instance and...
2678         (make_pass_lower_tm): ...new function.
2679         (pass_tm_init): Convert from a global struct to a subclass of
2680         gimple_opt_pass along with...
2681         (pass_data_tm_init): ...new pass_data instance and...
2682         (make_pass_tm_init): ...new function.
2683         (pass_tm_mark): Convert from a global struct to a subclass of
2684         gimple_opt_pass along with...
2685         (pass_data_tm_mark): ...new pass_data instance and...
2686         (make_pass_tm_mark): ...new function.
2687         (pass_tm_edges): Convert from a global struct to a subclass of
2688         gimple_opt_pass along with...
2689         (pass_data_tm_edges): ...new pass_data instance and...
2690         (make_pass_tm_edges): ...new function.
2691         (pass_tm_memopt): Convert from a global struct to a subclass of
2692         gimple_opt_pass along with...
2693         (pass_data_tm_memopt): ...new pass_data instance and...
2694         (make_pass_tm_memopt): ...new function.
2695         (pass_ipa_tm): Convert from a global struct to a subclass of
2696         simple_ipa_opt_pass along with...
2697         (pass_data_ipa_tm): ...new pass_data instance and...
2698         (make_pass_ipa_tm): ...new function.
2699         * tree-call-cdce.c (pass_call_cdce): Convert from a global struct to a
2700         subclass of gimple_opt_pass along with...
2701         (pass_data_call_cdce): ...new pass_data instance and...
2702         (make_pass_call_cdce): ...new function.
2703         * tree-cfg.c (pass_build_cfg): Convert from a global struct to a
2704         subclass of gimple_opt_pass along with...
2705         (pass_data_build_cfg): ...new pass_data instance and...
2706         (make_pass_build_cfg): ...new function.
2707         (pass_split_crit_edges): Convert from a global struct to a subclass of
2708         gimple_opt_pass along with...
2709         (pass_data_split_crit_edges): ...new pass_data instance and...
2710         (make_pass_split_crit_edges): ...new function.
2711         (pass_warn_function_return): Convert from a global struct to a
2712         subclass of gimple_opt_pass along with...
2713         (pass_data_warn_function_return): ...new pass_data instance and...
2714         (make_pass_warn_function_return): ...new function.
2715         (pass_warn_function_noreturn): Convert from a global struct to a
2716         subclass of gimple_opt_pass along with...
2717         (pass_data_warn_function_noreturn): ...new pass_data instance and...
2718         (make_pass_warn_function_noreturn): ...new function.
2719         (pass_warn_unused_result): Convert from a global struct to a subclass
2720         of gimple_opt_pass along with...
2721         (pass_data_warn_unused_result): ...new pass_data instance and...
2722         (make_pass_warn_unused_result): ...new function.
2723         * tree-cfgcleanup.c (pass_merge_phi): Convert from a global struct to
2724         a subclass of gimple_opt_pass along with...
2725         (pass_data_merge_phi): ...new pass_data instance and...
2726         (make_pass_merge_phi): ...new function.
2727         * tree-complex.c (pass_lower_complex): Convert from a global struct to
2728         a subclass of gimple_opt_pass along with...
2729         (pass_data_lower_complex): ...new pass_data instance and...
2730         (make_pass_lower_complex): ...new function.
2731         (pass_lower_complex_O0): Convert from a global struct to a subclass of
2732         gimple_opt_pass along with...
2733         (pass_data_lower_complex_O0): ...new pass_data instance and...
2734         (make_pass_lower_complex_O0): ...new function.
2735         * tree-eh.c (pass_lower_eh): Convert from a global struct to a
2736         subclass of gimple_opt_pass along with...
2737         (pass_data_lower_eh): ...new pass_data instance and...
2738         (make_pass_lower_eh): ...new function.
2739         (pass_refactor_eh): Convert from a global struct to a subclass of
2740         gimple_opt_pass along with...
2741         (pass_data_refactor_eh): ...new pass_data instance and...
2742         (make_pass_refactor_eh): ...new function.
2743         (pass_lower_resx): Convert from a global struct to a subclass of
2744         gimple_opt_pass along with...
2745         (pass_data_lower_resx): ...new pass_data instance and...
2746         (make_pass_lower_resx): ...new function.
2747         (pass_lower_eh_dispatch): Convert from a global struct to a subclass
2748         of gimple_opt_pass along with...
2749         (pass_data_lower_eh_dispatch): ...new pass_data instance and...
2750         (make_pass_lower_eh_dispatch): ...new function.
2751         (pass_cleanup_eh): Convert from a global struct to a subclass of
2752         gimple_opt_pass along with...
2753         (pass_data_cleanup_eh): ...new pass_data instance and...
2754         (make_pass_cleanup_eh): ...new function.
2755         * tree-emutls.c (pass_ipa_lower_emutls): Convert from a global struct
2756         to a subclass of simple_ipa_opt_pass along with...
2757         (pass_data_ipa_lower_emutls): ...new pass_data instance and...
2758         (make_pass_ipa_lower_emutls): ...new function.
2759         * tree-if-conv.c (pass_if_conversion): Convert from a global struct to
2760         a subclass of gimple_opt_pass along with...
2761         (pass_data_if_conversion): ...new pass_data instance and...
2762         (make_pass_if_conversion): ...new function.
2763         * tree-into-ssa.c (pass_build_ssa): Convert from a global struct to a
2764         subclass of gimple_opt_pass along with...
2765         (pass_data_build_ssa): ...new pass_data instance and...
2766         (make_pass_build_ssa): ...new function.
2767         * tree-loop-distribution.c (pass_loop_distribution): Convert from a
2768         global struct to a subclass of gimple_opt_pass along with...
2769         (pass_data_loop_distribution): ...new pass_data instance and...
2770         (make_pass_loop_distribution): ...new function.
2771         * tree-mudflap.c (pass_mudflap_1): Convert from a global struct to a
2772         subclass of gimple_opt_pass along with...
2773         (pass_data_mudflap_1): ...new pass_data instance and...
2774         (make_pass_mudflap_1): ...new function.
2775         (pass_mudflap_2): Convert from a global struct to a subclass of
2776         gimple_opt_pass along with...
2777         (pass_data_mudflap_2): ...new pass_data instance and...
2778         (make_pass_mudflap_2): ...new function.
2779         * tree-nomudflap.c (pass_mudflap_1): Convert from a global struct to a
2780         subclass of gimple_opt_pass along with...
2781         (pass_data_mudflap_1): ...new pass_data instance and...
2782         (make_pass_mudflap_1): ...new function.
2783         (pass_mudflap_2): Convert from a global struct to a subclass of
2784         gimple_opt_pass along with...
2785         (pass_data_mudflap_2): ...new pass_data instance and...
2786         (make_pass_mudflap_2): ...new function.
2787         * tree-nrv.c (pass_nrv): Convert from a global struct to a subclass of
2788         gimple_opt_pass along with...
2789         (pass_data_nrv): ...new pass_data instance and...
2790         (make_pass_nrv): ...new function.
2791         (pass_return_slot): Convert from a global struct to a subclass of
2792         gimple_opt_pass along with...
2793         (pass_data_return_slot): ...new pass_data instance and...
2794         (make_pass_return_slot): ...new function.
2795         * tree-object-size.c (pass_object_sizes): Convert from a global struct
2796         to a subclass of gimple_opt_pass along with...
2797         (pass_data_object_sizes): ...new pass_data instance and...
2798         (make_pass_object_sizes): ...new function.
2799         * tree-optimize.c (pass_cleanup_cfg_post_optimizing): Convert from a
2800         global struct to a subclass of gimple_opt_pass along with...
2801         (pass_data_cleanup_cfg_post_optimizing): ...new pass_data instance
2802         and...
2803         (make_pass_cleanup_cfg_post_optimizing): ...new function.
2804         (pass_fixup_cfg): Convert from a global struct to a subclass of
2805         gimple_opt_pass along with...
2806         (pass_data_fixup_cfg): ...new pass_data instance and...
2807         (make_pass_fixup_cfg): ...new function.
2808         * tree-pass.h (pass_mudflap_1): Replace declaration with that of...
2809         (make_pass_mudflap_1): ...new function.
2810         (pass_mudflap_2): Replace declaration with that of...
2811         (make_pass_mudflap_2): ...new function.
2812         (pass_asan): Replace declaration with that of...
2813         (make_pass_asan): ...new function.
2814         (pass_asan_O0): Replace declaration with that of...
2815         (make_pass_asan_O0): ...new function.
2816         (pass_tsan): Replace declaration with that of...
2817         (make_pass_tsan): ...new function.
2818         (pass_tsan_O0): Replace declaration with that of...
2819         (make_pass_tsan_O0): ...new function.
2820         (pass_lower_cf): Replace declaration with that of...
2821         (make_pass_lower_cf): ...new function.
2822         (pass_refactor_eh): Replace declaration with that of...
2823         (make_pass_refactor_eh): ...new function.
2824         (pass_lower_eh): Replace declaration with that of...
2825         (make_pass_lower_eh): ...new function.
2826         (pass_lower_eh_dispatch): Replace declaration with that of...
2827         (make_pass_lower_eh_dispatch): ...new function.
2828         (pass_lower_resx): Replace declaration with that of...
2829         (make_pass_lower_resx): ...new function.
2830         (pass_build_cfg): Replace declaration with that of...
2831         (make_pass_build_cfg): ...new function.
2832         (pass_early_tree_profile): Replace declaration with that of...
2833         (make_pass_early_tree_profile): ...new function.
2834         (pass_cleanup_eh): Replace declaration with that of...
2835         (make_pass_cleanup_eh): ...new function.
2836         (pass_sra): Replace declaration with that of...
2837         (make_pass_sra): ...new function.
2838         (pass_sra_early): Replace declaration with that of...
2839         (make_pass_sra_early): ...new function.
2840         (pass_early_ipa_sra): Replace declaration with that of...
2841         (make_pass_early_ipa_sra): ...new function.
2842         (pass_tail_recursion): Replace declaration with that of...
2843         (make_pass_tail_recursion): ...new function.
2844         (pass_tail_calls): Replace declaration with that of...
2845         (make_pass_tail_calls): ...new function.
2846         (pass_tree_loop): Replace declaration with that of...
2847         (make_pass_tree_loop): ...new function.
2848         (pass_tree_loop_init): Replace declaration with that of...
2849         (make_pass_tree_loop_init): ...new function.
2850         (pass_lim): Replace declaration with that of...
2851         (make_pass_lim): ...new function.
2852         (pass_tree_unswitch): Replace declaration with that of...
2853         (make_pass_tree_unswitch): ...new function.
2854         (pass_predcom): Replace declaration with that of...
2855         (make_pass_predcom): ...new function.
2856         (pass_iv_canon): Replace declaration with that of...
2857         (make_pass_iv_canon): ...new function.
2858         (pass_scev_cprop): Replace declaration with that of...
2859         (make_pass_scev_cprop): ...new function.
2860         (pass_empty_loop): Replace declaration with that of...
2861         (make_pass_empty_loop): ...new function.
2862         (pass_record_bounds): Replace declaration with that of...
2863         (make_pass_record_bounds): ...new function.
2864         (pass_graphite): Replace declaration with that of...
2865         (make_pass_graphite): ...new function.
2866         (pass_graphite_transforms): Replace declaration with that of...
2867         (make_pass_graphite_transforms): ...new function.
2868         (pass_if_conversion): Replace declaration with that of...
2869         (make_pass_if_conversion): ...new function.
2870         (pass_loop_distribution): Replace declaration with that of...
2871         (make_pass_loop_distribution): ...new function.
2872         (pass_vectorize): Replace declaration with that of...
2873         (make_pass_vectorize): ...new function.
2874         (pass_slp_vectorize): Replace declaration with that of...
2875         (make_pass_slp_vectorize): ...new function.
2876         (pass_complete_unroll): Replace declaration with that of...
2877         (make_pass_complete_unroll): ...new function.
2878         (pass_complete_unrolli): Replace declaration with that of...
2879         (make_pass_complete_unrolli): ...new function.
2880         (pass_parallelize_loops): Replace declaration with that of...
2881         (make_pass_parallelize_loops): ...new function.
2882         (pass_loop_prefetch): Replace declaration with that of...
2883         (make_pass_loop_prefetch): ...new function.
2884         (pass_iv_optimize): Replace declaration with that of...
2885         (make_pass_iv_optimize): ...new function.
2886         (pass_tree_loop_done): Replace declaration with that of...
2887         (make_pass_tree_loop_done): ...new function.
2888         (pass_ch): Replace declaration with that of...
2889         (make_pass_ch): ...new function.
2890         (pass_ccp): Replace declaration with that of...
2891         (make_pass_ccp): ...new function.
2892         (pass_phi_only_cprop): Replace declaration with that of...
2893         (make_pass_phi_only_cprop): ...new function.
2894         (pass_build_ssa): Replace declaration with that of...
2895         (make_pass_build_ssa): ...new function.
2896         (pass_build_alias): Replace declaration with that of...
2897         (make_pass_build_alias): ...new function.
2898         (pass_build_ealias): Replace declaration with that of...
2899         (make_pass_build_ealias): ...new function.
2900         (pass_dominator): Replace declaration with that of...
2901         (make_pass_dominator): ...new function.
2902         (pass_dce): Replace declaration with that of...
2903         (make_pass_dce): ...new function.
2904         (pass_dce_loop): Replace declaration with that of...
2905         (make_pass_dce_loop): ...new function.
2906         (pass_cd_dce): Replace declaration with that of...
2907         (make_pass_cd_dce): ...new function.
2908         (pass_call_cdce): Replace declaration with that of...
2909         (make_pass_call_cdce): ...new function.
2910         (pass_merge_phi): Replace declaration with that of...
2911         (make_pass_merge_phi): ...new function.
2912         (pass_split_crit_edges): Replace declaration with that of...
2913         (make_pass_split_crit_edges): ...new function.
2914         (pass_pre): Replace declaration with that of...
2915         (make_pass_pre): ...new function.
2916         (pass_profile): Replace declaration with that of...
2917         (make_pass_profile): ...new function.
2918         (pass_strip_predict_hints): Replace declaration with that of...
2919         (make_pass_strip_predict_hints): ...new function.
2920         (pass_lower_complex_O0): Replace declaration with that of...
2921         (make_pass_lower_complex_O0): ...new function.
2922         (pass_lower_complex): Replace declaration with that of...
2923         (make_pass_lower_complex): ...new function.
2924         (pass_lower_vector): Replace declaration with that of...
2925         (make_pass_lower_vector): ...new function.
2926         (pass_lower_vector_ssa): Replace declaration with that of...
2927         (make_pass_lower_vector_ssa): ...new function.
2928         (pass_lower_omp): Replace declaration with that of...
2929         (make_pass_lower_omp): ...new function.
2930         (pass_diagnose_omp_blocks): Replace declaration with that of...
2931         (make_pass_diagnose_omp_blocks): ...new function.
2932         (pass_expand_omp): Replace declaration with that of...
2933         (make_pass_expand_omp): ...new function.
2934         (pass_expand_omp_ssa): Replace declaration with that of...
2935         (make_pass_expand_omp_ssa): ...new function.
2936         (pass_object_sizes): Replace declaration with that of...
2937         (make_pass_object_sizes): ...new function.
2938         (pass_strlen): Replace declaration with that of...
2939         (make_pass_strlen): ...new function.
2940         (pass_fold_builtins): Replace declaration with that of...
2941         (make_pass_fold_builtins): ...new function.
2942         (pass_stdarg): Replace declaration with that of...
2943         (make_pass_stdarg): ...new function.
2944         (pass_early_warn_uninitialized): Replace declaration with that of...
2945         (make_pass_early_warn_uninitialized): ...new function.
2946         (pass_late_warn_uninitialized): Replace declaration with that of...
2947         (make_pass_late_warn_uninitialized): ...new function.
2948         (pass_cse_reciprocals): Replace declaration with that of...
2949         (make_pass_cse_reciprocals): ...new function.
2950         (pass_cse_sincos): Replace declaration with that of...
2951         (make_pass_cse_sincos): ...new function.
2952         (pass_optimize_bswap): Replace declaration with that of...
2953         (make_pass_optimize_bswap): ...new function.
2954         (pass_optimize_widening_mul): Replace declaration with that of...
2955         (make_pass_optimize_widening_mul): ...new function.
2956         (pass_warn_function_return): Replace declaration with that of...
2957         (make_pass_warn_function_return): ...new function.
2958         (pass_warn_function_noreturn): Replace declaration with that of...
2959         (make_pass_warn_function_noreturn): ...new function.
2960         (pass_cselim): Replace declaration with that of...
2961         (make_pass_cselim): ...new function.
2962         (pass_phiopt): Replace declaration with that of...
2963         (make_pass_phiopt): ...new function.
2964         (pass_forwprop): Replace declaration with that of...
2965         (make_pass_forwprop): ...new function.
2966         (pass_phiprop): Replace declaration with that of...
2967         (make_pass_phiprop): ...new function.
2968         (pass_tree_ifcombine): Replace declaration with that of...
2969         (make_pass_tree_ifcombine): ...new function.
2970         (pass_dse): Replace declaration with that of...
2971         (make_pass_dse): ...new function.
2972         (pass_nrv): Replace declaration with that of...
2973         (make_pass_nrv): ...new function.
2974         (pass_rename_ssa_copies): Replace declaration with that of...
2975         (make_pass_rename_ssa_copies): ...new function.
2976         (pass_sink_code): Replace declaration with that of...
2977         (make_pass_sink_code): ...new function.
2978         (pass_fre): Replace declaration with that of...
2979         (make_pass_fre): ...new function.
2980         (pass_check_data_deps): Replace declaration with that of...
2981         (make_pass_check_data_deps): ...new function.
2982         (pass_copy_prop): Replace declaration with that of...
2983         (make_pass_copy_prop): ...new function.
2984         (pass_vrp): Replace declaration with that of...
2985         (make_pass_vrp): ...new function.
2986         (pass_uncprop): Replace declaration with that of...
2987         (make_pass_uncprop): ...new function.
2988         (pass_return_slot): Replace declaration with that of...
2989         (make_pass_return_slot): ...new function.
2990         (pass_reassoc): Replace declaration with that of...
2991         (make_pass_reassoc): ...new function.
2992         (pass_rebuild_cgraph_edges): Replace declaration with that of...
2993         (make_pass_rebuild_cgraph_edges): ...new function.
2994         (pass_remove_cgraph_callee_edges): Replace declaration with that of...
2995         (make_pass_remove_cgraph_callee_edges): ...new function.
2996         (pass_build_cgraph_edges): Replace declaration with that of...
2997         (make_pass_build_cgraph_edges): ...new function.
2998         (pass_local_pure_const): Replace declaration with that of...
2999         (make_pass_local_pure_const): ...new function.
3000         (pass_tracer): Replace declaration with that of...
3001         (make_pass_tracer): ...new function.
3002         (pass_warn_unused_result): Replace declaration with that of...
3003         (make_pass_warn_unused_result): ...new function.
3004         (pass_diagnose_tm_blocks): Replace declaration with that of...
3005         (make_pass_diagnose_tm_blocks): ...new function.
3006         (pass_lower_tm): Replace declaration with that of...
3007         (make_pass_lower_tm): ...new function.
3008         (pass_tm_init): Replace declaration with that of...
3009         (make_pass_tm_init): ...new function.
3010         (pass_tm_mark): Replace declaration with that of...
3011         (make_pass_tm_mark): ...new function.
3012         (pass_tm_memopt): Replace declaration with that of...
3013         (make_pass_tm_memopt): ...new function.
3014         (pass_tm_edges): Replace declaration with that of...
3015         (make_pass_tm_edges): ...new function.
3016         (pass_split_functions): Replace declaration with that of...
3017         (make_pass_split_functions): ...new function.
3018         (pass_feedback_split_functions): Replace declaration with that of...
3019         (make_pass_feedback_split_functions): ...new function.
3020         (pass_strength_reduction): Replace declaration with that of...
3021         (make_pass_strength_reduction): ...new function.
3022         (pass_ipa_lower_emutls): Replace declaration with that of...
3023         (make_pass_ipa_lower_emutls): ...new function.
3024         (pass_ipa_function_and_variable_visibility): Replace declaration with
3025         that of...
3026         (make_pass_ipa_function_and_variable_visibility): ...new function.
3027         (pass_ipa_tree_profile): Replace declaration with that of...
3028         (make_pass_ipa_tree_profile): ...new function.
3029         (pass_early_local_passes): Replace declaration with that of...
3030         (make_pass_early_local_passes): ...new function.
3031         (pass_ipa_whole_program_visibility): Replace declaration with that
3032         of...
3033         (make_pass_ipa_whole_program_visibility): ...new function.
3034         (pass_ipa_lto_gimple_out): Replace declaration with that of...
3035         (make_pass_ipa_lto_gimple_out): ...new function.
3036         (pass_ipa_increase_alignment): Replace declaration with that of...
3037         (make_pass_ipa_increase_alignment): ...new function.
3038         (pass_ipa_inline): Replace declaration with that of...
3039         (make_pass_ipa_inline): ...new function.
3040         (pass_ipa_free_lang_data): Replace declaration with that of...
3041         (make_pass_ipa_free_lang_data): ...new function.
3042         (pass_ipa_free_inline_summary): Replace declaration with that of...
3043         (make_pass_ipa_free_inline_summary): ...new function.
3044         (pass_ipa_cp): Replace declaration with that of...
3045         (make_pass_ipa_cp): ...new function.
3046         (pass_ipa_reference): Replace declaration with that of...
3047         (make_pass_ipa_reference): ...new function.
3048         (pass_ipa_pure_const): Replace declaration with that of...
3049         (make_pass_ipa_pure_const): ...new function.
3050         (pass_ipa_pta): Replace declaration with that of...
3051         (make_pass_ipa_pta): ...new function.
3052         (pass_ipa_lto_finish_out): Replace declaration with that of...
3053         (make_pass_ipa_lto_finish_out): ...new function.
3054         (pass_ipa_tm): Replace declaration with that of...
3055         (make_pass_ipa_tm): ...new function.
3056         (pass_ipa_profile): Replace declaration with that of...
3057         (make_pass_ipa_profile): ...new function.
3058         (pass_ipa_cdtor_merge): Replace declaration with that of...
3059         (make_pass_ipa_cdtor_merge): ...new function.
3060         (pass_cleanup_cfg_post_optimizing): Replace declaration with that
3061         of...
3062         (make_pass_cleanup_cfg_post_optimizing): ...new function.
3063         (pass_init_datastructures): Replace declaration with that of...
3064         (make_pass_init_datastructures): ...new function.
3065         (pass_fixup_cfg): Replace declaration with that of...
3066         (make_pass_fixup_cfg): ...new function.
3067         (pass_expand): Replace declaration with that of...
3068         (make_pass_expand): ...new function.
3069         (pass_instantiate_virtual_regs): Replace declaration with that of...
3070         (make_pass_instantiate_virtual_regs): ...new function.
3071         (pass_rtl_fwprop): Replace declaration with that of...
3072         (make_pass_rtl_fwprop): ...new function.
3073         (pass_rtl_fwprop_addr): Replace declaration with that of...
3074         (make_pass_rtl_fwprop_addr): ...new function.
3075         (pass_jump): Replace declaration with that of...
3076         (make_pass_jump): ...new function.
3077         (pass_jump2): Replace declaration with that of...
3078         (make_pass_jump2): ...new function.
3079         (pass_lower_subreg): Replace declaration with that of...
3080         (make_pass_lower_subreg): ...new function.
3081         (pass_cse): Replace declaration with that of...
3082         (make_pass_cse): ...new function.
3083         (pass_fast_rtl_dce): Replace declaration with that of...
3084         (make_pass_fast_rtl_dce): ...new function.
3085         (pass_ud_rtl_dce): Replace declaration with that of...
3086         (make_pass_ud_rtl_dce): ...new function.
3087         (pass_rtl_dce): Replace declaration with that of...
3088         (make_pass_rtl_dce): ...new function.
3089         (pass_rtl_dse1): Replace declaration with that of...
3090         (make_pass_rtl_dse1): ...new function.
3091         (pass_rtl_dse2): Replace declaration with that of...
3092         (make_pass_rtl_dse2): ...new function.
3093         (pass_rtl_dse3): Replace declaration with that of...
3094         (make_pass_rtl_dse3): ...new function.
3095         (pass_rtl_cprop): Replace declaration with that of...
3096         (make_pass_rtl_cprop): ...new function.
3097         (pass_rtl_pre): Replace declaration with that of...
3098         (make_pass_rtl_pre): ...new function.
3099         (pass_rtl_hoist): Replace declaration with that of...
3100         (make_pass_rtl_hoist): ...new function.
3101         (pass_rtl_store_motion): Replace declaration with that of...
3102         (make_pass_rtl_store_motion): ...new function.
3103         (pass_cse_after_global_opts): Replace declaration with that of...
3104         (make_pass_cse_after_global_opts): ...new function.
3105         (pass_rtl_ifcvt): Replace declaration with that of...
3106         (make_pass_rtl_ifcvt): ...new function.
3107         (pass_into_cfg_layout_mode): Replace declaration with that of...
3108         (make_pass_into_cfg_layout_mode): ...new function.
3109         (pass_outof_cfg_layout_mode): Replace declaration with that of...
3110         (make_pass_outof_cfg_layout_mode): ...new function.
3111         (pass_loop2): Replace declaration with that of...
3112         (make_pass_loop2): ...new function.
3113         (pass_rtl_loop_init): Replace declaration with that of...
3114         (make_pass_rtl_loop_init): ...new function.
3115         (pass_rtl_move_loop_invariants): Replace declaration with that of...
3116         (make_pass_rtl_move_loop_invariants): ...new function.
3117         (pass_rtl_unswitch): Replace declaration with that of...
3118         (make_pass_rtl_unswitch): ...new function.
3119         (pass_rtl_unroll_and_peel_loops): Replace declaration with that of...
3120         (make_pass_rtl_unroll_and_peel_loops): ...new function.
3121         (pass_rtl_doloop): Replace declaration with that of...
3122         (make_pass_rtl_doloop): ...new function.
3123         (pass_rtl_loop_done): Replace declaration with that of...
3124         (make_pass_rtl_loop_done): ...new function.
3125         (pass_web): Replace declaration with that of...
3126         (make_pass_web): ...new function.
3127         (pass_cse2): Replace declaration with that of...
3128         (make_pass_cse2): ...new function.
3129         (pass_df_initialize_opt): Replace declaration with that of...
3130         (make_pass_df_initialize_opt): ...new function.
3131         (pass_df_initialize_no_opt): Replace declaration with that of...
3132         (make_pass_df_initialize_no_opt): ...new function.
3133         (pass_reginfo_init): Replace declaration with that of...
3134         (make_pass_reginfo_init): ...new function.
3135         (pass_inc_dec): Replace declaration with that of...
3136         (make_pass_inc_dec): ...new function.
3137         (pass_stack_ptr_mod): Replace declaration with that of...
3138         (make_pass_stack_ptr_mod): ...new function.
3139         (pass_initialize_regs): Replace declaration with that of...
3140         (make_pass_initialize_regs): ...new function.
3141         (pass_combine): Replace declaration with that of...
3142         (make_pass_combine): ...new function.
3143         (pass_if_after_combine): Replace declaration with that of...
3144         (make_pass_if_after_combine): ...new function.
3145         (pass_ree): Replace declaration with that of...
3146         (make_pass_ree): ...new function.
3147         (pass_partition_blocks): Replace declaration with that of...
3148         (make_pass_partition_blocks): ...new function.
3149         (pass_match_asm_constraints): Replace declaration with that of...
3150         (make_pass_match_asm_constraints): ...new function.
3151         (pass_regmove): Replace declaration with that of...
3152         (make_pass_regmove): ...new function.
3153         (pass_split_all_insns): Replace declaration with that of...
3154         (make_pass_split_all_insns): ...new function.
3155         (pass_fast_rtl_byte_dce): Replace declaration with that of...
3156         (make_pass_fast_rtl_byte_dce): ...new function.
3157         (pass_lower_subreg2): Replace declaration with that of...
3158         (make_pass_lower_subreg2): ...new function.
3159         (pass_mode_switching): Replace declaration with that of...
3160         (make_pass_mode_switching): ...new function.
3161         (pass_sms): Replace declaration with that of...
3162         (make_pass_sms): ...new function.
3163         (pass_sched): Replace declaration with that of...
3164         (make_pass_sched): ...new function.
3165         (pass_ira): Replace declaration with that of...
3166         (make_pass_ira): ...new function.
3167         (pass_reload): Replace declaration with that of...
3168         (make_pass_reload): ...new function.
3169         (pass_clean_state): Replace declaration with that of...
3170         (make_pass_clean_state): ...new function.
3171         (pass_branch_prob): Replace declaration with that of...
3172         (make_pass_branch_prob): ...new function.
3173         (pass_value_profile_transformations): Replace declaration with that
3174         of...
3175         (make_pass_value_profile_transformations): ...new function.
3176         (pass_postreload_cse): Replace declaration with that of...
3177         (make_pass_postreload_cse): ...new function.
3178         (pass_gcse2): Replace declaration with that of...
3179         (make_pass_gcse2): ...new function.
3180         (pass_split_after_reload): Replace declaration with that of...
3181         (make_pass_split_after_reload): ...new function.
3182         (pass_branch_target_load_optimize1): Replace declaration with that
3183         of...
3184         (make_pass_branch_target_load_optimize1): ...new function.
3185         (pass_thread_prologue_and_epilogue): Replace declaration with that
3186         of...
3187         (make_pass_thread_prologue_and_epilogue): ...new function.
3188         (pass_stack_adjustments): Replace declaration with that of...
3189         (make_pass_stack_adjustments): ...new function.
3190         (pass_peephole2): Replace declaration with that of...
3191         (make_pass_peephole2): ...new function.
3192         (pass_if_after_reload): Replace declaration with that of...
3193         (make_pass_if_after_reload): ...new function.
3194         (pass_regrename): Replace declaration with that of...
3195         (make_pass_regrename): ...new function.
3196         (pass_cprop_hardreg): Replace declaration with that of...
3197         (make_pass_cprop_hardreg): ...new function.
3198         (pass_reorder_blocks): Replace declaration with that of...
3199         (make_pass_reorder_blocks): ...new function.
3200         (pass_branch_target_load_optimize2): Replace declaration with that
3201         of...
3202         (make_pass_branch_target_load_optimize2): ...new function.
3203         (pass_leaf_regs): Replace declaration with that of...
3204         (make_pass_leaf_regs): ...new function.
3205         (pass_split_before_sched2): Replace declaration with that of...
3206         (make_pass_split_before_sched2): ...new function.
3207         (pass_compare_elim_after_reload): Replace declaration with that of...
3208         (make_pass_compare_elim_after_reload): ...new function.
3209         (pass_sched2): Replace declaration with that of...
3210         (make_pass_sched2): ...new function.
3211         (pass_stack_regs): Replace declaration with that of...
3212         (make_pass_stack_regs): ...new function.
3213         (pass_stack_regs_run): Replace declaration with that of...
3214         (make_pass_stack_regs_run): ...new function.
3215         (pass_df_finish): Replace declaration with that of...
3216         (make_pass_df_finish): ...new function.
3217         (pass_compute_alignments): Replace declaration with that of...
3218         (make_pass_compute_alignments): ...new function.
3219         (pass_duplicate_computed_gotos): Replace declaration with that of...
3220         (make_pass_duplicate_computed_gotos): ...new function.
3221         (pass_variable_tracking): Replace declaration with that of...
3222         (make_pass_variable_tracking): ...new function.
3223         (pass_free_cfg): Replace declaration with that of...
3224         (make_pass_free_cfg): ...new function.
3225         (pass_machine_reorg): Replace declaration with that of...
3226         (make_pass_machine_reorg): ...new function.
3227         (pass_cleanup_barriers): Replace declaration with that of...
3228         (make_pass_cleanup_barriers): ...new function.
3229         (pass_delay_slots): Replace declaration with that of...
3230         (make_pass_delay_slots): ...new function.
3231         (pass_split_for_shorten_branches): Replace declaration with that of...
3232         (make_pass_split_for_shorten_branches): ...new function.
3233         (pass_split_before_regstack): Replace declaration with that of...
3234         (make_pass_split_before_regstack): ...new function.
3235         (pass_convert_to_eh_region_ranges): Replace declaration with that
3236         of...
3237         (make_pass_convert_to_eh_region_ranges): ...new function.
3238         (pass_shorten_branches): Replace declaration with that of...
3239         (make_pass_shorten_branches): ...new function.
3240         (pass_set_nothrow_function_flags): Replace declaration with that of...
3241         (make_pass_set_nothrow_function_flags): ...new function.
3242         (pass_dwarf2_frame): Replace declaration with that of...
3243         (make_pass_dwarf2_frame): ...new function.
3244         (pass_final): Replace declaration with that of...
3245         (make_pass_final): ...new function.
3246         (pass_rtl_seqabstr): Replace declaration with that of...
3247         (make_pass_rtl_seqabstr): ...new function.
3248         (pass_release_ssa_names): Replace declaration with that of...
3249         (make_pass_release_ssa_names): ...new function.
3250         (pass_early_inline): Replace declaration with that of...
3251         (make_pass_early_inline): ...new function.
3252         (pass_inline_parameters): Replace declaration with that of...
3253         (make_pass_inline_parameters): ...new function.
3254         (pass_update_address_taken): Replace declaration with that of...
3255         (make_pass_update_address_taken): ...new function.
3256         (pass_convert_switch): Replace declaration with that of...
3257         (make_pass_convert_switch): ...new function.
3258         * tree-profile.c (pass_ipa_tree_profile): Convert from a global struct
3259         to a subclass of simple_ipa_opt_pass along with...
3260         (pass_data_ipa_tree_profile): ...new pass_data instance and...
3261         (make_pass_ipa_tree_profile): ...new function.
3262         * tree-sra.c (pass_sra_early): Convert from a global struct to a
3263         subclass of gimple_opt_pass along with...
3264         (pass_data_sra_early): ...new pass_data instance and...
3265         (make_pass_sra_early): ...new function.
3266         (pass_sra): Convert from a global struct to a subclass of
3267         gimple_opt_pass along with...
3268         (pass_data_sra): ...new pass_data instance and...
3269         (make_pass_sra): ...new function.
3270         (pass_early_ipa_sra): Convert from a global struct to a subclass of
3271         gimple_opt_pass along with...
3272         (pass_data_early_ipa_sra): ...new pass_data instance and...
3273         (make_pass_early_ipa_sra): ...new function.
3274         * tree-ssa-ccp.c (pass_ccp): Convert from a global struct to a
3275         subclass of gimple_opt_pass along with...
3276         (pass_data_ccp): ...new pass_data instance and...
3277         (make_pass_ccp): ...new function.
3278         (pass_fold_builtins): Convert from a global struct to a subclass of
3279         gimple_opt_pass along with...
3280         (pass_data_fold_builtins): ...new pass_data instance and...
3281         (make_pass_fold_builtins): ...new function.
3282         * tree-ssa-copy.c (pass_copy_prop): Convert from a global struct to a
3283         subclass of gimple_opt_pass along with...
3284         (pass_data_copy_prop): ...new pass_data instance and...
3285         (make_pass_copy_prop): ...new function.
3286         * tree-ssa-copyrename.c (pass_rename_ssa_copies): Convert from a
3287         global struct to a subclass of gimple_opt_pass along with...
3288         (pass_data_rename_ssa_copies): ...new pass_data instance and...
3289         (make_pass_rename_ssa_copies): ...new function.
3290         * tree-ssa-dce.c (pass_dce): Convert from a global struct to a
3291         subclass of gimple_opt_pass along with...
3292         (pass_data_dce): ...new pass_data instance and...
3293         (make_pass_dce): ...new function.
3294         (pass_dce_loop): Convert from a global struct to a subclass of
3295         gimple_opt_pass along with...
3296         (pass_data_dce_loop): ...new pass_data instance and...
3297         (make_pass_dce_loop): ...new function.
3298         (pass_cd_dce): Convert from a global struct to a subclass of
3299         gimple_opt_pass along with...
3300         (pass_data_cd_dce): ...new pass_data instance and...
3301         (make_pass_cd_dce): ...new function.
3302         * tree-ssa-dom.c (pass_dominator): Convert from a global struct to a
3303         subclass of gimple_opt_pass along with...
3304         (pass_data_dominator): ...new pass_data instance and...
3305         (make_pass_dominator): ...new function.
3306         (pass_phi_only_cprop): Convert from a global struct to a subclass of
3307         gimple_opt_pass along with...
3308         (pass_data_phi_only_cprop): ...new pass_data instance and...
3309         (make_pass_phi_only_cprop): ...new function.
3310         * tree-ssa-dse.c (pass_dse): Convert from a global struct to a
3311         subclass of gimple_opt_pass along with...
3312         (pass_data_dse): ...new pass_data instance and...
3313         (make_pass_dse): ...new function.
3314         * tree-ssa-forwprop.c (pass_forwprop): Convert from a global struct to
3315         a subclass of gimple_opt_pass along with...
3316         (pass_data_forwprop): ...new pass_data instance and...
3317         (make_pass_forwprop): ...new function.
3318         * tree-ssa-ifcombine.c (pass_tree_ifcombine): Convert from a global
3319         struct to a subclass of gimple_opt_pass along with...
3320         (pass_data_tree_ifcombine): ...new pass_data instance and...
3321         (make_pass_tree_ifcombine): ...new function.
3322         * tree-ssa-loop-ch.c (pass_ch): Convert from a global struct to a
3323         subclass of gimple_opt_pass along with...
3324         (pass_data_ch): ...new pass_data instance and...
3325         (make_pass_ch): ...new function.
3326         * tree-ssa-loop.c (pass_tree_loop): Convert from a global struct to a
3327         subclass of gimple_opt_pass along with...
3328         (pass_data_tree_loop): ...new pass_data instance and...
3329         (make_pass_tree_loop): ...new function.
3330         (pass_tree_loop_init): Convert from a global struct to a subclass of
3331         gimple_opt_pass along with...
3332         (pass_data_tree_loop_init): ...new pass_data instance and...
3333         (make_pass_tree_loop_init): ...new function.
3334         (pass_lim): Convert from a global struct to a subclass of
3335         gimple_opt_pass along with...
3336         (pass_data_lim): ...new pass_data instance and...
3337         (make_pass_lim): ...new function.
3338         (pass_tree_unswitch): Convert from a global struct to a subclass of
3339         gimple_opt_pass along with...
3340         (pass_data_tree_unswitch): ...new pass_data instance and...
3341         (make_pass_tree_unswitch): ...new function.
3342         (pass_predcom): Convert from a global struct to a subclass of
3343         gimple_opt_pass along with...
3344         (pass_data_predcom): ...new pass_data instance and...
3345         (make_pass_predcom): ...new function.
3346         (pass_vectorize): Convert from a global struct to a subclass of
3347         gimple_opt_pass along with...
3348         (pass_data_vectorize): ...new pass_data instance and...
3349         (make_pass_vectorize): ...new function.
3350         (pass_graphite): Convert from a global struct to a subclass of
3351         gimple_opt_pass along with...
3352         (pass_data_graphite): ...new pass_data instance and...
3353         (make_pass_graphite): ...new function.
3354         (pass_graphite_transforms): Convert from a global struct to a subclass
3355         of gimple_opt_pass along with...
3356         (pass_data_graphite_transforms): ...new pass_data instance and...
3357         (make_pass_graphite_transforms): ...new function.
3358         (pass_check_data_deps): Convert from a global struct to a subclass of
3359         gimple_opt_pass along with...
3360         (pass_data_check_data_deps): ...new pass_data instance and...
3361         (make_pass_check_data_deps): ...new function.
3362         (pass_iv_canon): Convert from a global struct to a subclass of
3363         gimple_opt_pass along with...
3364         (pass_data_iv_canon): ...new pass_data instance and...
3365         (make_pass_iv_canon): ...new function.
3366         (pass_scev_cprop): Convert from a global struct to a subclass of
3367         gimple_opt_pass along with...
3368         (pass_data_scev_cprop): ...new pass_data instance and...
3369         (make_pass_scev_cprop): ...new function.
3370         (pass_record_bounds): Convert from a global struct to a subclass of
3371         gimple_opt_pass along with...
3372         (pass_data_record_bounds): ...new pass_data instance and...
3373         (make_pass_record_bounds): ...new function.
3374         (pass_complete_unroll): Convert from a global struct to a subclass of
3375         gimple_opt_pass along with...
3376         (pass_data_complete_unroll): ...new pass_data instance and...
3377         (make_pass_complete_unroll): ...new function.
3378         (pass_complete_unrolli): Convert from a global struct to a subclass of
3379         gimple_opt_pass along with...
3380         (pass_data_complete_unrolli): ...new pass_data instance and...
3381         (make_pass_complete_unrolli): ...new function.
3382         (pass_parallelize_loops): Convert from a global struct to a subclass
3383         of gimple_opt_pass along with...
3384         (pass_data_parallelize_loops): ...new pass_data instance and...
3385         (make_pass_parallelize_loops): ...new function.
3386         (pass_loop_prefetch): Convert from a global struct to a subclass of
3387         gimple_opt_pass along with...
3388         (pass_data_loop_prefetch): ...new pass_data instance and...
3389         (make_pass_loop_prefetch): ...new function.
3390         (pass_iv_optimize): Convert from a global struct to a subclass of
3391         gimple_opt_pass along with...
3392         (pass_data_iv_optimize): ...new pass_data instance and...
3393         (make_pass_iv_optimize): ...new function.
3394         (pass_tree_loop_done): Convert from a global struct to a subclass of
3395         gimple_opt_pass along with...
3396         (pass_data_tree_loop_done): ...new pass_data instance and...
3397         (make_pass_tree_loop_done): ...new function.
3398         * tree-ssa-math-opts.c (pass_cse_reciprocals): Convert from a global
3399         struct to a subclass of gimple_opt_pass along with...
3400         (pass_data_cse_reciprocals): ...new pass_data instance and...
3401         (make_pass_cse_reciprocals): ...new function.
3402         (pass_cse_sincos): Convert from a global struct to a subclass of
3403         gimple_opt_pass along with...
3404         (pass_data_cse_sincos): ...new pass_data instance and...
3405         (make_pass_cse_sincos): ...new function.
3406         (pass_optimize_bswap): Convert from a global struct to a subclass of
3407         gimple_opt_pass along with...
3408         (pass_data_optimize_bswap): ...new pass_data instance and...
3409         (make_pass_optimize_bswap): ...new function.
3410         (pass_optimize_widening_mul): Convert from a global struct to a
3411         subclass of gimple_opt_pass along with...
3412         (pass_data_optimize_widening_mul): ...new pass_data instance and...
3413         (make_pass_optimize_widening_mul): ...new function.
3414         * tree-ssa-phiopt.c (pass_phiopt): Convert from a global struct to a
3415         subclass of gimple_opt_pass along with...
3416         (pass_data_phiopt): ...new pass_data instance and...
3417         (make_pass_phiopt): ...new function.
3418         (pass_cselim): Convert from a global struct to a subclass of
3419         gimple_opt_pass along with...
3420         (pass_data_cselim): ...new pass_data instance and...
3421         (make_pass_cselim): ...new function.
3422         * tree-ssa-phiprop.c (pass_phiprop): Convert from a global struct to a
3423         subclass of gimple_opt_pass along with...
3424         (pass_data_phiprop): ...new pass_data instance and...
3425         (make_pass_phiprop): ...new function.
3426         * tree-ssa-pre.c (pass_pre): Convert from a global struct to a
3427         subclass of gimple_opt_pass along with...
3428         (pass_data_pre): ...new pass_data instance and...
3429         (make_pass_pre): ...new function.
3430         (pass_fre): Convert from a global struct to a subclass of
3431         gimple_opt_pass along with...
3432         (pass_data_fre): ...new pass_data instance and...
3433         (make_pass_fre): ...new function.
3434         * tree-ssa-reassoc.c (pass_reassoc): Convert from a global struct to a
3435         subclass of gimple_opt_pass along with...
3436         (pass_data_reassoc): ...new pass_data instance and...
3437         (make_pass_reassoc): ...new function.
3438         * tree-ssa-sink.c (pass_sink_code): Convert from a global struct to a
3439         subclass of gimple_opt_pass along with...
3440         (pass_data_sink_code): ...new pass_data instance and...
3441         (make_pass_sink_code): ...new function.
3442         * tree-ssa-strlen.c (pass_strlen): Convert from a global struct to a
3443         subclass of gimple_opt_pass along with...
3444         (pass_data_strlen): ...new pass_data instance and...
3445         (make_pass_strlen): ...new function.
3446         * tree-ssa-structalias.c (pass_build_alias): Convert from a global
3447         struct to a subclass of gimple_opt_pass along with...
3448         (pass_data_build_alias): ...new pass_data instance and...
3449         (make_pass_build_alias): ...new function.
3450         (pass_build_ealias): Convert from a global struct to a subclass of
3451         gimple_opt_pass along with...
3452         (pass_data_build_ealias): ...new pass_data instance and...
3453         (make_pass_build_ealias): ...new function.
3454         (pass_ipa_pta): Convert from a global struct to a subclass of
3455         simple_ipa_opt_pass along with...
3456         (pass_data_ipa_pta): ...new pass_data instance and...
3457         (make_pass_ipa_pta): ...new function.
3458         * tree-ssa-uncprop.c (pass_uncprop): Convert from a global struct to a
3459         subclass of gimple_opt_pass along with...
3460         (pass_data_uncprop): ...new pass_data instance and...
3461         (make_pass_uncprop): ...new function.
3462         * tree-ssa-uninit.c (pass_late_warn_uninitialized): Convert from a
3463         global struct to a subclass of gimple_opt_pass along with...
3464         (pass_data_late_warn_uninitialized): ...new pass_data instance and...
3465         (make_pass_late_warn_uninitialized): ...new function.
3466         * tree-ssa.c (pass_init_datastructures): Convert from a global struct
3467         to a subclass of gimple_opt_pass along with...
3468         (pass_data_init_datastructures): ...new pass_data instance and...
3469         (make_pass_init_datastructures): ...new function.
3470         (pass_early_warn_uninitialized): Convert from a global struct to a
3471         subclass of gimple_opt_pass along with...
3472         (pass_data_early_warn_uninitialized): ...new pass_data instance and...
3473         (make_pass_early_warn_uninitialized): ...new function.
3474         (pass_update_address_taken): Convert from a global struct to a
3475         subclass of gimple_opt_pass along with...
3476         (pass_data_update_address_taken): ...new pass_data instance and...
3477         (make_pass_update_address_taken): ...new function.
3478         * tree-ssanames.c (pass_release_ssa_names): Convert from a global
3479         struct to a subclass of gimple_opt_pass along with...
3480         (pass_data_release_ssa_names): ...new pass_data instance and...
3481         (make_pass_release_ssa_names): ...new function.
3482         * tree-stdarg.c (pass_stdarg): Convert from a global struct to a
3483         subclass of gimple_opt_pass along with...
3484         (pass_data_stdarg): ...new pass_data instance and...
3485         (make_pass_stdarg): ...new function.
3486         * tree-switch-conversion.c (pass_convert_switch): Convert from a
3487         global struct to a subclass of gimple_opt_pass along with...
3488         (pass_data_convert_switch): ...new pass_data instance and...
3489         (make_pass_convert_switch): ...new function.
3490         * tree-tailcall.c (pass_tail_recursion): Convert from a global struct
3491         to a subclass of gimple_opt_pass along with...
3492         (pass_data_tail_recursion): ...new pass_data instance and...
3493         (make_pass_tail_recursion): ...new function.
3494         (pass_tail_calls): Convert from a global struct to a subclass of
3495         gimple_opt_pass along with...
3496         (pass_data_tail_calls): ...new pass_data instance and...
3497         (make_pass_tail_calls): ...new function.
3498         * tree-vect-generic.c (pass_lower_vector): Convert from a global
3499         struct to a subclass of gimple_opt_pass along with...
3500         (pass_data_lower_vector): ...new pass_data instance and...
3501         (make_pass_lower_vector): ...new function.
3502         (pass_lower_vector_ssa): Convert from a global struct to a subclass of
3503         gimple_opt_pass along with...
3504         (pass_data_lower_vector_ssa): ...new pass_data instance and...
3505         (make_pass_lower_vector_ssa): ...new function.
3506         * tree-vectorizer.c (pass_slp_vectorize): Convert from a global struct
3507         to a subclass of gimple_opt_pass along with...
3508         (pass_data_slp_vectorize): ...new pass_data instance and...
3509         (make_pass_slp_vectorize): ...new function.
3510         (pass_ipa_increase_alignment): Convert from a global struct to a
3511         subclass of simple_ipa_opt_pass along with...
3512         (pass_data_ipa_increase_alignment): ...new pass_data instance and...
3513         (make_pass_ipa_increase_alignment): ...new function.
3514         * tree-vrp.c (pass_vrp): Convert from a global struct to a subclass of
3515         gimple_opt_pass along with...
3516         (pass_data_vrp): ...new pass_data instance and...
3517         (make_pass_vrp): ...new function.
3518         * tree.c (pass_ipa_free_lang_data): Convert from a global struct to a
3519         subclass of simple_ipa_opt_pass along with...
3520         (pass_data_ipa_free_lang_data): ...new pass_data instance and...
3521         (make_pass_ipa_free_lang_data): ...new function.
3522         * tsan.c (pass_tsan): Convert from a global struct to a subclass of
3523         gimple_opt_pass along with...
3524         (pass_data_tsan): ...new pass_data instance and...
3525         (make_pass_tsan): ...new function.
3526         (pass_tsan_O0): Convert from a global struct to a subclass of
3527         gimple_opt_pass along with...
3528         (pass_data_tsan_O0): ...new pass_data instance and...
3529         (make_pass_tsan_O0): ...new function.
3530         * var-tracking.c (pass_variable_tracking): Convert from a global
3531         struct to a subclass of rtl_opt_pass along with...
3532         (pass_data_variable_tracking): ...new pass_data instance and...
3533         (make_pass_variable_tracking): ...new function.
3534         * web.c (pass_web): Convert from a global struct to a subclass of
3535         rtl_opt_pass along with...
3536         (pass_data_web): ...new pass_data instance and...
3537         (make_pass_web): ...new function.
3538         * config/epiphany/epiphany.h (pass_mode_switch_use): Replace
3539         declaration with that of...
3540         (make_pass_mode_switch_use): ...new function.
3541         (pass_resolve_sw_modes): Replace declaration with that of...
3542         (make_pass_resolve_sw_modes): ...new function.
3543         * config/epiphany/mode-switch-use.c (pass_mode_switch_use): Convert
3544         from a global struct to a subclass of rtl_opt_pass along with...
3545         (pass_data_mode_switch_use): ...new pass_data instance and...
3546         (make_pass_mode_switch_use): ...new function.
3547         * config/epiphany/resolve-sw-modes.c (pass_resolve_sw_modes): Convert
3548         from a global struct to a subclass of rtl_opt_pass along with...
3549         (pass_data_resolve_sw_modes): ...new pass_data instance and...
3550         (make_pass_resolve_sw_modes): ...new function.
3551         * config/i386/i386.c (pass_insert_vzeroupper): Convert from a global
3552         struct to a subclass of rtl_opt_pass along with...
3553         (pass_data_insert_vzeroupper): ...new pass_data instance and...
3554         (make_pass_insert_vzeroupper): ...new function.
3555         * config/sparc/sparc.c (pass_work_around_errata): Convert from a
3556         global struct to a subclass of rtl_opt_pass along with...
3557         (pass_data_work_around_errata): ...new pass_data instance and...
3558         (make_pass_work_around_errata): ...new function.
3559         * config/mips/mips.c (pass_mips_machine_reorg2): Convert from a global
3560         struct to a subclass of rtl_opt_pass along with...
3561         (pass_data_mips_machine_reorg2): ...new pass_data instance and...
3562         (make_pass_mips_machine_reorg2): ...new function.
3564 2013-08-05  David Malcolm  <dmalcolm@redhat.com>
3566         * passes.c (pass_manager::operator new): New.
3568 2013-08-05  David Malcolm  <dmalcolm@redhat.com>
3570         Handwritten part of conversion of passes to C++ classes.
3572         * Makefile.in (PASS_MANAGER_H): Add dep on pass-instances.def.
3573         (toplev.o): Add dep on PASS_MANAGER_H.
3574         * cgraphunit.c (cgraph_process_new_functions): Rework invocation
3575         of early local pases to reflect this moving from a global to a
3576         member of gcc::pass_manager.
3577         (cgraph_add_new_function): Likewise.
3578         * lto-cgraph.c (lto_output_node): Update for conversion of
3579         struct ipa_opt_pass_d to a C++ subclass of opt_pass.
3580         * passes.c (opt_pass::clone): New.
3581         (opt_pass::gate): New.
3582         (opt_pass::execute): New.
3583         (opt_pass::opt_pass): New.
3584         (pass_manager::execute_early_local_passes): New.
3585         (pass_manager::execute_pass_mode_switching): new.
3586         (finish_optimization_passes): Convert to...
3587         (pass_manager::finish_optimization_passes): ...this.
3588         (finish_optimization_passes): Update for conversion of passes to
3589         C++ classes.
3590         (register_dump_files_1): Use has_gate since we cannot portably
3591         check a vtable entry against NULL.
3592         (dump_one_pass): Likewise.
3593         (ipa_write_summaries_2): Likewise.
3594         (ipa_write_optimization_summaries_1): Likewise.
3595         (ipa_read_summaries_1): Likewise.
3596         (ipa_read_optimization_summaries_1): Likewise.
3597         (execute_ipa_stmt_fixups): Likewise.
3598         (pass_manager::pass_manager): Rewrite pass-creation, invoking
3599         pass-creation functions rather than wiring up globals, and
3600         storing the results in fields of pass_manager generated using
3601         pass-instances.def.
3602         (pass_manager::dump_profile_report): Update for conversion of
3603         passes to C++ classes.
3604         (pass_manager::execute_ipa_summary_passes): Likewise.
3605         (execute_one_ipa_transform_pass): Likewise.
3606         (execute_one_pass): Use has_gate and has_execute since we cannot
3607         portably check a vtable entry against NULL.
3608         * pass_manager.h (pass_manager::finish_optimization_passes): New.
3609         (pass_manager): Use pass-instances.def to add fields for the
3610         various pass instances.
3611         * toplev.c (finalize): Update for move of
3612         finish_optimization_passes to a method of gcc::pass_manager.
3613         * toplev.h (finish_optimization_passes): Move to method of class
3614         pass_manager.
3615         * tree-pass.h (struct pass_data): New.
3616         (opt_pass): Convert to C++ class, make it a subclass of pass_data.
3617         (opt_pass::gate): Convert to virtual function.
3618         (opt_pass::~opt_pass): New.
3619         (opt_pass::clone): New.
3620         (opt_pass::execute): Convert to virtual function.
3621         (opt_pass::opt_pass): New.
3622         (opt_pass::ctxt_): new.
3623         (gimple_opt_pass): Convert to subclass of opt_pass.
3624         (gimple_opt_pass::gimple_opt_pass): New.
3625         (rtl_opt_pass): Convert to subclass of opt_pass.
3626         (rtl_opt_pass::rtl_opt_pass): New.
3627         (ipa_opt_pass_d): Convert to subclass of opt_pass.
3628         (ipa_opt_pass_d::ipa_opt_pass_d): New.
3629         (simple_ipa_opt_pass): Convert to subclass of opt_pass.
3630         (simple_ipa_opt_pass::simple_ipa_opt_pass): New.
3631         * config/i386/i386.c (rest_of_handle_insert_vzeroupper): Rework
3632         invocation of pass_mode_switching to reflect this moving from a
3633         global to a member of gcc::pass_manager.
3634         (ix86_option_override): Rework how pass_insert_vzeroupper is
3635         added to the pass_manager to reflect autogenerated changes.
3636         * config/i386/t-i386 (i386.o) Add deps on CONTEXT_H and PASS_MANAGER_H.
3638 2013-08-05  Richard Earnshaw  <rearnsha@arm.com>
3640         PR rtl-optimization/57708
3641         * recog.c (peep2_find_free_register): Validate all regs in a
3642         multi-reg mode.
3644 2013-08-05  Jan Hubicka  <jh@suse.cz>
3646         PR lto/57602
3647         * cgraph.c (verify_cgraph_node): Accept local flags from other
3648         partitions.
3649         * ipa.c (symtab_remove_unreachable_nodes): Do not clear local flag.
3650         (function_and_variable_visibility): Likewise.
3651         * trans-mem.c (ipa_tm_create_version): TM versions are not local.
3653 2013-08-05  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3655         * graph.c (init_graph_slim_pretty_print): Remove.
3656         (print_graph_cfg): Do not call it.  Use local pretty printer.
3657         (start_graph_dump): Likewise.
3659 2013-08-05  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3661         * gimple-pretty-print.c (buffer): Remove.
3662         (initialized): Likewise.
3663         (maybe_init_pretty_print): Likewise.
3664         (print_gimple_stmt): Do not call it.  Use non-static local
3665         pretty_printer variable.
3666         (print_gimple_expr): Likewise.
3667         (print_gimple_seq): Likewise.
3668         (gimple_dump_bb): Likewise.
3670 2013-08-05  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3672         * asan.c (asan_pp): Remove.
3673         (asan_pp_initialized): Likewise.
3674         (asan_pp_initialize): Likewise.
3675         (asan_pp_string): Take a pretty_printer parameter.  Adjust callers.
3676         (asan_emit_stack_protection): Tidy.  Use local pretty printer.
3677         (asan_add_global): Likewise.
3679 2013-08-04  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3681         * pretty-print.h (pp_base): Remove.  Adjust dependent macros.
3682         * diagnostic.h (diagnostic_flush_buffer): Adjust.
3683         * pretty-print.c (pp_formatted_text_data): Likewise.
3684         (pp_indent): Rename from pp_base_indent.
3685         (pp_format): Rename from pp_base_format.
3686         (pp_output_formatted_text): Rename from pp_base_output_formatted_text.
3687         (pp_format_verbatim): Rename from pp_base_format_verbatim.
3688         (pp_flush): Rename from pp_base_flush.
3689         (pp_set_line_maximum_length): Rename from
3690         pp_base_set_line_maximum_length.
3691         (pp_clear_output_area): Rename from pp_base_clear_output_area.
3692         (pp_set_prefix): Rename from pp_base_set_prefix.
3693         (pp_destroy_prefix): Rename from pp_base_destroy_prefix.
3694         (pp_emit_prefix): Rename from pp_base_emit_prefix.
3695         (pp_append_text): Rename from pp_base_append_text.
3696         (pp_formatted_text): Rename from pp_base_formatted_text.
3697         (pp_last_position_in_text): Rename from pp_base_last_position_in_text.
3698         (pp_remaining_character_count_for_line): Rename from
3699         pp_base_remaining_character_count_for_line.
3700         (pp_newline): Rename from pp_base_newline.
3701         (pp_character): Rename from pp_base_character.
3702         (pp_string): Rename from pp_base_string.
3703         (pp_maybe_space): Rename from pp_base_maybe_space.
3704         * asan.c (asan_pp_string): Adjust.
3705         (asan_emit_stack_protection): Likewise.
3706         (asan_add_global): Likewise.
3707         * sched-vis.c (str_pattern_slim): Adjust pretty printer function call.
3708         * tree-mudflap.c (mf_varname_tree): Likewise.
3709         * tree-pretty-print.c (pp_tree_identifier): Rename from
3710         pp_base_tree_identifier.
3711         * tree-pretty-print.h (pp_tree_identifier): Remove macro definition.
3712         Declare as function.
3714 2013-08-03  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3716         * pretty-print.h (pp_bar_bar): New.
3717         (pp_ampersand_ampersand): Likewise.
3718         (pp_less_equal): Likewise.
3719         (pp_greater_equal): Likewise.
3720         * gimple-pretty-print.c (dump_ternary_rhs): Use specialized pretty
3721         printer functions instead of pp_string or operators and punctuators.
3722         (dump_gimple_call): Likewise.
3723         (dump_gimple_omp_for): Likewise.
3724         (dump_gimple_transaction): Likewise.
3725         (dump_gimple_phi): Likewise.
3726         (pp_gimple_stmt_1): Likewise.
3727         * sched-vis.c (print_insn): Likewise.
3728         * tree-mudflap.c (mf_varname_tree): Likewise.
3729         * tree-pretty-print.c (dump_block_node): Likewise.
3730         (dump_generic_node): Likewise.
3732 2013-08-02  Jan Hubicka  <jh@suse.cz>
3734         * lto-cgraph.c (compute_ltrans_boundary): Add abstract origins into
3735         boundaries.
3736         * lto-streamer-out.c (tree_is_indexable): Results decls and
3737         parm decls are not indexable.
3738         (DFS_write_tree_body): Do not follow args and results.
3739         (hash_tree): Likewise.
3740         (output_functions): Rearrange so struct function is needed
3741         only when real body is output; be able to also ouptut abstract
3742         functions; output DECL_ARGUMENTS and DECL_RESULT.
3743         (lto_output): When not in WPA, ale store abstract functions.
3744         (write_symbol): Do not care about RESULT_DECL.
3745         (output_symbol_p): Handle correctly sbtract decls.
3746         * lto-streamer-in.c (input_function): Rearrange so struct
3747         function can be NULL at entry; allow streaming of
3748         functions w/o body; store DECL_ARGUMENTS and DECL_RESULT.
3749         * ipa.c (symtab_remove_unreachable_nodes): Silence confused
3750         sanity check during LTO.
3751         * tree-streamer-out.c (write_ts_decl_non_common_tree_pointers): Skip
3752         RESULT_DECl and DECL_ARGUMENTS.
3753         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
3754         Likewise.
3756 2013-08-03  Gabriel Dos Reis  <gdr@integrable-solutions.net>
3758         * pretty-print.h (pp_underscore): New.
3759         (pp_comma): Tidy.
3760         * gimple-pretty-print.c (dump_unary_rhs): Use specialized pretty
3761         printer functions instead of pp_character.
3762         (dump_binary_rhs): Likewise.
3763         (dump_ternary_rhs): Likewise.
3764         (dump_gimple_call_args): Likewise.
3765         (pp_points_to_solution): Likewise.
3766         (dump_gimple_call): Likewise.
3767         (dump_gimple_switch): Likewise.
3768         (dump_gimple_cond): Likewise.
3769         (dump_gimple_bind): Likewise.
3770         (dump_gimple_try): Likewise.
3771         (dump_gimple_omp_for): Likewise.
3772         (dump_gimple_omp_continue): Likewise.
3773         (dump_gimple_omp_single): Likewise.
3774         (dump_gimple_omp_sections): Likewise.
3775         (dump_gimple_omp_block): Likewise.
3776         (dump_gimple_omp_critical): Likewise.
3777         (dump_gimple_transaction): Likewise.
3778         (dump_gimple_asm): Likewise.
3779         (dump_gimple_phi): Likewise.
3780         (dump_gimple_omp_parallel): Likewise.
3781         (dump_gimple_omp_task): Likewise.
3782         (dump_gimple_omp_atomic_load): Likewise.
3783         (dump_gimple_omp_atomic_store): Likewise.
3784         (dump_gimple_mem_ops): Likewise.
3785         (pp_gimple_stmt_1): Likewise.
3786         (pp_cfg_jump): Likewise.
3787         (dump_implicit_edges): Likewise.
3788         (gimple_dump_bb_for_graph): Likewise.
3789         * graph.c (draw_cfg_node): Likewise.
3790         * langhooks.c (lhd_print_error_function): Likewise.
3791         * sched-vis.c (print_exp): Likewise.
3792         (print_value): Likewise.
3793         (print_pattern): Likewise.
3794         (print_insn): Likewise.
3795         (rtl_dump_bb_for_graph): Likewise.
3796         * tree-pretty-print.c (dump_function_declaration): Likewise.
3797         (dump_array_domain): Likewise.
3798         (dump_omp_clause): Likewise.
3799         (dump_location): Likewise.
3800         (dump_generic_node): Likewise.
3801         (print_struct_decl): Likewise.
3802         * diagnostic.c (diagnostic_show_locus): Use pp_space.
3804 2013-08-03  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
3806         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Update
3807         candidate table when replacing a candidate statement.
3808         (replace_rhs_if_not_dup): Likewise.
3809         (replace_one_candidate): Likewise.
3811 2013-08-02  Jan Hubicka  <jh@suse.cz>
3812             Martin Liska  <marxin.liska@gmail.com>
3814         * cgraphunit.c (add_new_function): Fix logic when adding from
3815         late IPA pass.
3816         (assemble_thunk): Rename to ...
3817         (expand_thunk); .. this one; export; get it working with
3818         general functions; make produced gimple valid.
3819         * cgraph.h (expand_thunk): Declare.
3821 2013-08-02  Jan Hubicka  <jh@suse.cz>
3823         * ipa-cp.c (gather_context_independent_values): Use
3824         ipa_get_param_move_cost.
3825         (get_replacement_map): Remove PARAM; move parameter folding
3826         into tree-inline.c
3827         (create_specialized_node): Update.
3828         * ipa-prop.c (ipa_populate_param_decls): Do not look for origins;
3829         assert that we have gimple body; update move_cost.
3830         (count_formal_params): Assert that we have gimple body.
3831         (ipa_dump_param): New function.
3832         (ipa_alloc_node_params): Break out from ...
3833         (ipa_initialize_node_params): ... here.
3834         (ipa_get_vector_of_formal_parms): ICE when used in WPA.
3835         (ipa_write_node_info): Stream move costs.
3836         (ipa_read_node_info): Read move costs.
3837         (ipa_update_after_lto_read): Do not recompute node params.
3838         * ipa-prop.h (ipa_param_descriptor): Add move_cost.
3839         (ipa_get_param): Check we are not in WPA.
3840         (ipa_get_param_move_cost): New.
3841         * tree-inline.c (tree_function_versioning): Fold replacement as needed.
3842         * ipa-inline-analysis.c (inline_node_duplication_hook): Expect only
3843         parm numbers to be present.
3845 2013-08-02  Vladimir Makarov  <vmakarov@redhat.com>
3847         PR rtl-optimization/58048
3848         * lra-constraints.c (process_alt_operands): Don't check asm
3849         operand on register.
3851 2013-08-02  Eric Botcazou  <ebotcazou@adacore.com>
3853         * config/sparc/sparc.c (sparc_emit_membar_for_model) <SMM_TSO>: Add
3854         the implied StoreLoad barrier for atomic operations if before.
3856 2013-08-02  Jan Hubicka  <jh@suse.cz>
3857             Martin Liska  <marxin.liska@gmail.com>
3859         * cgraph.c (cgraph_function_body_availability): Do not check
3860         cgraph flags.
3861         * cgraph.h (symtab_for_node_and_aliases, symtab_nonoverwritable_alias,
3862         symtab_node_availability): Declare.
3863         * ipa.c (can_replace_by_local_alias): New.
3864         (function_and_variable_visibility): Use it.
3865         * symtab.c (symtab_for_node_and_aliases,
3866         symtab_nonoverwritable_alias_1, symtab_nonoverwritable_alias): New.
3868 2013-08-02  Vladimir Makarov  <vmakarov@redhat.com>
3870         PR rtl-optimization/57963
3871         * lra-constraints.c (reverse_equiv_p, contains_reloaded_insn_p): New.
3872         (lra_constraints): Use them.
3874 2013-08-02  Sofiane Naci  <sofiane.naci@arm.com>
3876         * config/arm/types.md (define_attr "type"): Add "load_acq"
3877         and "store_rel".
3878         * config/arm/cortex-a53.md (cortex_a53_load1): Update for attribute
3879         changes.
3880         (cortex_a53_store1): Likewise.
3882 2013-08-01  Jan Hubicka  <jh@suse.cz>
3884         * ipa.c (symtab_remove_unreachable_nodes): Nodes in other
3885         partitions are not needed.
3887 2013-08-01  Uros Bizjak  <ubizjak@gmail.com>
3889         * config/i386/i386.h (MAYBE_NON_Q_CLASS_P): New.
3890         * config/i386/i386.c (ix86_secondary_reload): Use INTEGER_CLASS_P and
3891         MAYBE_NON_Q_CLASS_P where appropriate.
3893 2013-08-01  Jan Hubicka  <jh@suse.cz>
3895         * cgraph.h (release_function_body): Declare.
3896         * tree.c (free_lang_data_in_decl): Free, parameters and return values
3897         of unused delcarations.
3899 2013-08-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3901         * config/arm/arm.md (minmax_arithsi_non_canon): Emit canonical
3902         RTL form when subtracting a constant.
3904 2013-08-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3906         * config/arm/arm.md (peepholes for eq (reg1) (reg2/imm)):
3907         Generate canonical plus rtx with negated immediate instead of minus
3908         where appropriate.
3909         * config/arm/arm.c (thumb2_reorg): Handle ADCS <Rd>, <Rn> case.
3911 2013-08-01  Jan Hubicka  <jh@suse.cz>
3913         * cgraph.c (cgraph_release_function_body): Use used_as_abstract_origin.
3914         (cgraph_release_function_body): Likewise.
3915         (cgraph_can_remove_if_no_direct_calls_p): Likewise.
3916         * cgraph.h (cgrpah_node): Rename abstract_and_needed
3917         to used_as_abstract_origin.
3918         * tree-inline-transfrom.c (can_remove_node_now_p_1): Do not remove
3919         symbols used as abstract origins.
3920         * cgraphunit.c (analyze_functions): Update.
3921         * ipa.c (symtab_remove_unreachable_nodes): Recompute
3922         used_as_abstract_origin.
3923         * tree-inline.c (tree_function_versioning): Update
3924         used_as_abstract_origin; be ready for DECL_RESULT and
3925         DECL_ARGUMENTS to be NULL.
3927         * lto-symtab.c (lto_symtab_merge_symbols): Merge duplicated nodes
3928         for abstract functions.
3929         * cgraph.h (symtab_real_symbol_p): Abstract declarations are not
3930         real symbols.
3932 2013-08-01  Jan Hubicka  <jh@suse.cz>
3934         * profile.c (compute_value_histograms): Fix thinko.
3936 2013-08-01  Sofiane Naci  <sofiane.naci@arm.com>
3938         * config.gcc (aarch64*-*-*): Add aarch-common.o to extra_objs.  Add
3939         aarch-common-protos.h to extra_headers.
3940         (aarch64*-*-*): Add arm/aarch-common-protos.h to tm_p_file.
3941         * config/aarch64/aarch64.md: Include "../arm/cortex-a53.md".
3942         * config/aarch64/t-aarch64 (aarch-common.o): Define.
3944 2013-08-01  Sofiane Naci  <sofiane.naci@arm.com>
3946         * config/aarch64/aarch64.md (define_attr "type"): Delete.
3947         Include "../arm/types.md".  Define "type" attribute for all patterns.
3948         * config/aarch64/aarch64-simd.md (move_lo_quad_<mode>): Update for
3949         attribute changes.
3951 2013-07-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
3953         * config/rs6000/predicates.md (fusion_gpr_addis): New predicates
3954         to support power8 load fusion.
3955         (fusion_gpr_mem_load): Likewise.
3957         * config/rs6000/rs6000-modes.def (PTImode): Update a comment.
3959         * config/rs6000/rs6000-protos.h (fusion_gpr_load_p): New
3960         declarations for power8 load fusion.
3961         (emit_fusion_gpr_load): Likewise.
3963         * config/rs6000/rs6000.c (rs6000_option_override_internal): If
3964         tuning for power8, turn on fusion mode by default.  Turn on sign
3965         extending fusion mode if normal fusion mode is on, and we are at
3966         -O2 or -O3.
3967         (fusion_gpr_load_p): New function, return true if we can fuse an
3968         addis instruction with a dependent load to a GPR.
3969         (emit_fusion_gpr_load): Emit the instructions for power8 load
3970         fusion to GPRs.
3972         * config/rs6000/vsx.md (VSX_M2): New iterator for fusion peepholes.
3973         (VSX load fusion peepholes): New peepholes to fuse together an
3974         addi instruction with a VSX load instruction.
3976         * config/rs6000/rs6000.md (GPR load fusion peepholes): New
3977         peepholes to fuse an addis instruction with a load to a GPR base
3978         register.  If we are supporting sign extending fusions, convert
3979         sign extending loads to zero extending loads and add an explicit
3980         sign extension.
3982 2013-07-31  Sofiane Naci  <sofiane.naci@arm.com>
3984         * config.gcc (arm*-*-*): Add aarch-common.o to extra_objs.  Add
3985         aarch-common-protos.h to extra_headers.
3986         (arm*-*-*): Add arm/aarch-common-protos.h to tm_p_file.
3987         * config/arm/arm.c (arm_early_load_addr_dep): Move from here to ...
3988         (arm_early_store_addr_dep): Likewise.
3989         (arm_no_early_alu_shift_dep): Likewise.
3990         (arm_no_early_alu_shift_value_dep): Likewise.
3991         (arm_no_early_mul_dep): Likewise.
3992         (arm_no_early_store_addr_dep): Likewise.
3993         (arm_mac_accumulator_is_mul_result): Likewise.
3994         (arm_mac_accumulator_is_result): Likewise.
3995         * config/arm/aarch-common.c: ... here.  New file.
3996         * config/arm/arm-protos.h (arm_early_load_addr_dep): Move from
3997         here to ...
3998         (arm_early_store_addr_dep): Likewise.
3999         (arm_no_early_alu_shift_dep): Likewise.
4000         (arm_no_early_alu_shift_value_dep): Likewise.
4001         (arm_no_early_mul_dep): Likewise.
4002         (arm_no_early_store_addr_dep): Likewise.
4003         (arm_mac_accumulator_is_mul_result): Likewise.
4004         (arm_mac_accumulator_is_result): Likewise.
4005         * config/arm/aarch-common-protos.h: ... here.  New file.
4006         * config/arm/t-arm (aarch-common.o): Define.
4008 2013-07-31  Sofiane Naci  <sofiane.naci@arm.com>
4010         * config/arm/arm.md: Include new file "types.md".
4011         (define_attr "type"): Move from here to ...
4012         (define_attr "mul32"): Likewise.
4013         (define_attr "mul64"): Likewise.
4014         * config/arm/types.md: ... here.  New file.
4016 2013-07-31  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4018         * config.gcc (*-*-rtems*): Use __cxa_atexit by default.
4019         * config/rs6000/rtems.h (TARGET_LIBGCC_SDATA_SECTION): Define.
4021 2013-07-31  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4023         * gen-pass-instances.awk: Fix offset of substr().
4025 2013-07-31  David Malcolm  <dmalcolm@redhat.com>
4027         * Makefile.in (pass-instances.def): New.
4028         (passes.o): Replace dependency on passes.def with one on
4029         pass-instances.def
4031         * gen-pass-instances.awk: New.
4033         * passes.c (pass_manager::pass_manager): Use pass-instances.def
4034         rather than passes.def, updating local definition of NEXT_PASS
4035         macro to add an extra NUM parameter (currently unused).
4037 2013-07-30  David Malcolm  <dmalcolm@redhat.com>
4039         * Makefile.in (PASS_MANAGER_H): New.
4040         (lto-cgraph.o): Depend on CONTEXT_H and PASS_MANAGER_H.
4041         (passes.o): Likewise.
4042         (statistics.o): Likewise.
4043         (cgraphunit.o): Likewise.
4044         (context.o): Depend on PASS_MANAGER_H.
4046         * pass_manager.h: New.
4048         * cgraphunit.c (cgraph_add_new_function): Update for moves
4049         of globals to fields of pass_manager.
4050         (analyze_function): Likewise.
4051         (expand_function): Likewise.
4052         (ipa_passes): Likewise.
4053         (compile): Likewise.
4055         * context.c (context::context): New.
4056         * context.h  (context::context): New.
4057         (context::get_passes): New.
4058         (context::passes_): New.
4060         * lto-cgraph.c (input_node): Update for moves of globals to
4061         fields of pass_manager.
4063         * passes.c (all_passes): Remove, in favor of a field of the
4064         same name within the new class pass_manager.
4065         (all_small_ipa_passes): Likewise.
4066         (all_lowering_passes): Likewise.
4067         (all_regular_ipa_passes): Likewise.
4068         (all_late_ipa_passes): Likewise.
4069         (all_lto_gen_passes): Likewise.
4070         (passes_by_id): Likewise.
4071         (passes_by_id_size): Likewise.
4072         (gcc_pass_lists): Remove, in favor of "pass_lists" field within
4073         the new class pass_manager.
4074         (set_pass_for_id): Convert to...
4075         (pass_manager::set_pass_for_id): ...method.
4076         (get_pass_for_id): Convert to...
4077         (pass_manager::get_pass_for_id): ...method.
4078         (register_one_dump_file): Move body of implementation into...
4079         (pass_manager::register_one_dump_file): ...here.
4080         (register_dump_files_1): Convert to...
4081         (pass_manager::register_dump_files_1): ...method.
4082         (register_dump_files): Convert to...
4083         (pass_manager::register_dump_files): ...method.
4084         (create_pass_tab): Update for moves of globals to fields of
4085         pass_manager.
4086         (dump_passes): Move body of implementation into...
4087         (pass_manager::dump_passes): ...here.
4088         (register_pass): Move body of implementation into...
4089         (pass_manager::register_pass): ...here.
4090         (init_optimization_passes): Convert into...
4091         (pass_manager::pass_manager): ...constructor for new
4092         pass_manager class, and initialize the pass_lists array.
4093         (check_profile_consistency): Update for moves of globals to
4094         fields of pass_manager.
4095         (dump_profile_report): Move body of implementation into...
4096         (pass_manager::dump_profile_report): ...here.
4097         (ipa_write_summaries_1): Update for moves of pass lists from
4098         being globals to fields of pass_manager.
4099         (ipa_write_optimization_summaries): Likewise.
4100         (ipa_read_summaries):  Likewise.
4101         (ipa_read_optimization_summaries): Likewise.
4102         (execute_all_ipa_stmt_fixups): Likewise.
4104         * statistics.c (statistics_fini): Update for moves of globals to
4105         fields of pass_manager.
4107         * toplev.c (general_init): Replace call to
4108         init_optimization_passes with construction of the pass_manager
4109         instance.
4111         * tree-pass.h (all_passes): Remove, in favor of a field of the
4112         same name within the new class pass_manager.
4113         (all_small_ipa_passes): Likewise.
4114         (all_lowering_passes): Likewise.
4115         (all_regular_ipa_passes): Likewise.
4116         (all_lto_gen_passes): Likewise.
4117         (all_late_ipa_passes): Likewise.
4118         (passes_by_id): Likewise.
4119         (passes_by_id_size): Likewise.
4120         (gcc_pass_lists): Remove, in favor of "pass_lists" field within
4121         the new class pass_manager.
4122         (get_pass_for_id): Remove.
4124 2013-07-30  Richard Earnshaw  <rearnsha@arm.com>
4126         * config.gcc (arm): Require 64-bit host-wide-int for all ARM target
4127         configs.
4129 2013-07-30  Richard Earnshaw  <rearnsha@arm.com>
4131         * arm.md (mulhi3): New expand pattern.
4133 2013-07-30  Jan Hubicka  <jh@suse.cz>
4134             Martin Liska  <marxin.liska@gmail.com>
4136         * profile.c (compute_value_histograms): Do not ICE when
4137         there is mismatch only on some counters.
4139 2013-07-30  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
4141         PR rtl-optimization/57637
4142         * function.c (move_insn_for_shrink_wrap): Also check the
4143         GEN set of the LIVE problem for the liveness analysis
4144         if it exists, otherwise give up.
4146 2013-07-29  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4148         PR tree-optimization/57993
4149         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Record
4150         replaced statement in the candidate table.
4151         (phi_add_costs): Return infinite cost when the hidden basis does
4152         not dominate all phis on which the candidate is dependent.
4153         (replace_one_candidate): Record replaced statement in the
4154         candidate table.
4156 2013-07-29  Joern Rennecke  <joern.rennecke@embecosm.com>
4158         * config/epiphany/epiphany.md (*isub_i+2): New peephole.
4159         (ashlv2si3): New expander.
4160         (*ashlv2si3_i): New define_insn_and_split.
4161         * predicates.md (float_operation): Allow patterns with three
4162         basic sub-patterns.
4164         PR rtl-optimization/58021
4165         * mode-switching.c (create_pre_exit): Always split off preceding
4166         insns if we are not at the basic block head.
4168 2013-07-29  Maciej W. Rozycki  <macro@codesourcery.com>
4170         * config/mips/linux.h (GLIBC_DYNAMIC_LINKER): Handle `-mnan=2008'.
4171         (UCLIBC_DYNAMIC_LINKER): New macro.
4172         * config/mips/linux64.h (GLIBC_DYNAMIC_LINKER32): Handle
4173         `-mnan=2008'.
4174         (GLIBC_DYNAMIC_LINKER64, GLIBC_DYNAMIC_LINKERN32): Likewise.
4175         (UCLIBC_DYNAMIC_LINKER32): Undefine macro first.  Handle
4176         `-mnan=2008'.
4177         (UCLIBC_DYNAMIC_LINKER64): Redefine macro.
4178         (UCLIBC_DYNAMIC_LINKERN32): Likewise.
4179         * config/mips/mips-modes.def: Remove RESET_FLOAT_FORMAT calls
4180         for SF and DF modes.  Use ieee_quad_format for TF mode.
4181         * config/mips/mips-opts.h (mips_ieee_754_setting): New enum.
4182         * config/mips/mips.c (mips_file_start): Output a `.nan' directive.
4183         (mips_option_override): Handle `-mnan=legacy'.
4184         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Handle
4185         `-mabs=2008' and `-mnan=2008'.
4186         (OPTION_DEFAULT_SPECS): Add "nan" default.
4187         (ASM_SPEC): Handle `-mnan='.
4188         [!HAVE_AS_NAN] (HAVE_AS_NAN): New macro.
4189         * config/mips/mips.md (abs<mode>2): Handle `-mabs=2008', update
4190         comment accordingly.
4191         (neg<mode>2): Likewise.
4192         * config/mips/mips.opt (mabs, mnan): New options.
4193         * doc/install.texi (Configuration): Document `--with-nan=' option.
4194         * doc/invoke.texi (Option Summary): List MIPS `-mabs=' and
4195         `-mnan=' options.
4196         (MIPS Options): Document them.
4197         * config.gcc <mips*-*-*>: Handle `--with-nan='.
4198         * configure.ac <mips*-*-*>: Check for GAS `-mnan=2008' support.
4199         * configure: Regenerate.
4200         * config.in: Regenerate.
4202 2013-07-29  Uros Bizjak  <ubizjak@gmail.com>
4204         * config/i386/i386.md (float post-reload splitters): Do not check
4205         for subregs of SSE registers.
4207 2013-07-29  Uros Bizjak  <ubizjak@gmail.com>
4208             H.J. Lu  <hongjiu.lu@intel.com>
4210         PR target/57954
4211         PR target/57988
4212         * config/i386/i386.md (post-reload splitter
4213         to avoid partial SSE reg dependency stalls): New pattern.
4215 2013-07-29  Dominik Vogt  <vogt@linux.vnet.ibm.com>
4217         * config/s390/s390.md ("movcc"): Swap load and store instructions.
4219 2013-07-27  Joern Rennecke  <joern.rennecke@embecosm.com>
4221         * config/epiphany/epiphany.c (epiphany_compute_frame_size):
4222         Also reserve space for saving UNKNOWN_REGNUM for leaf functions.
4224 2013-07-26  Cary Coutant  <ccoutant@google.com>
4226         * dwarf2out.c (die_checksum_ordered): Don't include template
4227         instantiations in signature.
4228         (is_template_parameter): New function.
4229         (is_template_instantiation): New function.
4230         (generate_skeleton_bottom_up): Don't include template instantiations
4231         in type unit DIE.
4232         (generate_skeleton): Likewise.
4233         (break_out_comdat_types): Move recursive call to break out nested
4234         types earlier.
4235         (prune_unused_types_mark_generic_parms_dies): Call
4236         is_template_parameter.
4238 2013-07-26  Ian Bolton  <ian.bolton@arm.com>
4240         * config/aarch64/aarch64.md (neg<mode>2): Offer alternative that
4241         uses vector registers.
4242         * config/aarch64/iterators.md: Add attributes rtn and vas.
4244 2013-07-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4245             Richard Earnshaw  <richard.earnshaw@arm.com>
4247         * combine.c (simplify_comparison): Re-canonicalize operands
4248         where appropriate.
4249         * config/arm/arm.md (movcond_addsi): New splitter.
4251 2013-07-25  Sterling Augustine  <saugustine@google.com>
4253         * dwarf2out.c (size_of_pubnames): Move code to...
4254         (include_pubname_in_output): ...here.  New.
4255         (want_pubnames): Rearrange.
4256         (output_pubnames): Call include_pubname_in_output.  Move assertion.
4258 2013-07-25  Cameron McInally  <cameron.mcinally@nyu.edu>
4260         * doc/extend.texi: Fix return types for __builtin_ia32_cmp*s builtins.
4262 2013-07-25  Cameron McInally  <cameron.mcinally@nyu.edu>
4264         PR target/38836
4265         * doc/extend.texi: Remove obsolete builtins. Fix
4266         typo for __builtin_ia32_loadss and __builtin_ia32_cmpnltss.
4268 2013-07-25  Jan Hubicka  <jh@suse.cz>
4270         * cgraph.c (release_function_body): Break out from ...
4271         (cgraph_release_function_body): ... this one; also release DECL_RESULT
4272         and DECL_ARGUMENTS.
4273         * ipa-cp.c (get_replacement_map): Add parm_num argument; do not set
4274         old_tree in the map.
4275         (create_specialized_node): Update.
4276         * lto-cgraph.c (output_node_opt_summary): Do not translate old_tree
4277         into index.
4278         * cgraphclones.c (cgraph_create_virtual_clone): Do not copy
4279         DECL_ARGUMENTS, DECL_INITIAL and DECL_RESULT.
4280         * ipa-prop.c (ipa_populate_param_decls): Look for origin of clones.
4281         * tree-inline.c (initialize_cfun): Initialize DECL_ARGUMENTS and
4282         DECL_RESULT.
4284 2013-07-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4286         * config/arm/arm.md (arm_addsi3, addsi3_carryin_<optab>,
4287         addsi3_carryin_alt2_<optab>): Correct output template.
4289 2013-07-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4291         * config/arm/arm-fixed.md (ssmulsa3, usmulusa3):
4292         Adjust for arm_restrict_it.
4293         Remove trailing whitespace.
4295 2013-07-25 Â Mark Kettenis Â <kettenis@openbsd.org>
4297         * config/pa/pa.c (pa_trampoline_init): Emit __enable_execute_stack
4298         libcall if HAVE_ENABLE_EXECUTE_STACK is defined.
4300         * config.gcc (hppa-*-openbsd*): Don't set tmake_file.
4302 2013-07-25  Vladimir Makarov  <vmakarov@redhat.com>
4304         PR rtl-optimization/57960
4305         * lra-constraints.c (process_alt_operands): Use the right mode
4306         when checking strict_low.
4308 2013-07-25  Jan Hubicka  <jh@suse.cz>
4310         * lto-symtab.c (lto_cgraph_replace_node): Release function body.
4311         * cgraph.c (cgraph_remove_node): Do not release function body
4312         when in cgraph streaming.
4313         * ipa.c (process_references, symtab_remove_unreachable_nodes): Objects
4314         in other partitions are not considered reachable; fix handling of
4315         clones.
4317 2013-07-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4319         * config/arm/arm.md (*sibcall_insn): Remove unnecessary space.
4321 2013-07-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4323         PR target/19599
4324         PR target/57731
4325         PR target/57837
4326         * config/arm/arm.md ("*sibcall_insn): Replace use of
4327         Ss with US. Adjust output for v5 and v4t.
4328         (*sibcall_value_insn): Likewise and loosen predicate on operand0.
4330         * config/arm/constraints.md ("Ss"): Rename to US.
4332 2013-07-25  Terry Guo  <terry.guo@arm.com>
4334         * config/arm/arm.c (thumb1_size_rtx_costs): Assign proper cost for
4335         shift_add/shift_sub0/shift_sub1 RTXs.
4337 2013-07-24  Bill Schmidt  <wschmidt@linux.ibm.com>
4338             Anton Blanchard  <anton@au1.ibm.com>
4340         * config/rs6000/altivec.md (altivec_vpkpx): Handle little endian.
4341         (altivec_vpks<VI_char>ss): Likewise.
4342         (altivec_vpks<VI_char>us): Likewise.
4343         (altivec_vpku<VI_char>us): Likewise.
4344         (altivec_vpku<VI_char>um): Likewise.
4346 2013-07-24  David Malcolm  <dmalcolm@redhat.com>
4348         Introduce context class.
4350         * Makefile.in (CONTEXT_H): New.
4351         (OBJS): Add context.o.
4352         (toplev.o): Add CONTEXT_H to dependencies.
4353         (context.o): New.
4355         * toplev.c (general_init): Create the singleton gcc::context instance.
4357         * context.c: New.
4359         * context.h: New.
4361 2013-07-24  Joern Rennecke  <joern.rennecke@embecosm.com>
4363         PR rtl-optimization/57968
4364         * mode-switching.c (create_pre_exit): Allow instructions that
4365         don't set a return register to need a non-exit mode.
4367 2013-07-24  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4368             Anton Blanchard  <anton@au1.ibm.com>
4370         * config/rs6000/vector.md (vec_realign_load_<mode>): Reorder input
4371         operands to vperm for little endian.
4372         * config/rs6000/rs6000.c (rs6000_expand_builtin): Use lvsr instead
4373         of lvsl to create the control mask for a vperm for little endian.
4375 2013-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4376             Anton Blanchard  <anton@au1.ibm.com>
4378         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Reverse
4379         two operands for little-endian.
4381 2013-07-23  Steve Ellcey  <sellcey@mips.com>
4383         * config/mips/mips.c (mips_case_values_threshold): New.
4384         (TARGET_CASE_VALUES_THRESHOLD): Define.
4386 2013-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4387             Anton Blanchard  <anton@au1.ibm.com>
4389         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Correct
4390         selection of field for vector splat in little endian mode.
4392 2013-07-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4394         * config/rs6000/vector.md (xor<mode>3): Move 128-bit boolean
4395         expanders to rs6000.md.
4396         (ior<mode>3): Likewise.
4397         (and<mode>3): Likewise.
4398         (one_cmpl<mode>2): Likewise.
4399         (nor<mode>3): Likewise.
4400         (andc<mode>3): Likewise.
4401         (eqv<mode>3): Likewise.
4402         (nand<mode>3): Likewise.
4403         (orc<mode>3): Likewise.
4405         * config/rs6000/rs6000-protos.h (rs6000_split_logical): New
4406         declaration.
4408         * config/rs6000/rs6000.c (rs6000_split_logical_inner): Add support
4409         to split multi-word logical operations.
4410         (rs6000_split_logical_di): Likewise.
4411         (rs6000_split_logical): Likewise.
4413         * config/rs6000/vsx.md (VSX_L2): Delete, no longer used.
4414         (vsx_and<mode>3_32bit): Move 128-bit logical insns to rs6000.md,
4415         and allow TImode operations in 32-bit.
4416         (vsx_and<mode>3_64bit): Likewise.
4417         (vsx_ior<mode>3_32bit): Likewise.
4418         (vsx_ior<mode>3_64bit): Likewise.
4419         (vsx_xor<mode>3_32bit): Likewise.
4420         (vsx_xor<mode>3_64bit): Likewise.
4421         (vsx_one_cmpl<mode>2_32bit): Likewise.
4422         (vsx_one_cmpl<mode>2_64bit): Likewise.
4423         (vsx_nor<mode>3_32bit): Likewise.
4424         (vsx_nor<mode>3_64bit): Likewise.
4425         (vsx_andc<mode>3_32bit): Likewise.
4426         (vsx_andc<mode>3_64bit): Likewise.
4427         (vsx_eqv<mode>3_32bit): Likewise.
4428         (vsx_eqv<mode>3_64bit): Likewise.
4429         (vsx_nand<mode>3_32bit): Likewise.
4430         (vsx_nand<mode>3_64bit): Likewise.
4431         (vsx_orc<mode>3_32bit): Likewise.
4432         (vsx_orc<mode>3_64bit): Likewise.
4434         * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Always allow vector
4435         logical types in GPRs.
4437         * config/rs6000/altivec.md (altivec_and<mode>3): Move 128-bit
4438         logical insns to rs6000.md, and allow TImode operations in
4439         32-bit.
4440         (altivec_ior<mode>3): Likewise.
4441         (altivec_xor<mode>3): Likewise.
4442         (altivec_one_cmpl<mode>2): Likewise.
4443         (altivec_nor<mode>3): Likewise.
4444         (altivec_andc<mode>3): Likewise.
4446         * config/rs6000/rs6000.md (BOOL_128): New mode iterators and mode
4447         attributes for moving the 128-bit logical operations into
4448         rs6000.md.
4449         (BOOL_REGS_OUTPUT): Likewise.
4450         (BOOL_REGS_OP1): Likewise.
4451         (BOOL_REGS_OP2): Likewise.
4452         (BOOL_REGS_UNARY): Likewise.
4453         (BOOL_REGS_AND_CR0): Likewise.
4454         (one_cmpl<mode>2): Add support for DI logical operations on
4455         32-bit, splitting the operations to 32-bit.
4456         (anddi3): Likewise.
4457         (iordi3): Likewise.
4458         (xordi3): Likewise.
4459         (and<mode>3, 128-bit types): Rewrite 2013-06-06 logical operator
4460         changes to combine the 32/64-bit code, allow logical operations on
4461         TI mode in 32-bit, and to use similar match_operator patterns like
4462         scalar mode uses.  Combine the Altivec and VSX code for logical
4463         operations, and move it here.
4464         (ior<mode>3, 128-bit types): Likewise.
4465         (xor<mode>3, 128-bit types): Likewise.
4466         (one_cmpl<mode>3, 128-bit types): Likewise.
4467         (nor<mode>3, 128-bit types): Likewise.
4468         (andc<mode>3, 128-bit types): Likewise.
4469         (eqv<mode>3, 128-bit types): Likewise.
4470         (nand<mode>3, 128-bit types): Likewise.
4471         (orc<mode>3, 128-bit types): Likewise.
4472         (and<mode>3_internal): Likewise.
4473         (bool<mode>3_internal): Likewise.
4474         (boolc<mode>3_internal1): Likewise.
4475         (boolc<mode>3_internal2): Likewise.
4476         (boolcc<mode>3_internal1): Likewise.
4477         (boolcc<mode>3_internal2): Likewise.
4478         (eqv<mode>3_internal1): Likewise.
4479         (eqv<mode>3_internal2): Likewise.
4480         (one_cmpl1<mode>3_internal): Likewise.
4482 2013-07-23  David Holsgrove  <david.holsgrove@xilinx.com>
4484         * config/microblaze/microblaze.c (microblaze_expand_prologue):
4485         Rename flag_stack_usage to flag_stack_usage_info.
4487 2013-07-23  David Holsgrove  <david.holsgrove@xilinx.com>
4489         * config/microblaze/sync.md: New file.
4490         * config/microblaze/microblaze.md: Include sync.md
4491         * config/microblaze/microblaze.c: Add print_operand 'y'.
4492         * config/microblaze/constraints.md: Add memory_contraint
4493         'Q' which is a single register.
4495 2013-07-23  Eric Botcazou  <ebotcazou@adacore.com>
4497         * doc/invoke.texi (SPARC Options): Document new leon3 processor value.
4499 2013-07-22  Po-Chun Chang  <pchang9@cs.wisc.edu>
4501         * reload.c (find_reloads): Exit loop once we find this operand
4502         cannot be reloaded somehow for this alternative.
4504         * reload.c (find_reloads): Exit loop once we find a hard register.
4506         * rtlanal.c (computed_jump_p): Exit loop once we find label
4507         reference is used.
4509         * i386.c (ix86_pad_returns): Exit loop after setting replace.
4511         * cfgloopmanip.c (remove_path): Exit loop after setting
4512         irred_invalidated.
4514         * gensupport.c (subst_dup): Avoid loop if code is not
4515         MATCH_DUP nor MATCH_OP_DUP.
4517 2013-07-23  Nicklas Bo Jensen  <nbjensen@gmail.com>
4519         * doc/md.texi (Machine-Specific Peephole Optimizers): Fix a typo.
4521 2013-07-23  Yufeng Zhang  <yufeng.zhang@arm.com>
4523         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Also return
4524         true for SP_REGNUM if mode == ptr_mode.
4525         * config/aarch64/aarch64.h (ADDITIONAL_REGISTER_NAMES): Add "wsp"
4526         with value R0_REGNUM + 31.
4528 2013-07-23  Yufeng Zhang  <yufeng.zhang@arm.com>
4530         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): In big-endian,
4531         pad pointer-typed argument downward.
4533 2013-07-23  Yufeng Zhang  <yufeng.zhang@arm.com>
4535         * config/aarch64/aarch64.h (TARGET_CPU_CPP_BUILTINS): Define _ILP32
4536         and __ILP32__ when the ILP32 model is in use.
4538 2013-07-23  Yufeng Zhang  <yufeng.zhang@arm.com>
4540         * config/aarch64/aarch64.c (POINTER_BYTES): New define.
4541         (aarch64_load_symref_appropriately): In the case of
4542         SYMBOL_SMALL_ABSOLUTE, use the mode of 'dest' instead of Pmode
4543         to generate new rtx; likewise to the case of SYMBOL_SMALL_GOT.
4544         (aarch64_expand_mov_immediate): In the case of SYMBOL_FORCE_TO_MEM,
4545         change to pass 'ptr_mode' to force_const_mem and zero-extend 'mem'
4546         if 'mode' doesn't equal to 'ptr_mode'.
4547         (aarch64_output_mi_thunk): Add an assertion on the alignment of
4548         'vcall_offset'; change to call aarch64_emit_move differently depending
4549         on whether 'Pmode' equals to 'ptr_mode' or not; use 'POINTER_BYTES'
4550         to calculate the upper bound of 'vcall_offset'.
4551         (aarch64_cannot_force_const_mem): Change to also return true if
4552         mode != ptr_mode.
4553         (aarch64_legitimize_reload_address): In the case of large
4554         displacements, add new local variable 'xmode' and an assertion
4555         based on it; change to use 'xmode' to generate the new rtx and
4556         reload.
4557         (aarch64_asm_trampoline_template): Change to generate the template
4558         differently depending on TARGET_ILP32 or not; change to use
4559         'POINTER_BYTES' in the argument passed to assemble_aligned_integer.
4560         (aarch64_trampoline_size): Removed.
4561         (aarch64_trampoline_init): Add new local constant 'tramp_code_sz'
4562         and replace immediate literals with it.  Change to use 'ptr_mode'
4563         instead of 'DImode' and call convert_memory_address if the mode
4564         of 'fnaddr' doesn't equal to 'ptr_mode'.
4565         (aarch64_elf_asm_constructor): Change to use assemble_aligned_integer
4566         to output symbol.
4567         (aarch64_elf_asm_destructor): Likewise.
4568         * config/aarch64/aarch64.h (TRAMPOLINE_SIZE): Change to be dependent
4569         on TARGET_ILP32 instead of aarch64_trampoline_size.
4570         * config/aarch64/aarch64.md (movsi_aarch64): Add new alternatives
4571         of 'mov' between WSP and W registers as well as 'adr' and 'adrp'.
4572         (loadwb_pair<GPI:mode>_<PTR:mode>): Rename to ...
4573         (loadwb_pair<GPI:mode>_<P:mode>): ... this.  Replace PTR with P.
4574         (storewb_pair<GPI:mode>_<PTR:mode>): Likewise; rename to ...
4575         (storewb_pair<GPI:mode>_<P:mode>): ... this.
4576         (add_losym): Change to 'define_expand' and call gen_add_losym_<mode>
4577         depending on the value of 'mode'.
4578         (add_losym_<mode>): New.
4579         (ldr_got_small_<mode>): New, based on ldr_got_small.
4580         (ldr_got_small): Remove.
4581         (ldr_got_small_sidi): New.
4582         * config/aarch64/iterators.md (P): New.
4583         (PTR): Change to 'ptr_mode' in the condition.
4585 2013-07-23  Yufeng Zhang  <yufeng.zhang@arm.com>
4587         * config.gcc (aarch64*-*-*): Support --with-abi.
4588         (aarch64*-*-elf): Support --with-multilib-list.
4589         (aarch64*-*-linux*): Likewise.
4590         (supported_defaults): Add abi to aarch64*-*-*.
4591         * configure.ac: Mention AArch64 for --with-multilib-list.
4592         * configure: Re-generated.
4593         * config/aarch64/biarchilp32.h: New file.
4594         * config/aarch64/biarchlp64.h: New file.
4595         * config/aarch64/aarch64-elf.h (ENDIAN_SPEC): New define.
4596         (ABI_SPEC): Ditto.
4597         (MULTILIB_DEFAULTS): Ditto.
4598         (DRIVER_SELF_SPECS): Ditto.
4599         (ASM_SPEC): Update to also substitute -mabi.
4600         * config/aarch64/aarch64-elf-raw.h (LINK_SPEC): Add linker script
4601         file whose name depends on -mabi= and -mbig-endian.
4602         * config/aarch64/aarch64.h (LONG_TYPE_SIZE): Change to depend on
4603         TARGET_ILP32.
4604         (POINTER_SIZE): New define.
4605         (POINTERS_EXTEND_UNSIGNED): Ditto.
4606         (enum aarch64_abi_type): New enumeration tag.
4607         (AARCH64_ABI_LP64, AARCH64_ABI_ILP32): New enumerators.
4608         (AARCH64_ABI_DEFAULT): Define to AARCH64_ABI_LP64 if undefined.
4609         (TARGET_ILP32): New define.
4610         * config/aarch64/aarch64.opt (mabi): New.
4611         (aarch64_abi): New.
4612         (ilp32, lp64): New values for -mabi.
4613         * config/aarch64/t-aarch64 (comma): New define.
4614         (MULTILIB_OPTIONS): Ditto.
4615         (MULTILIB_DIRNAMES): Ditto.
4616         * config/aarch64/t-aarch64-linux (MULTIARCH_DIRNAME): New define.
4617         * doc/invoke.texi: Document -mabi for AArch64.
4619 2013-07-23  Georg-Johann Lay  <avr@gjlay.de>
4621         * config/avr/avr.md: Explain asm print modifier 'r' for REG.
4623 2013-07-22  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4624             Anton Blanchard  <anton@au1.ibm.com>
4626         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Fix
4627         endianness when selecting field to splat.
4629 2013-07-22  Eric Christopher  <echristo@gmail.com>
4631         * dwarf2out.c (die_odr_checksum): New function to use
4632         CHECKSUM_ macros and ULEB128 for DIE tag.
4633         (generate_type_signature): Use.
4635 2013-07-22  Eric Botcazou  <ebotcazou@adacore.com>
4637         * config.gcc (sparc*-*-*): Accept leon3 processor.
4638         (sparc-leon*-*): Merge with sparc*-*-* and add leon3 support.
4639         * doc/invoke.texi (SPARC Options): Adjust -mfix-ut699 entry.
4640         * config/sparc/sparc-opts.h (enum processor_type): Add PROCESSOR_LEON3.
4641         * config/sparc/sparc.opt (enum processor_type): Add leon3.
4642         (mfix-ut699): Adjust comment.
4643         * config/sparc/sparc.h (TARGET_CPU_leon3): New define.
4644         (CPP_CPU32_DEFAULT_SPEC): Add leon3 support.
4645         (CPP_CPU_SPEC): Likewise.
4646         (ASM_CPU_SPEC): Likewise.
4647         * config/sparc/sparc.c (leon3_cost): New constant.
4648         (sparc_option_override): Add leon3 support.
4649         (mem_ref): New function.
4650         (sparc_gate_work_around_errata): Return true if -mfix-ut699 is enabled.
4651         (sparc_do_work_around_errata): Look into the instruction in the delay
4652         slot and adjust accordingly.  Add fix for the data cache nullify issues
4653         of the UT699.  Change insertion position for the NOP.
4654         * config/sparc/leon.md (leon_fpalu, leon_fpmds, write_buf): Delete.
4655         (leon3_load): New reservation.
4656         (leon_store): Bump latency to 2.
4657         (grfpu): New automaton.
4658         (grfpu_alu): New unit.
4659         (grfpu_ds): Likewise.
4660         (leon_fp_alu): Adjust.
4661         (leon_fp_mult): Delete.
4662         (leon_fp_div): Split into leon_fp_divs and leon_fp_divd.
4663         (leon_fp_sqrt): Split into leon_fp_sqrts and leon_fp_sqrtd.
4664         * config/sparc/sparc.md (cpu): Add leon3.
4665         * config/sparc/sync.md (atomic_exchangesi): Disable if -mfix-ut699.
4666         (swapsi): Likewise.
4667         (atomic_test_and_set): Likewise.
4668         (ldstub): Likewise.
4670 2013-07-22  Jürgen Urban  <JuergenUrban@gmx.de>
4672         * config.gcc (mips*-*-*): Add --with-fpu support.  Make single the
4673         default for R5900 targets.
4674         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Handle --with-fpu.
4675         (ISA_HAS_LDC1_SDC1): Set to false for TARGET_MIPS5900.
4676         * config/mips/mips.c (mips_option_override): Report an error for
4677         -march=r5900 -mhard-float -mdouble-float.  Use spu_single_format
4678         for -march=r5900 -mhard-float.
4680 2013-07-22  Po-Chun Chang  <pchang9@cs.wisc.edu>
4682         * df-problems.c (can_move_insns_across): Exit loop once we
4683         find a non-fixed, non-global register.
4685         * ipa-pure-const.c (propagate_nothrow): Exit loop after
4686         setting can_throw.
4688         * omega.c (omega_eliminate_red): Break after setting red_found.
4689         (omega_problem_has_red_equations): Similarly after setting found.
4690         (omega_query_variable): Similarly after setting coupled.
4692 2013-07-22  Marek Polacek  <polacek@redhat.com>
4694         * gimplify.c: Don't include gimple.h twice.
4696 2013-07-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4698         * config/arm/constraints.md (Pd): Allow TARGET_THUMB
4699         instead of TARGET_THUMB1.
4700         (Pz): New constraint.
4701         * config/arm/arm.md (arm_addsi3): Add alternatives for 16-bit
4702         encodings.
4703         (compare_negsi_si): Likewise.
4704         (compare_addsi2_op0): Likewise.
4705         (compare_addsi2_op1): Likewise.
4706         (addsi3_carryin_<optab>): Likewise.
4707         (addsi3_carryin_alt2_<optab>): Likewise.
4708         (addsi3_carryin_shift_<optab>): Disable cond_exec variant
4709         for arm_restrict_it.
4710         (subsi3_carryin): Likewise.
4711         (arm_subsi3_insn): Add alternatives for 16-bit encoding.
4712         (minmax_arithsi): Disable for arm_restrict_it.
4713         (minmax_arithsi_non_canon): Adjust for arm_restrict_it.
4714         (satsi_<SAT:code>): Disable cond_exec variant for arm_restrict_it.
4715         (satsi_<SAT:code>_shift): Likewise.
4716         (arm_shiftsi3): Add alternative for 16-bit encoding.
4717         (arm32_movhf): Disable for arm_restrict_it.
4718         (arm_cmpdi_unsigned): Add alternatives for 16-bit encoding.
4719         (arm_movtas_ze): Disable cond_exec variant for arm_restrict_it.
4721 2013-07-22  Sofiane Naci  <sofiane.naci@arm.com>
4723         * config/arm/arm.md (attribute "insn"): Delete.
4724         (attribute "type"): Add "mov_imm", "mov_reg", "mov_shift",
4725         "mov_shift_reg", "mvn_imm", "mvn_reg", "mvn_shift" and "mvn_shift_reg".
4726         (not_shiftsi): Update for attribute change.
4727         (not_shiftsi_compare0): Likewise.
4728         (not_shiftsi_compare0_scratch): Likewise.
4729         (arm_one_cmplsi2): Likewise.
4730         (thumb1_one_cmplsi2): Likewise.
4731         (notsi_compare0): Likewise.
4732         (notsi_compare0_scratch): Likewise.
4733         (thumb1_movdi_insn): Likewise.
4734         (arm_movsi_insn): Likewise.
4735         (movhi_insn_arch4): Likewise.
4736         (movhi_bytes): Likewise.
4737         (arm_movqi_insn): Likewise.
4738         (thumb1_movqi_insn): Likewise.
4739         (arm32_movhf): Likewise.
4740         (thumb1_movhf): Likewise.
4741         (arm_movsf_soft_insn): Likewise.
4742         (thumb1_movsf_insn): Likewise.
4743         (thumb_movdf_insn): Likewise.
4744         (movsicc_insn): Likewise.
4745         (movsfcc_soft_insn): Likewise.
4746         (and_scc): Likewise.
4747         (cond_move): Likewise.
4748         (if_move_not): Likewise.
4749         (if_not_move): Likewise.
4750         (if_shift_move): Likewise.
4751         (if_move_shift): Likewise.
4752         (if_shift_shift): Likewise.
4753         (if_not_arith): Likewise.
4754         (if_arith_not): Likewise.
4755         (cond_move_not): Likewise.
4756         * config/arm/neon.md (neon_mov<mode>): Update for attribute change.
4757         (neon_mov<mode>): Likewise.
4758         * config/arm/vfp.md (arm_movsi_vfp): Update for attribute change.
4759         (thumb2_movsi_vfp): Likewise.
4760         (movsf_vfp): Likewise.
4761         (thumb2_movsf_vfp): Likewise.
4762         * config/arm/arm.c (xscale_sched_adjust_cost): Update for attribute
4763         change.
4764         (cortexa7_older_only): Likewise.
4765         (cortexa7_younger): Likewise.
4766         * config/arm/arm1020e.md (1020alu_op): Update for attribute change.
4767         (1020alu_shift_op): Likewise.
4768         (1020alu_shift_reg_op): Likewise.
4769         * config/arm/arm1026ejs.md (alu_op): Update for attribute change.
4770         (alu_shift_op): Likewise.
4771         (alu_shift_reg_op): Likewise.
4772         * config/arm/arm1136jfs.md (11_alu_op): Update for attribute change.
4773         (11_alu_shift_op): Likewise.
4774         (11_alu_shift_reg_op): Likewise.
4775         * config/arm/arm926ejs.md (9_alu_op): Update for attribute change.
4776         (9_alu_shift_reg_op): Likewise.
4777         * config/arm/cortex-a15.md (cortex_a15_alu): Update for attribute
4778         change.
4779         (cortex_a15_alu_shift): Likewise.
4780         (cortex_a15_alu_shift_reg): Likewise.
4781         * config/arm/cortex-a5.md (cortex_a5_alu): Update for attribute change.
4782         (cortex_a5_alu_shift): Likewise.
4783         * config/arm/cortex-a53.md (cortex_a53_alu): Update for attribute
4784         change.
4785         (cortex_a53_alu_shift): Likewise.
4786         * config/arm/cortex-a7.md (cortex_a7_alu_imm): Update for attribute
4787         change.
4788         (cortex_a7_alu_reg): Likewise.
4789         (cortex_a7_alu_shift): Likewise.
4790         * config/arm/cortex-a8.md (cortex_a8_alu): Update for attribute change.
4791         (cortex_a8_alu_shift): Likewise.
4792         (cortex_a8_alu_shift_reg): Likewise.
4793         (cortex_a8_mov): Likewise.
4794         * config/arm/cortex-a9.md (cortex_a9_dp): Update for attribute change.
4795         (cortex_a9_dp_shift): Likewise.
4796         * config/arm/cortex-m4.md (cortex_m4_alu): Update for attribute change.
4797         * config/arm/cortex-r4.md (cortex_r4_alu): Update for attribute change.
4798         (cortex_r4_mov): Likewise.
4799         (cortex_r4_alu_shift): Likewise.
4800         (cortex_r4_alu_shift_reg): Likewise.
4801         * config/arm/fa526.md (526_alu_op): Update for attribute change.
4802         (526_alu_shift_op): Likewise.
4803         * config/arm/fa606te.md (606te_alu_op): Update for attribute change.
4804         * config/arm/fa626te.md (626te_alu_op): Update for attribute change.
4805         (626te_alu_shift_op): Likewise.
4806         * config/arm/fa726te.md (726te_shift_op): Update for attribute change.
4807         (726te_alu_op): Likewise.
4808         (726te_alu_shift_op): Likewise.
4809         (726te_alu_shift_reg_op): Likewise.
4810         * config/arm/fmp626.md (mp626_alu_op): Update for attribute change.
4811         (mp626_alu_shift_op): Likewise.
4812         * config/arm/marvell-pj4.md (pj4_alu_e1): Update for attribute change.
4813         (pj4_alu_e1_conds): Likewise.
4814         (pj4_alu): Likewise.
4815         (pj4_alu_conds): Likewise.
4816         (pj4_shift): Likewise.
4817         (pj4_shift_conds): Likewise.
4818         (pj4_alu_shift): Likewise.
4819         (pj4_alu_shift_conds): Likewise.
4821 2013-07-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4823         * config/arm/predicates.md (shiftable_operator_strict_it):
4824         New predicate.
4825         * config/arm/thumb2.md (thumb_andsi_not_shiftsi_si):
4826         Disable cond_exec version for arm_restrict_it.
4827         (thumb2_smaxsi3): Convert to generate cond_exec.
4828         (thumb2_sminsi3): Likewise.
4829         (thumb32_umaxsi3): Likewise.
4830         (thumb2_uminsi3): Likewise.
4831         (thumb2_abssi2): Adjust constraints for arm_restrict_it.
4832         (thumb2_neg_abssi2): Likewise.
4833         (thumb2_mov_scc): Add alternative for 16-bit encoding.
4834         (thumb2_movsicc_insn): Adjust alternatives.
4835         (thumb2_mov_negscc): Disable for arm_restrict_it.
4836         (thumb2_mov_negscc_strict_it): New pattern.
4837         (thumb2_mov_notscc_strict_it): New pattern.
4838         (thumb2_mov_notscc): Disable for arm_restrict_it.
4839         (thumb2_ior_scc): Likewise.
4840         (thumb2_ior_scc_strict_it): New pattern.
4841         (thumb2_cond_move): Adjust for arm_restrict_it.
4842         (thumb2_cond_arith): Disable for arm_restrict_it.
4843         (thumb2_cond_arith_strict_it): New pattern.
4844         (thumb2_cond_sub): Adjust for arm_restrict_it.
4845         (thumb2_movcond): Likewise.
4846         (thumb2_extendqisi_v6): Disable cond_exec variant for arm_restrict_it.
4847         (thumb2_zero_extendhisi2_v6): Likewise.
4848         (thumb2_zero_extendqisi2_v6): Likewise.
4849         (orsi_notsi_si): Likewise.
4850         (orsi_not_shiftsi_si): Likewise.
4852 2013-07-22  Georg-Johann Lay  <avr@gjlay.de>
4854         * config/avr/avr.c (avr_out_xload): No SBIS around LPM so that
4855         instruction sequence is 1 byte shorter.
4857 2013-07-22  Uros Bizjak  <ubizjak@gmail.com>
4859         * config/i386/i386.md (nonlocal_goto_receiver): Delete insn if
4860         it is not needed after split.
4862 2013-07-20  Iain Sandoe  <iain@codesourcery.com>
4864         PR target/51784
4865         * config/i386/i386.c (output_set_got) [TARGET_MACHO]: Adjust to emit a
4866         second label for nonlocal goto receivers. Don't output pic base labels
4867         unless we're producing PIC; mark that action unreachable().
4868         (ix86_save_reg): If the function contains a nonlocal label, save the
4869         PIC base reg.
4870         * config/darwin-protos.h (machopic_should_output_picbase_label): New.
4871         * gcc/config/darwin.c (emitted_pic_label_num): New GTY.
4872         (update_pic_label_number_if_needed): New.
4873         (machopic_output_function_base_name): Adjust for nonlocal receiver
4874         case.
4875         (machopic_should_output_picbase_label): New.
4876         * config/i386/i386.md (enum unspecv): UNSPECV_NLGR: New.
4877         (nonlocal_goto_receiver): New insn and split.
4879 2013-07-20  James Greenhalgh  <james.greenhalgh@arm.com>
4881         * config/aarch64/aarch64-builtins.c
4882         (aarch64_fold_builtin): Fold abs in all modes.
4883         * config/aarch64/aarch64-simd-builtins.def
4884         (abs): Enable for all modes.
4885         * config/aarch64/arm_neon.h
4886         (vabs<q>_s<8,16,32,64): Rewrite using builtins.
4887         (vabs_f64): Add missing intrinsic.
4889 2013-07-19  Ian Bolton  <ian.bolton@arm.com>
4891         * config/aarch64/arm_neon.h (vabs_s64): New function
4893 2013-07-19  Georg-Johann Lay  <avr@gjlay.de>
4895         PR target/57516
4896         * config/avr/avr-fixed.md (round<mode>3_const): Turn expander to insn.
4897         * config/avr/avr.md (adjust_len): Add `round'.
4898         * config/avr/avr-protos.h (avr_out_round): New prototype.
4899         (avr_out_plus): Add `out_label' argument.
4900         * config/avr/avr.c (avr_out_plus_1): Add `out_label' argument.
4901         (avr_out_plus): Pass down `out_label' to avr_out_plus_1.
4902         Handle the case where `insn' is just a pattern.
4903         (avr_out_bitop): Handle the case where `insn' is just a pattern.
4904         (avr_out_round): New function.
4905         (avr_adjust_insn_length): Handle ADJUST_LEN_ROUND.
4907 2013-07-18  David Holsgrove  <david.holsgrove@xilinx.com>
4909         * config/microblaze/microblaze.c (microblaze_expand_prologue):
4910         Add check for flag_stack_usage to handle -fstack-usage support
4912 2013-07-18  Pat Haugen  <pthaugen@us.ibm.com>
4914         * config/rs6000/rs6000.c (rs6000_option_override_internal): Adjust flag
4915         interaction for new Power8 flags and VSX.
4917 2013-07-18  Sriraman Tallam  <tmsriram@google.com>
4919         PR middle-end/57698
4920         * tree-inline.c (expand_call_inline): Emit errors during
4921         early_inlining only if optimization is not turned on.
4923 2013-07-18  David Malcolm  <dmalcolm@redhat.com>
4925         * passes.def: New.
4927         * passes.c (init_optimization_passes): Move the construction of
4928         the pass hierarchy into a new passes.def file.
4930         * Makefile.in (passes.o): Add dependency on passes.def.
4932 2013-07-18  David Malcolm  <dmalcolm@redhat.com>
4934         * passes.c (init_optimization_passes): Introduce macros for
4935         constructing the tree of passes (INSERT_PASSES_AFTER,
4936         PUSH_INSERT_PASSES_WITHIN, POP_INSERT_PASSES,
4937         TERMINATE_PASS_LIST).
4939 2013-07-18  Vladimir Makarov  <vmakarov@redhat.com>
4940             Wei Mi  <wmi@google.com>
4942         PR rtl-optimization/57878
4943         * lra-assigns.c (assign_by_spills): Move non_reload_pseudos to the
4944         top.
4945         (reload_pseudo_compare_func): Check nregs first for reload
4946         pseudos.
4948 2013-07-18  David Malcolm  <dmalcolm@redhat.com>
4950         * tree-pass.h (pass_ipa_lto_wpa_fixup): Remove redundant decl.
4952 2013-07-18  Po-Chun Chang  <pchang9@cs.wisc.edu>
4954         * read-rtl.c (validate_const_int): Once an invalid character is
4955         seen, quit the loop.
4957         * gengtype.c (write_roots): Similarly once we find the "deletable"
4958         or "if_marked" option.
4960 2013-07-18  Sofiane Naci  <sofiane.naci@arm.com>
4962         * config/arm/arm.md (attribute "insn"): Delete values "mrs", "msr",
4963         "xtab" and "sat".  Move value "clz" from here to ...
4964         (attriubte "type"): ... here.
4965         (satsi_<SAT:code>): Delete "insn" attribute.
4966         (satsi_<SAT:code>_shift): Likewise.
4967         (arm_zero_extendqisi2addsi): Likewise.
4968         (arm_extendqisi2addsi): Likewise.
4969         (clzsi2): Update for attribute changes.
4970         (rbitsi2): Likewise.
4971         * config/arm/arm-fixed.md (arm_ssatsihi_shift): Delete "insn"
4972         attribute.
4973         (arm_usatsihi): Likewise.
4974         * config/arm/cortex-a8.md (cortex_a8_alu): Update for attribute change.
4976 2013-07-18  Sofiane Naci  <sofiane.naci@arm.com>
4978         * config/arm/arm.md (attribute "type"): Rename "simple_alu_imm" to
4979         "arlo_imm".  Rename "alu_reg" to "arlo_reg".  Rename "simple_alu_shift"
4980         to "extend".  Split "alu_shift" into "shift" and "arlo_shift".  Split
4981         "alu_shift_reg" into "shift_reg" and "arlo_shift_reg".  List types
4982         in alphabetical order.
4983         (attribute "core_cycles"): Update for attribute changes.
4984         (arm_addsi3): Likewise.
4985         (addsi3_compare0): Likewise.
4986         (addsi3_compare0_scratch): Likewise.
4987         (addsi3_compare_op1): Likewise.
4988         (addsi3_compare_op2): Likewise.
4989         (compare_addsi2_op0): Likewise.
4990         (compare_addsi2_op1): Likewise.
4991         (addsi3_carryin_shift_<optab>): Likewise.
4992         (subsi3_carryin_shift): Likewise.
4993         (rsbsi3_carryin_shift): Likewise.
4994         (arm_subsi3_insn): Likewise.
4995         (subsi3_compare0): Likewise.
4996         (subsi3_compare): Likewise.
4997         (arm_andsi3_insn): Likewise.
4998         (thumb1_andsi3_insn): Likewise.
4999         (andsi3_compare0): Likewise.
5000         (andsi3_compare0_scratch): Likewise.
5001         (zeroextractsi_compare0_scratch
5002         (andsi_not_shiftsi_si): Likewise.
5003         (iorsi3_insn): Likewise.
5004         (iorsi3_compare0): Likewise.
5005         (iorsi3_compare0_scratch): Likewise.
5006         (arm_xorsi3): Likewise.
5007         (thumb1_xorsi3_insn): Likewise.
5008         (xorsi3_compare0): Likewise.
5009         (xorsi3_compare0_scratch): Likewise.
5010         (satsi_<SAT:code>_shift): Likewise.
5011         (rrx): Likewise.
5012         (arm_shiftsi3): Likewise.
5013         (shiftsi3_compare0): Likewise.
5014         (not_shiftsi): Likewise.
5015         (not_shiftsi_compare0): Likewise.
5016         (not_shiftsi_compare0_scratch): Likewise.
5017         (arm_one_cmplsi2): Likewise.
5018         (thumb_one_complsi2): Likewise.
5019         (notsi_compare0): Likewise.
5020         (notsi_compare0_scratch): Likewise.
5021         (thumb1_zero_extendhisi2): Likewise.
5022         (arm_zero_extendhisi2): Likewise.
5023         (arm_zero_extendhisi2_v6): Likewise.
5024         (arm_zero_extendhisi2addsi): Likewise.
5025         (thumb1_zero_extendqisi2): Likewise.
5026         (thumb1_zero_extendqisi2_v6): Likewise.
5027         (arm_zero_extendqisi2): Likewise.
5028         (arm_zero_extendqisi2_v6): Likewise.
5029         (arm_zero_extendqisi2addsi): Likewise.
5030         (thumb1_extendhisi2): Likewise.
5031         (arm_extendhisi2): Likewise.
5032         (arm_extendhisi2_v6): Likewise.
5033         (arm_extendqisi): Likewise.
5034         (arm_extendqisi_v6): Likewise.
5035         (arm_extendqisi2addsi): Likewise.
5036         (thumb1_extendqisi2): Likewise.
5037         (thumb1_movdi_insn): Likewise.
5038         (arm_movsi_insn): Likewise.
5039         (movsi_compare0): Likewise.
5040         (movhi_insn_arch4): Likewise.
5041         (movhi_bytes): Likewise.
5042         (arm_movqi_insn): Likewise.
5043         (thumb1_movqi_insn): Likewise.
5044         (arm32_movhf): Likewise.
5045         (thumb1_movhf): Likewise.
5046         (arm_movsf_soft_insn): Likewise.
5047         (thumb1_movsf_insn): Likewise.
5048         (movdf_soft_insn): Likewise.
5049         (thumb_movdf_insn): Likewise.
5050         (arm_cmpsi_insn): Likewise.
5051         (cmpsi_shiftsi): Likewise.
5052         (cmpsi_shiftsi_swp): Likewise.
5053         (arm_cmpsi_negshiftsi_si): Likewise.
5054         (movsicc_insn): Likewise.
5055         (movsfcc_soft_insn): Likewise.
5056         (arith_shiftsi): Likewise.
5057         (arith_shiftsi_compare0
5058         (arith_shiftsi_compare0_scratch
5059         (sub_shiftsi): Likewise.
5060         (sub_shiftsi_compare0
5061         (sub_shiftsi_compare0_scratch
5062         (and_scc): Likewise.
5063         (cond_move): Likewise.
5064         (if_plus_move): Likewise.
5065         (if_move_plus): Likewise.
5066         (if_move_not): Likewise.
5067         (if_not_move): Likewise.
5068         (if_shift_move): Likewise.
5069         (if_move_shift): Likewise.
5070         (if_shift_shift): Likewise.
5071         (if_not_arith): Likewise.
5072         (if_arith_not): Likewise.
5073         (cond_move_not): Likewise.
5074         (thumb1_ashlsi3): Set type attribute.
5075         (thumb1_ashrsi3): Likewise.
5076         (thumb1_lshrsi3): Likewise.
5077         (thumb1_rotrsi3): Likewise.
5078         (shiftsi3_compare0_scratch): Likewise.
5079         * config/arm/neon.md (neon_mov<mode>): Update for attribute changes.
5080         (neon_mov<mode>): Likewise.
5081         * config/arm/thumb2.md (thumb_andsi_not_shiftsi_si): Update for
5082         attribute changes.
5083         (thumb2_movsi_insn): Likewise.
5084         (thumb2_cmpsi_neg_shiftsi): Likewise.
5085         (thumb2_extendqisi_v6): Likewise.
5086         (thumb2_zero_extendhisi2_v6): Likewise.
5087         (thumb2_zero_extendqisi2_v6): Likewise.
5088         (thumb2_shiftsi3_short): Likewise.
5089         (thumb2_addsi3_compare0_scratch): Likewise.
5090         (orsi_not_shiftsi_si): Likewise.
5091         * config/arm/vfp.md (arm_movsi_vfp): Update for attribute changes.
5092         * config/arm/arm-fixed.md (arm_ssatsihi_shift): Update for attribute
5093         changes.
5094         * config/arm/arm1020e.md (1020alu_op): Update for attribute changes.
5095         (1020alu_shift_op): Likewise.
5096         (1020alu_shift_reg_op): Likewise.
5097         * config/arm/arm1026ejs.md (alu_op): Update for attribute changes.
5098         (alu_shift_op): Likewise.
5099         (alu_shift_reg_op): Likewise.
5100         * config/arm/arm1136jfs.md (11_alu_op): Update for attribute changes.
5101         (11_alu_shift_op): Likewise.
5102         (11_alu_shift_reg_op): Likewise.
5103         * config/arm/arm926ejs.md (9_alu_op): Update for attribute changes.
5104         (9_alu_shift_reg_op): Likewise.
5105         * config/arm/cortex-a15.md (cortex_a15_alu): Update for attribute
5106         changes.
5107         (cortex_a15_alu_shift): Likewise.
5108         (cortex_a15_alu_shift_reg): Likewise.
5109         * config/arm/cortex-a5.md (cortex_a5_alu): Update for attribute
5110         changes.
5111         (cortex_a5_alu_shift): Likewise.
5112         * config/arm/cortex-a53.md (cortex_a53_alu) : Update for attribute
5113         changes.
5114         (cortex_a53_alu_shift): Likewise.
5115         * config/arm/cortex-a7.md (cortex_a7_alu_imm): Update for attribute
5116         changes.
5117         (cortex_a7_alu_reg): Likewise.
5118         (cortex_a7_alu_shift): Likewise.
5119         * config/arm/cortex-a8.md (cortex_a8_alu): Update for attribute
5120         changes.
5121         (cortex_a8_alu_shift): Likewise.
5122         (cortex_a8_alu_shift_reg): Likewise.
5123         (cortex_a8_mov): Likewise.
5124         * config/arm/cortex-a9.md (cortex_a9_dp): Update for attribute changes.
5125         (cortex_a9_dp_shift): Likewise.
5126         * config/arm/cortex-m4.md (cortex_m4_alu): Update for attribute
5127         changes.
5128         * config/arm/cortex-r4.md (cortex_r4_alu): Update for attribute
5129         changes.
5130         (cortex_r4_mov): Likewise.
5131         (cortex_r4_alu_shift): Likewise.
5132         (cortex_r4_alu_shift_reg): Likewise.
5133         * config/arm/fa526.md (526_alu_op): Update for attribute changes.
5134         (526_alu_shift_op): Likewise.
5135         * config/arm/fa606te.md (606te_alu_op): Update for attribute changes.
5136         * config/arm/fa626te.md (626te_alu_op): Update for attribute changes.
5137         (626te_alu_shift_op): Likewise.
5138         * config/arm/fa726te.md (726te_shift_op): Update for attribute changes.
5139         (726te_alu_op): Likewise.
5140         (726te_alu_shift_op): Likewise.
5141         (726te_alu_shift_reg_op): Likewise.
5142         * config/arm/fmp626.md (mp626_alu_op): Update for attribute changes.
5143         (mp626_alu_shift_op): Likewise.
5144         * config/arm/marvell-pj4.md (pj4_alu_e1): Update for attribute changes.
5145         (pj4_alu_e1_conds): Likewise.
5146         (pj4_alu): Likewise.
5147         (pj4_alu_conds): Likewise.
5148         (pj4_shift): Likewise.
5149         (pj4_shift_conds): Likewise.
5150         (pj4_alu_shift): Likewise.
5151         (pj4_alu_shift_conds): Likewise.
5152         * config/arm/arm.c (xscale_sched_adjust_cost): Update for attribute
5153         changes.
5154         (cortexa7_older_only): Likewise.
5155         (cortexa7_younger): Likewise.
5157 2013-07-18  David Malcolm  <dmalcolm@redhat.com>
5159         * ipa-pure-const.c (generate_summary): Rename to...
5160         (pure_const_generate_summary): ... this.
5162 2013-07-17  Iain Sandoe  <iain@codesourcery.com>
5164         * config/rs6000/darwin.h (REGISTER_NAMES): Add HTM registers.
5166 2013-07-17  Yvan Roux  <yvan.roux@linaro.org>
5168         PR target/57909
5169         * config/arm/arm.c (gen_movmem_ldrd_strd): Fix unaligned load/store
5170         usage in HI mode.
5172 2013-07-17  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5174         * config/s390/s390.c: (s390_expand_builtin): Allow -mhtm to be
5175         enabled without -march=zEC12.
5176         * config/s390/s390.h (TARGET_HTM): Do not require EC12 machine
5177         flags to be set.
5179 2013-07-16  Maciej W. Rozycki  <macro@codesourcery.com>
5181         * config/mips/mips.h (ISA_HAS_FP4): Correct formatting.
5182         (ISA_HAS_FP_MADD4_MSUB4): Also enable for ISA_MIPS32R2.
5183         (ISA_HAS_NMADD4_NMSUB4): Remove the MODE argument; rewrite in
5184         terms of ISA_HAS_FP4, and also enable for ISA_MIPS32R2.
5185         (ISA_HAS_NMADD3_NMSUB3): Remove the MODE argument.
5186         * config/mips/mips.c (mips_rtx_costs) <PLUS>: Check for
5187         ISA_HAS_FP_MADD4_MSUB4 || ISA_HAS_FP_MADD3_MSUB3 rather than
5188         ISA_HAS_FP4.
5189         <MINUS, NEG>: Update according to changes to ISA_HAS_NMADD4_NMSUB4
5190         and ISA_HAS_NMADD3_NMSUB3.
5191         * config/mips/mips.md (nmadd4<mode>, nmadd3<mode>): Likewise.
5192         (nmadd4<mode>_fastmath, nmadd3<mode>_fastmath): Likewise.
5193         (nmsub4<mode>, nmsub3<mode>): Likewise.
5194         (nmsub4<mode>_fastmath, nmsub3<mode>_fastmath): Likewise.
5196 2013-07-16  Maciej W. Rozycki  <macro@codesourcery.com>
5198         * config/mips/mips.h (ISA_HAS_NMADD4_NMSUB4): Remove
5199         TARGET_MIPS5400 checking.
5201 2013-07-16  Jakub Jelinek  <jakub@redhat.com>
5202             Peter Bergner  <bergner@vnet.ibm.com>
5204         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTERS): Mention HTM
5205         registers in the comment.
5206         (DWARF_FRAME_REGISTERS): Subtract also the 3 HTM registers.
5207         (DWARF_REG_TO_UNWIND_COLUMN): Use DWARF_FRAME_REGISTERS
5208         rather than FIRST_PSEUDO_REGISTERS.
5210 2013-07-16  Peter Bergner  <bergner@vnet.ibm.com>
5212         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5213         enable extra ISA flags with TARGET_HTM.
5215 2013-07-16  Maciej W. Rozycki  <macro@codesourcery.com>
5217         * config/mips/mips.h (ISA_HAS_MULS, ISA_HAS_MSAC, ISA_HAS_MACC):
5218         Fix comment typos.
5220 2013-07-15  Cong Hou  <congh@google.com>
5222         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use hash function
5223         in compare function for sorting.
5225 2013-07-15  Peter Bergner  <bergner@vnet.ibm.com>
5227         * config.gcc (powerpc*-*-*): Install htmintrin.h and htmxlintrin.h.
5228         * config/rs6000/t-rs6000 (MD_INCLUDES): Add htm.md.
5229         * config/rs6000/rs6000.opt: Add -mhtm option.
5230         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Add OPTION_MASK_HTM.
5231         (ISA_2_7_MASKS_SERVER): Add OPTION_MASK_HTM.
5232         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
5233         __HTM__ if the HTM instructions are available.
5234         * config/rs6000/predicates.md (u3bit_cint_operand, u10bit_cint_operand,
5235         htm_spr_reg_operand): New define_predicates.
5236         * config/rs6000/rs6000.md (define_attr "type"): Add htm.
5237         (TFHAR_REGNO, TFIAR_REGNO, TEXASR_REGNO): New define_constants.
5238         Include htm.md.
5239         * config/rs6000/rs6000-builtin.def (BU_HTM_0, BU_HTM_1, BU_HTM_2,
5240         BU_HTM_3, BU_HTM_SPR0, BU_HTM_SPR1): Add support macros for defining
5241         HTM builtin functions.
5242         * config/rs6000/rs6000.c (RS6000_BUILTIN_H): New macro.
5243         (rs6000_reg_names, alt_reg_names): Add HTM SPR register names.
5244         (rs6000_init_hard_regno_mode_ok): Add support for HTM instructions.
5245         (rs6000_builtin_mask_calculate): Likewise.
5246         (rs6000_option_override_internal): Likewise.
5247         (bdesc_htm): Add new HTM builtin support.
5248         (htm_spr_num): New function.
5249         (htm_spr_regno): Likewise.
5250         (rs6000_htm_spr_icode): Likewise.
5251         (htm_expand_builtin): Likewise.
5252         (htm_init_builtins): Likewise.
5253         (rs6000_expand_builtin): Add support for HTM builtin functions.
5254         (rs6000_init_builtins): Likewise.
5255         (rs6000_invalid_builtin, rs6000_opt_mask): Add support for -mhtm
5256         option.
5257         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mhtm.
5258         (TARGET_HTM, MASK_HTM): Define macros.
5259         (FIRST_PSEUDO_REGISTER): Adjust for new HTM SPR registers.
5260         (FIXED_REGISTERS): Likewise.
5261         (CALL_USED_REGISTERS): Likewise.
5262         (CALL_REALLY_USED_REGISTERS): Likewise.
5263         (REG_ALLOC_ORDER): Likewise.
5264         (enum reg_class): Likewise.
5265         (REG_CLASS_NAMES): Likewise.
5266         (REG_CLASS_CONTENTS): Likewise.
5267         (REGISTER_NAMES): Likewise.
5268         (ADDITIONAL_REGISTER_NAMES): Likewise.
5269         (RS6000_BTC_SPR, RS6000_BTC_VOID, RS6000_BTC_32BIT, RS6000_BTC_64BIT,
5270         RS6000_BTC_MISC_MASK, RS6000_BTM_HTM): New macros.
5271         (RS6000_BTM_COMMON): Add RS6000_BTM_HTM.
5272         * config/rs6000/htm.md: New file.
5273         * config/rs6000/htmintrin.h: New file.
5274         * config/rs6000/htmxlintrin.h: New file.
5276 2013-07-15  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5278         * config/aarch64/aarch64-protos.h (aarch64_symbol_type):
5279         Define SYMBOL_TINY_GOT, update comment.
5280         * config/aarch64/aarch64.c
5281         (aarch64_load_symref_appropriately): Handle SYMBOL_TINY_GOT.
5282         (aarch64_expand_mov_immediate): Likewise.
5283         (aarch64_print_operand): Likewise.
5284         (aarch64_classify_symbol): Likewise.
5285         * config/aarch64/aarch64.md (UNSPEC_GOTTINYPIC): Define.
5286         (ldr_got_tiny): Define.
5288 2013-07-13  Tobias Grosser  <tobias@grosser.es>
5290         PR tree-optimization/54094
5291         * graphite-clast-to-gimple.c (translate_clast_for_loop): Derive the
5292           scheduling dimension for the parallelism check from the polyhedral
5293           information in the AST.
5294         * graphite-dependences.c (carries_deps): Do not assume the schedule is
5295           in 2D + 1 form.
5297 2013-07-13  Jason Merrill  <jason@redhat.com>
5299         * print-tree.c (debug_vec_tree): Use debug_raw.
5300         (debug_raw (vec<tree, va_gc> &)): New.
5301         (debug_raw (vec<tree, va_gc> *)): New.
5302         * tree.h: Declare them.
5304 2013-07-13  Bin Cheng  <bin.cheng@arm.com>
5306         * ifcvt.c (ifcvt_after_combine): New static variable.
5307         (cheap_bb_rtx_cost_p): Set scale to REG_BR_PROB_BASE when optimizing
5308         for size.
5309         (if_convert): New parameter after_combine.  Set ifcvt_after_combine.
5310         (rest_of_handle_if_conversion, rest_of_handle_if_after_combine,
5311         rest_of_handle_if_after_reload): Pass new argument for if_convert.
5313 2013-07-12  Maciej W. Rozycki  <macro@codesourcery.com>
5315         * config/mips/mips.c (mips_expand_call): Remove empty statement.
5317 2013-07-12  Michael Matz  <matz@suse.de>
5319         PR middle-end/55771
5320         * convert.c (convert_to_real): Reject non-float inner types.
5322 2013-07-12  Tejas Belagod  <tejas.belagod@arm.com>
5324         * config/aarch64/aarch64-protos.h
5325         (aarch64_simd_immediate_valid_for_move): Remove.
5326         * config/aarch64/aarch64.c (simd_immediate_info): New member.
5327         (aarch64_simd_valid_immediate): Recognize idioms for shifting ones
5328         cases.
5329         (aarch64_output_simd_mov_immediate): Print the correct shift specifier.
5331 2013-07-11  Steve Ellcey  <sellcey@mips.com>
5333         * config/mips/mips.c (mips_conditional_register_usage): Do not
5334         use t[0-7] registers in MIPS16 mode when optimizing for size.
5336 2013-07-11  Sriraman Tallam  <tmsriram@google.com>
5338         * config/i386/i386.c (dispatch_function_versions): Fix array
5339         indexing of function_version_info to match actual_versions.
5341 2013-07-11  Teresa Johnson  <tejohnson@google.com>
5343         * vec.h (struct va_gc): Move release out-of-line.
5344         (va_gc::release): Call ggc_free on released vec.
5346 2013-07-11  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
5348         * config/rs6000/rs6000.md (""*tls_gd_low<TLSmode:tls_abi_suffix>"):
5349         Require GOT register as additional operand in UNSPEC.
5350         ("*tls_ld_low<TLSmode:tls_abi_suffix>"): Likewise.
5351         ("*tls_got_dtprel_low<TLSmode:tls_abi_suffix>"): Likewise.
5352         ("*tls_got_tprel_low<TLSmode:tls_abi_suffix>"): Likewise.
5353         ("*tls_gd<TLSmode:tls_abi_suffix>"): Update splitter.
5354         ("*tls_ld<TLSmode:tls_abi_suffix>"): Likewise.
5355         ("tls_got_dtprel_<TLSmode:tls_abi_suffix>"): Likewise.
5356         ("tls_got_tprel_<TLSmode:tls_abi_suffix>"): Likewise.
5358 2013-07-11  Georg-Johann Lay  <avr@gjlay.de>
5360         PR target/57631
5361         * config/avr/avr.c (avr_set_current_function): Sanity-check signal
5362         name seen by assembler/linker rather if available.
5364 2013-07-11  Andreas Schwab  <schwab@suse.de>
5366         * config/aarch64/aarch64-linux.h (CPP_SPEC): Define.
5368 2013-07-10  Vladimir Makarov  <vmakarov@redhat.com>
5370         * lra-constraints.c (curr_insn_transform): Switch off optional reloads.
5372 2013-07-10  Joseph Myers  <joseph@codesourcery.com>
5374         * doc/tm.texi.in: Move hook documentation to ....
5375         * target.def: ... here.
5377         * doc/tm.texi.in (TARGET_CANONICALIZE_COMPARISON): Remove stray
5378         text on @hook line.
5379         * doc/tm.texi: Regenerate.
5381 2013-07-10  Paolo Carlini  <paolo.carlini@oracle.com>
5383         PR c++/57869
5384         * doc/invoke.texi: Document -Wconditionally-supported.
5386 2013-07-10  Georg-Johann Lay  <avr@gjlay.de>
5388         PR target/57844
5389         * config/avr/avr.c (avr_prologue_setup_frame): Trunk -size to mode
5390         of my_fp.
5392 2013-07-10  Georg-Johann Lay  <avr@gjlay.de>
5394         PR target/57506
5395         * config/avr/avr-mcus.def (atmega16hva, atmega16hva2, atmega16hvb)
5396         (atmega16m1, atmega16u4, atmega32a, atmega32c1, atmega32hvb)
5397         (atmega32m1, atmega32u4, atmega32u6, atmega64c1, atmega64m1):
5398         Remove duplicate devices.
5399         * config/avr/gen-avr-mmcu-texi.c (print_mcus): Fail on duplicate MCUs.
5400         * config/avr/t-multilib: Regenerate.
5401         * config/avr/avr-tables.opt: Regenerate.
5402         * doc/avr-mmcu.texi: Regenerate.
5404 2013-07-10  Georg-Johann Lay  <avr@gjlay.de>
5406         PR target/56987
5407         * config/avr/avr.opt (Waddr-space-convert): Fix typo.
5409 2013-07-10  Graham Stott  <graham.stott@btinternet.com>
5411         * config/mips/mips.c (mips_rtx_costs): Very slightly increase
5412         the cost of MULT when optimizing for size.
5414 2013-07-10  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5416         * config/cr16/cr16-protos.h: Don't include target.h.
5418 2013-07-09  Joseph Myers  <joseph@codesourcery.com>
5420         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Only
5421         adjust register size for TDmode and TFmode for VSX registers.
5423 2013-07-08  Kai Tietz  <ktietz@redhat.com>
5425         PR target/56892
5426         * config/i386/i386.c (TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P): Define as
5427         hook_bool_const_tree_true.
5429 2013-07-08  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5431         * config/s390/s390.c: Replace F*_REGNUM with FPR*_REGNUM.
5432         * config/s390/s390.h: Remove F*_REGNUM macro definitions.
5433         * config/s390/s390.md: Define FPR*_REGNUM constants.
5434         Fix FPR2_REGNUM constant (18 -> 17).
5435         ("*trunc<BFP:mode><DFP_ALL:mode>2")
5436         ("*trunc<DFP_ALL:mode><BFP:mode>2")
5437         ("trunc<BFP:mode><DFP_ALL:mode>2")
5438         ("trunc<DFP_ALL:mode><BFP:mode>2")
5439         ("*extend<BFP:mode><DFP_ALL:mode>2")
5440         ("*extend<DFP_ALL:mode><BFP:mode>2")
5441         ("extend<BFP:mode><DFP_ALL:mode>2")
5442         ("extend<DFP_ALL:mode><BFP:mode>2"): Replace FPR2_REGNUM with
5443         FPR4_REGNUM.
5445 2013-07-08  Graham Stott  <graham.stott@btinternet.com>
5447         * Makefile.in: (c-family-warn): Define to $(STRICT_WARN)
5449 2013-07-08  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5451         * config/s390/s390.c: Rename cfun_set_fpr_bit to cfun_set_fpr_save
5452         and cfun_fpr_bit_p to cfun_fpr_save_p.
5453         (s390_frame_area, s390_register_info, s390_frame_info)
5454         (s390_emit_prologue, s390_emit_epilogue)
5455         (s390_conditional_register_usage): Use the *_REGNUM macros for FPR
5456         register numbers.
5457         * config/s390/s390.h: Define *_REGNUM macros for floating point
5458         register numbers.
5460 2013-07-08  Eric Botcazou  <ebotcazou@adacore.com>
5462         * Makefile.in (tree-ssa-reassoc.o): Add dependency on $(PARAMS_H).
5464 2013-07-08  Po-Chun Chang  <pchang9@cs.wisc.edu>
5466         PR rtl-optimization/57786
5467         * combine.c (distribute_notes) <case REG_DEAD>: Change all_used to bool
5468         and break out of the loop when it is set to false.
5470 2013-07-08  Jakub Jelinek  <jakub@redhat.com>
5472         PR target/57819
5473         * simplify-rtx.c (simplify_unary_operation_1) <case ZERO_EXTEND>:
5474         Simplify (zero_extend:SI (subreg:QI (and:SI (reg:SI)
5475         (const_int 63)) 0)).
5476         * combine.c (make_extraction): Create ZERO_EXTEND or SIGN_EXTEND
5477         using simplify_gen_unary instead of gen_rtx_*_EXTEND.
5478         * config/i386/i386.md (*jcc_bt<mode>_1): New define_insn_and_split.
5480         PR rtl-optimization/57829
5481         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Ensure that
5482         mask bits outside of mode are just sign-extension from mode to HWI.
5484 2013-07-08  Michael Zolotukhin  <michael.v.zolotukhin@gmail.com>
5486         * config/i386/i386-opts.h (enum stringop_alg): Add vector_loop.
5487         * config/i386/i386.c (expand_set_or_movmem_via_loop): Use
5488         adjust_address instead of change_address to keep info about alignment.
5489         (emit_strmov): Remove.
5490         (emit_memmov): New function.
5491         (expand_movmem_epilogue): Refactor to properly handle bigger sizes.
5492         (expand_movmem_epilogue): Likewise and return updated rtx for
5493         destination.
5494         (expand_constant_movmem_prologue): Likewise and return updated rtx for
5495         destination and source.
5496         (decide_alignment): Refactor, handle vector_loop.
5497         (ix86_expand_movmem): Likewise.
5498         (ix86_expand_setmem): Likewise.
5499         * config/i386/i386.opt (Enum): Add vector_loop to option stringop_alg.
5501 2013-07-07  Uros Bizjak  <ubizjak@gmail.com>
5503         * config/i386/driver-i386.c (host_detect_local_cpu): Do not check
5504         signature_TM2_ebx, it interferes with signature_INTEL_ebx.
5506 2013-07-06  Uros Bizjak  <ubizjak@gmail.com>
5508         * config/i386/sse.md (sse_movlhps): Change alternative 3
5509         of operand 2 to "m".
5511 2013-07-06  Uros Bizjak  <ubizjak@gmail.com>
5513         PR target/57807
5514         * config/i386/sse.md (iptr): New mode attribute.
5515         (sse2_movq128): Add pointer size overrides for Intel asm dialect.
5516         (<sse>_vm<plusminus_insn><mode>3): Ditto.
5517         (<sse>_vmmul<mode>3): Ditto.
5518         (<sse>_vmdiv<mode>3): Ditto.
5519         (sse_vmrcpv4sf2): Ditto.
5520         (<sse>_vmsqrt<mode>2): Ditto.
5521         (sse_vmrsqrtv4sf2): Ditto.
5522         (<sse>_vm<code><mode>3): Ditto.
5523         (avx_vmcmp<mode>3): Ditto.
5524         (<sse>_vmmaskcmp<mode>3): Ditto.
5525         (<sse>_comi): Ditto.
5526         (<sse>_ucomi): Ditto.
5527         (*xop_vmfrcz_<mode>): Ditto.
5528         (*fmai_fmadd_<mode>): Ditto.
5529         (*fmai_fmsub_<mode>): Ditto.
5530         (*fmai_fnmadd_<mode>): Ditto.
5531         (*fmai_fnmsub_<mode>): Ditto.
5532         (*fma4i_vmfmadd_<mode>): Ditto.
5533         (*fma4i_vmfmsub_<mode>): Ditto.
5534         (*fma4i_vmfnmadd_<mode>): Ditto.
5535         (*fma4i_vmfnmsub_<mode>): Ditto.
5536         (*xop_vmfrcz_<mode>): Ditto.
5537         (sse_cvtps2pi): Ditto.
5538         (sse_cvttps2pi): Ditto.
5539         (sse_cvtss2si): Ditto.
5540         (sse_cvtss2si_2): Ditto.
5541         (sse_cvtss2siq_2): Ditto.
5542         (sse_cvttss2si): Ditto.
5543         (sse_cvttss2siq): Ditto.
5544         (sse_cvtsd2si): Ditto.
5545         (sse_cvtsd2si_2): Ditto.
5546         (sse_cvtsd2siq_2): Ditto.
5547         (sse_cvttsd2si): Ditto.
5548         (sse_cvttsd2siq): Ditto.
5549         (sse_cvtsd2ss): Ditto.
5550         (sse_cvtss2sd): Ditto.
5551         (avx2_pbroadcast<mode>): Ditto.
5552         (avx2_pbroadcast<mode>_1): Ditto.
5553         (*avx_vperm_broadcast_v4sf): Ditto.
5555         (sse_movhlps): Ditto for movlp[sd]/movhp[sd] alternatives.
5556         (sse_movlhps): Ditto.
5557         (sse_storehps): Ditto.
5558         (sse_loadhps): Ditto.
5559         (sse_storelps): Ditto.
5560         (sse_loadlps): Ditto.
5561         (*vec_concatv4sf): Ditto.
5562         (*vec_interleave_highv2df): Ditto.
5563         (*vec_interleave_lowv2df): Ditto.
5564         (*vec_extractv2df_1_sse): Ditto.
5565         (*vec_extractv2df_0_sse): Ditto.
5566         (sse2_storelpd): Ditto.
5567         (sse2_loadlpd): Ditto.
5568         (sse2_movsd): Ditto.
5569         (*vec_concatv4si): Ditto.
5570         (vec_concatv2di): Ditto.
5572         * config/i386/mmx.md (mmx_punpcklbw): Add pointer size overrides
5573         for Intel asm dialect.
5574         (mmx_punpcklwd): Ditto.
5575         (mmx_punpckldq): Ditto.
5577         * config/i386/i386.c (ix86_print_operand) ['H']: Output 'qword ptr'
5578         for intel assembler dialect.
5580 2013-07-06  Jakub Jelinek  <jakub@redhat.com>
5582         PR target/29776
5583         * fold-const.c (tree_call_nonnegative_warnv_p): Return true
5584         for BUILT_IN_C{LZ,LRSB}*.
5585         * tree.h (CASE_INT_FN): Add FN##IMAX case.
5586         * tree-vrp.c (extract_range_basic): Handle
5587         BUILT_IN_{FFS,PARITY,POPCOUNT,C{LZ,TZ,LRSB}}*.  For
5588         BUILT_IN_CONSTANT_P if argument isn't (D) of PARM_DECL,
5589         fall thru to code calling set_value*.
5590         * builtins.c (expand_builtin): Remove *IMAX cases.
5591         (fold_builtin_bitop): For BUILT_IN_CLRSB* return NULL_TREE
5592         if width is bigger than 2*HWI.
5594 2013-07-05  Vladimir Makarov  <vmakarov@redhat.com>
5596         PR rtl-optimization/55342
5597         * lra-int.h (lra_subreg_reload_pseudos): New.
5598         * lra.c: Add undoing optional reloads to the block diagram.
5599         (lra_subreg_reload_pseudos): New.
5600         (lra_optional_reload_pseudos): Change comments.
5601         (lra): Init and clear lra_subreg_reload_pseudos.  Clear
5602         lra_optional_reload_pseudos after undo transformations.
5603         * lra-assigns.c (pseudo_prefix_title): New.
5604         (lra_setup_reg_renumber): Use it.
5605         (spill_for): Ditto.  Check subreg reload pseudos too.
5606         (assign_by_spills): Consider subreg reload pseudos too.
5607         * lra-constraints.c (simplify_operand_subreg): Use
5608         lra_subreg_reload_pseudos instead of lra_optional_reload_pseudos.
5609         (curr_insn_transform): Recognize and do optional reloads.
5610         (undo_optional_reloads): New.
5611         (lra_undo_inheritance): Call undo_optional_reloads.
5613 2013-07-05  Thomas Quinot  <quinot@adacore.com>
5615         * tree-complex.c (expand_complex_operations_1): Fix typo.
5617 2013-07-04  Tejas Belagod  <tejas.belagod@arm.com>
5619         * config/aarch64/aarch64-protos.h (cpu_vector_cost): New.
5620         (tune_params): New member 'const vec_costs'.
5621         * config/aarch64/aarch64.c (generic_vector_cost): New.
5622         (generic_tunings): New member 'generic_vector_cost'.
5623         (aarch64_builtin_vectorization_cost): New.
5624         (aarch64_add_stmt_cost): New.
5625         (TARGET_VECTORIZE_ADD_STMT_COST): New.
5626         (TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST): New.
5628 2013-07-03  Jakub Jelinek  <jakub@redhat.com>
5630         PR target/57777
5631         * config/i386/predicates.md (vsib_address_operand): Disallow
5632         SYMBOL_REF or LABEL_REF in parts.disp if TARGET_64BIT && flag_pic.
5634 2013-07-03  Hans-Peter Nilsson  <hp@bitrange.com>
5636         PR middle-end/55030
5637         * stmt.c (expand_nl_goto_receiver): Remove almost-copy of
5638         expand_builtin_setjmp_receiver.
5639         (expand_label): Adjust, call expand_builtin_setjmp_receiver
5640         with NULL for the label parameter.
5641         * builtins.c (expand_builtin_setjmp_receiver): Don't clobber
5642         the frame-pointer.  Adjust comments.
5643         [HAVE_builtin_setjmp_receiver]: Emit builtin_setjmp_receiver
5644         only if LABEL is non-NULL.
5646 2013-07-03  Yufeng Zhang  <yufeng.zhang@arm.com>
5648         * config/aarch64/aarch64.h (enum arm_abi_type): Remove.
5649         (ARM_ABI_AAPCS64): Ditto.
5650         (arm_abi): Ditto.
5651         (ARM_DEFAULT_ABI): Ditto.
5653 2013-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
5655         * config/aarch64/aarch64-builtins.c
5656         (aarch64_simd_expand_builtin): Handle AARCH64_SIMD_STORE1.
5657         * config/aarch64/aarch64-simd-builtins.def (ld1): New.
5658         (st1): Likewise.
5659         * config/aarch64/aarch64-simd.md
5660         (aarch64_ld1<VALL:mode>): New.
5661         (aarch64_st1<VALL:mode>): Likewise.
5662         * config/aarch64/arm_neon.h
5663         (vld1<q>_<fpsu><8, 16, 32, 64>): Convert to RTL builtins.
5665 2013-07-02  Sriraman Tallam  <tmsriram@google.com>
5667         * config/i386/i386.c (gate_insert_vzeroupper): Check if
5668         target ISA is AVX.
5669         (ix86_option_override_internal):Turn on all -mavx target flags by
5670         default as they are dependent on AVX anyway.
5672 2013-07-02  Cary Coutant  <ccoutant@google.com>
5674         * dwarf2out.c (loc_checksum): Call hash_loc_operands for a
5675         deterministic hash.
5676         (loc_checksum_ordered): Likewise.
5677         (hash_loc_operands): Remove inline keyword.
5679 2013-07-02  Jakub Jelinek  <jakub@redhat.com>
5681         PR tree-optimization/57741
5682         * tree-vect-loop.c (vect_is_simple_iv_evolution): Disallow
5683         non-INTEGRAL_TYPE_P non-SCALAR_FLOAT_TYPE_P SSA_NAME step_exprs,
5684         or SCALAR_FLOAT_TYPE_P SSA_NAMEs if !flag_associative_math.
5685         Allow REAL_CST step_exprs if flag_associative_math.
5686         (get_initial_def_for_induction): Handle SCALAR_FLOAT_TYPE_P step_expr.
5688 2013-07-02  Ian Bolton  <ian.bolton@arm.com>
5690         * config/aarch64/aarch64-simd.md (absdi2): Support abs for DI mode.
5692 2013-07-02  Ian Bolton  <ian.bolton@arm.com>
5694         * config/aarch64/aarch64.md (*extr_insv_reg<mode>): New pattern.
5696 2013-07-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5698         * config/arm/arm.md (arm_andsi3_insn): Add alternatives for 16-bit
5699         encoding.
5700         (iorsi3_insn): Likewise.
5701         (arm_xorsi3): Likewise.
5703 2013-07-01  Sofiane Naci  <sofiane.naci@arm.com>
5705         * arm.md (attribute "wtype"): Delete.  Move attribute values from here
5706         to ...
5707         (attribute "type"): ... here, and prefix with "wmmx_".
5708         (attribute "core_cycles"): Update for attribute changes.
5709         * iwmmxt.md (tbcstv8qi): Update for attribute changes.
5710         (tbcstv4hi): Likewise.
5711         (tbcstv2si): Likewise.
5712         (iwmmxt_iordi3): Likewise.
5713         (iwmmxt_xordi3): Likewise.
5714         (iwmmxt_anddi3): Likewise.
5715         (iwmmxt_nanddi3): Likewise.
5716         (iwmmxt_arm_movdi): Likewise.
5717         (iwmmxt_movsi_insn): Likewise.
5718         (mov<mode>_internal): Likewise.
5719         (and<mode>3_iwmmxt): Likewise.
5720         (ior<mode>3_iwmmxt): Likewise.
5721         (xor<mode>3_iwmmxt): Likewise.
5722         (add<mode>3_iwmmxt): Likewise.
5723         (ssaddv8qi3): Likewise.
5724         (ssaddv4hi3): Likewise.
5725         (ssaddv2si3): Likewise.
5726         (usaddv8qi3): Likewise.
5727         (usaddv4hi3): Likewise.
5728         (usaddv2si3): Likewise.
5729         (sub<mode>3_iwmmxt): Likewise.
5730         (sssubv8qi3): Likewise.
5731         (sssubv4hi3): Likewise.
5732         (sssubv2si3): Likewise.
5733         (ussubv8qi3): Likewise.
5734         (ussubv4hi3): Likewise.
5735         (ussubv2si3): Likewise.
5736         (mulv4hi3_iwmmxt): Likewise.
5737         (smulv4hi3_highpart): Likewise.
5738         (umulv4hi3_highpart): Likewise.
5739         (iwmmxt_wmacs): Likewise.
5740         (iwmmxt_wmacsz): Likewise.
5741         (iwmmxt_wmacu): Likewise.
5742         (iwmmxt_wmacuz): Likewise.
5743         (iwmmxt_clrdi): Likewise.
5744         (iwmmxt_clrv8qi): Likewise.
5745         (iwmmxt_clr4hi): Likewise.
5746         (iwmmxt_clr2si): Likewise.
5747         (iwmmxt_uavgrndv8qi3): Likewise.
5748         (iwmmxt_uavgrndv4hi3): Likewise.
5749         (iwmmxt_uavgv8qi3): Likewise.
5750         (iwmmxt_uavgv4hi3): Likewise.
5751         (iwmmxt_tinsrb): Likewise.
5752         (iwmmxt_tinsrh): Likewise.
5753         (iwmmxt_tinsrw): Likewise.
5754         (iwmmxt_textrmub): Likewise.
5755         (iwmmxt_textrmsb): Likewise.
5756         (iwmmxt_textrmuh): Likewise.
5757         (iwmmxt_textrmsh): Likewise.
5758         (iwmmxt_textrmw): Likewise.
5759         (iwmxxt_wshufh): Likewise.
5760         (eqv8qi3): Likewise.
5761         (eqv4hi3): Likewise.
5762         (eqv2si3): Likewise.
5763         (gtuv8qi3): Likewise.
5764         (gtuv4hi3): Likewise.
5765         (gtuv2si3): Likewise.
5766         (gtv8qi3): Likewise.
5767         (gtv4hi3): Likewise.
5768         (gtv2si3): Likewise.
5769         (smax<mode>3_iwmmxt): Likewise.
5770         (umax<mode>3_iwmmxt): Likewise.
5771         (smin<mode>3_iwmmxt): Likewise.
5772         (umin<mode>3_iwmmxt): Likewise.
5773         (iwmmxt_wpackhss): Likewise.
5774         (iwmmxt_wpackwss): Likewise.
5775         (iwmmxt_wpackdss): Likewise.
5776         (iwmmxt_wpackhus): Likewise.
5777         (iwmmxt_wpackwus): Likewise.
5778         (iwmmxt_wpackdus): Likewise.
5779         (iwmmxt_wunpckihb): Likewise.
5780         (iwmmxt_wunpckihh): Likewise.
5781         (iwmmxt_wunpckihw): Likewise.
5782         (iwmmxt_wunpckilb): Likewise.
5783         (iwmmxt_wunpckilh): Likewise.
5784         (iwmmxt_wunpckilw): Likewise.
5785         (iwmmxt_wunpckehub): Likewise.
5786         (iwmmxt_wunpckehuh): Likewise.
5787         (iwmmxt_wunpckehuw): Likewise.
5788         (iwmmxt_wunpckehsb): Likewise.
5789         (iwmmxt_wunpckehsh): Likewise.
5790         (iwmmxt_wunpckehsw): Likewise.
5791         (iwmmxt_wunpckelub): Likewise.
5792         (iwmmxt_wunpckeluh): Likewise.
5793         (iwmmxt_wunpckeluw): Likewise.
5794         (iwmmxt_wunpckelsb): Likewise.
5795         (iwmmxt_wunpckelsh): Likewise.
5796         (iwmmxt_wunpckelsw): Likewise.
5797         (ror<mode>3): Likewise.
5798         (ashr<mode>3_iwmmxt): Likewise.
5799         (lshr<mode>3_iwmmxt): Likewise.
5800         (ashl<mode>3_iwmmxt): Likewise.
5801         (ror<mode>3_di): Likewise.
5802         (ashr<mode>3_di): Likewise.
5803         (lshr<mode>3_di): Likewise.
5804         (ashl<mode>3_di): Likewise.
5805         (iwmmxt_wmadds): Likewise.
5806         (iwmmxt_wmaddu): Likewise.
5807         (iwmmxt_tmia): Likewise.
5808         (iwmmxt_tmiaph): Likewise.
5809         (iwmmxt_tmiabb): Likewise.
5810         (iwmmxt_tmiatb): Likewise.
5811         (iwmmxt_tmiabt): Likewise.
5812         (iwmmxt_tmiatt): Likewise.
5813         (iwmmxt_tmovmskb): Likewise.
5814         (iwmmxt_tmovmskh): Likewise.
5815         (iwmmxt_tmovmskw): Likewise.
5816         (iwmmxt_waccb): Likewise.
5817         (iwmmxt_wacch): Likewise.
5818         (iwmmxt_waccw): Likewise.
5819         (iwmmxt_waligni): Likewise.
5820         (iwmmxt_walignr): Likewise.
5821         (iwmmxt_walignr0): Likewise.
5822         (iwmmxt_walignr1): Likewise.
5823         (iwmmxt_walignr2): Likewise.
5824         (iwmmxt_walignr3): Likewise.
5825         (iwmmxt_wsadb): Likewise.
5826         (iwmmxt_wsadh): Likewise.
5827         (iwmmxt_wsadbz): Likewise.
5828         (iwmmxt_wsadhz): Likewise.
5829         * iwmmxt2.md (iwmmxt_wabs<mode>3): Update for attribute changes.
5830         (iwmmxt_wabsdiffb): Likewise.
5831         (iwmmxt_wabsdiffh): Likewise.
5832         (iwmmxt_wabsdiffw): Likewise.
5833         (iwmmxt_waddsubhx): Likewise
5834         (iwmmxt_wsubaddhx): Likewise.
5835         (addc<mode>3): Likewise.
5836         (iwmmxt_avg4): Likewise.
5837         (iwmmxt_avg4r): Likewise.
5838         (iwmmxt_wmaddsx): Likewise.
5839         (iwmmxt_wmaddux): Likewise.
5840         (iwmmxt_wmaddsn): Likewise.
5841         (iwmmxt_wmaddun): Likewise.
5842         (iwmmxt_wmulwsm): Likewise.
5843         (iwmmxt_wmulwum): Likewise.
5844         (iwmmxt_wmulsmr): Likewise.
5845         (iwmmxt_wmulumr): Likewise.
5846         (iwmmxt_wmulwsmr): Likewise.
5847         (iwmmxt_wmulwumr): Likewise.
5848         (iwmmxt_wmulwl): Likewise.
5849         (iwmmxt_wqmulm): Likewise.
5850         (iwmmxt_wqmulwm): Likewise.
5851         (iwmmxt_wqmulmr): Likewise.
5852         (iwmmxt_wqmulwmr): Likewise.
5853         (iwmmxt_waddbhusm): Likewise.
5854         (iwmmxt_waddbhusl): Likewise.
5855         (iwmmxt_wqmiabb): Likewise.
5856         (iwmmxt_wqmiabt): Likewise.
5857         (iwmmxt_wqmiatb): Likewise.
5858         (iwmmxt_wqmiatt): Likewise.
5859         (iwmmxt_wqmiabbn): Likewise.
5860         (iwmmxt_wqmiabtn): Likewise.
5861         (iwmmxt_wqmiatbn): Likewise.
5862         (iwmmxt_wqmiattn): Likewise.
5863         (iwmmxt_wmiabb): Likewise.
5864         (iwmmxt_wmiabt): Likewise.
5865         (iwmmxt_wmiatb): Likewise.
5866         (iwmmxt_wmiatt): Likewise.
5867         (iwmmxt_wmiabbn): Likewise.
5868         (iwmmxt_wmiabtn): Likewise.
5869         (iwmmxt_wmiatbn): Likewise.
5870         (iwmmxt_wmiattn): Likewise.
5871         (iwmmxt_wmiawbb): Likewise.
5872         (iwmmxt_wmiawbt): Likewise.
5873         (iwmmxt_wmiawtb): Likewise.
5874         (iwmmxt_wmiawtt): Likewise.
5875         (iwmmxt_wmiawbbn): Likewise.
5876         (iwmmxt_wmiawbtn): Likewise.
5877         (iwmmxt_wmiawtbn): Likewise.
5878         (iwmmxt_wmiawttn): Likewise.
5879         (iwmmxt_wmerge): Likewise.
5880         (iwmmxt_tandc<mode>3): Likewise.
5881         (iwmmxt_torc<mode>3): Likewise.
5882         (iwmmxt_torvsc<mode>3): Likewise.
5883         (iwmmxt_textrc<mode>3): Likewise.
5884         * marvell-f-iwmmxt.md (wmmxt_shift): Update for attribute changes.
5885         (wmmxt_pack): Likewise.
5886         (wmmxt_mult_c1): Likewise.
5887         (wmmxt_mult_c2): Likewise.
5888         (wmmxt_alu_c1): Likewise.
5889         (wmmxt_alu_c2): Likewise.
5890         (wmmxt_alu_c3): Likewise.
5891         (wmmxt_transfer_c1): Likewise.
5892         (wmmxt_transfer_c2): Likewise.
5893         (wmmxt_transfer_c3): Likewise.
5894         (marvell_f_iwmmxt_wstr): Likewise.
5895         (marvell_f_iwmmxt_wldr): Likewise.
5897 2013-06-29  Yufeng Zhang  <yufeng.zhang@arm.com>
5899         * config/aarch64/aarch64.c: Remove junk from the beginning of the file.
5901 2013-06-28  Vladimir Makarov  <vmakarov@redhat.com>
5903         Revert:
5904         2013-06-28  Vladimir Makarov  <vmakarov@redhat.com>
5905         * lra-constraints.c (need_for_split_p): Check call used hard regs
5906         living through calls.
5908         * lra-constraints.c (inherit_in_ebb): Reset live_hard_regs for
5909         call used regs for call insn.
5911 2013-06-28  Jakub Jelinek  <jakub@redhat.com>
5913         PR target/57736
5914         * config/i386/i386.c (ix86_expand_builtin): If target == NULL and
5915         mode is VOIDmode, don't create a VOIDmode pseudo to copy result into.
5917 2013-06-28  Balaji V. Iyer  <balaji.v.iyer@intel.com>
5919         * builtins.def: Fixed the function type of CILKPLUS_BUILTIN.
5921 2013-06-28  Vladimir Makarov  <vmakarov@redhat.com>
5923         * lra-constraints.c (need_for_split_p): Check call used hard regs
5924         living through calls.
5926 2013-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
5928         PR target/57744
5929         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Do not allow PTImode
5930         to tie with any other modes.  Eliminate Altivec vector mode tests,
5931         since these are a subset of ALTIVEC or VSX vector modes.  Simplify
5932         code, to return 0 if testing MODE2 for a condition, if we've
5933         already tested MODE1 for the same condition.
5935 2013-06-28  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5937         * config/aarch64/aarch64.c (aarch64_cannot_force_const_mem): Adjust
5938         layout.
5940 2013-06-28  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5942         * config/aarch64/aarch64-protos.h (aarch64_symbol_type):
5943         Update comment w.r.t SYMBOL_TINY_ABSOLUTE.
5945 2013-06-28  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5947         * config/aarch64/aarch64-protos.h (aarch64_classify_symbol_expression):
5948         Define.
5949         (aarch64_symbolic_constant_p): Remove.
5950         * config/aarch64/aarch64.c (aarch64_classify_symbol_expression): Remove
5951         static.  Fix line length and white space.
5952         (aarch64_symbolic_constant_p): Remove.
5953         * config/aarch64/predicates.md (aarch64_valid_symref):
5954         Use aarch64_classify_symbol_expression.
5956 2013-06-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5958         * config/arm/constraints.md (Ts): New constraint.
5959         * config/arm/arm.md (arm_movqi_insn): Add alternatives for
5960         16-bit encodings.
5961         (compare_scc): Use "Ts" constraint for operand 0.
5962         (ior_scc_scc): Likewise.
5963         (and_scc_scc): Likewise.
5964         (and_scc_scc_nodom): Likewise.
5965         (ior_scc_scc_cmp): Likewise for operand 7.
5966         (and_scc_scc_cmp): Likewise.
5967         * config/arm/thumb2.md (thumb2_movsi_insn):
5968         Add alternatives for 16-bit encodings.
5969         (thumb2_movhi_insn): Likewise.
5970         (thumb2_movsicc_insn): Likewise.
5971         (thumb2_and_scc): Take 'and' outside cond_exec.  Use "Ts" constraint.
5972         (thumb2_negscc): Use "Ts" constraint.
5973         Move mvn instruction outside cond_exec block.
5974         * config/arm/vfp.md (thumb2_movsi_vfp): Add alternatives
5975         for 16-bit encodings.
5977 2013-06-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5979         * config/arm/arm.md (arm_mulsi3_v6): Add alternative for 16-bit
5980         encoding.
5981         (mulsi3addsi_v6): Disable predicable variant for arm_restrict_it.
5982         (mulsi3subsi): Likewise.
5983         (mulsidi3adddi): Likewise.
5984         (mulsidi3_v6): Likewise.
5985         (umulsidi3_v6): Likewise.
5986         (umulsidi3adddi_v6): Likewise.
5987         (smulsi3_highpart_v6): Likewise.
5988         (umulsi3_highpart_v6): Likewise.
5989         (mulhisi3tb): Likewise.
5990         (mulhisi3bt): Likewise.
5991         (mulhisi3tt): Likewise.
5992         (maddhisi4): Likewise.
5993         (maddhisi4tb): Likewise.
5994         (maddhisi4tt): Likewise.
5995         (maddhidi4): Likewise.
5996         (maddhidi4tb): Likewise.
5997         (maddhidi4tt): Likewise.
5998         (zeroextractsi_compare0_scratch): Likewise.
5999         (insv_zero): Likewise.
6000         (insv_t2): Likewise.
6001         (anddi_notzesidi_di): Likewise.
6002         (anddi_notsesidi_di): Likewise.
6003         (andsi_notsi_si): Likewise.
6004         (iordi_zesidi_di): Likewise.
6005         (xordi_zesidi_di): Likewise.
6006         (andsi_iorsi3_notsi): Likewise.
6007         (smax_0): Likewise.
6008         (smax_m1): Likewise.
6009         (smin_0): Likewise.
6010         (not_shiftsi): Likewise.
6011         (unaligned_loadsi): Likewise.
6012         (unaligned_loadhis): Likewise.
6013         (unaligned_loadhiu): Likewise.
6014         (unaligned_storesi): Likewise.
6015         (unaligned_storehi): Likewise.
6016         (extv_reg): Likewise.
6017         (extzv_t2): Likewise.
6018         (divsi3): Likewise.
6019         (udivsi3): Likewise.
6020         (arm_zero_extendhisi2addsi): Likewise.
6021         (arm_zero_extendqisi2addsi): Likewise.
6022         (compareqi_eq0): Likewise.
6023         (arm_extendhisi2_v6): Likewise.
6024         (arm_extendqisi2addsi): Likewise.
6025         (arm_movt): Likewise.
6026         (thumb2_ldrd): Likewise.
6027         (thumb2_ldrd_base): Likewise.
6028         (thumb2_ldrd_base_neg): Likewise.
6029         (thumb2_strd): Likewise.
6030         (thumb2_strd_base): Likewise.
6031         (thumb2_strd_base_neg): Likewise.
6032         (arm_negsi2): Add alternative for 16-bit encoding.
6033         (arm_one_cmplsi2): Likewise.
6035 2013-06-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
6037         * config/arm/predicates.md (arm_cond_move_operator): New predicate.
6038         * config/arm/arm.md (movsfcc): Use arm_cond_move_operator predicate.
6039         (movdfcc): Likewise.
6040         * config/arm/vfp.md (*thumb2_movsf_vfp):
6041         Disable predication for arm_restrict_it.
6042         (*thumb2_movsfcc_vfp): Disable for arm_restrict_it.
6043         (*thumb2_movdfcc_vfp): Likewise.
6044         (*abssf2_vfp, *absdf2_vfp, *negsf2_vfp, *negdf2_vfp,*addsf3_vfp,
6045         *adddf3_vfp, *subsf3_vfp, *subdf3_vfpc, *divsf3_vfp,*divdf3_vfp,
6046         *mulsf3_vfp, *muldf3_vfp, *mulsf3negsf_vfp, *muldf3negdf_vfp,
6047         *mulsf3addsf_vfp, *muldf3adddf_vfp, *mulsf3subsf_vfp,
6048         *muldf3subdf_vfp, *mulsf3negsfaddsf_vfp, *fmuldf3negdfadddf_vfp,
6049         *mulsf3negsfsubsf_vfp, *muldf3negdfsubdf_vfp, *fma<SDF:mode>4,
6050         *fmsub<SDF:mode>4, *fnmsub<SDF:mode>4, *fnmadd<SDF:mode>4,
6051         *extendsfdf2_vfp, *truncdfsf2_vfp, *extendhfsf2, *truncsfhf2,
6052         *truncsisf2_vfp, *truncsidf2_vfp, fixuns_truncsfsi2, fixuns_truncdfsi2,
6053         *floatsisf2_vfp, *floatsidf2_vfp, floatunssisf2, floatunssidf2,
6054         *sqrtsf2_vfp, *sqrtdf2_vfp, *cmpsf_vfp, *cmpsf_trap_vfp, *cmpdf_vfp,
6055         *cmpdf_trap_vfp, <vrint_pattern><SDF:mode>2):
6056         Disable predication for arm_restrict_it.
6058 2013-06-28  Kirill Yukhin  <kirill.yukhin@intel.com>
6060         * config/i386/bmiintrin.h (_bextr_u32): New.
6061         (_bextr_u64): Ditto.
6063 2013-06-27  Richard Sandiford  <rdsandiford@googlemail.com>
6065         * config.gcc (mips*-mti-elf*, mips*-sde-elf*, mips64r5900-*-elf*)
6066         (mips64r5900el-*-elf*): Include mips/n32-elf.h.
6067         * config/mips/sde.h (LOCAL_LABEL_PREFIX, NO_DOLLAR_IN_LABEL)
6068         (LONG_DOUBLE_TYPE_SIZE, LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Move to...
6069         * config/mips/n32-elf.h: ...this new file.
6071 2013-06-27  Marc Glisse  <marc.glisse@inria.fr>
6073         PR target/57224
6074         * config/i386/i386.c (enum ix86_builtins, bdesc_args): Remove
6075         IX86_BUILTIN_CMPNGTSS and IX86_BUILTIN_CMPNGESS.
6077 2013-06-27  Catherine Moore  <clm@codesourcery.com>
6079         * config/mips/mips-tables.opt: Regenerate.
6080         * config/mips/mips-cpus.def: Add m14ke and m14kec.
6081         * config/mips/mips.h (BASE_DRIVER_SELF_SPECS): m14ke* implies -mdspr2.
6082         * doc/invoke.texi: Add -m14kc.
6084 2013-06-27  Jakub Jelinek  <jakub@redhat.com>
6086         PR target/57623
6087         * config/i386/i386.md (bmi_bextr_<mode>): Swap predicates and
6088         constraints of operand 1 and 2.
6090         PR target/57623
6091         * config/i386/i386.md (bmi2_bzhi_<mode>3): Swap AND arguments
6092         to match RTL canonicalization.  Swap predicates and
6093         constraints of operand 1 and 2.
6095 2013-06-27  Vladimir Makarov  <vmakarov@redhat.com>
6097         * lra-constraints.c (inherit_in_ebb): Process static hard regs too.
6098         Process OP_INOUT regs for splitting too.
6100 2013-06-27  Jakub Jelinek  <jakub@redhat.com>
6102         * tree-vect-stmts.c (vectorizable_store): Move ptr_incr var
6103         decl before the loop, initialize to NULL.
6104         (vectorizable_load): Initialize ptr_incr to NULL.
6106 2013-06-27  Martin Jambor  <mjambor@suse.cz>
6108         PR lto/57208
6109         * ipa-ref.h (ipa_maybe_record_reference): Declare.
6110         * ipa-ref.c (ipa_maybe_record_reference): New function.
6111         * cgraphclones.c (cgraph_create_virtual_clone): Use it.
6112         * ipa-cp.c (create_specialized_node): Record potential references from
6113         aggvals.
6114         * Makefile.in (ipa-ref.o): Add IPA_REF_H to dependencies.
6116 2013-06-27  Yufeng Zhang  <yufeng.zhang@arm.com>
6118         * config/aarch64/aarch64.c (aarch64_force_temporary): Add an extra
6119         parameter 'mode' of type 'enum machine_mode mode'; change to pass
6120         'mode' to force_reg.
6121         (aarch64_add_offset): Update calls to aarch64_force_temporary.
6122         (aarch64_expand_mov_immediate): Likewise.
6124 2013-06-27  Yufeng Zhang  <yufeng.zhang@arm.com>
6126         * config/aarch64/aarch64.c (aarch64_add_offset): Change to pass
6127         'mode' to aarch64_plus_immediate and gen_rtx_PLUS.
6129 2013-06-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
6131         * config/s390/s390.c: Rename UNSPEC_CCU_TO_INT to
6132         UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
6133         (struct machine_function): Add tbegin_p.
6134         (s390_canonicalize_comparison): Fold CC mode compares to
6135         conditional jump if possible.
6136         (s390_emit_jump): Return the emitted jump.
6137         (s390_branch_condition_mask, s390_branch_condition_mnemonic):
6138         Handle CCRAWmode compares.
6139         (s390_option_override): Default to -mhtm if available.
6140         (s390_reg_clobbered_rtx): Handle floating point regs as well.
6141         (s390_regs_ever_clobbered): Use s390_regs_ever_clobbered also for
6142         FPRs instead of df_regs_ever_live_p.
6143         (s390_optimize_nonescaping_tx): New function.
6144         (s390_init_frame_layout): Extend clobbered_regs array to cover
6145         FPRs as well.
6146         (s390_emit_prologue): Call s390_optimize_nonescaping_tx.
6147         (s390_expand_tbegin): New function.
6148         (enum s390_builtin): New enum definition.
6149         (code_for_builtin): New array definition.
6150         (s390_init_builtins): New function.
6151         (s390_expand_builtin): New function.
6152         (TARGET_INIT_BUILTINS): Define.
6153         (TARGET_EXPAND_BUILTIN): Define.
6154         * common/config/s390/s390-common.c (processor_flags_table): Add PF_TX.
6155         * config/s390/predicates.md (s390_comparison): Handle CCRAWmode.
6156         (s390_alc_comparison): Likewise.
6157         * config/s390/s390-modes.def: Add CCRAWmode.
6158         * config/s390/s390.h (processor_flags): Add PF_TX.
6159         (TARGET_CPU_HTM): Define macro.
6160         (TARGET_HTM): Define macro.
6161         (TARGET_CPU_CPP_BUILTINS): Define __HTM__ for htm.
6162         * config/s390/s390.md: Rename UNSPEC_CCU_TO_INT to
6163         UNSPEC_STRCMPCC_TO_INT and UNSPEC_CCZ_TO_INT to UNSPEC_CC_TO_INT.
6164         (UNSPECV_TBEGIN, UNSPECV_TBEGINC, UNSPECV_TEND, UNSPECV_TABORT)
6165         (UNSPECV_ETND, UNSPECV_NTSTG, UNSPECV_PPA): New unspecv enum values.
6166         (TBEGIN_MASK, TBEGINC_MASK): New constants.
6167         ("*cc_to_int"): Move up.
6168         ("*mov<mode>cc", "*cjump_64", "*cjump_31"): Accept integer
6169         constants other than 0.
6170         ("*ccraw_to_int"): New insn and splitter definition.
6171         ("tbegin", "tbegin_nofloat", "tbegin_retry")
6172         ("tbegin_retry_nofloat", "tbeginc", "tend", "tabort")
6173         ("tx_assist"): New expander.
6174         ("tbegin_1", "tbegin_nofloat_1", "*tbeginc_1", "*tend_1")
6175         ("*tabort_1", "etnd", "ntstg", "*ppa"): New insn definition.
6176         * config/s390/s390.opt: Add -mhtm option.
6177         * config/s390/s390-protos.h (s390_emit_jump): Add return type.
6178         * config/s390/htmxlintrin.h: New file.
6179         * config/s390/htmintrin.h: New file.
6180         * config/s390/s390intrin.h: New file.
6181         * doc/extend.texi: Document htm builtins.
6182         * config.gcc: Add the new header files to extra_headers.
6184 2013-06-26  Thomas Schwinge  <thomas@codesourcery.com>
6186         * config/i386/gnu.h [TARGET_LIBC_PROVIDES_SSP]
6187         (TARGET_CAN_SPLIT_STACK, TARGET_THREAD_SPLIT_STACK_OFFSET): Undefine.
6189 2013-06-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
6190             Pat Haugen  <pthaugen@us.ibm.com>
6191             Peter Bergner  <bergner@vnet.ibm.com>
6193         * config/rs6000/power8.md: New.
6194         * config/rs6000/rs6000-cpus.def (RS6000_CPU table): Adjust processor
6195         setting for power8 entry.
6196         * config/rs6000/t-rs6000 (MD_INCLUDES): Add power8.md.
6197         * config/rs6000/rs6000.c (is_microcoded_insn, is_cracked_insn): Adjust
6198         test for Power4/Power5 only.
6199         (insn_must_be_first_in_group, insn_must_be_last_in_group): Add Power8
6200         support.
6201         (force_new_group): Adjust comment.
6202         * config/rs6000/rs6000.md: Include power8.md.
6204 2013-06-26  Greta Yorsh  <Greta.Yorsh@arm.com>
6206         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Define macro.
6207         * config/arm/arm-protos.h (arm_max_conditional_execute): New
6208         declaration.
6209         (tune_params): Update comment.
6210         * config/arm/arm.c (arm_cortex_a15_tune): Set max_cond_insns to 2.
6211         (arm_max_conditional_execute): New function.
6212         (thumb2_final_prescan_insn): Use max_insn_skipped and
6213         MAX_INSN_PER_IT_BLOCK to compute maximum instructions in a block.
6215 2013-06-25  Jakub Jelinek  <jakub@redhat.com>
6217         PR tree-optimization/57705
6218         * tree-vect-loop.c (vect_is_simple_iv_evolution): Allow
6219         SSA_NAME step, provided that it is not defined inside the loop.
6220         (vect_analyze_scalar_cycles_1): Disallow SSA_NAME step in nested loop.
6221         (get_initial_def_for_induction): Handle SSA_NAME IV step.
6223 2013-06-25  Martin Jambor  <mjambor@suse.cz>
6225         PR middle-end/57670
6226         * cgraph.h (cgraph_indirect_call_info): New flag member_ptr.
6227         * ipa-prop.c (ipa_print_node_jump_functions): Mark member pointer
6228         calls in the dump.
6229         (ipa_note_param_call): Initialize member_ptr flag.
6230         (ipa_analyze_indirect_call_uses): Set member_ptr flag.
6231         (ipa_make_edge_direct_to_target): Bail out if member_ptr is set.
6232         (ipa_write_indirect_edge_info): Stream member_ptr flag.
6233         (ipa_read_indirect_edge_info): Likewise.
6235 2013-06-25  Richard Biener  <rguenther@suse.de>
6237         PR middle-end/56977
6238         * passes.c (init_optimization_passes): Move pass_fold_builtins
6239         and pass_dce earlier with -Og.
6241 2013-06-25  Eric Botcazou  <ebotcazou@adacore.com>
6243         * expr.c (expand_expr_real_1) <ARRAY_REF>: Fix formatting glitches.
6244         <BIT_FIELD_REF>: Remove trailing TAB.
6245         * varasm.c (output_constructor_bitfield): Fix formatting glitch and
6246         remove blank line.
6248 2013-06-24  Martin Jambor  <mjambor@suse.cz>
6250         PR tree-optimization/57358
6251         * ipa-prop.c (ipa_func_spec_opts_forbid_analysis_p): New function.
6252         (ipa_compute_jump_functions_for_edge): Bail out if it returns true.
6253         (ipa_analyze_params_uses): Generate pessimistic info when true.
6255 2013-06-24  Martin Jambor  <mjambor@suse.cz>
6257         PR tree-optimization/57539
6258         * cgraphclones.c (cgraph_clone_node): Add parameter new_inlined_to, set
6259         global.inlined_to of the new node to it.  All callers changed.
6260         * ipa-inline-transform.c (clone_inlined_nodes): New variable
6261         inlining_into, pass it to cgraph_clone_node.
6262         * ipa-prop.c (ipa_propagate_indirect_call_infos): Do not call
6263         ipa_free_edge_args_substructures.
6264         (ipa_edge_duplication_hook): Only add edges from inlined nodes to
6265         rdesc linked list.  Do not assert rdesc edges have inlined caller.
6266         Assert we have found an rdesc in the rdesc list.
6268 2013-06-24  Richard Biener  <rguenther@suse.de>
6270         * pointer-set.h (struct pointer_set_t): Move here from pointer-set.c.
6271         (pointer_set_lookup): Declare.
6272         (class pointer_map): New template class implementing a
6273         generic pointer to T map.
6274         (pointer_map<T>::pointer_map, pointer_map<T>::~pointer_map,
6275         pointer_map<T>::contains, pointer_map<T>::insert,
6276         pointer_map<T>::traverse): New functions.
6277         * pointer-set.c (struct pointer_set_t): Moved to pointer-set.h.
6278         (pointer_set_lookup): New function.
6279         (pointer_set_contains): Use pointer_set_lookup.
6280         (pointer_set_insert): Likewise.
6281         (insert_aux): Remove.
6282         (struct pointer_map_t): Embed a pointer_set_t.
6283         (pointer_map_create): Adjust.
6284         (pointer_map_destroy): Likewise.
6285         (pointer_map_contains): Likewise.
6286         (pointer_map_insert): Likewise.
6287         (pointer_map_traverse): Likewise.
6288         * tree-streamer.h (struct streamer_tree_cache_d): Use a
6289         pointer_map<unsigned> instead of a pointer_map_t.
6290         * tree-streamer.c (streamer_tree_cache_insert_1): Adjust.
6291         (streamer_tree_cache_lookup): Likewise.
6292         (streamer_tree_cache_create): Likewise.
6293         (streamer_tree_cache_delete): Likewise.
6294         * lto-streamer.h (struct lto_tree_ref_encoder): Use a
6295         pointer_map<unsigned> instead of a pointer_map_t.
6296         (lto_init_tree_ref_encoder): Adjust.
6297         (lto_destroy_tree_ref_encoder): Likewise.
6298         * lto-section-out.c (lto_output_decl_index): Likewise.
6299         (lto_record_function_out_decl_state): Likewise.
6300         * dominance.c (iterate_fix_dominators): Use pointer_map<int>.
6302 2013-06-24  Richard Biener  <rguenther@suse.de>
6304         PR tree-optimization/57488
6305         * tree-ssa-pre.c (insert): Clear NEW sets before each iteration.
6307 2013-06-24  Alan Modra  <amodra@gmail.com>
6309         * config/rs6000/rs6000.c (vspltis_constant): Correct for little-endian.
6310         (gen_easy_altivec_constant): Likewise.
6311         * config/rs6000/predicates.md (easy_vector_constant_add_self,
6312         easy_vector_constant_msb): Likewise.
6314 2013-06-23  Jakub Jelinek  <jakub@redhat.com>
6316         PR target/57688
6317         * common/config/i386/i386-common.c (ix86_handle_option): For OPT_mlzcnt
6318         add missing return true.
6320 2013-06-23  Oleg Endo  <olegendo@gcc.gnu.org>
6322         PR target/52483
6323         * config/sh/predicates.md (general_extend_operand): Invoke
6324         general_movsrc_operand for memory operands.
6325         (general_movsrc_operand): Allow reg+reg addressing, do not use
6326         general_operand for memory operands.
6328 2013-06-23  Sriraman Tallam  <tmsriram@google.com>
6330         * config/i386/i386.c (ix86_pragma_target_parse):  Restore target
6331         when current target options does not apply.
6332         * config/i386/i386-protos.h (ix86_reset_previous_fndecl): New function.
6333         * config/i386/i386.c (ix86_reset_previous_fndecl): Ditto.
6334         * config/i386/bmiintrin.h: Pass appropriate target
6335         attributes to header.
6336         * config/i386/mmintrin.h: Ditto.
6337         * config/i386/nmmintrin.h: Ditto.
6338         * config/i386/avx2intrin.h: Ditto.
6339         * config/i386/fxsrintrin.h: Ditto.
6340         * config/i386/tbmintrin.h: Ditto.
6341         * config/i386/xsaveintrin.h: Ditto.
6342         * config/i386/f16cintrin.h: Ditto.
6343         * config/i386/xtestintrin.h: Ditto.
6344         * config/i386/xsaveoptintrin.h: Ditto.
6345         * config/i386/bmi2intrin.h: Ditto.
6346         * config/i386/lzcntintrin.h: Ditto.
6347         * config/i386/smmintrin.h: Ditto.
6348         * config/i386/wmmintrin.h: Ditto.
6349         * config/i386/x86intrin.h: Remove all header include guards.
6350         * config/i386/prfchwintrin.h: Ditto.
6351         * config/i386/pmmintrin.h: Ditto.
6352         * config/i386/tmmintrin.h: Ditto.
6353         * config/i386/xmmintrin.h: Ditto.
6354         * config/i386/popcntintrin.h: Ditto.
6355         * config/i386/rdseedintrin.h: Ditto.
6356         * config/i386/ammintrin.h: Ditto.
6357         * config/i386/emmintrin.h: Ditto.
6358         * config/i386/immintrin.h: Remove all header include guards.
6359         * config/i386/fma4intrin.h: Ditto.
6360         * config/i386/lwpintrin.h: Ditto.
6361         * config/i386/xopintrin.h: Ditto.
6362         * config/i386/ia32intrin.h: Ditto.
6363         * config/i386/avxintrin.h: Ditto.
6364         * config/i386/rtmintrin.h: Ditto.
6365         * config/i386/fmaintrin.h: Ditto.
6366         * config/i386/mm3dnow.h: Ditto.
6368 2013-06-22  Sriraman Tallam  <tmsriram@google.com>
6370         * common/config/i386/i386-common.c: Handle LZCNT.
6372 2013-06-22  Andi Kleen  <ak@linux.intel.com>
6374         * doc/extend.texi: Use __atomic_store_n instead of
6375         __atomic_store in HLE example.
6377 2013-06-22  Oleg Endo  <olegendo@gcc.gnu.org>
6379         * config/sh/sh.c: Remove <cstdlib> workaround.
6381 2013-06-21  Andi Kleen  <ak@linux.intel.com>
6383         * doc/extend.texi: Dont use __atomic_clear in HLE example.  Fix typo.
6385 2013-06-21  Andi Kleen  <ak@linux.intel.com>
6387         * doc/extend.texi: Document that __atomic_clear and
6388           __atomic_test_and_set should only be used with bool.
6390 2013-06-20  Jan Hubicka  <jh@suse.cz>
6392         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Use
6393         types_same_for_odr.
6394         * tree.c (decls_same_for_odr): New function.
6395         (same_for_edr): New function.
6396         (types_same_for_odr): New function.
6397         (get_binfo_at_offset): Use it.
6398         * tree.h (types_same_for_odr): Declare.
6400 2013-06-20  Oleg Endo  <olegendo@gcc.gnu.org>
6401             Jason Merrill  <jason@redhat.com>
6403         * system.h: Include <cstdlib> as well as <stdlib.h>.
6405 2013-06-20  Uros Bizjak  <ubizjak@gmail.com>
6407         PR target/57655
6408         * config/i386/i386.c (construct_container): Report error if
6409         long double is used with disabled x87 float returns.
6411 2013-06-20  Jan Hubicka  <jh@suse.cz>
6413         * lto-cgraph.c (input_symtab): Do not set cgraph state.
6415 2013-06-20  Joern Rennecke  <joern.rennecke@embecosm.com>
6417         PR rtl-optimization/57425
6418         PR rtl-optimization/57569
6419         * alias.c (write_dependence_p): Remove parameters mem_mode and
6420         canon_mem_addr.  Add parameters x_mode, x_addr and x_canonicalized.
6421         Changed all callers.
6422         (canon_anti_dependence): Get comments and semantics in sync.
6423         Add parameter mem_canonicalized.  Changed all callers.
6424         * rtl.h (canon_anti_dependence): Update prototype.
6426 2013-06-20  Richard Biener  <rguenther@suse.de>
6428         * data-streamer-in.c (streamer_read_uhwi): Optimize single
6429         byte case, inline streamer_read_uchar and defer section
6430         overrun check.
6432 2013-06-20  Richard Biener  <rguenther@suse.de>
6434         PR tree-optimization/57584
6435         * tree-ssa-loop-niter.c (expand_simple_operations): Avoid including
6436         SSA names into the expanded expression that take part in
6437         abnormal coalescing.
6439 2013-06-19  Sharad Singhai  <singhai@google.com>
6441         * gcov.c (print_usage): Handle new option.
6442         (process_args): Ditto.
6443         (get_gcov_intermediate_filename): New function.
6444         (output_intermediate_file): New function.
6445         (output_gcov_file): New function
6446         (generate_results): Handle new option.
6447         (release_function): Relase demangled name.
6448         (read_graph_file): Handle demangled name.
6449         (output_lines): Ditto.
6450         * doc/gcov.texi: Document gcov intermediate format.
6452 2013-06-19  Vladimir Makarov  <vmakarov@redhat.com>
6454         PR bootstrap/57604
6455         * lra.c (emit_add3_insn, emit_add2_insn): New functions.
6456         (lra_emit_add): Use the functions.  Add comment about Y as an
6457         address segment.
6459 2013-06-19  David Edelsohn  <dje.gcc@gmail.com>
6461         PR driver/57652
6462         * collect2.c (collect_atexit): New.
6463         (collect_exit): Delete.
6464         (main): Register collect_atexit with atexit.
6465         (collect_wait): Change collect_exit to exit.
6466         (do_wait): Same.
6467         * collect2.h (collect_exit): Delete.
6468         * tlink.c (do_tlink): Rename exit to ret. Change collect_exit to exit.
6470 2013-06-19  Wei Mi  <wmi@google.com>
6472         PR rtl-optimization/57518
6473         * ira.c (set_paradoxical_subreg): Set pdx_subregs[regno]
6474         if regno is used in paradoxical subreg.
6475         (update_equiv_regs): Check pdx_subregs[regno] before
6476         set a reg to be equivalent with a mem.
6478 2013-06-19  Matthias Klose  <doko@ubuntu.com>
6480         PR driver/57651
6481         * file-find.h (find_a_file): Add a mode parameter.
6482         * file-find.c (find_a_file): Likewise.
6483         * gcc-ar.c (main): Call find_a_file with R_OK for the plugin,
6484         with X_OK for the executables.
6485         * collect2.c (main): Call find_a_file with X_OK.
6487 2013-06-19  Steve Ellcey  <sellcey@mips.com>
6489         PR target/56942
6490         * config/mips/mips.md (casesi_internal_mips16_<mode>):
6491         Use NEXT_INSN instead of next_real_insn.
6493 2013-06-19  Jan Hubicka  <jh@suse.cz>
6495         * cgraph.h (const_value_known_p): Replace by ...
6496         (ctor_for_folding): .. this one.
6497         * cgraphunit.c (process_function_and_variable_attributes): Use it.
6498         * lto-cgraph.c (compute_ltrans_boundary): Use ctor_for_folding.
6499         * expr.c (expand_expr_real_1): Likewise.
6500         (string_constant): Likewise.
6501         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Likewise.
6502         * ipa.c (process_references): Likewise.
6503         (symtab_remove_unreachable_nodes): Likewise.
6504         * ipa-inline-analysis.c (param_change_prob): Likewise.
6505         * gimple-fold.c (canonicalize_constructor_val): Likewise.
6506         (get_base_constructor): Likwise.
6507         * varpool.c (varpool_remove_node): Likewise.
6508         (varpool_remove_initializer): LIkewise.
6509         (dump_varpool_node): LIkwise.
6510         (const_value_known_p): Rewrite to ...
6511         (ctor_for_folding): ... this one.
6513 2013-06-19  Jakub Jelinek  <jakub@redhat.com>
6515         PR driver/57651
6516         * gcc-ar.c (main): If not CROSS_DIRECTORY_STRUCTURE, look for
6517         PERSONALITY in $PATH derived prefixes.
6519 2013-06-19  Jeff Law  <law@redhat.com>
6521         * tree-ssa-forwprop.c (simplify_bitwise_binary_boolean): Fix typo
6522         in comment.
6524         * tree-ssa-forwprop.c (simplify_bitwise_binary_boolean): New function.
6525         (simplify_bitwise_binary): Use it to simpify certain binary ops on
6526         booleans.
6528 2013-06-19  Sofiane Naci  <sofiane.naci@arm.com>
6530         * config/arm/vfp.md: Move VFP instruction classification documentation
6531         to ...
6532         * config/arm/arm.md: ... here.  Update instruction classification
6533         documentation.
6535 2013-06-19  Richard Earnshaw  <rearnsha@arm.com>
6537         arm.md (split for eq(reg, 0)): Add variants for ARMv5 and Thumb2.
6538         (peepholes for eq(reg, not-0)): Ensure condition register is dead after
6539         pattern.  Use more efficient sequences on ARMv5 and Thumb2.
6541 2013-06-19  Steven Bosscher  <steven@gcc.gnu.org>
6543         PR target/57609
6544         * config/s390/s390.c (s390_chunkify_start): Replace next_real_insn
6545         with NEXT_INSN.  Use tablejump_p to check for jump table data
6546         insns.
6548 2013-06-19  Paolo Carlini  <paolo.carlini@oracle.com>
6550         PR c++/56544
6551         * doc/cpp.texi [Standard Predefined Macros, __cplusplus]: Document
6552         that now in C++ the value is correct per the C++ standards.
6554 2013-06-19  Richard Biener  <rguenther@suse.de>
6556         * expr.c (expand_expr_real_1): Use SCOPE_FILE_SCOPE_P to check
6557         for global context.
6559 2013-06-19  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
6561         Revert:
6562         2013-06-18  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
6564         PR target/57609
6565         * config/s390/s390.c (s390_chunkify_start): Replace next_real_insn
6566         with next_active_insn.
6568 2013-06-18  Sriraman Tallam  <tmsriram@google.com>
6570         * ipa-inline.c (inline_always_inline_functions): Pretend always_inline
6571         functions are inlined during failures to flag an error.
6572         * tree-inline.c (expand_call_inline): Allow the error to be flagged
6573         in early inline pass.
6575 2013-06-18  H.J. Lu  <hongjiu.lu@intel.com>
6577         * config/i386/i386.c (initial_ix86_tune_features): Fix a typo
6578         in comments.
6580 2013-06-18  Julian Brown  <julian@codesourcery.com>
6582         * config/arm/arm.c (neon_vector_mem_operand): Add strict argument.
6583         Permit virtual register pre-reload if !strict.
6584         (coproc_secondary_reload_class): Adjust for neon_vector_mem_operand
6585         change.
6586         * config/arm/arm-protos.h (neon_vector_mem_operand): Adjust
6587         prototype.
6588         * config/arm/neon.md (movmisalign<mode>): Use
6589         neon_perm_struct_or_reg_operand instead of
6590         neon_struct_or_register_operand.
6591         (*movmisalign<mode>_neon_load, *movmisalign<mode>_neon_store): Use
6592         neon_permissive_struct_operand instead of neon_struct_operand.
6593         * config/arm/constraints.md (Un, Um, Us): Adjust calls to
6594         neon_vector_mem_operand.
6595         * config/arm/predicates.md (neon_struct_operand): Adjust call to
6596         neon_vector_mem_operand.
6597         (neon_permissive_struct_operand): New.
6598         (neon_struct_or_register_operand): Rename to...
6599         (neon_perm_struct_or_reg_operand): This. Adjust call to
6600         neon_vector_mem_operand.
6602 2013-06-18  Richard Biener  <rguenther@suse.de>
6604         * Makefile.in (LTO_STREAMER_H): Add pointer-set.h dependency.
6605         * lto-streamer.h: Include pointer-set.h.
6606         (struct lto_decl_slot): Remove.
6607         (struct lto_tree_ref_encoder): Make tree_hash_table a pointer-map.
6608         Remove next_index entry.
6609         (lto_hash_decl_slot_node, lto_eq_decl_slot_node,
6610         lto_hash_type_slot_node, lto_eq_type_slot_node): Remove.
6611         (lto_init_tree_ref_encoder): Adjust.
6612         (lto_destroy_tree_ref_encoder): Likewise.
6613         * lto-section-out.c (lto_hash_decl_slot_node, lto_eq_decl_slot_node,
6614         lto_hash_type_slot_node, lto_eq_type_slot_node): Remove.
6615         (lto_output_decl_index): Adjust.
6616         (lto_new_out_decl_state): Likewise.
6617         (lto_record_function_out_decl_state): Likewise.
6618         * lto-streamer-out.c (copy_function): Likewise.
6620 2013-06-18  Richard Biener  <rguenther@suse.de>
6622         * Makefile.in (cgraphunit.o): Add $(CFGLOOP_H) dependency.
6623         * cgraphunit.c: Include cfgloop.h.
6624         (init_lowered_empty_function): Initialize the loop tree.
6625         (assemble_thunk): Insert new BBs into loops.
6627 2013-06-18  Richard Biener  <rguenther@suse.de>
6629         * tree-streamer.h (streamer_tree_cache_create): Adjust prototype.
6630         * tree-streamer.c (streamer_tree_cache_create): Make maintaining
6631         the map from cache entry to cache index optional.
6632         (streamer_tree_cache_replace_tree): Adjust accordingly.
6633         (streamer_tree_cache_append): Likewise.
6634         (streamer_tree_cache_delete): Likewise.
6635         * lto-streamer-in.c (lto_data_in_create): Do not maintain the
6636         streamer cache map from cache entry to cache index.
6637         * lto-streamer-out.c (create_output_block): Adjust.
6639 2013-06-18  Sofiane Naci  <sofiane.naci@arm.com>
6641         * config/arm/arm.md (attribute "insn"): Move multiplication and
6642         division attributes to...
6643         (attribute "type"): ... here.  Remove mult.
6644         (attribute "mul32"): New attribute.
6645         (attribute "mul64"): Add umaal.
6646         (*arm_mulsi3): Update attributes.
6647         (*arm_mulsi3_v6): Likewise.
6648         (*thumb_mulsi3): Likewise.
6649         (*thumb_mulsi3_v6): Likewise.
6650         (*mulsi3_compare0): Likewise.
6651         (*mulsi3_compare0_v6): Likewise.
6652         (*mulsi_compare0_scratch): Likewise.
6653         (*mulsi_compare0_scratch_v6): Likewise.
6654         (*mulsi3addsi): Likewise.
6655         (*mulsi3addsi_v6): Likewise.
6656         (*mulsi3addsi_compare0): Likewise.
6657         (*mulsi3addsi_compare0_v6): Likewise.
6658         (*mulsi3addsi_compare0_scratch): Likewise.
6659         (*mulsi3addsi_compare0_scratch_v6): Likewise.
6660         (*mulsi3subsi): Likewise.
6661         (*mulsidi3adddi): Likewise.
6662         (*mulsi3addsi_v6): Likewise.
6663         (*mulsidi3adddi_v6): Likewise.
6664         (*mulsidi3_nov6): Likewise.
6665         (*mulsidi3_v6): Likewise.
6666         (*umulsidi3_nov6): Likewise.
6667         (*umulsidi3_v6): Likewise.
6668         (*umulsidi3adddi): Likewise.
6669         (*umulsidi3adddi_v6): Likewise.
6670         (*smulsi3_highpart_nov6): Likewise.
6671         (*smulsi3_highpart_v6): Likewise.
6672         (*umulsi3_highpart_nov6): Likewise.
6673         (*umulsi3_highpart_v6): Likewise.
6674         (mulhisi3): Likewise.
6675         (*mulhisi3tb): Likewise.
6676         (*mulhisi3bt): Likewise.
6677         (*mulhisi3tt): Likewise.
6678         (maddhisi4): Likewise.
6679         (*maddhisi4tb): Likewise.
6680         (*maddhisi4tt): Likewise.
6681         (maddhidi4): Likewise.
6682         (*maddhidi4tb): Likewise.
6683         (*maddhidi4tt): Likewise.
6684         (divsi3): Likewise.
6685         (udivsi3): Likewise.
6686         * config/arm/thumb2.md (thumb2_mulsi_short): Update attributes.
6687         (thumb2_mulsi_short_compare0): Likewise.
6688         (thumb2_mulsi_short_compare0_scratch): Likewise.
6689         * config/arm/arm1020e.md (1020mult1): Update attribute change.
6690         (1020mult2): Likewise.
6691         (1020mult3): Likewise.
6692         (1020mult4): Likewise.
6693         (1020mult5): Likewise.
6694         (1020mult6): Likewise.
6695         * config/arm/cortex-a15.md (cortex_a15_mult32): Update attribute
6696         change.
6697         (cortex_a15_mult64): Likewise.
6698         (cortex_a15_sdiv): Likewise.
6699         (cortex_a15_udiv): Likewise.
6700         * config/arm/arm1026ejs.md (mult1): Update attribute change.
6701         (mult2): Likewise.
6702         (mult3): Likewise.
6703         (mult4): Likewise.
6704         (mult5): Likewise.
6705         (mult6): Likewise.
6706         * config/arm/marvell-pj4.md (pj4_ir_mul): Update attribute change.
6707         (pj4_ir_div): Likewise.
6708         * config/arm/arm1136jfs.md (11_mult1): Update attribute change.
6709         (11_mult2): Likewise.
6710         (11_mult3): Likewise.
6711         (11_mult4): Likewise.
6712         (11_mult5): Likewise.
6713         (11_mult6): Likewise.
6714         (11_mult7): Likewise.
6715         * config/arm/cortex-a8.md (cortex_a8_mul): Update attribute change.
6716         (cortex_a8_mla): Likewise.
6717         (cortex_a8_mull): Likewise.
6718         (cortex_a8_smulwy): Likewise.
6719         (cortex_a8_smlald): Likewise.
6720         * config/arm/cortex-m4.md (cortex_m4_alu): Update attribute change.
6721         * config/arm/cortex-r4.md (cortex_r4_mul_4): Update attribute change.
6722         (cortex_r4_mul_3): Likewise.
6723         (cortex_r4_mla_4): Likewise.
6724         (cortex_r4_mla_3): Likewise.
6725         (cortex_r4_smlald): Likewise.
6726         (cortex_r4_mull): Likewise.
6727         (cortex_r4_sdiv): Likewise.
6728         (cortex_r4_udiv): Likewise.
6729         * config/arm/cortex-a7.md (cortex_a7_mul): Update attribute change.
6730         (cortex_a7_idiv): Likewise.
6731         * config/arm/arm926ejs.md (9_mult1): Update attribute change.
6732         (9_mult2): Likewise.
6733         (9_mult3): Likewise.
6734         (9_mult4): Likewise.
6735         (9_mult5): Likewise.
6736         (9_mult6): Likewise.
6737         * config/arm/cortex-a53.md (cortex_a53_mul): Update attribute change.
6738         (cortex_a53_sdiv): Likewise.
6739         (cortex_a53_udiv): Likewise.
6740         * config/arm/fa726te.md (726te_mult_op): Update attribute change.
6741         * config/arm/fmp626.md (mp626_mult1): Update attribute change.
6742         (mp626_mult2): Likewise.
6743         (mp626_mult3): Likewise.
6744         (mp626_mult4): Likewise.
6745         * config/arm/fa526.md (526_mult1): Update attribute change.
6746         (526_mult2): Likewise.
6747         * config/arm/arm-generic.md (mult): Update attribute change.
6748         (mult_ldsched_strongarm): Likewise.
6749         (mult_ldsched): Likewise.
6750         (multi_cycle): Likewise.
6751         * config/arm/cortex-a5.md (cortex_a5_mul): Update attribute change.
6752         * config/arm/fa606te.md (606te_mult1): Update attribute change.
6753         (606te_mult2): Likewise.
6754         (606te_mult3): Likewise.
6755         (606te_mult4): Likewise.
6756         * config/arm/cortex-a9.md (cortex_a9_mult16): Update attribute change.
6757         (cortex_a9_mac16): Likewise.
6758         (cortex_a9_multiply): Likewise.
6759         (cortex_a9_mac): Likewise.
6760         (cortex_a9_multiply_long): Likewise.
6761         * config/arm/fa626te.md (626te_mult1): Update attribute change.
6762         (626te_mult2): Likewise.
6763         (626te_mult3): Likewise.
6764         (626te_mult4): Likewise.
6766 2013-06-18  Richard Biener  <rguenther@suse.de>
6768         PR lto/57334
6769         * lto-symtab.c (lto_symtab_merge_decls): Process nodes properly.
6771 2013-06-18  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
6773         PR target/57609
6774         * config/s390/s390.c (s390_chunkify_start): Replace next_real_insn
6775         with next_active_insn.
6777 2013-06-18  Alan Modra  <amodra@gmail.com>
6779         * config/rs6000/rs6000.h (enum data_align): New.
6780         (LOCAL_ALIGNMENT, DATA_ALIGNMENT): Use rs6000_data_alignment.
6781         (DATA_ABI_ALIGNMENT): Define.
6782         (CONSTANT_ALIGNMENT): Correct comment.
6783         * config/rs6000/rs6000-protos.h (rs6000_data_alignment): Declare.
6784         * config/rs6000/rs6000.c (rs6000_data_alignment): New function.
6786 2013-06-17  David Malcolm  <dmalcolm@redhat.com>
6788         * ggc-page.c (ggc_pch_write_object) <d>: Remove erroneous
6789         ATTRIBUTE_UNUSED marking.
6791 2013-06-17  Sofiane Naci  <sofiane.naci@arm.com>
6793         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>): Add r<-w
6794         alternative and update.
6795         (aarch64_dup_lanedi): Delete.
6796         * config/aarch64/arm_neon.h (vdup<bhsd>_lane_*): Update.
6797         * config/aarch64/aarch64-simd-builtins.def: Update.
6799 2013-06-17  Richard Biener  <rguenther@suse.de>
6801         * lto-streamer.h (enum LTO_tags): Add LTO_tree_scc.
6802         (lto_input_scc): Declare.
6803         (lto_input_tree_1): Likewise.
6804         (struct lto_stats_d): Add num_tree_bodies_output and
6805         num_pickle_refs_output.
6806         * lto-streamer-in.c (lto_read_body): Use streamer_tree_cache_get_tree.
6807         (lto_read_tree_1): Split out from ...
6808         (lto_read_tree): ... this.
6809         (lto_input_scc): New function.
6810         (lto_input_tree_1): Split out from ...
6811         (lto_input_tree): ... this.  Handle LTO_tree_scc.
6812         (lto_data_in_create): Create the streamer cache without hashes.
6813         * lto-streamer-out.c (create_output_block): Create the streamer
6814         cache with hashes when not doing WPA.
6815         (lto_write_tree_1): Split out from ...
6816         (lto_write_tree): ... this.
6817         (get_symbol_initial_value): New function.
6818         (lto_output_tree_1): Split out from ...
6819         (lto_output_tree): ... this.  Write trees as series of SCCs
6820         using a DFS walk via DFS_write_tree.
6821         (struct sccs, struct scc_entry): New types.
6822         (next_dfs_num, sccstack, sccstate, sccstate_obstack): New globals.
6823         (DFS_write_tree_body): New function.
6824         (DFS_write_tree): Likewise.
6825         (hash_tree): Likewise.
6826         (scc_entry_compare): Likewise.
6827         (hash_scc): Likewise.
6828         (tree_is_indexable): DEBUG_EXPR_DECLs are local entities.
6829         * tree-streamer-in.c (lto_input_ts_list_tree_pointers): Stream
6830         TREE_CHAIN as regular reference.
6831         (streamer_read_integer_cst): Remove.
6832         (streamer_get_pickled_tree): Adjust.
6833         * tree-streamer-out.c (streamer_write_chain): Disable streaming
6834         of DECL_EXTERNALs in BLOCK_VARS for now.
6835         (write_ts_list_tree_pointers): Stream TREE_CHAIN as regular
6836         reference.
6837         * tree-streamer.c (streamer_tree_cache_add_to_node_array):
6838         Add hash value argument and record that if hashes are recorded
6839         in the cache.
6840         (streamer_tree_cache_insert_1): Adjust.
6841         (streamer_tree_cache_insert): Likewise.
6842         (streamer_tree_cache_insert_at): Rename to ...
6843         (streamer_tree_cache_replace_tree): ... this and adjust.
6844         (streamer_tree_cache_append): Adjust.
6845         (record_common_node): Likewise.
6846         (streamer_tree_cache_create): Add argument whether to
6847         record hash values together with trees.
6848         (streamer_tree_cache_delete): Adjust.
6849         * tree-streamer.h (struct streamer_tree_cache_d): Add
6850         vector of hashes.
6851         (streamer_read_integer_cst): Remove.
6852         (streamer_tree_cache_insert): Adjust.
6853         (streamer_tree_cache_append): Likewise.
6854         (streamer_tree_cache_insert_at): Rename to ...
6855         (streamer_tree_cache_replace_tree): ... this and adjust.
6856         (streamer_tree_cache_create): Add argument whether to record hashes.
6857         (streamer_tree_cache_get): Rename to ...
6858         (streamer_tree_cache_get_tree): ... this.
6859         (streamer_tree_cache_get_hash): New function.
6860         * tree.c (cache_integer_cst): New function.
6861         * tree.h (cache_integer_cst): Declare.
6862         (ANON_AGGRNAME_FORMAT, ANON_AGGRNAME_P): Move here from cp/cp-tree.h.
6863         * lto-symtab.c (lto_varpool_replace_node): Only release
6864         DECL_INITIAL of non-prevailing decls.
6865         * varpool.c (varpool_remove_initializer): Do not release
6866         DECL_INITIAL when we are still in CGRAPH_LTO_STREAMING.
6868 2013-06-16  Jürgen Urban  <JuergenUrban@gmx.de>
6870         * config/mips/mips.h (ISA_HAS_MUL3): Include TARGET_MIPS5900.
6871         (ISA_HAS_MULT, ISA_HAS_DMULT, ISA_HAS_DIV, ISA_HAS_DDIV): New macros.
6872         * config/mips/mips.md (mul<mode>3, mul<mode>3_internal)
6873         (mul<mode>3_r4000): Require ISA_HAS_<D>MULT.
6874         (mul<mode>3_mul3): Handle TARGET_MIPS5900.
6875         (mulsidi3_64bit_dmul): Remove redundant TARGET_64BIT test.
6876         (<su>muldi3_highpart, <su>muldi3_highpart_internal, <u>mulditi3)
6877         (<u>mulditi3_internal, <u>mulditi3_r4000): Require ISA_HAS_DMULT
6878         instead of TARGET_64BIT.
6879         (divmod<mode>4, udivmod<mode>4, <u>divmod<GPR:mode>4_hilo_<HILO:mode>):
6880         Require ISA_HAS_<D>DIV.
6882 2013-06-16  Richard Sandiford  <rdsandiford@googlemail.com>
6884         * config.gcc (mips*-mti-linux*, mips64*-*-linux*, mipsisa64*-*-linux*)
6885         (mips*-*-linux*): Move default with_llsc setting to where other
6886         defaults are set.
6887         (mips*-*-vxworks*): Move with_arch default from with_cpu block to
6888         with_arch block.
6889         (mips64r5900-*-*, mips64r5900el-*-*, mipsr5900-*-*, mipsr5900el-*-*):
6890         Likewise.  Remove default with_tune setting.  Move default float
6891         setting to its own block.  Handle with_llsc in the same block as above.
6893 2013-06-16  Joern Rennecke  <joern.rennecke@embecosm.com>
6895         PR rtl-optimization/57425
6896         PR rtl-optimization/57569
6897         * alias.c (write_dependence_p): Add new parameters mem_mode,
6898         canon_mem_addr and mem_canonicalized.  Change type of writep to bool.
6899         Changed all callers.
6900         (canon_anti_dependence): New function.
6901         * cse.c (check_dependence): Use canon_anti_dependence.
6902         * cselib.c (cselib_invalidate_mem): Likewise.
6903         * rtl.h (canon_anti_dependence): Declare.
6905 2013-06-16  Jürgen Urban  <JuergenUrban@gmx.de>
6907         * config/mips/mips.h (ISA_HAS_LL_SC): Exclude TARGET_MIPS5900.
6908         * config/mips/mips.c (mips_start_ll_sc_sync_block): Output
6909         ".set mips3" for 64-bit targets.
6911 2013-06-15  Dehao Chen  <dehao@google.com>
6913         * tree-flow.h (gimple_check_call_matching_types): Add new argument.
6914         * gimple-low.c (gimple_check_call_matching_types): Likewise.
6915         (gimple_check_call_args): Likewise.
6916         * value-prof.c (check_ic_target): Likewise.
6917         * ipa-inline.c (early_inliner): Likewise.
6918         * ipa-prop.c (update_indirect_edges_after_inlining): Likewise.
6919         * cgraph.c (cgraph_create_edge_1): Likewise.
6920         (cgraph_make_edge_direct): Likewise.
6922 2013-06-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
6924         PR target/57615
6925         * config/rs6000/rs6000.md (mov<mode>_ppc64): Call
6926         rs6000_output_move_128bit to handle emitting quad memory
6927         operations.  Set attribute length to 8 bytes.
6929 2013-06-14  Vidya Praveen  <vidyapraveen@arm.com>
6931         * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_lo<mode>):
6932         New pattern.
6933         (aarch64_<su>mlal_hi<mode>, aarch64_<su>mlsl_lo<mode>): Likewise.
6934         (aarch64_<su>mlsl_hi<mode>, aarch64_<su>mlal<mode>): Likewise.
6935         (aarch64_<su>mlsl<mode>): Likewise.
6937 2013-06-14  Mike Stump  <mikestump@comcast.net>
6939         * Makefile.in (TARGET_H): Add insn-codes.h.
6941 2013-06-14  Alan Modra  <amodra@gmail.com>
6943         PR middle-end/57134
6944         PR middle-end/57586
6945         * expr.c (expand_expr_real_1 <normal_inner_ref>): Pass
6946         EXPAND_MEMORY and EXPAND_WRITE to recursive call.  Don't use
6947         bitfield expansion when EXPAND_MEMORY.
6948         (expand_expr_real_1 <VIEW_CONVERT_EXPR>): Pass modifier likewise.
6950 2013-06-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
6952         * config/rs6000/rs6000.c (rs6000_option_override_internal): Move
6953         test for clearing quad memory on 32-bit later.
6955 2013-06-13  Marc Glisse  <marc.glisse@inria.fr>
6957         * fold-const.c (negate_expr_p): Handle VECTOR_CST.
6958         (fold_negate_expr): Likewise.
6959         (fold_real_zero_addition_p): Handle vectors.
6960         (fold_binary_loc) <PLUS_EXPR, MINUS_EXPR>: Likewise.
6962 2013-06-14  Alan Modra  <amodra@gmail.com>
6964         * varasm.c (force_const_mem): Revert 2013-06-07 change.
6966 2013-06-13  Jan Hubicka  <jh@suse.cz>
6968         * ipa.c (cgraph_externally_visible_p, varpool_externally_visible_p):
6969         Local comdats are not externally visible.
6970         * symtab.c (dump_symtab_base): Dump externally visible.
6971         (verify_symtab_base): Verify back links in the symtab hash.
6973 2013-06-13  Bin Cheng  <bin.cheng@arm.com>
6975         * fold-const.c (operand_equal_p): Consider NOP_EXPR and
6976         CONVERT_EXPR as equal nodes.
6978 2013-06-13  Bin Cheng  <bin.cheng@arm.com>
6980         * rtlanal.c (noop_move_p): Check the code to be executed for COND_EXEC.
6982 2013-06-13  Marc Glisse  <marc.glisse@inria.fr>
6984         * tree-ssa-forwprop.c (simplify_bitwise_binary, associate_plusminus):
6985         Generalize to complex and vector.
6986         * tree.c (build_all_ones_cst): New function.
6987         * tree.h (build_all_ones_cst): Declare it.
6989 2013-06-13  Alan Modra  <amodra@gmail.com>
6991         * config/rs6000/rs6000.h (LONG_DOUBLE_LARGE_FIRST): Define.
6992         * config/rs6000/rs6000.md (signbittf2): New insn.
6993         (extenddftf2_internal): Use LONG_DOUBLE_LARGE_FIRST.
6994         (abstf2_internal, cmptf_internal2): Likewise.
6995         * config/rs6000/spe.md (spe_abstf2_cmp, spe_abstf2_tst): Likewise.
6997 2013-06-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
6998             Pat Haugen  <pthaugen@us.ibm.com>
6999             Peter Bergner  <bergner@vnet.ibm.com>
7001         * config/rs6000/rs6000.c (emit_load_locked): Add support for
7002         power8 byte, half-word, and quad-word atomic instructions.
7003         (emit_store_conditional): Likewise.
7004         (rs6000_expand_atomic_compare_and_swap): Likewise.
7005         (rs6000_expand_atomic_op): Likewise.
7007         * config/rs6000/sync.md (larx): Add new modes for power8.
7008         (stcx): Likewise.
7009         (AINT): New mode iterator to include TImode as well as normal
7010         integer modes on power8.
7011         (fetchop_pred): Use int_reg_operand instead of gpc_reg_operand so
7012         that VSX registers are not considered.  Use AINT mode iterator
7013         instead of INT1 to allow inclusion of quad word atomic operations
7014         on power8.
7015         (load_locked<mode>): Likewise.
7016         (store_conditional<mode>): Likewise.
7017         (atomic_compare_and_swap<mode>): Likewise.
7018         (atomic_exchange<mode>): Likewise.
7019         (atomic_nand<mode>): Likewise.
7020         (atomic_fetch_<fetchop_name><mode>): Likewise.
7021         (atomic_nand_fetch<mode>): Likewise.
7022         (mem_thread_fence): Use gen_loadsync_<mode> instead of enumerating
7023         each type.
7024         (ATOMIC): On power8, add QImode, HImode modes.
7025         (load_locked<QHI:mode>_si): Varients of load_locked for QI/HI
7026         modes that promote to SImode.
7027         (load_lockedti): Convert TImode arguments to PTImode, so that we
7028         get a guaranteed even/odd register pair.
7029         (load_lockedpti): Likewise.
7030         (store_conditionalti): Likewise.
7031         (store_conditionalpti): Likewise.
7033         * config/rs6000/rs6000.md (QHI): New mode iterator for power8
7034         atomic load/store instructions.
7035         (HSI): Likewise.
7037 2013-06-12  Richard Sandiford  <rdsandiford@googlemail.com>
7039         * config/mips/mips.md (extended_mips16): Include GOT and constant-pool
7040         loads.
7041         (insn_count): New attribute, with most cases extracted from...
7042         (length): ...here.  Redefine most cases in terms of insn_count.
7043         (single_insn): Delete.
7044         (can_delay): Use insn_count to check for single instructions.
7045         (*mul<mode>3_r4300, mul<mode>3_r4000, *mul_acc_si, *mul_acc_si_r3900)
7046         (*msac_using_macc, *mul_sub_si, <u>mulsidi3_32bit_r4000)
7047         (<u>mulsidi3_64bit_r4000, <su>muldi3_highpart_internal)
7048         (<su>mulsi3_highpart_split, <su>muldi3_highpart_internal)
7049         (<u>mulditi3_r4000, *div<mode>3, *recip<mode>3, divmod<mode>4)
7050         (udivmod<mode>4, sqrt<mode>2, *rsqrt<mode>a, *rsqrt<mode>b)
7051         (fix_truncdfsi2_macro, fix_truncsfsi2_macro, *lea_high64)
7052         (*lea64, cprestore_<mode>, clear_hazard_<mode>, <unnamed insn>)
7053         (casesi_internal_mips16_<mode>, *tls_get_tp_<mode>_split)
7054         (tls_get_tp_mips16, *tls_get_tp_mips16_call_<mode>): Use "insn_count"
7055         rather than "length".
7056         (tls_get_tp_<mode>): Likewise.  Remove redundant "no_delay" attribute.
7057         * config/mips/mips-ps-3d.md (mips_c_cond_4s, mips_cabs_cond_4s):
7058         Use "insn_count" rather than "length".
7059         * config/mips/mips-dsp.md
7060         (mips_l<SHORT:size><u>x_ext<GPR:mode>_<P:mode>)
7061         (mips_l<GPR:size>x_<P:mode>, *mips_lw<u>x_<P:mode>_ext): Remove
7062         length attributes.
7064 2013-06-12  Marc Glisse  <marc.glisse@inria.fr>
7066         PR tree-optimization/57361
7067         * tree-ssa-dse.c (dse_possible_dead_store_p): Handle self-assignment.
7069 2013-06-12  Sofiane Naci  <sofiane.naci@arm.com>
7071         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Convert
7072         to split.
7073         (aarch64_simd_combine<mode>): New instruction expansion.
7074         * config/aarch64/aarch64-protos.h (aarch64_split_simd_combine): New
7075         function prototype.
7076         * config/aarch64/aarch64.c (aarch64_split_combine): New function.
7077         * config/aarch64/iterators.md (Vdbl): Add entry for DF.
7079 2013-06-12  Jan Hubicka  <jh@suse.cz>
7081         * cgraph.c (verify_edge_corresponds_to_fndecl): Be lax about
7082         decl has when in streaming stage.
7083         * lto-symtab.c (lto_symtab_merge_symbols): Likewise.
7084         * cgraph.h (cgraph_state): Add CGRAPH_LTO_STREAMING.
7086 2013-06-12  Roland Stigge  <stigge@antcom.de>
7088         PR target/57578
7089         * config/rs6000/t-linux (MULTIARCH_DIRNAME): Fix SPE version detection.
7091 2013-06-12  Jakub Jelinek  <jakub@redhat.com>
7093         PR tree-optimization/57537
7094         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): If
7095         vect_handle_widen_op_by_const, convert oprnd1 to half_type1.
7097 2013-06-12  Richard Biener  <rguenther@suse.de>
7099         * data-streamer.h (streamer_write_char_stream): CSE
7100         obs->current_pointer.
7101         * data-streamer-out.c (streamer_write_uhwi_stream): Inline
7102         streamer_write_char_stream manually and optimize the resulting loop.
7103         (streamer_write_hwi_stream): Likewise.
7105 2013-06-12  Jan Hubicka  <jh@suse.cz>
7107         * lto-symtab.c (lto_symtab_merge_symbols): Populate symtab hashtable.
7108         * cgraph.h (varpool_create_empty_node): Declare.
7109         * lto-cgraph.c (input_node, input_varpool_node): Forcingly create
7110         duplicated nodes.
7111         * symtab.c (symtab_unregister_node): Be lax about missin entries
7112         in node hash.
7113         (symtab_get_node): Update comment.
7114         * varpool.c (varpool_create_empty_node): Break out from ...
7115         (varpool_node_for_decl): ... here.
7116         * lto-streamer.h (lto_file_decl_data): Add RESOLUTION_MAP.
7118 2013-06-12  Eric Botcazou  <ebotcazou@adacore.com>
7120         * expr.c (expand_expr_real_1) <TARGET_MEM_REF>: Use straight-line flow.
7121         <MEM_REF>: Use 'type' instead of TREE_TYPE (exp) and tidy up the first
7122         part.  Use straight-line flow at the end.
7123         <COMPONENT_REF>: Remove superfluous else.
7124         <VIEW_CONVERT_EXPR>: Use 'type' instead of TREE_TYPE (exp).
7126 2013-06-12  Jakub Jelinek  <jakub@redhat.com>
7128         PR target/56564
7129         * varasm.c (decl_binds_to_current_def_p): Call binds_local_p
7130         target hook even for !TREE_PUBLIC decls.  If no resolution info
7131         is available, return false for common and external decls.
7133 2013-06-12  Kaushik Phatak  <kaushik.phatak@kpitcummins.com>
7135         * config/rl78/constraints.md (U): New constraint.
7136         * config/rl78/rl78.md (*mulqi3_rl78,*mulhi3_rl78,*mulhi3_g13): Add
7137         valloc attribute.
7139 2013-06-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
7141         PR target/57589
7142         * config/rs6000/driver-rs6000.c (elf_platform): Make buffer static
7143         to allow returning address to AT_PLATFORM name.
7145 2013-06-11  Jan Hubicka  <jh@suse.cz>
7147         * cgraph.c (cgraph_create_function_alias): Set weakref flag.
7148         * cgraph.h (symtab_node_base): Add weakref flag.
7149         * cgraphunit.c (cgraph_reset_node): Clear weakref flag.
7150         (handle_alias_pairs): Set weakref flag, do not set DECL_EXTERNAL.
7151         (output_weakrefs): Use weakref flag.
7152         * fold-const.c (simple_operand_p): Handle WEAK.
7153         * gimple-fold.c (can_refer_decl_in_current_unit_p): Drop weakref.
7154         * ipa.c (varpool_externally_visible_p): Drop weakref.
7155         (function_and_variable_visibility): Update comment; fix weakref
7156         sanity checks; do not clear DECL_WEAK on them.
7157         * lto-cgraph.c (lto_output_node): update.
7158         (lto_output_varpool_node): Update.
7159         (input_overwrite_node): Update.
7160         (input_node): Update.
7161         (input_varpool_node): Update.
7162         * lto-symtab.c (lto_symtab_symbol_p): Do not special case weakrefs.
7163         (lto_symtab_merge_symbols): Add sanity check.
7164         (lto_symtab_prevailing_decl): Do not special case weakrefs.
7165         * passes.c (rest_of_decl_compilation): Set static flag, too.
7166         * symtab.c (dump_symtab_base): Dump weakref.
7167         (verify_symtab_base): Sanity check weakrefs.
7168         (symtab_make_decl_local): Remove duplicated code.
7169         (symtab_alias_ultimate_target): Simplify.
7170         * varpool.c (varpool_create_variable_alias): Set weakref flag.
7172 2013-06-11  DJ Delorie  <dj@redhat.com>
7174         * config/rl78/rl78.c (TARGET_UNWIND_WORD_MODE): Define.
7175         (rl78_unwind_word_mode): New.
7177 2013-06-11  David Malcolm  <dmalcolm@redhat.com>
7179         * final.c (debug_prefix_maps): Make static.
7181 2013-06-11  David Malcolm  <dmalcolm@redhat.com>
7183         * function.c (initial_trampoline): Remove stray copy.
7185 2013-06-11  Sofiane Naci  <sofiane.naci@arm.com>
7187         * config/aarch64/aarch64-simd.md (move_lo_quad_<mode>): Update.
7189 2013-06-11  Martin Jambor  <mjambor@suse.cz>
7191         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that param_index is
7192         within bounds at the beginning of the function.
7194 2013-06-11  Alan Modra  <amodra@gmail.com>
7196         * varasm.c (get_section): Don't die on !DECL_P decl.  Tidy error
7197         reporting.
7198         (get_named_section): Don't NULL !DECL_P decl.
7200 2013-06-11  Igor Zamyatin  <igor.zamyatin@intel.com>
7202         * doc/invoke.texi (core-avx2): Document.
7203         (slm): Likewise.
7204         (atom): Updated with MOVBE.
7206 2013-06-11  Richard Biener  <rguenther@suse.de>
7208         * collect2.c (main): Do not redirect ld stdout/stderr when debugging.
7210 2013-06-11  Anton Blanchard  <anton@samba.org>
7212         * config/rs6000/rs6000.c (rs6000_adjust_atomic_subword): Calculate
7213         correct shift value in little-endian mode.
7215 2013-06-11  Jakub Jelinek  <jakub@redhat.com>
7217         PR target/56564
7218         * varasm.c (get_variable_align): Move #endif to the right place.
7220 2013-06-10  Cary Coutant  <ccoutant@google.com>
7222         * dwarf2out.c (hash_external_ref): Use die_symbol or signature
7223         for hash so that hash table traversal order is deterministic.
7225 2013-06-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7226             Pat Haugen  <pthaugen@us.ibm.com>
7227             Peter Bergner  <bergner@vnet.ibm.com>
7229         * config/rs6000/vector.md (GPR move splitter): Do not split moves
7230         of vectors in GPRS if they are direct moves or quad word load or
7231         store moves.
7233         * config/rs6000/rs6000-protos.h (rs6000_output_move_128bit): Add
7234         declaration.
7235         (direct_move_p): Likewise.
7236         (quad_load_store_p): Likewise.
7238         * config/rs6000/rs6000.c (enum rs6000_reg_type): Simplify register
7239         classes into bins based on the physical register type.
7240         (reg_class_to_reg_type): Likewise.
7241         (IS_STD_REG_TYPE): Likewise.
7242         (IS_FP_VECT_REG_TYPE): Likewise.
7243         (reload_fpr_gpr): Arrays to determine what insn to use if we can
7244         use direct move instructions.
7245         (reload_gpr_vsx): Likewise.
7246         (reload_vsx_gpr): Likewise.
7247         (rs6000_init_hard_regno_mode_ok): Precalculate the register type
7248         information that is a simplification of register classes.  Also
7249         precalculate direct move reload helpers.
7250         (direct_move_p): New function to return true if the operation can
7251         be done as a direct move instruciton.
7252         (quad_load_store_p): New function to return true if the operation
7253         is a quad memory operation.
7254         (rs6000_legitimize_address): If quad memory, only allow register
7255         indirect for TImode addresses.
7256         (rs6000_legitimate_address_p): Likewise.
7257         (enum reload_reg_type): Delete, replace with rs6000_reg_type.
7258         (rs6000_reload_register_type): Likewise.
7259         (register_to_reg_type): Return register type.
7260         (rs6000_secondary_reload_simple_move): New helper function for
7261         secondary reload and secondary memory needed to identify anything
7262         that is a simple move, and does not need reloading.
7263         (rs6000_secondary_reload_direct_move): New helper function for
7264         secondary reload to identify cases that can be done with several
7265         instructions via the direct move instructions.
7266         (rs6000_secondary_reload_move): New helper function for secondary
7267         reload to identify moves between register types that can be done.
7268         (rs6000_secondary_reload): Add support for quad memory operations
7269         and for direct move.
7270         (rs6000_secondary_memory_needed): Likewise.
7271         (rs6000_debug_secondary_memory_needed): Change argument names.
7272         (rs6000_output_move_128bit): New function to return the move to
7273         use for 128-bit moves, including knowing about the various
7274         limitations of quad memory operations.
7276         * config/rs6000/vsx.md (vsx_mov<mode>): Add support for quad
7277         memory operations.  call rs6000_output_move_128bit for the actual
7278         instruciton(s) to generate.
7279         (vsx_movti_64bit): Likewise.
7281         * config/rs6000/rs6000.md (UNSPEC_P8V_FMRGOW): New unspec values.
7282         (UNSPEC_P8V_MTVSRWZ): Likewise.
7283         (UNSPEC_P8V_RELOAD_FROM_GPR): Likewise.
7284         (UNSPEC_P8V_MTVSRD): Likewise.
7285         (UNSPEC_P8V_XXPERMDI): Likewise.
7286         (UNSPEC_P8V_RELOAD_FROM_VSX): Likewise.
7287         (UNSPEC_FUSION_GPR): Likewise.
7288         (FMOVE128_GPR): New iterator for direct move.
7289         (f32_lv): New mode attribute for load/store of SFmode/SDmode values.
7290         (f32_sv): Likewise.
7291         (f32_dm): Likewise.
7292         (zero_extend<mode>di2_internal1): Add support for power8 32-bit
7293         loads and direct move instructions.
7294         (zero_extendsidi2_lfiwzx): Likewise.
7295         (extendsidi2_lfiwax): Likewise.
7296         (extendsidi2_nocell): Likewise.
7297         (floatsi<mode>2_lfiwax): Likewise.
7298         (lfiwax): Likewise.
7299         (floatunssi<mode>2_lfiwzx): Likewise.
7300         (lfiwzx): Likewise.
7301         (fix_trunc<mode>_stfiwx): Likewise.
7302         (fixuns_trunc<mode>_stfiwx): Likewise.
7303         (mov<mode>_hardfloat, 32-bit floating point): Likewise.
7304         (mov<move>_hardfloat64, 64-bit floating point): Likewise.
7305         (parity<mode>2_cmpb): Set length/type attr.
7306         (unnamed shift right patterns, mov<mode>_internal2): Change type attr
7307         for 'mr.' to fast_compare.
7308         (bpermd_<mode>): Change type attr to popcnt.
7309         (p8_fmrgow_<mode>): New insns for power8 direct move support.
7310         (p8_mtvsrwz_1): Likewise.
7311         (p8_mtvsrwz_2): Likewise.
7312         (reload_fpr_from_gpr<mode>): Likewise.
7313         (p8_mtvsrd_1): Likewise.
7314         (p8_mtvsrd_2): Likewise.
7315         (p8_xxpermdi_<mode>): Likewise.
7316         (reload_vsx_from_gpr<mode>): Likewise.
7317         (reload_vsx_from_gprsf): Likewise.
7318         (p8_mfvsrd_3_<mode>): LIkewise.
7319         (reload_gpr_from_vsx<mode>): Likewise.
7320         (reload_gpr_from_vsxsf): Likewise.
7321         (p8_mfvsrd_4_disf): Likewise.
7322         (multi-word GPR splits): Do not split direct moves or quad memory
7323         operations.
7325 2013-06-10  David Malcolm  <dmalcolm@redhat.com>
7327         * tree-into-ssa.c (interesting_blocks): Make static.
7329 2013-06-10  Jakub Jelinek  <jakub@redhat.com>
7331         PR target/56564
7332         * varasm.c (align_variable): Don't use DATA_ALIGNMENT or
7333         CONSTANT_ALIGNMENT if !decl_binds_to_current_def_p (decl).
7334         Use DATA_ABI_ALIGNMENT for that case instead if defined.
7335         (get_variable_align): New function.
7336         (get_variable_section, emit_bss, emit_common,
7337         assemble_variable_contents, place_block_symbol): Use
7338         get_variable_align instead of DECL_ALIGN.
7339         (assemble_noswitch_variable): Add align argument, use it
7340         instead of DECL_ALIGN.
7341         (assemble_variable): Adjust caller.  Use get_variable_align
7342         instead of DECL_ALIGN.
7343         * config/i386/i386.h (DATA_ALIGNMENT): Adjust x86_data_alignment
7344         caller.
7345         (DATA_ABI_ALIGNMENT): Define.
7346         * config/i386/i386-protos.h (x86_data_alignment): Adjust prototype.
7347         * config/i386/i386.c (x86_data_alignment): Add opt argument.  If
7348         opt is false, only return the psABI mandated alignment increase.
7349         * config/c6x/c6x.h (DATA_ALIGNMENT): Renamed to...
7350         (DATA_ABI_ALIGNMENT): ... this.
7351         * config/mmix/mmix.h (DATA_ALIGNMENT): Renamed to...
7352         (DATA_ABI_ALIGNMENT): ... this.
7353         * config/mmix/mmix.c (mmix_data_alignment): Adjust function comment.
7354         * config/s390/s390.h (DATA_ALIGNMENT): Renamed to...
7355         (DATA_ABI_ALIGNMENT): ... this.
7356         * doc/tm.texi.in (DATA_ABI_ALIGNMENT): Document.
7357         * doc/tm.texi: Regenerated.
7359 2013-06-10  Uros Bizjak  <ubizjak@gmail.com>
7361         * config/alpha/alpha.c (alpha_emit_xfloating_compare): Also use
7362         cmp_code to construct REG_EQUAL note.
7364 2013-06-09  Jakub Jelinek  <jakub@redhat.com>
7366         PR target/57568
7367         * config/i386/i386.md (TARGET_READ_MODIFY_WRITE peepholes): Ensure
7368         that operands[2] doesn't overlap with operands[0].
7370 2013-06-09  David Edelsohn  <dje.gcc@gmail.com>
7371             Jan Hubicka  <jh@suse.cz>
7373         * config/rs6000/rs6000.c (print_operand, 'z'): Remove historical
7374         hack to mark symbols as used.
7376 2013-06-08  Vladimir Makarov  <vmakarov@redhat.com>
7378         PR rtl-optimization/57559
7379         * lra-constraints.c (process_alt_operands): Don't discourage
7380         memory with known offset for offsetable memory constraint.
7381         * lra.c (lra_emit_add): Exchange y and z for 2-op add insn.
7383 2013-06-08  Eric Botcazou  <ebotcazou@adacore.com>
7385         * varasm.c (struct oc_local_state): Reorder fields.
7386         (output_constructor_bitfield): Replace OUTER parameter with BIT_OFFSET
7387         and adjust accordingly.
7388         (output_constructor): Reorder initialization code and adjust call to
7389         output_constructor_bitfield.
7391 2013-06-07  Jan Hubicka  <jh@suse.cz>
7393         * symtab.c (symtab_resolve_alias): Do not remove alias attribute.
7395 2013-06-07  David Malcolm  <dmalcolm@redhat.com>
7397         * tree-object-size.c (unknown): Make const.
7399 2013-06-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
7401         * config/s390/s390.md (cpu_facility): Add cpu_zarch.
7402         ("*movmem_short", "*clrmem_short", "*cmpmem_short): Use cpu_zarch
7403         for last alternative in the cpu_facility attribute.
7405 2013-06-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7407         PR target/56315
7408         * config/arm/arm.md (*xordi3_insn): Change to insn_and_split.
7409         (xordi3): Change operand 2 constraint to arm_xordi_operand.
7410         * config/arm/arm.c (const_ok_for_dimode_op): Handle XOR.
7411         * config/arm/constraints.md (Dg): New constraint.
7412         * config/arm/neon.md (xordi3_neon): Remove.
7413         (neon_veor<mode>): Generate xordi3 instead of xordi3_neon.
7414         * config/arm/predicates.md (arm_xordi_operand): New predicate.
7416 2013-06-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7418         * config/arm/arm.md (anddi3_insn): Remove duplicate alternatives.
7419         Clean up alternatives.
7421 2013-06-07  Alan Modra  <amodra@gmail.com>
7423         * config/rs6000/rs6000.c (setup_incoming_varargs): Round up
7424         va_list_gpr_size.
7426 2013-06-07  Alan Modra  <amodra@gmail.com>
7428         * varasm.c (force_const_mem): Assert mode is not VOID or BLK.
7430 2013-06-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7432         * config/arm/constraints.md (Df): New constraint.
7433         * config/arm/arm.md (iordi3_insn): Use Df constraint instead of De.
7434         Correct length attribute for last two alternatives.
7436 2013-06-07  Alan Modra  <amodra@gmail.com>
7438         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7439         override user -mfp-in-toc.
7440         (offsettable_ok_by_alignment): Consider just the current access
7441         rather than the whole object, unless BLKmode.  Handle
7442         CONSTANT_POOL_ADDRESS_P constants that lack a decl too.
7443         (use_toc_relative_ref): Allow CONSTANT_POOL_ADDRESS_P constants
7444         for -mcmodel=medium.
7445         * config/rs6000/linux64.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Don't
7446         override user -mfp-in-toc or -msum-in-toc.  Default to
7447         -mno-fp-in-toc for -mcmodel=medium.
7449 2013-06-06  DJ Delorie  <dj@redhat.com>
7451         * config/rl78/rl78.c (rl78_valid_pointer_mode): New, implements
7452         TARGET_VALID_POINTER_MODE.
7454 2013-06-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
7455             Pat Haugen  <pthaugen@us.ibm.com>
7456             Peter Bergner  <bergner@vnet.ibm.com>
7458         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
7459         Document new power8 builtins.
7461         * config/rs6000/vector.md (and<mode>3): Add a clobber/scratch of a
7462         condition code register, to allow 128-bit logical operations to be
7463         done in the VSX or GPR registers.
7464         (nor<mode>3): Use the canonical form for nor.
7465         (eqv<mode>3): Add expanders for power8 xxleqv, xxlnand, xxlorc,
7466         vclz*, and vpopcnt* vector instructions.
7467         (nand<mode>3): Likewise.
7468         (orc<mode>3): Likewise.
7469         (clz<mode>2): LIkewise.
7470         (popcount<mode>2): Likewise.
7472         * config/rs6000/predicates.md (int_reg_operand): Rework tests so
7473         that only the GPRs are recognized.
7475         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7476         support for new power8 builtins.
7478         * config/rs6000/rs6000-builtin.def (xscvspdpn): Add new power8
7479         builtin functions.
7480         (xscvdpspn): Likewise.
7481         (vclz): Likewise.
7482         (vclzb): Likewise.
7483         (vclzh): Likewise.
7484         (vclzw): Likewise.
7485         (vclzd): Likewise.
7486         (vpopcnt): Likewise.
7487         (vpopcntb): Likewise.
7488         (vpopcnth): Likewise.
7489         (vpopcntw): Likewise.
7490         (vpopcntd): Likewise.
7491         (vgbbd): Likewise.
7492         (vmrgew): Likewise.
7493         (vmrgow): Likewise.
7494         (eqv): Likewise.
7495         (eqv_v16qi3): Likewise.
7496         (eqv_v8hi3): Likewise.
7497         (eqv_v4si3): Likewise.
7498         (eqv_v2di3): Likewise.
7499         (eqv_v4sf3): Likewise.
7500         (eqv_v2df3): Likewise.
7501         (nand): Likewise.
7502         (nand_v16qi3): Likewise.
7503         (nand_v8hi3): Likewise.
7504         (nand_v4si3): Likewise.
7505         (nand_v2di3): Likewise.
7506         (nand_v4sf3): Likewise.
7507         (nand_v2df3): Likewise.
7508         (orc): Likewise.
7509         (orc_v16qi3): Likewise.
7510         (orc_v8hi3): Likewise.
7511         (orc_v4si3): Likewise.
7512         (orc_v2di3): Likewise.
7513         (orc_v4sf3): Likewise.
7514         (orc_v2df3): Likewise.
7516         * config/rs6000/rs6000.c (rs6000_option_override_internal): Only
7517         allow power8 quad mode in 64-bit.
7518         (rs6000_builtin_vectorized_function): Add support to vectorize
7519         ISA 2.07 count leading zeros, population count builtins.
7520         (rs6000_expand_vector_init): On ISA 2.07 use xscvdpspn to form
7521         V4SF vectors instead of xscvdpsp to avoid IEEE related traps.
7522         (builtin_function_type): Add vgbbd builtin function which takes an
7523         unsigned argument.
7524         (altivec_expand_vec_perm_const): Add support for new power8 merge
7525         instructions.
7527         * config/rs6000/vsx.md (VSX_L2): New iterator for 128-bit types,
7528         that does not include TImdoe for use with 32-bit.
7529         (UNSPEC_VSX_CVSPDPN): Support for power8 xscvdpspn and xscvspdpn
7530         instructions.
7531         (UNSPEC_VSX_CVDPSPN): Likewise.
7532         (vsx_xscvdpspn): Likewise.
7533         (vsx_xscvspdpn): Likewise.
7534         (vsx_xscvdpspn_scalar): Likewise.
7535         (vsx_xscvspdpn_directmove): Likewise.
7536         (vsx_and<mode>3): Split logical operations into 32-bit and
7537         64-bit. Add support to do logical operations on TImode as well as
7538         VSX vector types.  Allow logical operations to be done in either
7539         VSX registers or in general purpose registers in 64-bit mode.  Add
7540         splitters if GPRs were used. For AND, add clobber of CCmode to
7541         allow use of ANDI on GPRs.  Rewrite nor to use the canonical RTL
7542         encoding.
7543         (vsx_and<mode>3_32bit): Likewise.
7544         (vsx_and<mode>3_64bit): Likewise.
7545         (vsx_ior<mode>3): Likewise.
7546         (vsx_ior<mode>3_32bit): Likewise.
7547         (vsx_ior<mode>3_64bit): Likewise.
7548         (vsx_xor<mode>3): Likewise.
7549         (vsx_xor<mode>3_32bit): Likewise.
7550         (vsx_xor<mode>3_64bit): Likewise.
7551         (vsx_one_cmpl<mode>2): Likewise.
7552         (vsx_one_cmpl<mode>2_32bit): Likewise.
7553         (vsx_one_cmpl<mode>2_64bit): Likewise.
7554         (vsx_nor<mode>3): Likewise.
7555         (vsx_nor<mode>3_32bit): Likewise.
7556         (vsx_nor<mode>3_64bit): Likewise.
7557         (vsx_andc<mode>3): Likewise.
7558         (vsx_andc<mode>3_32bit): Likewise.
7559         (vsx_andc<mode>3_64bit): Likewise.
7560         (vsx_eqv<mode>3_32bit): Add support for power8 xxleqv, xxlnand,
7561         and xxlorc instructions.
7562         (vsx_eqv<mode>3_64bit): Likewise.
7563         (vsx_nand<mode>3_32bit): Likewise.
7564         (vsx_nand<mode>3_64bit): Likewise.
7565         (vsx_orc<mode>3_32bit): Likewise.
7566         (vsx_orc<mode>3_64bit): Likewise.
7568         * config/rs6000/rs6000.h (VLOGICAL_REGNO_P): Update comment.
7570         * config/rs6000/altivec.md (UNSPEC_VGBBD): Add power8 vgbbd
7571         instruction.
7572         (p8_vmrgew): Add power8 vmrgew and vmrgow instructions.
7573         (p8_vmrgow): Likewise.
7574         (altivec_and<mode>3): Add clobber of CCmode to allow AND using
7575         GPRs to be split under VSX.
7576         (p8v_clz<mode>2): Add power8 count leading zero support.
7577         (p8v_popcount<mode>2): Add power8 population count support.
7578         (p8v_vgbbd): Add power8 gather bits by bytes by doubleword
7579         support.
7581         * config/rs6000/rs6000.md (eqv<mode>3): Add support for powerp eqv
7582         instruction.
7584         * config/rs6000/altivec.h (vec_eqv): Add defines to export power8
7585         builtin functions.
7586         (vec_nand): Likewise.
7587         (vec_vclz): Likewise.
7588         (vec_vclzb): Likewise.
7589         (vec_vclzd): Likewise.
7590         (vec_vclzh): Likewise.
7591         (vec_vclzw): Likewise.
7592         (vec_vgbbd): Likewise.
7593         (vec_vmrgew): Likewise.
7594         (vec_vmrgow): Likewise.
7595         (vec_vpopcnt): Likewise.
7596         (vec_vpopcntb): Likewise.
7597         (vec_vpopcntd): Likewise.
7598         (vec_vpopcnth): Likewise.
7599         (vec_vpopcntw): Likewise.
7601 2013-06-06  Vladimir Makarov  <vmakarov@redhat.com>
7603         PR rtl-optimization/57468
7604         * config/i386/i386.c (inline_secondary_memory_needed): Ignore
7605         spilled pseudos.
7607 2013-06-06  Vladimir Makarov  <vmakarov@redhat.com>
7609         PR rtl-optimization/57459
7610         * lra-constraints.c (update_ebb_live_info): Fix typo for operand
7611         type when setting live regs.
7613 2013-06-06  Vladimir Makarov  <vmakarov@redhat.com>
7615         * config/s390/s390.opt (mlra): New option.
7616         * config/s390/s390.c (s390_decompose_address): Check displacement
7617         for all registers for LRA.
7618         (s390_secondary_reload): Don't used secondary reloads for LRA.
7619         (s390_lra_p): New function.
7620         (TARGET_LRA_P): Define.
7621         * config/s390/s390.md (*movmem_short, *clrmem_short): Change value
7622         of attribute cpu_facility to zarch for the last alternative.
7623         (*cmpmem_short): Ditto.
7625 2013-06-06  Eric Botcazou  <ebotcazou@adacore.com>
7627         * config/arm/arm.c (arm_r3_live_at_start_p): New predicate.
7628         (arm_compute_static_chain_stack_bytes): Use it.  Tidy up.
7629         (arm_expand_prologue): Likewise.
7631 2013-06-06  Teresa Johnson  <tejohnson@google.com>
7633         PR c++/53743
7634         * ifcvt.c (find_if_case_1): Replace BB_COPY_PARTITION with assert
7635         as this is now done by redirect_edge_and_branch_force.
7636         * function.c (thread_prologue_and_epilogue_insns): Insert new bb after
7637         barriers, and fix interaction with splitting.
7638         * emit-rtl.c (try_split): Copy REG_CROSSING_JUMP notes.
7639         * cfgcleanup.c (try_forward_edges): Fix early return value to properly
7640         reflect changes made in the routine.
7641         * bb-reorder.c (emit_barrier_after_bb): Move to cfgrtl.c.
7642         (fix_up_fall_thru_edges): Remove incorrect check for bb layout order
7643         since this is called in cfglayout mode, and replace partition fixup
7644         with assert as that is now done by force_nonfallthru_and_redirect.
7645         (add_reg_crossing_jump_notes): Handle the fact that some jumps may
7646         already be marked with region crossing note.
7647         (insert_section_boundary_note): Make non-static, gate on flag
7648         has_bb_partition, rewrite to also check for multiple partitions.
7649         (rest_of_handle_reorder_blocks): Remove call to
7650         insert_section_boundary_note, now done later during free_cfg.
7651         (duplicate_computed_gotos): Don't duplicate partition crossing edge.
7652         * bb-reorder.h (insert_section_boundary_note): Declare.
7653         * Makefile.in (cfgrtl.o): Depend on bb-reorder.h
7654         * cfgrtl.c (rest_of_pass_free_cfg): If partitions exist
7655         invoke insert_section_boundary_note.
7656         (try_redirect_by_replacing_jump): Remove unnecessary
7657         check for region crossing note.
7658         (fixup_partition_crossing): New function.
7659         (rtl_redirect_edge_and_branch): Fixup partition boundaries.
7660         (emit_barrier_after_bb): Move here from bb-reorder.c, handle insertion
7661         in non-cfglayout mode.
7662         (force_nonfallthru_and_redirect): Fixup partition boundaries,
7663         remove old code that tried to do this. Emit barrier correctly
7664         when we are in cfglayout mode.
7665         (last_bb_in_partition): New function.
7666         (rtl_split_edge): Correctly fixup partition boundaries.
7667         (commit_one_edge_insertion): Remove old code that tried to
7668         fixup region crossing edge since this is now handled in
7669         split_block, and set up insertion point correctly since
7670         block may now end in a jump.
7671         (verify_hot_cold_block_grouping): Guard against checking when not in
7672         linearized RTL mode.
7673         (rtl_verify_edges): Add checks for incorrect/missing REG_CROSSING_JUMP
7674         notes.
7675         (rtl_verify_flow_info_1): Move verify_hot_cold_block_grouping to
7676         rtl_verify_flow_info, so not called in cfglayout mode.
7677         (rtl_verify_flow_info): Move verify_hot_cold_block_grouping here.
7678         (fixup_reorder_chain): Remove old code that attempted to fixup region
7679         crossing note as this is now handled in force_nonfallthru_and_redirect.
7680         (duplicate_insn_chain): Don't duplicate switch section notes.
7681         (rtl_can_remove_branch_p): Remove unnecessary check for region crossing
7682         note.
7683         * basic-block.h (emit_barrier_after_bb): Declare.
7685 2013-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7687         * config/arm/arm-fixed.md (add<mode>3,usadd<mode>3,ssadd<mode>3,
7688         sub<mode>3, ussub<mode>3, sssub<mode>3, arm_ssatsihi_shift,
7689         arm_usatsihi): Adjust alternatives for arm_restrict_it.
7691 2013-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7693         * config/arm/arm-ldmstm.ml: Set "predicable_short_it" to "no"
7694         where appropriate.
7695         * config/arm/ldmstm.md: Regenerate.
7697 2013-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7699         * config/arm/sync.md (atomic_loaddi_1):
7700         Disable predication for arm_restrict_it.
7701         (arm_load_exclusive<mode>): Likewise.
7702         (arm_load_exclusivesi): Likewise.
7703         (arm_load_exclusivedi): Likewise.
7704         (arm_load_acquire_exclusive<mode>): Likewise.
7705         (arm_load_acquire_exclusivesi): Likewise.
7706         (arm_load_acquire_exclusivedi): Likewise.
7707         (arm_store_exclusive<mode>): Likewise.
7708         (arm_store_exclusive<mode>): Likewise.
7709         (arm_store_release_exclusivedi): Likewise.
7710         (arm_store_release_exclusive<mode>): Likewise.
7712 2013-06-06  Richard Biener  <rguenther@suse.de>
7714         * lto-streamer.h (enum LTO_tags): Move LTO_tree_pickle_reference
7715         after LTO_null.
7716         (lto_tag_is_tree_code_p): Adjust.
7717         (lto_tag_is_gimple_code_p): Likewise.
7718         (lto_gimple_code_to_tag): Likewise.
7719         (lto_tag_to_gimple_code): Likewise.
7720         (lto_tree_code_to_tag): Likewise.
7721         (lto_tag_to_tree_code): Likewise.
7722         * data-streamer.h (streamer_write_hwi_in_range): Use
7723         uhwi streaming to stream the normalized range.
7724         (streamer_read_hwi_in_range): Likewise.
7726 2013-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7728         * config/arm/arm.md (enabled_for_depr_it): New attribute.
7729         (predicable_short_it): Likewise.
7730         (predicated): Likewise.
7731         (enabled): Handle above.
7732         (define_cond_exec): Set predicated attribute to yes.
7734 2013-06-05  Mike Stump  <mikestump@comcast.net>
7736         * gdbinit.in (__FUNCTION__): Add.
7738 2013-06-05  Uros Bizjak  <ubizjak@gmail.com>
7740         * config/alpha/alpha.c (alpha_emit_conditional_move): Swap all
7741         GE, GT, GEU and GTU compares, modulo DImode compares with zero.
7743 2013-06-05  Jan Hubicka  <jh@suse.cz>
7745         * varasm.c (mark_decl_referenced): Revert the removal until targets
7746         are fixed.
7748 2013-06-05  David Edelsohn  <dje.gcc@gmail.com>
7750         * config/rs6000/rs6000.c (print_operand, 'z'): Use DECL_PRESERVE_P
7751         instead of mark_decl_referenced.
7753 2013-06-05  Jan Hubicka  <jh@suse.cz>
7755         * cgraph.c (cgraph_remove_node): Clear forced_by_abi.
7756         (cgraph_node_cannot_be_local_p_1): Honnor symbol.forced_by_abi
7757         and symtab_used_from_object_file_p.
7758         (cgraph_make_node_local_1): Clear forced_by_abi.
7759         (cgraph_can_remove_if_no_direct_calls_and): Use forced_by_abi
7760         * cgraph.h (symtab_node_base): Add forced_by_abi.
7761         (decide_is_variable_needed): Remove.
7762         (varpool_can_remove_if_no_refs): Honnor symbol.forced_by_abi.
7763         * cgraphunit.c (cgraph_decide_is_function_needed): Rename to ..
7764         (decide_is_symbol_needed): ... this one; handle symbols in general;
7765         always analyze virtuals; honnor forced_by_abi.
7766         (cgraph_finalize_function): Update.
7767         (varpool_finalize_decl): Update.
7768         (symbol_defined_and_needed): Remove.
7769         (analyze_functions): Update.
7770         * lto-cgraph.c (lto_output_node, lto_output_varpool_node,
7771         output_refs, input_overwrite_node): Handle forced_by_abi.
7772         * ipa.c (cgraph_address_taken_from_non_vtable_p): Rename to ...
7773         (address_taken_from_non_vtable_p): ... this one.
7774         (comdat_can_be_unshared_p_1): New function.
7775         (cgraph_comdat_can_be_unshared_p): Rename to ...
7776         (comdat_can_be_unshared_p): ... this one; handle symbols in general.
7777         (varpool_externally_visible_p): Use comdat_can_be_unshared_p.
7778         (function_and_variable_visibility): Clear forced_by_abi as needed.
7779         * trans-mem.c (ipa_tm_mark_forced_by_abi_node): New functoin.
7780         (ipa_tm_create_version_alias, ipa_tm_create_version): Update.
7781         * symtab.c (dump_symtab_base): Dump forced_by_abi.
7782         * varpool.c (decide_is_variable_needed): Remove.
7784 2013-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7786         * config/arm/arm.c (MAX_INSN_PER_IT_BLOCK): New macro.
7787         (arm_option_override): Override arm_restrict_it where appropriate.
7788         (thumb2_final_prescan_insn): Use MAX_INSN_PER_IT_BLOCK.
7789         * config/arm/arm.opt (mrestrict-it): New command-line option.
7790         * doc/invoke.texi: Document -mrestrict-it.
7792 2013-06-05  David Malcolm  <dmalcolm@redhat.com>
7794         * tsan.c (tsan_atomic_table): Make const.
7796 2013-06-05  Richard Biener  <rguenther@suse.de>
7798         * tree-streamer.c (streamer_tree_cache_insert_1): Update the
7799         index associated with the tree we are supposed to replace.
7800         * tree-streamer-out.c (pack_ts_base_value_fields): Output
7801         TREE_ASM_WRITTEN as zero for everything but SSA names.
7803 2013-06-05  David Malcolm  <dmalcolm@redhat.com>
7805         * tree-ssa-structalias.c (call_stmt_vars): Make static.
7807 2013-06-04  Jan Hubicka  <jh@suse.cz>
7809         * lto-cgraph.c (get_alias_symbol): Remove weakref sanity check.
7810         (input_node, input_varpool_node): Handle correctly external same
7811         body aliases.
7812         * ipa.c (symtab_remove_unreachable_nodes): Do not remove external
7813         nodes at ltrans stage.
7815 2013-06-04  Jan Hubicka  <jh@suse.cz>
7817         * ipa-inline.c (update_caller_keys): Fix availability test.
7818         (update_callee_keys): Likewise.
7819         * symtab.c (symtab_alias_ultimate_target): Make availaiblity logic
7820         to follow ELF standard.
7822 2013-06-04  Jürgen Urban  <JuergenUrban@gmx.de>
7824         * config.gcc (mipsr5900-*-elf*, mipsr5900el-*-elf*, mips64r5900-*-elf*)
7825         (mips64r5900el-*-elf*): New configurations.
7826         * config/mips/mips-cpus.def (r5900): New processor.
7827         * config/mips/mips-tables.opt: Regenerate.
7828         * config/mips/mips.c (mips_rtx_cost_data): Add an R5900 entry.
7829         (mips_issue_rate): Handle PROCESSOR_R5900.
7830         (mips_reorg_process_insns): Force reorder mode for the R5900.
7831         * config/mips/mips.h (TARGET_MIPS5900): Define.
7832         (ISA_HAS_CONDMOVE, ISA_HAS_PREFETCH, ISA_HAS_HILO_INTERLOCKS): Include
7833         TARGET_MIPS5900.
7834         (ISA_HAS_LOAD_DELAY, ISA_HAS_XFER_DELAY, ISA_HAS_FCMP_DELAY): Exclude
7835         TARGET_MIPS5900.
7836         * config/mips/mips.md (processor): Add r5900.
7837         (MOVECC): Disallow CCmode conditions for TARGET_MIPS5900.
7839 2013-06-04  Ian Bolton  <ian.bolton@arm.com>
7841         * config/aarch64/aarch64.md (*mov<mode>_aarch64): Call
7842         into function to generate MOVI instruction.
7843         * config/aarch64/aarch64.c (aarch64_simd_container_mode): New function.
7844         (aarch64_preferred_simd_mode): Turn into wrapper.
7845         (aarch64_output_scalar_simd_mov_immediate): New function.
7846         * config/aarch64/aarch64-protos.h: Add prototype for above.
7848 2013-06-04  Ian Bolton  <ian.bolton@arm.com>
7850         * config/aarch64/aarch64.c (simd_immediate_info): Remove
7851         element_char member.
7852         (sizetochar): Return signed char.
7853         (aarch64_simd_valid_immediate): Remove elchar and other
7854         unnecessary variables.
7855         (aarch64_output_simd_mov_immediate): Take rtx instead of &rtx.
7856         Calculate element_char as required.
7857         * config/aarch64/aarch64-protos.h: Update and move prototype
7858         for aarch64_output_simd_mov_immediate.
7859         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
7860         Update arguments.
7862 2013-06-04  Ian Bolton  <ian.bolton@arm.com>
7864         * config/aarch64/aarch64.c (simd_immediate_info): Struct to hold
7865         information completed by aarch64_simd_valid_immediate.
7866         (aarch64_legitimate_constant_p): Update arguments.
7867         (aarch64_simd_valid_immediate): Work with struct rather than many
7868         pointers.
7869         (aarch64_simd_scalar_immediate_valid_for_move): Update arguments.
7870         (aarch64_simd_make_constant): Update arguments.
7871         (aarch64_output_simd_mov_immediate): Work with struct rather than
7872         many pointers.  Output immediate directly rather than as operand.
7873         * config/aarch64/aarch64-protos.h (aarch64_simd_valid_immediate):
7874         Update prototype.
7875         * config/aarch64/constraints.md (Dn): Update arguments.
7877 2013-06-04  Ian Bolton  <ian.bolton@arm.com>
7879         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): No
7880         longer static.
7881         (aarch64_simd_immediate_valid_for_move): Remove.
7882         (aarch64_simd_scalar_immediate_valid_for_move): Update call.
7883         (aarch64_simd_make_constant): Update call.
7884         (aarch64_output_simd_mov_immediate): Update call.
7885         * config/aarch64/aarch64-protos.h (aarch64_simd_valid_immediate):
7886         Add prototype.
7887         * config/aarch64/constraints.md (Dn): Update call.
7889 2013-06-04  Ian Bolton  <ian.bolton@arm.com>
7891         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Change
7892         return type to bool for prototype.
7893         (aarch64_legitimate_constant_p): Check for true instead of not -1.
7894         (aarch64_simd_valid_immediate): Fix up each return to return a bool.
7895         (aarch64_simd_immediate_valid_for_move): Update retval for bool.
7897 2013-06-04  Catherine Moore  <clm@codesourcery.com>
7899         * config/mips/mips.opt (meva): New.
7900         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define __mips_eva.
7901         (ASM_SPEC): Handle -meva.
7902         * doc/invoke.texi (meva):  Document.
7904 2013-06-04  Alan Modra  <amodra@gmail.com>
7906         * config/rs6000/rs6000.c (output_toc): Correct little-endian float
7907         constant output.
7909 2013-06-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7911         * rtl.def: Add extra fourth optional field to define_cond_exec.
7912         * gensupport.c (process_one_cond_exec): Process attributes from
7913         define_cond_exec.
7914         * doc/md.texi: Document fourth field in define_cond_exec.
7916 2013-06-04  Eric Botcazou  <ebotcazou@adacore.com>
7918         * expmed.c (extract_bit_field_1): In the larger-than-a-word case, factor
7919         out the processing order as in store_bit_field_1.
7921 2013-06-04  Jan Hubicka  <jh@suse.cz>
7923         PR middle-end/57500
7924         * cgraphunit.c (cgraph_process_same_body_aliases): Create
7925         non-VAR_DECL node if it does not exist yet.
7927 2013-06-03  Richard Sandiford  <rdsandiford@googlemail.com>
7929         * config.gcc (mipsisa64sr71k-*-elf*, mipsisa64sb1-*-elf*)
7930         (mipsisa64sb1el-*-elf*, mips64-*-elf*, mips64el-*-elf*)
7931         (mips64orion-*-elf*, mips64orionel-*-elf*): Remove
7932         target_cpu_default setting.
7934 2013-06-03  Teresa Johnson  <tejohnson@google.com>
7936         * dumpfile.c (opt_info_switch_p): Change -fopt-info
7937         default to -fopt-info=optimized instead of all.
7938         * doc/invoke.texi: Ditto.
7939         * tree-vectorizer.c (vectorize_loops): Emit loop vectorization
7940         success under MSG_OPTIMIZED_LOCATIONS, and use dump_printf_loc.
7941         (execute_vect_slp): Emit BB vectorization success under
7942         MSG_OPTIMIZED_LOCATIONS.
7943         * tree-vect-slp.c (vect_slp_transform_bb): Change
7944         MSG_OPTIMIZED_LOCATIONS to MSG_NOTE.
7945         * tree-vect-loop.c (vect_transform_loop): Ditto.
7947 2013-06-03  Jason Merrill  <jason@redhat.com>
7949         PR c++/57415
7950         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
7951         Use TARGET_EXPR for C++.
7953 2013-06-03  Jakub Jelinek  <jakub@redhat.com>
7955         PR rtl-optimization/57268
7956         * sched-deps.c (sched_analyze_2): Don't flush_pending_lists
7957         if DEBUG_INSN_P (insn).
7959         Reapply
7960         2013-05-31  Dinar Temirbulatov  <dinar@kugelworks.com>
7962         PR rtl-optimization/57268
7963         * sched-deps.c (sched_analyze_2): Flush dependence lists if
7964         the sum of the read and write lists exceeds MAX_PENDING_LIST_LENGTH.
7966 2013-06-03  Yuri Rumyantsev  <yuri.s.rumyantsev@intel.com>
7968         * config/i386/i386.c (ix86_lea_outperforms): Fix formatting.
7969         (ix86_avoid_lea_for_addr): Likewise.
7970         (exact_dependency_1): Likewise.
7971         (ix86_adjust_cost): Likewise.
7972         (swap_top_of_ready_list): Fix formatting and !reload_completed check
7973         removed.
7974         (do_reorder_for_imul): Fix typo, formatting and
7975         !reload_completed check removed.
7976         (ix86_sched_reorder): Fix typo and formatting.
7977         (fold_builtin_cpu): Move M_INTEL_SLM at the end of processor types
7978         list.
7980 2013-06-03  Sofiane Naci  <sofiane.naci@arm.com>
7982         * config/aarch64/aarch64.md (*movdi_aarch64): Define "simd" attribute.
7984 2013-06-03  Eric Botcazou  <ebotcazou@adacore.com>
7986         * varasm.c (output_constant) <CONSTRUCTOR>: Minor formatting tweak.
7987         <STRING_CST>: Likewise.
7988         <VECTOR_CST>: Likewise.
7990 2013-06-01  Janus Weil  <janus@gcc.gnu.org>
7991             Mikael Morin  <mikael@gcc.gnu.org>
7993         * configure.ac: Add AC_HEADER_TIOCGWINSZ macro.
7994         * config.in: Regenerated.
7995         * configure: Regenerated.
7997 2013-06-01  Jan Hubicka  <jh@suse.cz>
7999         PR middle-end/57366
8000         * cgraphunit.c (compile): When weakref is not supported,
8001         set up transparent aliases before final output pass.
8002         * varasm.c (assemble_alias): Do not try to do it here.
8004 2013-06-01  Jan Hubicka  <jh@suse.cz>
8006         PR middle-end/57467
8007         * passes.c (for_per_function): Skip unanalyzed functions.
8009 2013-06-01  Jan Hubicka  <jh@suse.cz>
8011         * lto-symtab.c (lto_symtab_merge_cgraph_nodes_1): Rename to ...
8012         (lto_symtab_merge_symbols_1): ... this one.
8013         (lto_symtab_merge_cgraph_nodes): Rename to ...
8014         (lto_symtab_merge_symbols): ... this one; simplify.
8015         * cgraph.c (same_body_aliases_done): Rename to ...
8016         (cpp_implicit_aliases_done): ... this one.
8017         (cgraph_create_function_alias): Update.
8018         (cgraph_same_body_alias): Update.
8019         (dump_cgraph_node): Remove alias dumping; simplify thunk dumping.
8020         (verify_edge_corresponds_to_fndecl): Simplify.
8021         * cgraph.h (symtab_node_base): Add cpp_implicit_alias, alias_target.
8022         (cgraph_node): Remove same_body_alias.
8023         (varpool_node): Remove alias_of and extra_name_alias.
8024         (same_body_aliases_done): Rename to ..
8025         (cpp_implicit_aliases_done): ... this one.
8026         (symtab_alias_ultimate_target): Add default parameter.
8027         (symtab_resolve_alias): New function.
8028         (fixup_same_cpp_alias_visibility): Declare.
8029         (cgraph_function_node): Add default parameter.
8030         (cgraph_node_asm_name): Likewise.
8031         (cgraph_function_or_thunk_node): Add default parameter; do
8032         not ICE when it is NULL.
8033         (varpool_variable_node): Likewise.
8034         * tree-emutls.c (create_emultls_var): Update.
8035         (ipa_lower_emutls): Update.
8036         * cgraphunit.c (cgraph_decide_is_function_needed): Update.
8037         (cgraph_reset_node): Reset alias info.
8038         (cgraph_finalize_function): Update.
8039         (fixup_same_cpp_alias_visibility): Move to symtab.c.
8040         (analyze_function): Simplify.
8041         (cgraph_process_same_body_aliases): Simplify.
8042         (analyze_functions): Fixup same body aliases.
8043         (handle_alias_pairs): Simplify.
8044         (assemble_thunk): Update.
8045         (assemble_thunks_and_aliases): Update.
8046         (output_weakrefs): Rewrite.
8047         * lto-cgraph.c (lto_output_node): Rewrite alias handling.
8048         (lto_output_varpool_node): Likewise.
8049         (compute_ltrans_boundary): Remve assert.
8050         (get_alias_symbol): New functoin.
8051         (input_node): Rewrite alias handling.
8052         (input_varpool_node): Likewise.
8053         * ipa-pure-const.c (propagate_pure_const): Fix formating.
8054         * ipa.c (process_references): Handle weakrefs correctly.
8055         (symtab_remove_unreachable_nodes): Likewise.
8056         * trans-mem.c (get_cg_data): Update.
8057         (ipa_tm_create_version_alias): Update.
8058         (ipa_tm_execute): Update.
8059         * symtab.c (dump_symtab_base): Dump aliases.
8060         (verify_symtab_base): Verify aliases.
8061         (symtab_node_availability): New function.
8062         (symtab_alias_ultimate_target): Simplify.
8063         (fixup_same_cpp_alias_visibility): Move here from cgraphunit.c;
8064         handle all the fixup cases.
8065         (symtab_resolve_alias): New function.
8066         * passes.c (ipa_write_summaries): Handle weakrefs.
8067         * varpool.c (varpool_analyze_node): Simplify.
8068         (assemble_aliases): Update.
8069         (varpool_create_variable_alias): Simplify.
8070         (varpool_extra_name_alias): Simplify.
8071         * lto-streamer.h (lto_symtab_merge_cgraph_nodes): Rename to...
8072         (lto_symtab_merge_symbols): ... this one.
8074 2013-06-01  Dinar Temirbulatov  <dinar@kugelworks.com>
8076         Revert
8077         PR rtl-optimization/57268
8078         * sched-deps.c (sched_analyze_2): Flush dependence lists if
8079         the sum of the read and write lists exceeds MAX_PENDING_LIST_LENGTH.
8081 2013-06-01  Tobias Burnus  <burnus@net-b.de>
8083         Partially reverted:
8084         2013-05-31  Tobias Burnus  <burnus@net-b.de>
8086         PR middle-end/57073
8087         * tree-ssa-math-opts.c (execute_cse_sincos): Move check
8088         further up.
8090 2013-05-31  Dinar Temirbulatov  <dinar@kugelworks.com>
8092         PR rtl-optimization/57268
8093         * sched-deps.c (sched_analyze_2): Flush dependence lists if
8094         the sum of the read and write lists exceeds MAX_PENDING_LIST_LENGTH.
8096 2013-05-31  Eric Botcazou  <ebotcazou@adacore.com>
8098         * config/rs6000/predicates.md (rs6000_cbranch_operator): Accept some
8099         unordered comparison operators when -fno-trapping-math is in effect
8100         on the e500.
8101         * config/rs6000/rs6000.c (rs6000_generate_compare): Remove dead code
8102         and implement unordered comparison operators properly on the e500.
8104 2013-05-31  Eric Botcazou  <ebotcazou@adacore.com>
8106         * simplify-rtx.c (simplify_byte_swapping_operation): Use proper macro
8107         for constant scalar integers.
8108         (simplify_relational_operation_1): Likewise.
8110 2013-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
8112         * config/rs6000/rs6000-opts.h (enum processor_type): Reorder.
8113         * config/rs6000/rs6000.md (cpu): Reorder.  Split long line.
8114         Fix comment.
8116 2013-05-31  Yuri Rumyantsev  <yuri.s.rumyantsev@intel.com>
8117             Igor Zamyatin  <igor.zamyatin@intel.com>
8119         Silvermont (SLM) architecture performance tuning.
8120         * config/i386/i386.h (enum ix86_tune_indices): Add
8121         X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS.
8122         (TARGET_SPLIT_MEM_OPND_FOR_FP_CONVERTS): New define.
8124         * config/i386/i386.c (initial_ix86_tune_features)
8125         <X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS>: Initialize.
8126         (ix86_lea_outperforms): Handle Silvermont tuning.
8127         (ix86_avoid_lea_for_add): Add new argument to ix86_lea_outperforms
8128         call.
8129         (ix86_use_lea_for_mov): Likewise.
8130         (ix86_avoid_lea_for_addr): Likewise.
8131         (ix86_lea_for_add_ok): Likewise.
8132         (exact_dependency_1): New function.
8133         (exact_store_load_dependency): Likewise.
8134         (ix86_adjust_cost): Handle Silvermont tuning.
8135         (do_reoder_for_imul): Likewise.
8136         (swap_top_of_ready_list): New function.
8137         (ix86_sched_reorder): Changed to handle Silvermont tuning.
8139         * config/i386/i386.md (peepholes that split memory operand in fp
8140         converts): New.
8142 2013-05-31  Marcus Shawcroft  <marcus.shawcroft@arm.com>
8144         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
8145         Remove un-necessary braces.
8147 2013-05-31  Marcus Shawcroft  <marcus.shawcroft@arm.com>
8149         * config/aarch64/aarch64.c (aarch64_classify_symbol):
8150         Use SYMBOL_TINY_ABSOLUTE for AARCH64_CMODEL_TINY_PIC.
8152 2013-05-31  Tobias Burnus  <burnus@net-b.de>
8154         PR middle-end/57073
8155         * tree-ssa-math-opts.c (execute_cse_sincos): Move check further up.
8157 2013-05-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8159         PR target/56315
8160         * config/arm/arm.c (const_ok_for_dimode_op): Handle IOR.
8161         * config/arm/arm.md (*iordi3_insn): Change to insn_and_split.
8162         * config/arm/neon.md (iordi3_neon): Remove.
8163         (neon_vorr<mode>): Generate iordi3 instead of iordi3_neon.
8164         * config/arm/predicates.md (imm_for_neon_logic_operand):
8165         Move to earlier in the file.
8166         (neon_logic_op2): Likewise.
8167         (arm_iordi_operand_neon): New predicate.
8169 2013-05-31  Richard Biener  <rguenther@suse.de>
8171         PR tree-optimization/57478
8172         PR tree-optimization/57453
8173         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Uses in PHI nodes
8174         are life as well.
8176 2013-05-31  Kaushik Phatak  <kaushik.phatak@kpitcummins.com>
8178         * config/rl78/rl78.md (mulqi3,mulhi3): New define_expands.
8179         (*mulqi3_rl78,*mulhi3_rl78,*mulhi3_g13): New define_insns.
8181 2013-05-30  Tobias Burnus  <burnus@net-b.de>
8182             Thomas Koenig  <tkoenig@gcc.gnu.org>
8184         PR middle-end/57073
8185         * tree-ssa-math-opts.c (execute_cse_sincos): Optimize
8186         powi (-1.0, k) to (k & 1) ? -1.0 : 1.0.
8188 2013-05-30  Steven Bosscher  <steven@gcc.gnu.org>
8190         * rtlanal.c (tablejump_p): Expect table and label to be adjacent.
8192 2013-05-30  Vladimir Makarov  <vmakarov@redhat.com>
8194         * target.def (register_usage_leveling_p): New hook.
8195         * targhooks.c (default_register_usage_leveling_p): New.
8196         * targhooks.h (default_register_usage_leveling_p): New prototype.
8197         * lra-assigns.c (register_usage_leveling_p): Use the hook.
8198         * doc/tm.texi.in (TARGET_REGISTER_USAGE_LEVELING_P): New hook.
8199         * doc/tm.texi: Update.
8200         * config/i386/i386.c (TARGET_REGISTER_USAGE_LEVELING_P): Define.
8202 2013-05-30  Ian Bolton  <ian.bolton@arm.com>
8204         * config/aarch64/aarch64.md (insv<mode>): New define_expand.
8205         (*insv_reg<mode>): New define_insn.
8207 2013-05-30  Joern Rennecke  <joern.rennecke@embecosm.com>
8209         PR rtl-optimization/57439
8210         * postreload.c (move2add_valid_value_p): Check that we have
8211         a zero subreg_regno_offset when accessing the register in
8212         the requested mode.
8214 2013-05-30  Yuri Rumyantsev  <yuri.s.rumyantsev@intel.com>
8215             Igor Zamyatin  <igor.zamyatin@intel.com>
8217         Silvermont (SLM) architecture pipeline model, tuning and
8218         insn selection.
8219         * config.gcc: Add slm config options and target.
8221         * config/i386/slm.md: New.
8223         * config/i386/driver-i386.c (host_detect_local_cpu): Check movbe.
8225         * gcc/config/i386/i386-c.c (ix86_target_macros_internal):  New case
8226         PROCESSOR_SLM.
8227         (ix86_target_macros_internal): Likewise.
8229         * gcc/config/i386/i386.c (slm_cost): New cost.
8230         (m_SLM): New macro flag.
8231         (initial_ix86_tune_features): Set m_SLM.
8232         (x86_accumulate_outgoing_args): Likewise.
8233         (x86_arch_always_fancy_math_387): Likewise.
8234         (processor_target_table): Add slm cost.
8235         (cpu_names): Add slm cpu name.
8236         (x86_option_override_internal): Set SLM ISA.
8237         (ix86_issue_rate): New case PROCESSOR_SLM.
8238         (ia32_multipass_dfa_lookahead): Likewise.
8239         (fold_builtin_cpu): Add slm.
8241         * config/i386/i386.h (TARGET_SLM): New target macro.
8242         (target_cpu_default): Add TARGET_CPU_DEFAULT_slm.
8243         (processor_type): Add PROCESSOR_SLM.
8245         * config/i386/i386.md (cpu): Add new value "slm".
8246         (slm.md): Include slm.md.
8248 2013-05-30  Bernd Schmidt  <bernds@codesourcery.com>
8249             Zhenqiang Chen  <zhenqiang.chen@linaro.org>
8251         * config/arm/arm-protos.h: Add and update function protos.
8252         * config/arm/arm.c (use_simple_return_p): New added.
8253         (thumb2_expand_return): Check simple_return flag.
8254         * config/arm/arm.md: Add simple_return and conditional simple_return.
8255         * config/arm/iterators.md: Add iterator for return and simple_return.
8257 2013-05-30  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
8259         * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): New added.
8260         (arm_emit_multi_reg_pop): Add REG_CFA_ADJUST_CFA notes.
8261         (arm_emit_vfp_multi_reg_pop): Likewise.
8262         (thumb2_emit_ldrd_pop): Likewise.
8263         (arm_expand_epilogue): Add misc REG_CFA notes.
8264         (arm_unwind_emit): Skip REG_CFA_ADJUST_CFA and REG_CFA_RESTORE.
8266 2013-05-29  Lawrence Crowl  <crowl@google.com>
8268         * config/arm/t-arm: Update for below.
8270         * config/arm/arm.c (arm_libcall_uses_aapcs_base::libcall_htab):
8271         Change type to hash_table.  Update dependent calls and types.
8273         * config/i386/t-cygming: Update for below.
8275         * config/i386/t-interix: Update for below.
8277         * config/i386/winnt.c (i386_pe_section_type_flags::htab):
8278         Change type to hash_table.  Update dependent calls and types.
8279         (i386_find_on_wrapper_list::wrappers): Likewise.
8281         * config/ia64/t-ia64: Update for below.
8283         * config/ia64/ia64.c (bundle_state_table):
8284         Change type to hash_table.  Update dependent calls and types.
8286         * config/mips/mips.c (mips_reorg_process_insns::htab):
8287         Change type to hash_table.  Update dependent calls and types.
8289         * config/sol2.c (solaris_comdat_htab):
8290         Change type to hash_table.  Update dependent calls and types.
8292         * config/t-sol2: Update for above.
8294 2013-05-29  Teresa Johnson  <tejohnson@google.com>
8296         * passes.c (dump_passes): Use FOR_EACH_FUNCTION since
8297         functions are not yet marked as defined.
8299 2013-05-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
8300             Pat Haugen  <pthaugen@us.ibm.com>
8301             Peter Bergner  <bergner@vnet.ibm.com>
8303         * config/rs6000/vector.md (VEC_I): Add support for new power8 V2DI
8304         instructions.
8305         (VEC_A): Likewise.
8306         (VEC_C): Likewise.
8307         (vrotl<mode>3): Likewise.
8308         (vashl<mode>3): Likewise.
8309         (vlshr<mode>3): Likewise.
8310         (vashr<mode>3): Likewise.
8312         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8313         support for power8 V2DI builtins.
8315         * config/rs6000/rs6000-builtin.def (abs_v2di): Add support for
8316         power8 V2DI builtins.
8317         (vupkhsw): Likewise.
8318         (vupklsw): Likewise.
8319         (vaddudm): Likewise.
8320         (vminsd): Likewise.
8321         (vmaxsd): Likewise.
8322         (vminud): Likewise.
8323         (vmaxud): Likewise.
8324         (vpkudum): Likewise.
8325         (vpksdss): Likewise.
8326         (vpkudus): Likewise.
8327         (vpksdus): Likewise.
8328         (vrld): Likewise.
8329         (vsld): Likewise.
8330         (vsrd): Likewise.
8331         (vsrad): Likewise.
8332         (vsubudm): Likewise.
8333         (vcmpequd): Likewise.
8334         (vcmpgtsd): Likewise.
8335         (vcmpgtud): Likewise.
8336         (vcmpequd_p): Likewise.
8337         (vcmpgtsd_p): Likewise.
8338         (vcmpgtud_p): Likewise.
8339         (vupkhsw): Likewise.
8340         (vupklsw): Likewise.
8341         (vaddudm): Likewise.
8342         (vmaxsd): Likewise.
8343         (vmaxud): Likewise.
8344         (vminsd): Likewise.
8345         (vminud): Likewise.
8346         (vpksdss): Likewise.
8347         (vpksdus): Likewise.
8348         (vpkudum): Likewise.
8349         (vpkudus): Likewise.
8350         (vrld): Likewise.
8351         (vsld): Likewise.
8352         (vsrad): Likewise.
8353         (vsrd): Likewise.
8354         (vsubudm): Likewise.
8356         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Add
8357         support for power8 V2DI instructions.
8359         * config/rs6000/altivec.md (UNSPEC_VPKUHUM): Add support for
8360         power8 V2DI instructions.  Combine pack and unpack insns to use an
8361         iterator for each mode.  Check whether a particular mode supports
8362         Altivec instructions instead of just checking TARGET_ALTIVEC.
8363         (UNSPEC_VPKUWUM): Likewise.
8364         (UNSPEC_VPKSHSS): Likewise.
8365         (UNSPEC_VPKSWSS): Likewise.
8366         (UNSPEC_VPKUHUS): Likewise.
8367         (UNSPEC_VPKSHUS): Likewise.
8368         (UNSPEC_VPKUWUS): Likewise.
8369         (UNSPEC_VPKSWUS): Likewise.
8370         (UNSPEC_VPACK_SIGN_SIGN_SAT): Likewise.
8371         (UNSPEC_VPACK_SIGN_UNS_SAT): Likewise.
8372         (UNSPEC_VPACK_UNS_UNS_SAT): Likewise.
8373         (UNSPEC_VPACK_UNS_UNS_MOD): Likewise.
8374         (UNSPEC_VUPKHSB): Likewise.
8375         (UNSPEC_VUNPACK_HI_SIGN): Likewise.
8376         (UNSPEC_VUNPACK_LO_SIGN): Likewise.
8377         (UNSPEC_VUPKHSH): Likewise.
8378         (UNSPEC_VUPKLSB): Likewise.
8379         (UNSPEC_VUPKLSH): Likewise.
8380         (VI2): Likewise.
8381         (VI_char): Likewise.
8382         (VI_scalar): Likewise.
8383         (VI_unit): Likewise.
8384         (VP): Likewise.
8385         (VP_small): Likewise.
8386         (VP_small_lc): Likewise.
8387         (VU_char): Likewise.
8388         (add<mode>3): Likewise.
8389         (altivec_vaddcuw): Likewise.
8390         (altivec_vaddu<VI_char>s): Likewise.
8391         (altivec_vadds<VI_char>s): Likewise.
8392         (sub<mode>3): Likewise.
8393         (altivec_vsubcuw): Likewise.
8394         (altivec_vsubu<VI_char>s): Likewise.
8395         (altivec_vsubs<VI_char>s): Likewise.
8396         (altivec_vavgs<VI_char>): Likewise.
8397         (altivec_vcmpbfp): Likewise.
8398         (altivec_eq<mode>): Likewise.
8399         (altivec_gt<mode>): Likewise.
8400         (altivec_gtu<mode>): Likewise.
8401         (umax<mode>3): Likewise.
8402         (smax<mode>3): Likewise.
8403         (umin<mode>3): Likewise.
8404         (smin<mode>3): Likewise.
8405         (altivec_vpkuhum): Likewise.
8406         (altivec_vpkuwum): Likewise.
8407         (altivec_vpkshss): Likewise.
8408         (altivec_vpkswss): Likewise.
8409         (altivec_vpkuhus): Likewise.
8410         (altivec_vpkshus): Likewise.
8411         (altivec_vpkuwus): Likewise.
8412         (altivec_vpkswus): Likewise.
8413         (altivec_vpks<VI_char>ss): Likewise.
8414         (altivec_vpks<VI_char>us): Likewise.
8415         (altivec_vpku<VI_char>us): Likewise.
8416         (altivec_vpku<VI_char>um): Likewise.
8417         (altivec_vrl<VI_char>): Likewise.
8418         (altivec_vsl<VI_char>): Likewise.
8419         (altivec_vsr<VI_char>): Likewise.
8420         (altivec_vsra<VI_char>): Likewise.
8421         (altivec_vsldoi_<mode>): Likewise.
8422         (altivec_vupkhsb): Likewise.
8423         (altivec_vupkhs<VU_char>): Likewise.
8424         (altivec_vupkls<VU_char>): Likewise.
8425         (altivec_vupkhsh): Likewise.
8426         (altivec_vupklsb): Likewise.
8427         (altivec_vupklsh): Likewise.
8428         (altivec_vcmpequ<VI_char>_p): Likewise.
8429         (altivec_vcmpgts<VI_char>_p): Likewise.
8430         (altivec_vcmpgtu<VI_char>_p): Likewise.
8431         (abs<mode>2): Likewise.
8432         (vec_unpacks_hi_v16qi): Likewise.
8433         (vec_unpacks_hi_v8hi): Likewise.
8434         (vec_unpacks_lo_v16qi): Likewise.
8435         (vec_unpacks_hi_<VP_small_lc>): Likewise.
8436         (vec_unpacks_lo_v8hi): Likewise.
8437         (vec_unpacks_lo_<VP_small_lc>): Likewise.
8438         (vec_pack_trunc_v8h): Likewise.
8439         (vec_pack_trunc_v4si): Likewise.
8440         (vec_pack_trunc_<mode>): Likewise.
8442         * config/rs6000/altivec.h (vec_vaddudm): Add defines for power8
8443         V2DI builtins.
8444         (vec_vmaxsd): Likewise.
8445         (vec_vmaxud): Likewise.
8446         (vec_vminsd): Likewise.
8447         (vec_vminud): Likewise.
8448         (vec_vpksdss): Likewise.
8449         (vec_vpksdus): Likewise.
8450         (vec_vpkudum): Likewise.
8451         (vec_vpkudus): Likewise.
8452         (vec_vrld): Likewise.
8453         (vec_vsld): Likewise.
8454         (vec_vsrad): Likewise.
8455         (vec_vsrd): Likewise.
8456         (vec_vsubudm): Likewise.
8457         (vec_vupkhsw): Likewise.
8458         (vec_vupklsw): Likewise.
8460 2013-05-29  Jan Hubicka  <jh@suse.cz>
8462         * cgraph.h (symtab_node_base): Add definition, alias and analyzed
8463         flags; reorder rest of fields in more consistent way.
8464         (varpool_node): Remove analyzed, finalized and alias.
8465         (cgraph_ndoe): Likewise.
8466         (symtab_alias_ultimate_target): New function.
8467         (cgraph_function_node): Move offline.
8468         (cgraph_reset_node): Declare.
8469         (cgraph_comdat_can_be_unshared_p): Remove.
8470         (varpool_remove_initializer): Declare.
8471         (varpool_first_defined_variable, varpool_next_defined_variable
8472         cgraph_first_defined_function, cgraph_next_defined_function): Update.
8473         (cgraph_function_with_gimple_body_p): Update.
8474         (varpool_all_refs_explicit_p): Update.
8475         (symtab_alias_target): New function.
8476         (cgraph_alias_aliased_node, varpool_alias_aliased_node): Rename to ...
8477         (cgraph_alias_target, varpool_alias_target): .. this one; simplify.
8478         (cgraph_function_or_thunk_node): Simplify using
8479         symtab_alias_ultimate_target.
8480         (varpool_variable_node): Likewise.
8481         * cgraph.c (cgraph_create_function_alias): Update.
8482         (cgraph_add_thunk): Update.
8483         (cgraph_remove_node): Update.
8484         (dump_cgraph_node): Do not dump removed flags.
8485         (cgraph_function_body_availability): Update.
8486         (cgraph_propagate_frequency): Update.
8487         (verify_cgraph_node): Check sanity of local flag.
8488         (cgraph_function_node): Move here from cgraph.h; revamp for
8489         cgraph_function_or_thunk_node.
8490         * lto-symtab.c (lto_varpool_replace_node): Update.
8491         (lto_symtab_resolve_can_prevail_p): Update.
8492         (lto_symtab_merge_cgraph_nodes): Update.
8493         * ipa-cp.c (determine_versionability, initialize_node_lattices,
8494         propagate_constants_accross_call, devirtualization_time_bonus,
8495         ipcp_propagate_stage): Update.
8496         * tree-emutls.c (create_emultls_var, ipa_lower_emutls): Update.
8497         * ipa-inline-transform.c (clone_inlined_nodes,
8498         preserve_function_body_p): Update.
8499         * ipa-reference.c (propagate): Update.
8500         (write_node_summary_p): Update.
8501         * toplev.c (wrapup_global_declaration_2): Update.
8502         * cgraphunit.c (cgraph_analyze_function): Rename to ...
8503         (analyze_function) ... this one.
8504         (cgraph_process_new_functions): Update.
8505         (cgraph_reset_node): Export.
8506         (cgraph_finalize_function): Update.
8507         (cgraph_add_new_function): Update.
8508         (process_function_and_variable_attributes): Update.
8509         (varpool_finalize_decl): Update.
8510         (symbol_finalized): Remove.
8511         (symbol_finalized_and_needed): Rename to ...
8512         (symbol_defined_and_needed): ... update.
8513         (cgraph_analyze_functions): Update.
8514         (handle_alias_pairs): Update.
8515         (mark_functions_to_output): Update.
8516         (assemble_thunk): Update.
8517         (output_in_order): Update.
8518         (output_weakrefs): Update.
8519         (finalize_compilation_unit): Update.
8520         * lto-cgraph.c (reachable_from_other_partition_p, lto_output_node,
8521         lto_output_varpool_node, compute_ltrans_boundary, input_overwrite_node,
8522         input_node, input_varpool_node): Update.
8523         * dbxout.c (dbxout_expand_expr): Update.
8524         * cgraphclones.c (cgraph_clone_node): Update.
8525         (cgraph_copy_node_for_versioning): Update.
8526         (cgraph_materialize_clone): Update.
8527         (cgraph_materialize_all_clones): Update.
8528         * ipa-pure-const.c (analyze_function, pure_const_write_summary,
8529         propagate_pure_const, propagate_nothrow): Update.
8530         * lto-streamer-out.c (lto_output, write_symbol): Update.
8531         * ipa-utils.c (ipa_reverse_postorder): Update.
8532         * ipa-inline.c (can_inline_edge_p): Update.
8533         (update_caller_keys, ipa_inline): Update.
8534         * dwarf2out.c (reference_to_unused,
8535         premark_types_used_by_global_vars_helper): Update.
8536         * tree-eh.c (tree_could_trap_p): Update.
8537         * ipa-split.c (consider_split, execute_split_functions): Update.
8538         * ipa.c (cgraph_non_local_node_p_1, cgraph_local_node_p,
8539         has_addr_references_p): Update; move ahead in file for better
8540         readability.
8541         (process_references): Simplify.
8542         (symtab_remove_unreachable_nodes): Update; cleanup way function/var
8543         bodies are removed.
8544         (cgraph_comdat_can_be_unshared_p): Make static.
8545         (cgraph_externally_visible_p): Update.
8546         (varpool_externally_visible_p): Update.
8547         (function_and_variable_visibility): Update.
8548         * trans-mem.c (get_cg_data, ipa_tm_mayenterirr_function,
8549         ipa_tm_mark_force_output_node): Update.
8550         * ipa-inline-analysis.c (dump_inline_summary, initialize_inline_failed,
8551         estimate_edge_devirt_benefit, inline_generate_summary,
8552         inline_write_summary): Update.
8553         * gimple-fold.c (can_refer_decl_in_current_unit_p): Update.
8554         * ipa-prop.c (ipa_compute_jump_functions): Update.
8555         (ipa_print_node_params, ipa_prop_read_section,
8556         ipa_update_after_lto_read, read_replacements_section): Update.
8557         * varasm.c (mark_decl_referenced): Update.
8558         (assemble_alias, dump_tm_clone_pairs): Update.
8559         * tree-inline.c (copy_bb): Update.
8560         (estimate_num_insns, optimize_inline_calls, tree_function_versioning):
8561         Update.
8562         * symtab.c (dump_symtab_base): Print new flags.
8563         (verify_symtab_base): Verify new flags.
8564         (symtab_alias_ultimate_target): New function.
8565         * tree-ssa-structalias.c (get_constraint_for_ssa_var,
8566         create_variable_info_for, associate_varinfo_to_alias, ipa_pta_execute):
8567         Update.
8568         * passes.c (ipa_write_summaries, ipa_write_optimization_summaries):
8569         Update.
8570         * i386.c (ix86_get_function_versions_dispatcher,
8571         ix86_generate_version_dispatcher_body): Update.
8572         (fold_builtin_cpu): Use varpool_add_new_variable.
8573         * varpool.c (varpool_remove_initializer): Break out from ...
8574         (varpool_remove_node): ... this one.
8575         (dump_varpool_node, varpool_node_for_asm,
8576         cgraph_variable_initializer_availability, varpool_analyze_node,
8577         varpool_assemble_decl, varpool_remove_unreferenced_decls,
8578         varpool_finalize_named_section_flags, varpool_create_variable_alias):
8579         Update.
8581 2013-05-29  Jan Hubicka  <jh@suse.cz>
8583         * passes.c (init_optimization_passes): Move OMP expansion into lowering.
8585 2013-05-29  Easwaran Raman  <eraman@google.com>
8587         PR tree-optimization/57442
8588         * tree-ssa-reassoc.c (appears_later_in_bb): Return correct value
8589         when control exits the main loop.
8591 2013-05-29  Sandeep Kumar Singh  <Sandeep.Singh2@kpitcummins.com>
8593         * rx/rx.h (TARGET_CPU_CPP_BUILTINS): Add macros for RX100, RX200,
8594         and RX600.
8595         * rx/rx.opt: Add macro for rx100 with string rx100 and value RX100.
8596         * rx/rx-opts.h (rx_cpu_types): Add new cpu type rx100.
8597         * rx/t-rx: Add rx100 under multi library matches option for nofpu
8598         option.
8600 2013-05-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8602         PR tree-optimization/57441
8603         * gimple-ssa-strength-reduction.c (analyze_candidates_and_replace):
8604         Don't limit size of incr_vec to number of candidates.
8606 2013-05-29  Steve Ellcey  <sellcey@imgtec.com>
8608         * config/mips/mti-linux.h (SYSROOT_SUFFIX_SPEC): Add micromips
8609         and mips16 directories.
8610         * config/mips/t-mti-linux (MULTILIB_OPTIONS): Add micromips and mips16.
8611         (MULTILIB_DIRNAMES): Ditto.
8612         (MULTILIB_EXCEPTIONS): Add new exceptions.
8613         * config/mips/t-mti-elf (MULTILIB_OPTIONS): Add micromips.
8614         (MULTILIB_DIRNAMES): Ditto.
8615         (MULTILIB_EXCEPTIONS): Add new exceptions.
8617 2012-05-29  Chris Schlumberger-Socha  <chris.schlumberger-socha@arm.com>
8618             Marcus Shawcroft  <marcus.shawcroft@arm.com>
8620         * config/aarch64/aarch64-protos.h (aarch64_symbol_type): Define
8621         SYMBOL_TINY_ABSOLUTE.
8622         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Handle
8623         SYMBOL_TINY_ABSOLUTE.
8624         (aarch64_expand_mov_immediate): Likewise.
8625         (aarch64_classify_symbol): Likewise.
8626         (aarch64_mov_operand_p): Remove ATTRIBUTE_UNUSED.
8627         Permit SYMBOL_TINY_ABSOLUTE.
8628         * config/aarch64/predicates.md (aarch64_mov_operand): Permit CONST.
8630 2013-05-29  Chris Schlumberger-Socha  <chris.schlumberger-socha@arm.com>
8631             Marcus Shawcroft  <marcus.shawcroft@arm.com>
8633         * config/aarch64/aarch64.c (aarch64_classify_symbol): Remove comment.
8634         Refactor if/switch.  Replace gcc_assert with if.
8636 2013-05-29  Ganesh Gopalasubramanian  <Ganesh.Gopalasubramanian@amd.com>
8638         * config/i386/i386.c (initial_ix86_tune_features): Enable
8639         FP Reassociation for AMD bdver1 and bdver2.
8641 2013-05-29  Martin Jambor  <mjambor@suse.cz>
8643         * tree-cfg.c (verify_expr): Verify that BIT_FIELD_REF, REALPART_EXPR
8644         and IMAGPART_EXPR do not occur within other handled_components.
8646 2013-05-29  Richard Biener  <rguenther@suse.de>
8648         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Guard vinfo
8649         access on whether the use is in the BB we currently try to
8650         vectorize.
8651         (vect_bb_vectorization_profitable_p): Pass the BB we currently
8652         vectorize to vect_bb_slp_scalar_cost.
8654 2013-05-29  Richard Biener  <rguenther@suse.de>
8656         * tree-vect-slp.c (vect_bb_slp_scalar_cost): New function
8657         computing scalar cost offsetted by stmts that are kept live
8658         by scalar uses.
8659         (vect_bb_vectorization_profitable_p): Use vect_bb_slp_scalar_cost
8660         for computation of scalar cost.
8662 2013-05-28  Steve Ellcey  <sellcey@mips.com>
8664         * config/mips/mips-cpus.def (mips32r2): Change processor type.
8666 2013-05-28  Balaji V. Iyer  <balaji.v.iyer@intel.com>
8668         * doc/extend.texi (C Extensions): Added documentation about Cilk Plus
8669         array notation built-in reduction functions.
8670         * doc/passes.texi (Passes): Added documentation about changes done
8671         for Cilk Plus.
8672         * doc/invoke.texi (C Dialect Options): Added documentation about
8673         the -fcilkplus flag.
8674         * Makefile.in (C_COMMON_OBJS): Added c-family/array-notation-common.o.
8675         (BUILTINS_DEF): Depend on cilkplus.def.
8676         * builtins.def: Include cilkplus.def.  Define DEF_CILKPLUS_BUILTIN.
8677         * builtin-types.def: Define BT_FN_INT_PTR_PTR_PTR.
8678         * cilkplus.def: New file.
8680 2013-05-28  Joern Rennecke  <joern.rennecke@embecosm.com>
8682         PR rtl-optimization/57439
8683         * postreload.c (move2add_use_add2_insn): Use gen_lowpart_common.
8685 2013-05-28  Easwaran Raman  <eraman@google.com>
8687         PR tree-optimization/57337
8688         * tree-ssa-reassoc.c (appears_later_in_bb): New function.
8689         (find_insert_point): Correctly identify the insertion point
8690         when two statements with the same UID is compared.
8692 2013-05-28  Richard Biener  <rguenther@suse.de>
8694         PR tree-optimization/56787
8695         * tree-vect-data-refs.c (vect_analyze_data_refs): Drop clobbers
8696         from the list of data references.
8697         * tree-vect-loop.c (vect_determine_vectorization_factor): Skip
8698         clobbers.
8699         (vect_analyze_loop_operations): Likewise.
8700         (vect_transform_loop): Remove clobbers.
8702 2013-05-28  Martin Jambor  <mjambor@suse.cz>
8704         * tree-cfg.c (verify_expr): Verify that BIT_FIELD_REFs, IMAGPART_EXPRs
8705         and REALPART_EXPRs have scalar type.
8707 2013-05-28  Richard Biener  <rguenther@suse.de>
8709         PR tree-optimization/57411
8710         * tree-ssa-copy.c (may_propagate_copy): Cannot propagate
8711         virtual operands.
8712         * tree-ssa-dom.c (eliminate_const_or_copy): Special-case
8713         virtual operand propagation.
8715 2013-05-28  Eric Botcazou  <ebotcazou@adacore.com>
8717         * config/sparc/sparc.c (sparc_expand_vec_perm_bmask): Use %g0 as
8718         destination register for bmasksi_vis.
8719         (vector_init_bshuffle): Likewise.
8720         * config/sparc/sparc.md (vec_perm_constv8qi): Likewise.
8722 2013-05-28  Eric Botcazou  <ebotcazou@adacore.com>
8724         * doc/invoke.texi (SPARC Options): Document -mfix-ut699.
8725         * builtins.c (expand_builtin_mathfn) <BUILT_IN_SQRT>: Try to widen the
8726         mode if the instruction isn't available in the original mode.
8727         * config/sparc/sparc.opt (mfix-ut699): New option.
8728         * config/sparc/sparc.md (muldf3_extend): Disable if -mfix-ut699.
8729         (divdf3): Turn into expander.
8730         (divdf3_nofix): New insn.
8731         (divdf3_fix): Likewise.
8732         (divsf3): Disable if -mfix-ut699.
8733         (sqrtdf2): Turn into expander.
8734         (sqrtdf2_nofix): New insn.
8735         (sqrtdf2_fix): Likewise.
8736         (sqrtsf2): Disable if -mfix-ut699.
8738 2013-05-27  Richard Biener  <rguenther@suse.de>
8740         PR middle-end/57412
8741         * omp-low.c (expand_omp_atomic_pipeline): Use the correct latch
8742         block for the new loop.
8744 2013-05-27  Richard Biener  <rguenther@suse.de>
8746         PR tree-optimization/57343
8747         * tree-ssa-loop-niter.c (number_of_iterations_ne_max): Do not
8748         use multiple_of_p if not TYPE_OVERFLOW_UNDEFINED.
8749         (number_of_iterations_cond): Do not build the folded tree.
8751 2013-05-27  Richard Biener  <rguenther@suse.de>
8753         Revert
8754         PR middle-end/57381
8755         * fold-const.c (operand_equal_p): Compare FIELD_DECLs with
8756         OEP_CONSTANT_ADDRESS_OF retained.
8758         PR tree-optimization/57417
8759         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Fix test
8760         for unchanged base.
8761         (set_ssa_val_to): Compare addresses using
8762         get_addr_base_and_unit_offset.
8764 2013-05-27  Joern Rennecke  <joern.rennecke@embecosm.com>
8766         PR rtl-optimization/56833
8767         * postreload.c (move2add_record_mode): New function.
8768         (move2add_record_sym_value, move2add_valid_value_p): Likewise.
8769         (move2add_use_add2_insn): Use move2add_record_sym_value.
8770         (move2add_use_add3_insn): Likewise.
8771         (reload_cse_move2add): Use move2add_valid_value_p and
8772         move2add_record_mode.  Invalidate call-clobbered and REG_INC
8773         affected regs by setting reg_mode to VOIDmode.
8774         (move2add_note_store): Don't pretend the inside of a SUBREG is
8775         the actual destination.  Invalidate single/leading registers by
8776         setting reg_mode to VOIDmode.
8777         Use move2add_record_sym_value, move2add_valid_value_p and
8778         move2add_record_mode.
8780 2013-05-27  Richard Biener  <rguenther@suse.de>
8782         PR tree-optimization/57396
8783         * tree-affine.c (double_int_constant_multiple_p): Properly
8784         return false for val == 0 and div != 0.
8786 2013-05-25  Richard Sandiford  <rdsandiford@googlemail.com>
8788         * config/mips/mips.h: Use #elif in preprocessor conditions.
8790 2013-05-25  Richard Sandiford  <rdsandiford@googlemail.com>
8792         PR target/53916
8793         * config/mips/constraints.md (kl): New constraint.
8794         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Delete.
8795         (divmod<mode>4_internal): Rename to divmod<mode>4.  Use "kl" as the
8796         constraint for operand 0.  Split after CSE for MIPS16.  Emit a move
8797         from LO for MIPS16.
8798         (udivmod<mode>4_internal): Likewise udivmod<mode>4.
8800 2013-05-25  Richard Sandiford  <rdsandiford@googlemail.com>
8802         PR target/55777
8803         * config/mips/mips.c (mips_can_inline_p): New function.
8804         (TARGET_CAN_INLINE_P): Define.
8806 2013-05-25  Steven Bosscher  <steven@gcc.gnu.org>
8808         * sched-int.h (ds_t, dw_t): Make unsigned int.
8809         Fix documentation that describes how all the ds_t bits are used.
8810         Reserve the last bit for delayed-branch scheduling.
8811         (BITS_PER_DEP_STATUS): Move to ds_t typedef.
8812         (BITS_PER_DEP_WEAK): Fix definition and documentation.
8813         (gen_dep_weak_1): Remove prototype.
8814         * sched-deps.c (get_dep_weak_1): Make static.
8815         * target.def (speculate_insn, needs_block_p, gen_spec_check,
8816         get_insn_spec_ds, get_insn_checked_ds): Adjust hook prototypes.
8817         * doc/tm.texi: Regenerate.
8818         * config/ia64/ia64.c (ia64_needs_block_p): Update prototype.
8820 2013-05-24  Steven Bosscher  <steven@gcc.gnu.org>
8822         PR debug/56950
8823         * haifa-sched.c (sched_extend_bb): Ignore DEBUG_INSNs.
8825 2013-05-24  Nathan Sidwell  <nathan@codesourcery.com>
8826             Sandra Loosemore  <sandra@codesourcery.com>
8828         * config.gcc (powerpc-*): Allow native for with-cpu.
8830 2013-05-24  Jeff Law  <law@redhat.com>
8832         PR tree-optimization/57124
8833         * tree-vrp.c (simplify_cond_using_ranges): Only simplify a
8834         conversion feeding a condition if the range has an overflow
8835         if -fstrict-overflow.  Add warnings for when we do make the
8836         transformation.
8838 2013-05-24  Dehao Chen  <dehao@google.com>
8840         * tree-cfg.c (locus_discrim_map): Fix the typo.
8841         (locus_discrim_hasher): Likewise.
8842         (locus_discrim_hasher::hash): Likewise.
8843         (locus_discrim_hasher::equal): Likewise.
8845 2013-05-24  Martin Jambor  <mjambor@suse.cz>
8847         PR tree-optimization/57294
8848         * cgraph.h (ipa_record_stmt_references): Declare.
8849         * cgraphbuild.c (ipa_record_stmt_references): New function.
8850         (build_cgraph_edges): Use ipa_record_stmt_references.
8851         (rebuild_cgraph_edges): Likewise.
8852         (cgraph_rebuild_references): Likewise.
8853         * ipa-prop.c (ipa_modify_call_arguments): Discard references
8854         associated with the old statement and build references from the
8855         newly built statements.
8856         * ipa-ref.c (ipa_remove_stmt_references): New function.
8857         * ipa-ref.h (ipa_remove_stmt_references): Declare.
8859 2013-05-24  Vladimir Makarov  <vmakarov@redhat.com>
8861         * lra-constraints.c (emit_spill_move): Use smaller mode for
8862         mem-mem moves.
8863         (check_and_process_move): Consider mem-reg moves for secondary
8864         too.
8865         (curr_insn_transform): Don't lose insns emitted before for
8866         secondary memory moves.
8867         (inherit_in_ebb): Mark defined reg.  Add usage only if it is not a
8868         reg set up in the current insn.
8870 2013-05-24  Dehao Chen  <dehao@google.com>
8872         * tree-cfg.c (locus_descrim_hasher::hash): Change discriminator
8873         hash function.
8874         (locus_descrim_hasher::equal): Likewise.
8875         (build_gimple_cfg): New discriminator assignment algorithm.
8876         (make_edges): Likewise.
8877         (next_discriminator_for_locus): Likewise.
8878         (same_line_p): Likewise.
8879         (assign_discriminators): Likewise.
8880         (make_cond_expr_edges): Likewise.
8881         (make_gimple_switch_edges): Likewise.
8882         (make_goto_expr_edges): Likewise.
8883         (make_gimple_asm_edges): Likewise.
8885 2013-05-24  Ian Bolton  <ian.bolton@arm.com>
8887         * config/aarch64/aarch64.c (aarch64_print_operand): Change the
8888         X format specifier to only display bottom 16 bits.
8889         * config/aarch64/aarch64.md (insv_imm<mode>): Allow any size of
8890         immediate to match for operand 2, since it will be masked.
8892 2013-05-24  Richard Biener  <rguenther@suse.de>
8894         PR tree-optimization/57287
8895         * tree-ssa-uninit.c (compute_uninit_opnds_pos): Disregard
8896         all SSA names that occur in abnormal PHIs.
8898 2013-05-24  Alexander Ivchenko  <alexander.ivchenko@intel.com>
8900         PR tree-ssa/57385
8901         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Check
8902         that index is not negative.
8904 2013-05-24  Eric Botcazou  <ebotcazou@adacore.com>
8906         PR rtl-optimization/55177
8907         * simplify-rtx.c (simplify_unary_operation_1) <NOT>: Deal with BSWAP.
8908         (simplify_byte_swapping_operation): New.
8909         (simplify_binary_operation_1): Call it for AND, IOR and XOR.
8910         (simplify_relational_operation_1): Deal with BSWAP.
8912 2013-05-23  Richard Henderson  <rth@redhat.com>
8914         PR target/56742
8915         * config/i386/i386.c (ix86_seh_fixup_eh_fallthru): New.
8916         (ix86_reorg): Call it.
8918 2013-05-23  Uros Bizjak  <ubizjak@gmail.com>
8920         PR target/57379
8921         * config/alpha/alpha.md (unspec): Add UNSPEC_XFLT_COMPARE.
8922         * config/alpha/alpha.c (alpha_emit_xfloating_compare): Construct
8923         REG_EQUAL note as UNSPEC_XFLT_COMPARE unspec.
8925 2013-05-23  Christian Bruel  <christian.bruel@st.com>
8927         PR debug/57351
8928         * config/arm/arm.c (arm_dwarf_register_span): Do not use dbx number.
8930 2013-05-23  Chris Schlumberger-Socha  <chris.schlumberger-socha@arm.com>
8931             Marcus Shawcroft  <marcus.shawcroft@arm.com>
8933         * config/aarch64/aarch64.md (*movdi_aarch64): Replace Usa with S.
8934         * config/aarch64/constraints.md (Usa): Remove.
8935         * doc/md.texi (AArch64 Usa): Remove.
8937 2013-05-23  Chris Schlumberger-Socha  <chris.schlumberger-socha@arm.com>
8938             Marcus Shawcroft  <marcus.shawcroft@arm.com>
8940         * config/aarch64/aarch64-protos.h (aarch64_mov_operand_p): Define.
8941         * config/aarch64/aarch64.c (aarch64_mov_operand_p): Define.
8942         * config/aarch64/predicates.md (aarch64_const_address): Remove.
8943         (aarch64_mov_operand): Use aarch64_mov_operand_p.
8945 2013-05-23  Vidya Praveen  <vidyapraveen@arm.com>
8947         * config/aarch64/aarch64-simd.md (clzv4si2): Support for CLZ
8948         instruction (AdvSIMD).
8949         * config/aarch64/aarch64-builtins.c
8950         (aarch64_builtin_vectorized_function): Handler for BUILT_IN_CLZ.
8951         * config/aarch64/aarch-simd-builtins.def: Entry for CLZ.
8953 2013-05-23  Martin Jambor  <mjambor@suse.cz>
8955         PR middle-end/57347
8956         * tree.h (contains_bitfld_component_ref_p): Declare.
8957         * tree-sra.c (contains_bitfld_comp_ref_p): Move...
8958         * tree.c (contains_bitfld_component_ref_p): ...here.  Adjust its
8959         caller.
8960         * ipa-prop.c (determine_known_aggregate_parts): Check that LHS does
8961         not access a bit-field.  Assert all final offsets are byte-aligned.
8963 2013-05-23  Richard Biener  <rguenther@suse.de>
8965         PR tree-optimization/57380
8966         * tree-ssa-phiprop.c (propagate_with_phi): Do not require at
8967         least one invariant or re-used load.
8968         * passes.c (init_optimization_passes): Move pass_phiprop before
8969         pass_forwprop.
8971 2013-05-23  James Greenhalgh  <james.greenhalgh@arm.com>
8973         * config/aarch64/aarch64-simd.md
8974         (aarch64_cm<optab>di): Add clobber of CC_REGNUM to unsplit pattern.
8976 2013-05-23  Richard Biener  <rguenther@suse.de>
8978         PR middle-end/57381
8979         * fold-const.c (operand_equal_p): Compare FIELD_DECLs with
8980         OEP_CONSTANT_ADDRESS_OF retained.
8982 2013-05-23  Jakub Jelinek  <jakub@redhat.com>
8984         PR middle-end/57344
8985         * expmed.c (store_split_bit_field): If op0 is a REG or SUBREG of a REG,
8986         don't lower unit.  Handle unit not being always BITS_PER_WORD.
8988 2013-05-23  Richard Biener  <rguenther@suse.de>
8990         PR rtl-optimization/57341
8991         * ira.c (validate_equiv_mem_from_store): Use anti_dependence
8992         instead of true_dependence.
8994 2013-05-22  David Malcolm  <dmalcolm@redhat.com>
8996         * bb-reorder.c (branch_threshold): Make const.
8997         (exec_threshold): Ditto.
8999 2013-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
9000             Pat Haugen  <pthaugen@us.ibm.com>
9001             Peter Bergner  <bergner@vnet.ibm.com>
9003         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Add
9004         documentation for the power8 crypto builtins.
9006         * config/rs6000/t-rs6000 (MD_INCLUDES): Add crypto.md.
9008         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_1): Add support
9009         macros for defining power8 builtin functions.
9010         (BU_P8V_AV_2): Likewise.
9011         (BU_P8V_AV_P): Likewise.
9012         (BU_P8V_VSX_1): Likewise.
9013         (BU_P8V_OVERLOAD_1): Likewise.
9014         (BU_P8V_OVERLOAD_2): Likewise.
9015         (BU_CRYPTO_1): Likewise.
9016         (BU_CRYPTO_2): Likewise.
9017         (BU_CRYPTO_3): Likewise.
9018         (BU_CRYPTO_OVERLOAD_1): Likewise.
9019         (BU_CRYPTO_OVERLOAD_2): Likewise.
9020         (XSCVSPDP): Fix typo, point to the correct instruction.
9021         (VCIPHER): Add power8 crypto builtins.
9022         (VCIPHERLAST): Likewise.
9023         (VNCIPHER): Likewise.
9024         (VNCIPHERLAST): Likewise.
9025         (VPMSUMB): Likewise.
9026         (VPMSUMH): Likewise.
9027         (VPMSUMW): Likewise.
9028         (VPERMXOR_V2DI): Likewise.
9029         (VPERMXOR_V4SI: Likewise.
9030         (VPERMXOR_V8HI: Likewise.
9031         (VPERMXOR_V16QI: Likewise.
9032         (VSHASIGMAW): Likewise.
9033         (VSHASIGMAD): Likewise.
9034         (VPMSUM): Likewise.
9035         (VPERMXOR): Likewise.
9036         (VSHASIGMA): Likewise.
9038         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
9039         __CRYPTO__ if the crypto instructions are available.
9040         (altivec_overloaded_builtins): Add support for overloaded power8
9041         builtins.
9043         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
9044         support for power8 crypto builtins.
9045         (builtin_function_type): Likewise.
9046         (altivec_init_builtins): Add support for builtins that take vector
9047         long long (V2DI) arguments.
9049         * config/rs6000/crypto.md: New file, define power8 crypto
9050         instructions.
9052 2013-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
9053             Pat Haugen  <pthaugen@us.ibm.com>
9054             Peter Bergner  <bergner@vnet.ibm.com>
9056         * doc/invoke.texi (Option Summary): Add power8 options.
9057         (RS/6000 and PowerPC Options): Likewise.
9059         * doc/md.texi (PowerPC and IBM RS6000 constraints): Update to use
9060         constraints.md instead of rs6000.h.  Reorder w* constraints.  Add
9061         wm, wn, wr documentation.
9063         * gcc/config/rs6000/constraints.md (wm): New constraint for VSX
9064         registers if direct move instructions are enabled.
9065         (wn): New constraint for no registers.
9066         (wq): New constraint for quad word even GPR registers.
9067         (wr): New constraint if 64-bit instructions are enabled.
9068         (wv): New constraint if power8 vector instructions are enabled.
9069         (wQ): New constraint for quad word memory locations.
9071         * gcc/config/rs6000/predicates.md (const_0_to_15_operand): New
9072         constraint for 0..15 for crypto instructions.
9073         (gpc_reg_operand): If VSX allow registers in VSX registers as well
9074         as GPR and floating point registers.
9075         (int_reg_operand): New predicate to match only GPR registers.
9076         (base_reg_operand): New predicate to match base registers.
9077         (quad_int_reg_operand): New predicate to match even GPR registers
9078         for quad memory operations.
9079         (vsx_reg_or_cint_operand): New predicate to allow vector logical
9080         operations in both GPR and VSX registers.
9081         (quad_memory_operand): New predicate for quad memory operations.
9082         (reg_or_indexed_operand): New predicate for direct move support.
9084         * gcc/config/rs6000/rs6000-cpus.def (ISA_2_5_MASKS_EMBEDDED):
9085         Inherit from ISA_2_4_MASKS, not ISA_2_2_MASKS.
9086         (ISA_2_7_MASKS_SERVER): New mask for ISA 2.07 (i.e. power8).
9087         (POWERPC_MASKS): Add power8 options.
9088         (power8 cpu): Use ISA_2_7_MASKS_SERVER instead of specifying the
9089         various options.
9091         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
9092         Define _ARCH_PWR8 and __POWER8_VECTOR__ for power8.
9094         * gcc/config/rs6000/rs6000.opt (-mvsx-timode): Add documentation.
9095         (-mpower8-fusion): New power8 options.
9096         (-mpower8-fusion-sign): Likewise.
9097         (-mpower8-vector): Likewise.
9098         (-mcrypto): Likewise.
9099         (-mdirect-move): Likewise.
9100         (-mquad-memory): Likewise.
9102         * gcc/config/rs6000/rs6000.c (power8_cost): Initial definition for
9103         power8.
9104         (rs6000_hard_regno_mode_ok): Make PTImode only match even GPR
9105         registers.
9106         (rs6000_debug_reg_print): Print the base register class if -mdebug=reg.
9107         (rs6000_debug_vector_unit): Add p8_vector.
9108         (rs6000_debug_reg_global): If -mdebug=reg, print power8 constraint
9109         definitions.  Also print fusion state.
9110         (rs6000_init_hard_regno_mode_ok): Set up power8 constraints.
9111         (rs6000_builtin_mask_calculate): Add power8 builtin support.
9112         (rs6000_option_override_internal): Add support for power8.
9113         (rs6000_common_init_builtins): Add debugging for skipped builtins
9114         if -mdebug=builtin.
9115         (rs6000_adjust_cost): Add power8 support.
9116         (rs6000_issue_rate): Likewise.
9117         (insn_must_be_first_in_group): Likewise.
9118         (insn_must_be_last_in_group): Likewise.
9119         (force_new_group): Likewise.
9120         (rs6000_register_move_cost): Likewise.
9121         (rs6000_opt_masks): Likewise.
9123         * config/rs6000/rs6000.h (ASM_CPU_POWER8_SPEC): If we don't have a
9124         power8 capable assembler, default to power7 options.
9125         (TARGET_DIRECT_MOVE): Likewise.
9126         (TARGET_CRYPTO): Likewise.
9127         (TARGET_P8_VECTOR): Likewise.
9128         (VECTOR_UNIT_P8_VECTOR_P): Define power8 vector support.
9129         (VECTOR_UNIT_VSX_OR_P8_VECTOR_P): Likewise.
9130         (VECTOR_MEM_P8_VECTOR_P): Likewise.
9131         (VECTOR_MEM_VSX_OR_P8_VECTOR_P): Likewise.
9132         (VECTOR_MEM_ALTIVEC_OR_VSX_P): Likewise.
9133         (TARGET_XSCVDPSPN): Likewise.
9134         (TARGET_XSCVSPDPN): Likewsie.
9135         (TARGET_SYNC_HI_QI): Likewise.
9136         (TARGET_SYNC_TI): Likewise.
9137         (MASK_CRYPTO): Likewise.
9138         (MASK_DIRECT_MOVE): Likewise.
9139         (MASK_P8_FUSION): Likewise.
9140         (MASK_P8_VECTOR): Likewise.
9141         (REG_ALLOC_ORDER): Move fr13 to be lower in priority so that the TFmode
9142         temporary used by some of the direct move instructions to get two FP
9143         temporary registers does not force creation of a stack frame.
9144         (VLOGICAL_REGNO_P): Allow vector logical operations in GPRs.
9145         (MODES_TIEABLE_P): Move the VSX tests above the Altivec tests so
9146         that any VSX registers are tieable, even if they are also an
9147         Altivec vector mode.
9148         (r6000_reg_class_enum): Add wm, wr, wv constraints.
9149         (RS6000_BTM_P8_VECTOR): Power8 builtin support.
9150         (RS6000_BTM_CRYPTO): Likewise.
9151         (RS6000_BTM_COMMON): Likewise.
9153         * config/rs6000/rs6000.md (cpu attribute): Add power8.
9154         * config/rs6000/rs6000-opts.h (PROCESSOR_POWER8): Likewise.
9155         (enum rs6000_vector): Add power8 vector support.
9157 2013-05-22  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9159         PR target/19599
9160         PR target/57340
9161         * config/arm/arm.c (any_sibcall_uses_r3): Rename to ..
9162         (any_sibcall_could_use_r3): this and handle indirect calls.
9163         (arm_get_frame_offsets): Rename use of any_sibcall_uses_r3.
9165 2013-05-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9167         * config/rs6000/rs6000.h (MALLOC_ABI_ALIGNMENT): New #define.
9169 2013-05-22  Richard Biener  <rguenther@suse.de>
9171         PR middle-end/57349
9172         * profile.c (branch_prob): Do not split blocks that are
9173         abnormally receiving from ECF_RETURNS_TWICE functions.
9175 2013-05-22  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
9177         * recog.c (offsettable_address_addr_space_p): Fix calculation of
9178         address mode.  Move pointer mode initialization to the same place.
9180 2013-05-22  Michael Zolotukhin  <michael.v.zolotukhin@intel.com>
9182         * read-rtl.c (copy_rtx_for_iterators): Continue applying iterators
9183         while it has any effect.
9185 2013-05-21  Easwaran Raman  <eraman@google.com>
9187         PR tree-optimization/57322
9188         * tree-ssa-reassoc.c (build_and_add_sum): If a BB is empty, set the
9189         UID of the statement added to the BB to be 1.
9191 2013-05-21  Jakub Jelinek  <jakub@redhat.com>
9193         PR tree-optimization/57331
9194         * tree-vrp.c (simplify_cond_using_ranges): Don't optimize comparison
9195         of conversion from pointer type to integral type with integer.
9197 2013-05-21  Martin Jambor  <mjambor@suse.cz>
9199         PR lto/57289
9200         * ipa-prop.c (ipa_read_node_info): Process param_used and
9201         controlled_uses in the same order as when writing.
9203 2013-05-21  Magnus Granberg  <baldrick@free.fr>
9205         PR plugins/56754
9206         * Makefile.in (PLUGIN_HEADERS): Add $(TARGET_H).
9208 2013-05-21  Richard Biener  <rguenther@suse.de>
9210         PR tree-optimization/57318
9211         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Do not
9212         estimate stmts with side-effects as likely eliminated.
9214 2013-05-21  Richard Biener  <rguenther@suse.de>
9216         PR tree-optimization/57330
9217         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Properly
9218         preserve the call stmts fntype.
9220 2013-05-21  Richard Biener  <rguenther@suse.de>
9222         PR tree-optimization/57303
9223         * tree-ssa-sink.c (statement_sink_location): Improve killing
9224         stmt detection and properly handle self-assignments.
9226 2013-05-21  Christian Bruel  <christian.bruel@st.com>
9228         * dwarf2out.c (multiple_reg_loc_descriptor): Use dbx_reg_number for
9229         spanning registers. LEAF_REG_REMAP is supported only for contiguous
9230         registers. Set register size out of the PARALLEL loop.
9232 2013-05-20  Oleg Endo  <olegendo@gcc.gnu.org>
9234         PR target/56547
9235         * config/sh/sh.md (fmasf4): Remove empty constraints strings.
9236         (*fmasf4, *fmasf4_media): New insns.
9238 2013-05-19  Richard Sandiford  <rdsandiford@googlemail.com>
9240         * config/mips/mips.h (BASE_INSN_LENGTH, NOP_INSN_LENGTH): New macros.
9241         * config/mips/mips.c (mips_symbol_insns, mips_address_insns)
9242         (mips_const_insns, mips_split_const_insns, mips_load_store_insns)
9243         (mips_idiv_insns): Update the comments to say that the returned
9244         instruction counts are in units of BASE_INSN_LENGTH.
9245         (mips_adjust_insn_length): Multiply the mips_load_label_num_insns
9246         by BASE_INSN_LENGTH rather than 4.  Add the jump separately,
9247         using 2 rather than 4 as the length of indirect MIPS16 and
9248         microMIPS jumps.  Use NOP_INSN_LENGTH rather than 4 as the
9249         length of a NOP.  Don't divide MIPS16 lengths by 2.
9250         (mips16_split_long_branches): Assume a branch is long if the
9251         length is greater than 4 rather than 8.
9252         * config/mips/mips.md (length): Give MIPS16 lengths directly,
9253         rather than multiplying them by 2.  Multiply instruction counts
9254         by BASE_INSN_LENGTH rather than 4.
9255         (*jump_mips16, tls_get_tp_mips16_<mode>)
9256         (*tls_get_tp_mips16_call_<mode>): Divide lengths by 2.
9258 2013-05-19  Richard Sandiford  <rdsandiford@googlemail.com>
9260         * config/mips/mips.md (extended_mips16): Remove branch case.
9261         (length): Remove duplicated extended_mips16 test.
9263 2013-05-19  Richard Sandiford  <rdsandiford@googlemail.com>
9265         * config/mips/t-sde: Don't build 64-bit microMIPS multilibs.
9267 2013-05-18  Richard Sandiford  <rdsandiford@googlemail.com>
9269         * recog.h (Recog_data): Rename to...
9270         (recog_data_d): ...this.
9271         (recog_data): Update accordingly.
9272         * recog.c (recog_data): Likewise.
9273         * reload.c (save_recog_data): Likewise.
9274         * config/picochip/picochip.c (picochip_saved_recog_data): Likewise.
9275         (picochip_save_recog_data, picochip_restore_recog_data): Likewise.
9277 2013-05-17  Julian Brown  <julian@codesourcery.com>
9279         * gcse.c (compute_ld_motion_mems): If a non-simple MEM is
9280         found in a REG_EQUAL note, invalidate it.
9282 2013-05-17  Easwaran Raman  <eraman@google.com>
9284         * tree-ssa-reassoc.c (find_insert_point): New function.
9285         (insert_stmt_after): Likewise.
9286         (get_def_stmt): Likewise.
9287         (ensure_ops_are_available): Likewise.
9288         (not_dominated_by): Likewise.
9289         (rewrite_expr_tree): Do not move statements beyond what is
9290         necessary. Remove call to swap_ops_for_binary_stmt...
9291         (reassociate_bb): ... and move it here.
9292         (build_and_add_sum): Assign UIDs for new statements.
9293         (linearize_expr): Likewise.
9294         (do_reassoc): Renumber gimple statement UIDs.
9296 2013-05-17  Jan Hubicka  <jh@suse.cz>
9298         * lto-symtab.c (lto_symtab_merge_cgraph_nodes): Resolve cross module
9299         weakrefs.
9300         * cgraph.c (dump_cgraph_node): Do not ice on unresolved alias.
9301         * cgraphunit.c (handle_alias_pairs): Store target of unresolved
9302         weakrefs.
9303         (output_weakrefs): Update.
9305 2013-05-17  Po-Chun Chang  <pchang9@cs.wisc.edu>
9306             Martin Jambor  <mjambor@suse.cz>
9308         PR middle-end/57276
9309         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Break when a
9310         value that corresponds to the given aggval is found in values vector.
9312 2013-05-17  Uros Bizjak  <ubizjak@gmail.com>
9314         * config/i386/driver-i386.c (host_detect_local_cpu): Pass mmx, 3dnow,
9315         sse, sse2, sse3, ssse3 and sse4a flags to options.
9317 2013-05-17  David Malcolm  <dmalcolm@redhat.com>
9319         * gengtype-state.c: (s_expr_writer): New class, to handle
9320         prettifying of output layout of s-expressions.
9321         (state_writer): New class, to write out gtype.state.
9322         (state_written_type_count): Move this variable into member data of
9323         state_writer.
9324         (s_expr_writer::s_expr_writer): New code: constructor for new class
9325         (state_writer::state_writer(): ditto
9326         (s_expr_writer::write_new_line): New function
9327         (s_expr_writer::write_any_indent): ditto
9328         (s_expr_writer::begin_s_expr): ditto
9329         (s_expr_writer::end_s_expr): ditto
9330         (write_state_fileloc): convert to method of state_writer...
9331         (state_writer:: write_state_fileloc): ...and use methods of
9332         s_expr_writer to write indentation into the gtype.state output file
9333         to visually represent the hierarchical structure of the list
9334         structures
9335         (write_state_fields): ditto, renaming to...
9336         (state_writer::write_state_fields)
9337         (write_state_a_string): ditto, renaming to...
9338         (state_writer::write_state_a_string)
9339         (write_state_string_option): ditto, renaming to...
9340         (state_writer::write_state_string_option)
9341         (write_state_type_option): ditto, renaming to...
9342         (state_writer::write_state_type_option)
9343         (write_state_nested_option): ditto, renaming to...
9344         (state_writer::write_state_nested_option)
9345         (write_state_option): ditto, renaming to...
9346         (state_writer::write_state_option)
9347         (write_state_options): ditto, renaming to...
9348         (state_writer::write_state_options)
9349         (write_state_lang_bitmap): ditto, renaming to...
9350         (state_writer::write_state_lang_bitmap)
9351         (write_state_version): ditto, renaming to...
9352         (state_writer::write_state_version)
9353         (write_state_scalar_type): ditto, renaming to...
9354         (state_writer::write_state_scalar_type)
9355         (write_state_string_type): ditto, renaming to...
9356         (state_writer::write_state_string_type)
9357         (write_state_undefined_type): ditto, renaming to...
9358         (state_writer::write_state_undefined_type)
9359         (write_state_struct_union_type): ditto, renaming to...
9360         (state_writer::write_state_struct_union_type)
9361         (write_state_struct_type): ditto, renaming to...
9362         (state_writer::write_state_struct_type)
9363         (write_state_user_struct_type): ditto, renaming to...
9364         (state_writer::write_state_user_struct_type)
9365         (write_state_lang_struct_type): ditto, renaming to...
9366         (state_writer::write_state_lang_struct_type)
9367         (write_state_param_struct_type): ditto, renaming to...
9368         (state_writer::write_state_param_struct_type)
9369         (write_state_pointer_type): ditto, renaming to...
9370         (state_writer::write_state_pointer_type)
9371         (write_state_array_type): ditto, renaming to...
9372         (state_writer::write_state_array_type)
9373         (write_state_gc_used): ditto, renaming to...
9374         (state_writer::write_state_gc_used)
9375         (write_state_common_type_content): ditto, renaming to...
9376         (state_writer::write_state_common_type_content)
9377         (write_state_type): ditto, renaming to...
9378         (state_writer::write_state_type)
9379         (write_state_pair_list): ditto, renaming to...
9380         (state_writer::write_state_pair_list)
9381         (write_state_pair): ditto, renaming to...
9382         (state_writer::write_state_pair)
9383         (write_state_typedefs): ditto, renaming to...
9384         (state_writer::write_state_typedefs)
9385         (write_state_structures): ditto, renaming to...
9386         (state_writer::write_state_structures)
9387         (write_state_param_structs): ditto, renaming to...
9388         (state_writer::write_state_param_structs)
9389         (write_state_variables): ditto, renaming to...
9390         (state_writer::write_state_variables)
9391         (write_state_srcdir): ditto, renaming to...
9392         (state_writer::write_state_srcdir)
9393         (write_state_files_list): ditto, renaming to...
9394         (state_writer::write_state_files_list)
9395         (write_state_languages): ditto, renaming to...
9396         (state_writer::write_state_languages)
9397         (write_state): create a state_writer instance and use it when
9398         writing out the state file
9400 2013-05-17  Mike Stump  <mikestump@comcast.net>
9402         PR rtl-optimization/57304
9403         * web.c (union_match_dups): Ensure that DF_REF_LOC exists before
9404         accessing DF_REF_REAL_LOC.
9406 2013-05-17  Jakub Jelinek  <jakub@redhat.com>
9408         PR rtl-optimization/57281
9409         PR rtl-optimization/57300
9410         * config/i386/i386.md (extendsidi2_1 dead reg splitter): Remove.
9411         (extendsidi2_1 peephole2s): Add instead 2 new peephole2s, that undo
9412         what the other splitter did if the registers are dead.
9414 2013-05-17  Richard Biener  <rguenther@suse.de>
9416         * tree-ssa-alias.c (stmt_kills_ref_p_1): Properly compare
9417         MEM_REF offsets.
9419 2013-05-17  Jakub Jelinek  <jakub@redhat.com>
9421         * gcc.c (SANITIZER_SPEC): Reject -fsanitize=address -fsanitize=thread
9422         linking.
9424 2013-05-17  Marek Polacek  <polacek@redhat.com>
9426         * tree-ssa-strlen.c (handle_char_store): Don't invalidate cached
9427         length when doing non-zero store of storing '\0' to '\0'.
9429 2013-05-17  Jakub Jelinek  <jakub@redhat.com>
9431         * tree-vect-patterns.c (vect_recog_rotate_pattern): For
9432         vect_external_def oprnd1 with loop_vinfo, try to emit
9433         optional cast, negation and and stmts on the loop preheader
9434         edge instead of into the pattern def seq.
9436         PR tree-optimization/57051
9437         * fold-const.c (const_binop) <case VEC_LSHIFT_EXPR,
9438         case VEC_RSHIFT_EXPR>: Fix BYTES_BIG_ENDIAN handling.
9440 2013-05-16  Nick Clifton  <nickc@redhat.com>
9442         * config/rl78/rl78.c (rl78_attribute_table): Add naked.
9443         (rl78_is_naked_func): New function.
9444         (rl78_expand_prologue): Skip prologue generation for naked functions.
9445         (rl78_expand_epilogue): Skip epilogue generation for naked functions.
9446         * doc/extend.texi (naked): Add RL78 to the list of processors
9447         that supports this attribute.
9449 2013-05-16  Jeff Law  <law@redhat.com>
9451         * Makefile.in (tree-switch-conversion.o): Depend on $(OPTABS_H).
9453 2013-05-16  Uros Bizjak  <ubizjak@gmail.com>
9455         * config/i386/driver-i386.c (host_detect_local_cpu): Determine
9456         cache parameters using detect_caches_amd also for CYRIX,
9457         NSC and TM2 signatures.
9459 2013-05-16  Uros Bizjak  <ubizjak@gmail.com>
9460             Dzianis Kahanovich  <mahatma@eu.by>
9462         PR target/45359
9463         PR target/46396
9464         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
9465         VIA/Centaur processors and determine their cache parameters
9466         using detect_caches_amd.
9468 2013-05-16  Teresa Johnson  <tejohnson@google.com>
9470         * cfgrtl.c (verify_hot_cold_block_grouping): Return err.
9471         (rtl_verify_edges): New function.
9472         (rtl_verify_bb_insns): Ditto.
9473         (rtl_verify_bb_pointers): Ditto.
9474         (rtl_verify_bb_insn_chain): Ditto.
9475         (rtl_verify_fallthru): Ditto.
9476         (rtl_verify_bb_layout): Ditto.
9477         (rtl_verify_flow_info_1): Outline checks into new functions.
9478         (rtl_verify_flow_info): Ditto.
9480 2013-05-16  Steve Ellcey  <sellcey@imgtec.com>
9482         * cfghooks.c (copy_bbs): Add update_dominance argument.
9483         * cfghooks.h (copy_bbs): Update prototype.
9484         * tree-cfg.c (gimple_duplicate_sese_region):
9485         Add update_dominance argument.
9486         * tree-flow.h (gimple_duplicate_sese_region): Update prototype.
9487         * tree-ssa-loop-ch.c (copy_loop_headers): Update
9488         gimple_duplicate_sese_region call.
9489         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg):
9490         Update copy_bbs call.
9491         * cfgloopmanip.c (duplicate_loop_to_header_edge): Ditto.
9492         * trans-mem.c (ipa_uninstrument_transaction): Ditto.
9494 2013-05-16  Jakub Jelinek  <jakub@redhat.com>
9496         * tree-vectorizer.h (NUM_PATTERNS): Increment.
9497         * tree-vect-patterns.c (vect_vect_recog_func_ptrs): Add
9498         vect_recog_rotate_pattern.
9499         (vect_recog_rotate_pattern): New function.
9501 2013-05-16  Jason Merrill  <jason@redhat.com>
9503         * Makefile.in (LLINKER): New variable.
9504         (mostlyclean): Remove link mutex.
9505         * configure.ac: Handle --enable-link-mutex.
9506         * lock-and-run.sh: New script.
9508 2013-05-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9510         PR target/19599
9511         * config/arm/arm.c (arm_function_ok_for_sibcall): Add check
9512         for NULL decl.
9514 2013-05-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9516         * reorg.c (link_cc0_insns): Wrap in #ifdef HAVE_cc0.
9518 2013-05-16  Greta Yorsh  <Greta.Yorsh@arm.com>
9520         * config/arm/arm-protos.h (gen_movmem_ldrd_strd): New declaration.
9521         * config/arm/arm.c (next_consecutive_mem): New function.
9522         (gen_movmem_ldrd_strd): Likewise.
9523         * config/arm/arm.md (movmemqi): Update condition and code.
9524         (unaligned_loaddi, unaligned_storedi): New patterns.
9526 2013-05-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9528         * config.gcc: Obsolete *-*-solaris2.9*.
9529         * doc/install.texi (Specific, *-*-solaris2*): Document it.
9531 2013-05-16  Richard Biener  <rguenther@suse.de>
9533         * passes.c (init_optimization_passes): Move pass_parallelize_loops
9534         earlier, after GRAPHITE transforms and IV canonicalization.
9536 2013-05-16  Jakub Jelinek  <jakub@redhat.com>
9538         * omp-low.c (extract_omp_for_data): For collapsed loops,
9539         if at least one of the loops is known at compile time to
9540         iterate zero times, set count to 0.
9541         (expand_omp_regimplify_p): New function.
9542         (expand_omp_for_generic): For collapsed loops, if at least
9543         one of the loops isn't known to iterate at least once,
9544         add runtime check with setting count to 0.
9545         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk):
9546         For unsigned types if it isn't known at compile time that
9547         the loop will iterate at least once, add runtime check to bypass
9548         the whole loop if initial condition isn't true.
9550 2013-05-16  Nathan Sidwell  <nathan@codesourcery.com>
9552         * varasm.c (default_use_anchors_for_symbol_p): Use decl_replaceable_p.
9554 2013-05-16  Marc Glisse  <marc.glisse@inria.fr>
9556         PR middle-end/57286
9557         * fold-const.c (fold_ternary_loc) <VEC_COND_EXPR>: Disable some
9558         transformations to avoid an infinite loop.
9560 2013-05-16  Marek Polacek  <polacek@redhat.com>
9562         * tree-scalar-evolution.c (scev_const_prop): Add more dumps.
9564 2013-05-15  Leif Ekblad  <leif@rdos.net>
9566         * config/i386/i386.c (ix86_decompose_address): Use
9567         DEFAULT_TLS_SEG_REG to access TLS segment register.
9568         * config/i386/i386.h (DEFAULT_TLS_SEG_REG): New define.
9569         * config/i386/rdos.h (DEFAULT_TLS_SEG_REG): Ditto.
9570         (TARGET_TLS_DIRECT_SEG_REFS_DEFAULT): Ditto.
9572 2013-05-15  Richard Sandiford  <rdsandiford@googlemail.com>
9574         PR target/57260
9575         * config/mips/mips.c (mips_function_ok_for_sibcall): Don't allow
9576         sibling calls to functions that would normally be lazily bound,
9577         unless $gp is call-clobbered.
9579 2013-05-15  Uros Bizjak  <ubizjak@gmail.com>
9581         * config/i386/i386.c (ix86_option_override_internal): Update
9582         processor_alias_table for missing PTA_PRFCHW and PTA_FXSR flags.  Add
9583         PTA_POPCNT to corei7 entry. Do not enable SSE prefetch on
9584         non-SSE 3dNow! targets.  Enable TARGET_PRFCHW for TARGET_3DNOW targets.
9585         * config/i386/i386.md (prefetch): Enable for TARGET_PRFCHW instead
9586         of TARGET_3DNOW.
9587         (*prefetch_3dnow): Enable for TARGET_PRFCHW only.
9589 2013-05-15  Andreas Schwab  <schwab@suse.de>
9591         * config/m68k/m68k.md (*rotlhi3_lowpart, *rotlqi3_lowpart): Name
9592         for rotlhi3+1 and rotlqi3+1, resp.  Fix reference to non-existing
9593         third operand.
9595 2013-05-15  Teresa Johnson  <tejohnson@google.com>
9597         * loop-unroll.c (report_unroll_peel): Check decision before
9598         emitting unroll/peel message.
9600 2013-05-15  Teresa Johnson  <tejohnson@google.com>
9602         * function.h (has_bb_partition): New rtl_data flag.
9603         (bb_reorder_complete): Ditto.
9604         * cfgcleanup.c (try_crossjump_to_edge): Check for has_bb_partition
9605         instead of flag_reorder_blocks_and_partition.
9606         * cfgrtl.c (verify_hot_cold_block_grouping): Moved from bb-reorder.c,
9607         with some enhancements.
9608         (rtl_verify_flow_info_1): Call verify_hot_cold_block_grouping.
9609         * bb-reorder.c (connect_traces): Check for has_bb_partition
9610         instead of flag_reorder_blocks_and_partition.
9611         (verify_hot_cold_block_grouping): Moved to cfgrtl.c.
9612         (reorder_basic_blocks): Set bb_reorder_complete flag, remove call to
9613         verify_hot_cold_block_grouping.
9614         (partition_hot_cold_basic_blocks): Set has_bb_partition.
9616 2013-05-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
9618         PR target/19599
9619         * config/arm/predicates.md (call_insn_operand): New predicate.
9620         * config/arm/constraints.md ("Cs", "Ss"):  New constraints.
9621         * config/arm/arm.md (*call_insn, *call_value_insn): Match only
9622         if insn is not a tail call.
9623         (*sibcall_insn, *sibcall_value_insn): Adjust for tailcalling through
9624         registers.
9625         * config/arm/arm.h (enum reg_class): New caller save register class.
9626         (REG_CLASS_NAMES): Likewise.
9627         (REG_CLASS_CONTENTS): Likewise.
9628         * config/arm/arm.c (arm_function_ok_for_sibcall): Allow tailcalling
9629         without decls.
9631 2013-05-15  Richard Biener  <rguenther@suse.de>
9633         * tree-vect-loop.c (vect_transform_loop): Use MSG_NOTE instead
9634         of MSG_OPTIMIZED_LOCATIONS.
9635         * tree-vect-slp.c (vect_make_slp_decision): Likewise.
9636         (vect_slp_transform_bb): Indicate location in MSG_OPTIMIZED_LOCATIONS
9637         message.
9638         * tree-vectorizer.c (vectorize_loops): Use MSG_NOTE instead
9639         of MSG_OPTIMIZED_LOCATIONS.
9640         (execute_vect_slp): Likewise.
9641         * tree-vect-loop-manip.c (vect_do_peeling_for_loop_bound): Likewise.
9642         (vect_create_cond_for_alias_checks): Likewise.
9643         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
9644         (vect_recog_widen_mult_pattern): Likewise.
9645         (vect_recog_widen_sum_pattern): Likewise.
9646         (vect_recog_over_widening_pattern): Likewise.
9647         (vect_recog_widen_shift_pattern): Likewise.
9648         (vect_recog_vector_vector_shift_pattern): Likewise.
9649         (vect_recog_divmod_pattern): Likewise.
9650         (vect_recog_mixed_size_cond_pattern): Likewise.
9651         (vect_recog_bool_pattern): Likewise.
9652         (vect_pattern_recog_1): Likewise.
9654 2013-05-15  Martin Jambor  <mjambor@suse.cz>
9656         * ipa-prop.c (ipa_make_edge_direct_to_target): Redirect calls to
9657         non-functions to builtin_unreachable.
9658         * ipa-inline-transform.c (inline_call): Do not assert estimates were
9659         correct when new direct edges were discovered.
9661 2013-05-15  Martin Jambor  <mjambor@suse.cz>
9663         * ipa-prop.c (ipa_print_node_jump_functions): Print symbol order in
9664         header, print symbol order instead of node uid, print more information
9665         about indirect edge targets.
9666         (ipa_make_edge_direct_to_target): Print symbol order instead of node
9667         uids.
9668         (ipa_make_edge_direct_to_target): Likewise.
9669         (remove_described_reference): Likewise.
9670         (propagate_controlled_uses): Likewise.
9671         (ipa_print_node_params): Also print symbol order.
9672         (ipcp_transform_function): Print symbol order instead of node uids.
9673         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Likewise.
9674         (cgraph_get_create_real_symbol_node): Likewise.
9675         * ipa-cp.c (print_lattice): Likewise.
9676         (print_all_lattices): Likewise.
9677         (determine_versionability): Likewise.
9678         (initialize_node_lattices): Likewise.
9679         (estimate_local_effects): Likewise.
9680         (update_profiling_info): Likewise.
9681         (create_specialized_node): Likewise.
9682         (perhaps_add_new_callers): Likewise.
9683         (decide_about_value): Likewise.
9684         (decide_whether_version_node): Likewise.
9685         (identify_dead_nodes): Likewise.
9686         * ipa-inline-analysis.c (dump_inline_edge_summary): Likewise.
9687         (dump_inline_summary): Likewise.
9688         (estimate_node_size_and_time): Likewise.
9689         (inline_analyze_function): Likewise.
9690         * ipa-inline.c (report_inline_failed_reason): Likewise.
9691         (want_early_inline_function_p): Likewise.
9692         (edge_badness): Likewise.
9693         (update_edge_key): Likewise.
9694         (inline_small_functions): Likewise.  Add dumping of order to two other
9695         dumps.
9696         * ipa-pure-const.c (pure_const_read_summary): Print symbol order
9697         instead of node uids.
9698         (propagate_pure_const): Likewise.
9699         (propagate_pure_const): Likewise.
9700         * ipa-utils.c (dump_cgraph_node_set): Likewise.
9701         * lto-cgraph.c (input_node): Explicitly specify we dump uid.
9702         * lto-symtab.c (lto_cgraph_replace_node): Print symbol order instead
9703         of node uids.
9704         * tree-pretty-print.c (dump_function_header): Likewise.
9705         * tree-sra.c (convert_callers_for_node): Dump in traditional format.
9706         Print symbol order instead of node uids.
9708 2013-05-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
9710         * config/s390/s390.c (s390_register_move_cost): Don't impose the
9711         FPR<->GPR move cost penalty if ldgr/lgdr can be used.
9713 2013-05-15  Richard Biener  <rguenther@suse.de>
9715         PR tree-optimization/57275
9716         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Fix
9717         return value for fail to do runtime alias checks for gather loads.
9719 2013-05-15  Jan Hubicka  <jh@suse.cz>
9721         PR lto/57038
9722         PR lto/47375
9723         * lto-symtab.c (lto_symtab_symbol_p): Add external symbol;
9724         weakrefs are not external.
9725         (lto_symtab_merge_decls): Fix thinko when dealing with
9726         non-lto_symtab decls.
9727         (lto_symtab_merge_cgraph_nodes): Use lto_symtab_symbol_p.
9728         (lto_symtab_prevailing_decl): Get int sync with lto_symtab_symbol_p.
9729         * varpool.c (dump_varpool_node): Dump more flags.
9731 2013-05-15  Ganesh Gopalasubramanian  <Ganesh.Gopalasubramanian@amd.com>
9733         * config/i386/i386.c (processor_alias_table): Add instruction
9734         FSGSBASE for AMD bdver3 architecture.
9736 2013-05-14  Jakub Jelinek  <jakub@redhat.com>
9738         * tree.c (warn_deprecated_use): Print file:line using locus color.
9739         * diagnostic.c (diagnostic_report_current_module): Print file:line
9740         and file:line:column using locus color.
9742 2013-05-14  Mike Stump  <mikestump@comcast.net>
9744         * gdbinit.in: Add __null.
9746 2013-05-14  Mike Stump  <mikestump@comcast.net>
9748         * recog.h: Rename struct recog_data to Recog_data.
9749         * recog.c: Likewise.
9750         * reload.c (can_reload_into): Likewise.
9751         * config/picochip/picochip.c: Likewise.
9753 2013-05-14  Mike Stump  <mikestump@comcast.net>
9755         * web.c (union_match_dups): Also check DF_REF_REAL_LOC.
9757 2013-05-14  Steven Bosscher  <steven@gcc.gnu.org>
9759         * resource.h (struct resources): Remove unch_memory member.
9760         (CLEAR_RESOURCE): Don't clear unch_memory.
9761         * resource.c (mark_referenced_resources): Don't set it.
9762         (mark_set_resources): Likewise.
9763         (mark_target_live_regs): Don't clear it.
9764         (init_resource_info): Likewise.
9765         * reorg.c (resource_conflicts_p): Don't compare it.
9766         (redundant_insn): Don't set it.
9768         * rtl.h (next_label, skip_consecutive_labels, link_cc0_insns):
9769         Remove prototypes.
9770         * emit-rtl.c (next_label): Remove unused function.
9771         (skip_consecutive_labels, link_cc0_insns): Move to ...
9772         * reorg.c (skip_consecutive_labels, link_cc0_insns): ... here, the
9773         only place where these functions are used, and make them static.
9775 2013-05-14  Marc Glisse  <marc.glisse@inria.fr>
9777         * fold-const.c (fold_negate_expr): Handle vectors.
9778         (fold_truth_not_expr): Make it static.
9779         (fold_invert_truthvalue): New static function.
9780         (invert_truthvalue_loc): Handle vectors. Do not call
9781         fold_truth_not_expr directly.
9782         (fold_unary_loc) <BIT_NOT_EXPR>: Handle comparisons.
9783         <TRUTH_NOT_EXPR>: Do not cast to boolean.
9784         (fold_comparison): Handle vector constants.
9785         (fold_binary_loc) <TRUTH_XOR_EXPR>: Remove redundant code.
9786         (fold_ternary_loc) <VEC_COND_EXPR>: Adapt more COND_EXPR optimizations.
9787         * tree.h (fold_truth_not_expr): Remove declaration.
9789 2013-05-14  James Greenhalgh  <james.greenhalgh@arm.com>
9791         * config/aarch64/aarch64-simd.md
9792         (aarch64_vcond_internal<mode>): Rename to...
9793         (aarch64_vcond_internal<mode><mode>): ...This, for integer modes.
9794         (aarch64_vcond_internal<VDQF_COND:mode><VDQF:mode>): ...This for
9795         float modes. Clarify all iterator modes.
9796         (vcond<mode><mode>): Use new name for vcond expanders.
9797         (vcond<v_cmp_result><mode>): Likewise.
9798         (vcondu<mode><mode>: Likewise.
9799         * config/aarch64/iterators.md (VDQF_COND): New.
9801 2013-05-14  Marc Glisse  <marc.glisse@inria.fr>
9803         PR bootstrap/57266
9804         * fold-const.c (fold_binary_loc) <shift>: Use an unsigned
9805         variable for the shift amount. Check that we shift by non-negative
9806         amounts.
9808 2013-05-14  Chung-Lin Tang  <cltang@codesourcery.com>
9810         PR target/42017
9811         * config/arm/arm.h (EPILOGUE_USES): Only return true
9812         for LR_REGNUM after epilogue_completed.
9814 2013-05-14  Joern Rennecke  <joern.rennecke@embecosm.com>
9816         * config/avr/avr.c (avr_encode_section_info): Bail out if the type
9817         is error_mark_node.
9819 2013-05-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9821         PR target/57261
9822         * configure.ac (gcc_cv_ld_as_needed): Disable before Solaris 11
9823         and Solaris 11+/x86 with gld.
9824         * configure: Regenerate.
9826 2013-05-14  Jakub Jelinek  <jakub@redhat.com>
9828         * expmed.c (expand_shift_1): Canonicalize rotates by
9829         constant bitsize / 2 to bitsize - 1.
9830         * simplify-rtx.c (simplify_binary_operation_1) <case ROTATE,
9831         case ROTATERT>: Likewise.
9833         Revert:
9834         2013-05-10  Jakub Jelinek  <jakub@redhat.com>
9836         * config/i386/i386.md (rotateinv): New code attr.
9837         (*<rotate_insn><mode>3_1, *<rotate_insn>si3_1_zext,
9838         *<rotate_insn>qi3_1_slp): Emit rorl %eax instead of
9839         roll $31, %eax, etc.
9841 2013-05-14  Richard Biener  <rguenther@suse.de>
9843         PR middle-end/57235
9844         * tree-eh.c (sink_clobbers): Give up for successors with
9845         multiple predecessors and no virtual uses.
9847 2013-05-14  Eric Botcazou  <ebotcazou@adacore.com>
9849         * config/sparc/sp64-elf.h (CPP_SUBTARGET_SPEC): Delete.
9850         * config/sparc/openbsd64.h (CPP_SUBTARGET_SPEC): Likewise.
9852 2013-05-14  Jakub Jelinek  <jakub@redhat.com>
9854         PR middle-end/57251
9855         * expr.c (expand_expr_real_2) <case WIDEN_MULT_EXPR>: Handle
9856         the case when both op0 and op1 have VOIDmode.
9858 2013-05-14  Kaushik Phatak  <kaushik.phatak@kpitcummins.com>
9860         * config/rl78/rl78.md(mulsi3_g13): Add additional 'nop' required
9861         in multiply-accumulate mode.
9863 2013-05-13  Guozhi Wei  <carrot@google.com>
9865         * dwarf2asm.c (dw2_output_indirect_constant_1): Mark new decl STATIC.
9867 2013-05-13  Kai Tietz  <ktietz@redhat.com>
9869         PR target/56975
9870         * config/i386/cygming.h (TARGET_PECOFF): Define as true.
9871         * config/i386/i386.h (TARGET_PECOFF): Define by default as false.
9872         (PIC_OFFSET_TABLE_REGNUM): Use TARGET_PECOFF.
9873         * config/i386/i386.c (ix86_option_override_internal): Likewise.
9874         (ix86_expand_prologue): Likewise.
9875         (ix86_expand_split_stack_prologue): Likewise.
9876         (legitimate_pic_address_disp_p): Likewise.
9877         (legitimize_pic_address): Likewise.
9878         (legitimize_tls_address): Likewise.
9879         (legitimize_pe_coff_symbol): Likewise.
9880         (output_pic_addr_const): Likewise.
9881         (construct_plt_address): Likewise.
9882         (ix86_expand_call): Likewise.
9883         (x86_output_mi_thunk): Likewise.
9884         (x86_function_profiler): Likewise.
9886 2013-05-13  Sofiane Naci  <sofiane.naci@arm.com>
9888         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<mode>): Group
9889         similar switch cases.
9890         (aarch64_simd_mov): Rename to aarch64_split_simd_mov. Update.
9891         (aarch64_simd_mov_to_<mode>low): Delete.
9892         (aarch64_simd_mov_to_<mode>high): Delete.
9893         (move_lo_quad_<mode>): Add w<-r alternative.
9894         (aarch64_simd_move_hi_quad_<mode>): Likewise.
9895         (aarch64_simd_mov_from_*): Update type attribute.
9896         * config/aarch64/aarch64.c (aarch64_split_simd_move): Refacror switch
9897         statement.
9899 2013-05-13  Jan Hubicka  <jh@suse.cz>
9901         * mode-switching.c (optimize_mode_switching): Set correct RTL profile.
9902         * config/i386/i386.c (ix86_compute_frame_layout,
9903         ix86_expand_epilogue, emit_i387_cw_initialization,
9904         ix86_expand_vector_move_misalign, ix86_fp_comparison_strategy,
9905         ix86_local_alignment): Fix use of size/speed predicates.
9907 2013-05-13  Jakub Jelinek  <jakub@redhat.com>
9909         PR tree-optimization/45216
9910         PR tree-optimization/57157
9911         * tree-ssa-forwprop.c (simplify_rotate): Only recognize
9912         the (-Y) & (B - 1) variant if OP is |.
9913         * expmed.c (expand_shift_1): For rotations by const0_rtx just
9914         return shifted.  Use (-op1) & (prec - 1) as other_amount
9915         instead of prec - op1.
9917 2013-05-13  Martin Jambor  <mjambor@suse.cz>
9919         PR middle-end/42371
9920         * ipa-prop.h (IPA_UNDESCRIBED_USE): New macro.
9921         (ipa_constant_data): New type.
9922         (ipa_jump_func): Use ipa_constant_data to hold information about
9923         constant jump functions.
9924         (ipa_get_jf_constant): Adjust to jump function type changes.
9925         (ipa_get_jf_constant_rdesc): New function.
9926         (ipa_param_descriptor): New field controlled_uses.
9927         (ipa_get_controlled_uses): New function.
9928         (ipa_set_controlled_uses): Likewise.
9929         * ipa-ref.h (ipa_find_reference): Declare.
9930         * ipa-prop.c (ipa_cst_ref_desc): New type.
9931         (ipa_print_node_jump_functions_for_edge): Adjust for jump function type
9932         changes.
9933         (ipa_set_jf_constant): Likewise.  Also create reference descriptions.
9934         New parameter cs.  Adjust all callers.
9935         (ipa_analyze_params_uses): Detect uncontrolled and controlled uses.
9936         (remove_described_reference): New function.
9937         (jfunc_rdesc_usable): Likewise.
9938         (try_make_edge_direct_simple_call): Decrement controlled use count,
9939         attempt to remove reference if it hits zero.
9940         (combine_controlled_uses_counters): New function.
9941         (propagate_controlled_uses): Likewise.
9942         (ipa_propagate_indirect_call_infos): Call propagate_controlled_uses.
9943         (ipa_edge_duplication_hook): Duplicate reference descriptions.
9944         (ipa_print_node_params): Print described use counter.
9945         (ipa_write_jump_function): Adjust to jump function type changes.
9946         (ipa_read_jump_function): New parameter CS, pass it to
9947         ipa_set_jf_constant.  Adjust caller.
9948         (ipa_write_node_info): Stream controlled use count
9949         (ipa_read_node_info): Likewise.
9950         * cgraph.c (cgraph_mark_address_taken_node): Bail out instead of
9951         asserting.
9952         * ipa-cp.c (ipcp_discover_new_direct_edges): Decrement controlled use
9953         count.  Remove cloning-added reference if it reaches zero.
9954         * ipa-ref.c (ipa_find_reference): New function.
9956 2013-05-13  Ganesh Gopalasubramanian  <Ganesh.Gopalasubramanian@amd.com>
9958         * config/i386/i386.c (processor_target_table): Modified default
9959         alignment values for AMD BD and BT architectures.
9961 2013-05-13  Marc Glisse  <marc.glisse@inria.fr>
9963         * tree-vect-generic.c (uniform_vector_p): Move ...
9964         * tree.c (uniform_vector_p): ... here.
9965         * tree.h (uniform_vector_p): Declare it.
9966         * fold-const.c (fold_binary_loc) <shift>: Turn the second argument
9967         into a scalar.
9969 2013-05-13  Jakub Jelinek  <jakub@redhat.com>
9971         PR tree-optimization/57230
9972         * tree-ssa-strlen.c (handle_char_store): Record length for
9973         array store from STRING_CST.
9975         PR tree-optimization/57230
9976         * tree-ssa-strlen.c (handle_char_store): Add missing integer_zerop
9977         check.
9979 2013-05-12  Joern Rennecke  <joern.rennecke@embecosm.com>
9981         * config/epiphany/epiphany.c (epiphany_init): Check size of
9982         NUM_MODES_FOR_MODE_SWITCHING.
9983         (epiphany_expand_prologue):
9984         Remove CONFIG_REGNUM initial value handling code.
9985         (epiphany_optimize_mode_switching): Handle EPIPHANY_MSW_ENTITY_CONFIG.
9986         (epiphany_mode_needed, epiphany_mode_entry_exit): Likewise.
9987         (emit_set_fp_mode, epiphany_mode_after): Likewise.
9988         (epiphany_mode_needed) <Handle EPIPHANY_MSW_ENTITY_AND>:
9989         Don't return 1 for FP_MODE_NONE.
9990         * config/epiphany/epiphany.h (NUM_MODES_FOR_MODE_SWITCHING):
9991         Add value for EPIPHANY_MSW_ENTITY_CONFIG.
9992         (EPIPHANY_MSW_ENTITY_CONFIG, EPIPHANY_MSW_ENTITY_NUM): Define.
9993         * config/epiphany/epiphany.md (save_config): New pattern.
9995 2013-05-12  Uros Bizjak  <ubizjak@gmail.com>
9997         * config/i386/i386.md (*zero_extendsidi2): Add *x->?r alternative.
9999 2013-05-10  Uros Bizjak  <ubizjak@gmail.com>
10001         * config/i386/i386.md (memory): Handle sseishft1.
10002         * config/i386/sse.md (*vec_extractv4si): Remove memory attribute.
10003         (*vec_extractv2di_1): Ditto.
10005 2013-05-10  Vladimir Makarov  <vmakarov@redhat.com>
10007         * lra-assigns.c (find_hard_regno_for): Add 1 to the cost of call
10008         saved registers.
10010 2013-05-10  Sebastian Huber  <sebastian.huber@embedded-brains.de>
10012         * config/arm/t-rtems-eabi: Remove mthumb/march=armv7 multilib.
10013         Add mthumb/march=armv7-a multilib.
10014         Add mthumb/march=armv7-r multilib.
10015         Add mthumb/march=armv7-a/mfpu=neon/mfloat-abi=hard multilib.
10017 2013-05-10  Ralf Corsépius  <ralf.corsepius@rtems.org>
10019         * config/v850/t-rtems: Add more multilibs.
10021 2013-05-10  Richard Biener  <rguenther@suse.de>
10023         PR tree-optimization/57214
10024         * tree-ssa-loop-ivcanon.c (propagate_constants_for_unrolling): Do
10025         not propagate from SSA names that occur in abnormal PHI nodes.
10027 2013-05-10  Marc Glisse  <marc.glisse@inria.fr>
10029         * stor-layout.c (element_precision): New function.
10030         * machmode.h (element_precision): Declare it.
10031         * tree.c (build_minus_one_cst): New function.
10032         (element_precision): Likewise.
10033         * tree.h (build_minus_one_cst): Declare new function.
10034         (element_precision): Likewise.
10035         * fold-const.c (operand_equal_p): Use element_precision.
10036         (fold_binary_loc): Handle vector types.
10037         * convert.c (convert_to_integer): Use element_precision.
10038         * gimple.c (iterative_hash_canonical_type): Handle complex and vectors
10039         separately.
10041 2013-05-10  Richard Sandiford  <rdsandiford@googlemail.com>
10043         * config/mips/mips-protos.h (m16_uimm3_b, m16_simm4_1, m16_nsimm4_1)
10044         (m16_simm5_1, m16_nsimm5_1, m16_uimm5_4, m16_nuimm5_4, m16_simm8_1)
10045         (m16_nsimm8_1, m16_uimm8_1, m16_nuimm8_1, m16_uimm8_m1_1, m16_uimm8_4)
10046         (m16_nuimm8_4, m16_simm8_8, m16_nsimm8_8): Delete.
10047         * config/mips/mips.c (m16_check_op, m16_uimm3_b, m16_simm4_1)
10048         (m16_nsimm4_1, m16_simm5_1, m16_nsimm5_1, m16_uimm5_4, m16_nuimm5_4)
10049         (m16_simm8_1, m16_nsimm8_1, m16_uimm8_1, m16_nuimm8_1, m16_uimm8_m1_1)
10050         (m16_uimm8_4, m16_nuimm8_4, m16_simm8_8, m16_nsimm8_8): Delete.
10051         * config/mips/constraints.md (Udb8, Usb5, Usb8, Usd8, Uub8, Uuw5)
10052         (Uuw8): New constraints.
10053         (Usb4): Move into alphabetical order.
10054         * config/mips/predicates.md (db8_operand, sb5_operand, sb8_operand)
10055         (sd8_operand, ub8_operand, uw8_operand): New predicates.
10056         * config/mips/mips.md (*xor<mode>3, *xor<mode>3_mips16): Name
10057         previously unnamed patterns.
10058         (*add<mode>3_mips16, *xor<mode>3_mips16, *<optab>si3_mips16)
10059         (*ashldi3_mips16, *ashrdi3_mips16, *lshrdi3_mips16)
10060         (*slt<u>_<GPR:mode><GPR2:mode>_mips16)
10061         (*sle<u>_<GPR:mode><GPR2:mode>_mips16): Use constraints instead
10062         of set_attr_alternative/if_then_else.  Use extended_mips16 instead
10063         of specific lengths.
10065 2013-05-10  Jakub Jelinek  <jakub@redhat.com>
10067         * config/i386/i386.md (rotateinv): New code attr.
10068         (*<rotate_insn><mode>3_1, *<rotate_insn>si3_1_zext,
10069         *<rotate_insn>qi3_1_slp): Emit rorl %eax instead of
10070         roll $31, %eax, etc.
10072         PR tree-optimization/45216
10073         PR tree-optimization/57157
10074         * tree-ssa-forwprop.c (simplify_rotate): New function.
10075         (ssa_forward_propagate_and_combine): Call it.
10077 2013-05-10  Richard Biener  <rguenther@suse.de>
10079         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Do not
10080         disable peeling when we version for aliasing.
10081         (vector_alignment_reachable_p): Honor explicit user alignment.
10082         (vect_supportable_dr_alignment): Likewise.
10083         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Use
10084         STMT_VINFO_LOOP_PHI_EVOLUTION_PART instead of recomputing it.
10085         * tree-vect-loop.c (vect_transform_loop): First apply versioning,
10086         then peeling to arrange for the cost-model check to come first.
10088 2013-05-10  Alan Modra  <amodra@gmail.com>
10090         * configure.ac (HAVE_AS_TLS): Swap powerpc64 and powerpc cases.
10091         (HAVE_LD_LARGE_TOC): Don't mention AIX in help text.
10092         * configure: Regenerate.
10094 2013-05-10  Alan Modra  <amodra@gmail.com>
10096         PR target/55033
10097         * varasm.c (default_elf_select_section): Move !DECL_P check..
10098         (get_named_section): ..to here before calling get_section_name.
10099         Adjust assertion.
10100         (default_section_type_flags): Add DECL_P check.
10101         * config/i386/winnt.c (i386_pe_section_type_flags): Likewise.
10102         * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags): Likewise.
10104 2013-05-09  Joern Rennecke  <joern.rennecke@embecosm.com>
10106         * config/epiphany/epiphany.c (epiphany_expand_prologue):
10107         When using gen_stack_adjust_str with a register offset, add a
10108         REG_FRAME_RELATED_EXPR note.
10110 2013-05-09  Uros Bizjak  <ubizjak@gmail.com>
10112         * config/i386/sse.md (*vec_extractv4si_0_zext): New pattern.
10113         (*vec_extractv4si_zext_mem): Ditto.
10114         (*vec_extractv2di): Add 0->x and x->x alternatives.
10115         * config/i386/mmx.md (*vec_extractv2si_zext_mem): New pattern.
10116         * config/i386/i386.md (*zero_extendsidi2): Add *Yj->?r alternative.
10118 2013-05-09  Jason Merrill  <jason@redhat.com>
10120         N3639 C++1y VLA support
10121         * gimplify.c (gimplify_vla_decl): Don't touch an existing
10122         DECL_VALUE_EXPR.
10124         * tree.c (build_constructor_va): New.
10125         * tree.h: Declare it.
10127 2013-05-09  Martin Jambor  <mjambor@suse.cz>
10129         PR lto/57084
10130         * gimple-fold.c (canonicalize_constructor_val): Call
10131         cgraph_get_create_real_symbol_node instead of cgraph_get_create_node.
10133 2013-05-09  Jan Hubicka  <jh@suse.cz>
10134             Richard Biener  <rguenther@suse.de>
10136         PR lto/54095
10137         * symtab.c (symtab_make_decl_local): Do not add private names.
10139 2013-05-09  Jan Hubicka  <jh@suse.cz>
10141         PR lto/54095
10142         * symtab.c (insert_to_assembler_name_hash): Handle clones.
10143         (unlink_from_assembler_name_hash): Likewise.
10144         (symtab_prevail_in_asm_name_hash, symtab_register_node,
10145         symtab_unregister_node, symtab_initialize_asm_name_hash,
10146         change_decl_assembler_name): Update.
10148 2013-05-09  Sofiane Naci  <sofiane.naci@arm.com>
10150         * config/aarch64/aarch64.md: New movtf split.
10151         (*movtf_aarch64): Update.
10152         (aarch64_movdi_tilow): Handle TF modes and rename to
10153         aarch64_movdi_<mode>low.
10154         (aarch64_movdi_tihigh): Handle TF modes and rename to
10155         aarch64_movdi_<mode>high
10156         (aarch64_movtihigh_di): Handle TF modes and rename to
10157         aarch64_mov<mode>high_di
10158         (aarch64_movtilow_di): Handle TF modes and rename to
10159         aarch64_mov<mode>low_di
10160         (aarch64_movtilow_tilow): Remove spurious whitespace.
10161         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Handle TFmode
10162         splits.
10163         (aarch64_print_operand): Update.
10165 2013-05-09  Alan Modra  <amodra@gmail.com>
10167         * configure.ac (HAVE_AS_TLS): Enable tests for powerpcle and
10168         powerpc64le.
10169         * configure: Regenerate.
10171 2013-05-08  Uros Bizjak  <ubizjak@gmail.com>
10173         * config/i386/mmx.md (*vec_extract* splitters): Simplify post-reload
10174         splitter preparation statements.
10175         * config/i386/sse.md (*vec_extract* splitters): Ditto.
10176         (*avx_vperm_broadcast_<mode>): Use adjust_address instead of
10177         adjust_address_nv.
10179 2013-05-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10181         * gimple-ssa-strength-reduction.c (count_candidates): Change
10182         return value to int.
10183         (analyze_candidates_and_replace): Change type of length to int.
10185 2013-05-08  Uros Bizjak  <ubizjak@gmail.com>
10187         * config/i386/sse.md (PEXTR_MODE, PEXTR_MODEx): Remove.
10188         (*vec_extract<mode>): Use VI12_128 mode iterator.
10189         (*vec_extract<mode>_mem): Ditto.
10190         (*vec_extract*_mem splitters): Merge splitters using VI_128 mode
10191         attribute.
10193 2013-05-08  Diego Novillo  <dnovillo@google.com>
10195         PR bootstrap/54659
10197         Revert:
10198         2012-08-17  Diego Novillo  <dnovillo@google.com>
10200         PR bootstrap/54281
10201         * configure.ac: Add libintl.h to AC_CHECK_HEADERS list.
10202         * config.in: Regenerate.
10203         * configure: Regenerate.
10204         * intl.h: Always include libintl.h if HAVE_LIBINTL_H is set.
10206 2013-05-08  Jan Hubicka  <jh@suse.cz>
10208         PR lto/54095
10209         * cgraph.c (cgraph_make_node_local_1): Se unique_name.
10210         * cgraph.h (symtab_node_base): Add unique_name.
10211         * lto-cgraph.c (lto_output_node, lto_output_varpool_node,
10212         input_overwrite_node, input_varpool_node): Stream unique_name.
10213         * cgraphclones.c (cgraph_create_virtual_clone,
10214         cgraph_function_versioning): Set unique_name.
10215         * ipa.c (function_and_variable_visibility): Set unique_name.
10217 2013-05-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10219         * gimple-ssa-strength-reduction.c (find_phi_def): Revert former "fix."
10220         (alloc_cand_and_find_basis): Restrict conditional candidate
10221         processing to CAND_MULTs.
10223 2013-05-08  Jan Hubicka  <jh@suse.cz>
10225         PR lto/54095
10226         lto-symtab.c (lto_symtab_symbol_p): New function.
10227         (lto_symtab_resolve_can_prevail_p, lto_symtab_resolve_symbols,
10228         lto_symtab_resolve_symbols, lto_symtab_merge_decls_2,
10229         lto_symtab_merge_decls_1, lto_symtab_merge_cgraph_nodes_1):
10230         Skip static symbols.
10232 2013-05-08  Paolo Carlini  <paolo.carlini@oracle.com>
10234         PR tree-optimization/57200
10235         * tree-ssa-loop-niter.c (do_warn_aggressive_loop_optimizations):
10236         Only call inform if the preceding warning_at returns true.
10238 2013-05-07  Han Shen  <shenhan@google.com>
10240         * cfgexpand.c (record_or_union_type_has_array_p): New function.
10241         (expand_used_vars): Add logic handling '-fstack-protector-strong'.
10242         * common.opt (fstack-protector-strong): New option.
10243         * doc/cpp.texi (__SSP_STRONG__): New builtin "__SSP_STRONG__".
10244         * doc/invoke.texi (Optimization Options): Document
10245         "-fstack-protector-strong".
10246         * gcc.c (LINK_SSP_SPEC): Add 'fstack-protector-strong'.
10248 2013-05-06  Steven Bosscher  <steven@gcc.gnu.org>
10250         * config/mips/mips.c (mips_machine_reorg2): Return 0.
10252 2013-05-07  Vladimir Makarov  <vmakarov@redhat.com>
10254         * ira.c (update_equiv_regs): Add insn having equiv memory even if
10255         it is not lhs of the insn.
10256         (setup_reg_equiv): Remove insn having equiv memory which it is not
10257         lhs of the insn.
10258         * lra-constraints.c (process_address): Try to improve generation
10259         code for address base + disp.
10260         (lra_constraints): Make correct the code for checking insn setting
10261         up backward equivalence.  Remove insn only if it is in the init
10262         insn list.
10263         * lra-eliminations.c (update_reg_eliminate): Change return value.
10264         (lra_eliminate): Use the result.
10266 2013-05-07  Uros Bizjak  <ubizjak@gmail.com>
10268         * config/i386/sse.md (ssescalarnummask): New mode attribute.
10269         (PEXTR_MODE, PEXTR_MODEx): New mode iterators.
10270         (*vec_extract<mode>): Merge from *sse4_1_pextrb_memory and
10271         *sse4_1_pextrw_memory using PEXTR_MODE mode iterator. Handle
10272         register target operands.
10273         (*vec_extractv8hi_sse2): New pattern.
10274         (*vec_extractv16qi_zext): Rename from *sse4_1_pextrb_<mode>.
10275         (*vec_extractv8hi_zext): Rename from *sse2_pextrw_<mode>.
10276         (*vec_extract<mode>_mem): New insn and split pattern.
10278 2013-05-07  Christophe Lyon  <christophe.lyon@linaro.org>
10280         * config/arm/arm.c (arm_asan_shadow_offset): New function.
10281         (TARGET_ASAN_SHADOW_OFFSET): Define.
10282         * config/arm/linux-eabi.h (ASAN_CC1_SPEC): Define.
10283         (LINUX_OR_ANDROID_CC): Add ASAN_CC1_SPEC.
10285 2013-05-07  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10287         * gimple-ssa-strength-reduction.c (MAX_INCR_VEC_LEN): New constant.
10288         (incr_vec_index): Return -1 if increment not found.
10289         (create_add_on_incoming_edge): Assert if increment not found.
10290         (record_increment): Limit number of increments recorded.
10291         (all_phi_incrs_profitable): Return false if an increment not found.
10292         (replace_profitable_candidates): Don't process increments that were
10293         not recorded.
10294         (analyze_candidates_and_replace): Limit size of incr_vec.
10296 2013-05-07  Richard Biener  <rguenther@suse.de>
10298         * calls.c (special_function_p): setjmp-like functions are leaf.
10299         * builtins.def (BUILT_IN_SETJMP): setjmp is leaf.
10300         * tree-inline.c (update_ssa_across_abnormal_edges): Remove assert.
10302 2013-05-07  Sofiane Naci  <sofiane.naci@arm.com>
10304         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>): call splitter.
10305         (aarch64_simd_mov<mode>): New expander.
10306         (aarch64_simd_mov_to_<mode>low): New instruction pattern.
10307         (aarch64_simd_mov_to_<mode>high): Likewise.
10308         (aarch64_simd_mov_from_<mode>low): Likewise.
10309         (aarch64_simd_mov_from_<mode>high): Likewise.
10310         (aarch64_dup_lane<mode>): Update.
10311         (aarch64_dup_lanedi): New instruction pattern.
10312         * config/aarch64/aarch64-protos.h (aarch64_split_simd_move): New prototype.
10313         * config/aarch64/aarch64.c (aarch64_split_simd_move): New function.
10315 2013-05-07  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10317         * gimple-ssa-strength-reduction.c (lazy_create_slsr_reg): Remove.
10318         (replace_mult_candidate): Remove unnecessary argument; remove
10319         unnecessary parameter from call to introduce_cast_before_cand.
10320         (replace_unconditional_candidate): Remove unnecessary parameter
10321         from call to replace_mult_candidate.
10322         (replace_conditional_candidate): Likewise.
10323         (insert_initializers): Use make_temp_ssa_name.
10324         (introduce_cast_before_cand): Remove unnecessary argument; use
10325         make_temp_ssa_name.
10326         (replace_one_candidate): Remove unnecessary argument; remove
10327         unnecessary parameter from calls to introduce_cast_before_cand.
10328         (replace_profitable_candidates): Remove unnecessary parameters
10329         from calls to replace_one_candidate.
10331 2013-05-07  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10333         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
10334         phi def as possibly hiding a basis for a CAND_ADD whose operands
10335         have been commuted in the analysis.
10336         (alloc_cand_and_find_basis): Add parms to call to find_phi_def.
10338 2013-05-07  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
10340         * config/aarch64/aarch64.md
10341         (cmp_swp_<optab><ALLX:mode>_shft_<GPI:mode>): Restrict the
10342         shift value between 0-4.
10344 2013-05-07  Richard Biener  <rguenther@suse.de>
10346         * double-int.h (rshift): New overload.
10347         * double-int.c (rshift): New function.
10348         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Optimize.
10349         (create_reference_ops_from_ref): Remove.
10350         (vn_reference_insert): Use shared ops for constructing the
10351         reference and copy it.
10353 2013-05-07  Richard Biener  <rguenther@suse.de>
10355         PR middle-end/57190
10356         * tree-eh.c (sink_clobbers): Properly propagate
10357         SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
10359 2013-05-07  Jakub Jelinek  <jakub@redhat.com>
10361         PR tree-optimization/57149
10362         * tree-ssa-uninit.c (uninit_undefined_value_p): New inline.
10363         (can_skip_redundant_opnd, compute_uninit_opnds_pos,
10364         collect_phi_def_edges, execute_late_warn_uninitialized): Use
10365         uninit_undefined_value_p instead of ssa_undefined_value_p.
10367         PR debug/57184
10368         * expr.c (expand_expr_addr_expr_1): Handle COMPOUND_LITERAL_EXPR
10369         for modifier == EXPAND_INITIALIZER.
10371 2013-05-07  Anton Blanchard  <anton@samba.org>
10373         * configure.ac (HAVE_LD_LARGE_TOC): Use correct linker emulation
10374         for powerpc64 little endian.
10375         * configure: Regenerate.
10377 2013-05-06  Graham Stott  <grahams@btinternet.com>
10379         * expmed.c (init_expmed_rtl): Remove unused fields reg_fld, plus_fld,
10380         mult_fld, sdiv_fld1, udiv_fld1, sdiv_32_fld1, smod_32_fld1,
10381         wide_mult_fld1, wide_lshr_fld1, shift_fld1, shift_mult_fld1,
10382         shift_add_fld1, shift_sub0_fld1, shift_sub1_fld1.
10384 2013-05-06  Graham Stott  <grahams@btinternet.com>
10386         * gensupport.c (add_predicate_code): Also exclude SCRATCH from rtx
10387         codes which allow non-lvalues.
10389 2013-05-06  Marc Glisse  <marc.glisse@inria.fr>
10391         * tree.c (integer_all_onesp) <COMPLEX_CST>: Test that both
10392         components are all 1s.
10393         (integer_minus_onep): New function.
10394         * tree.h (integer_minus_onep): Declare it.
10395         * fold-const.c (fold_binary_loc) <MULT_EXPR>: Test
10396         integer_minus_onep instead of integer_all_onesp.
10398 2013-05-06  Oleg Endo  <olegendo@gcc.gnu.org>
10400         PR target/52933
10401         * config/sh/sh.md (*cmp_div0s_0, *cmp_div0s_1, *movsicc_div0s): Add
10402         variations of these patterns.
10404 2013-05-06  Uros Bizjak  <ubizjak@gmail.com>
10406         * config/i386/i386.md (isa): Add x64_sse4 member.
10407         (enabled): Handle x64_sse4.
10408         (*movdi_internal): Add *x->?r alternative to emit pextrq $0,%xmm,%reg
10409         instruction for 64bit SSE4_1 targets.  Update insn attributes.
10410         (*movsi_internal): Add *x->?r alternative to emit pextrd $0,%xmm,%reg
10411         instruction for SSE4_1 targets.  Update insn attributes.
10412         * config/i386/sse.md (*vec_extract<ssevecmodelower>_0): Merge
10413         with *sse4_1_pextrd and *sse4_1_pextrq having const_0 selector.
10414         (*vec_extractv2di_1): Merge with *sse4_1_pextrq having
10415         const_1 selector.
10416         (*vec_extractv4si): Rename from *sse4_1_pextrd.
10417         (*vec_extractv4si_zext): Rename from *sse4_1_pextrd_zext.
10418         (*vec_extract<ssevecmodelower>_0 splitters): Merge splitters together.
10420 2013-05-06  Oleg Endo  <olegendo@gcc.gnu.org>
10422         PR target/57108
10423         * config/sh/sh.md (tstsi_t_zero_extract_eq): Use QIHISIDI mode iterator.
10425 2013-05-06  Maxim Kuznetsov  <maks.kuznetsov@gmail.com>
10427         * final.c (do_assembler_dialects): Don't handle curly braces and
10428         vertical bar escaped by % as dialect delimiters.
10429         (output_asm_insn): Print curly braces and vertical bar if escaped
10430         by % and ASSEMBLER_DIALECT defined.
10431         * doc/tm.texi.in (ASSEMBLER_DIALECT): Document new standard escapes.
10432         * doc/tm.texi: Regenerated.
10434 2013-05-06  Steven Bosscher  <steven@gcc.gnu.org>
10436         * config/mips/mips.c: Include tree-pass.h.
10437         (mips_reorg): Split in pre- and post-dbr_schedule parts.
10438         (mips_machine_reorg2): Move mips_reorg post-dbr_schedule parts here.
10439         (pass_mips_machine_reorg2): New machine specific pass.
10440         (insert_pass_mips_machine_reorg2): New pass plugin definition.
10441         (mips_option_override): Register the new pass.
10442         * rtl.h (cleanup_barriers): Remove prototype.
10443         (dbr_schedule): Likewise.
10444         * jump.c (cleanup_barriers): Make static.
10445         * reorg.c (dbr_schedule): Likewise.
10447 2013-05-06  Richard Biener  <rguenther@suse.de>
10449         PR tree-optimization/57185
10450         * tree-parloops.c (add_field_for_reduction): Handle anonymous
10451         SSA names properly.
10453 2013-05-06  Uros Bizjak  <ubizjak@gmail.com>
10455         PR target/57106
10456         * config/i386/i386.c (add_parameter_dependencies): Add dependence
10457         between "first_arg" and "insn", not "last" and "insn".
10459 2013-05-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10461         * gimple-ssa-strength-reduction.c (slsr_process_phi): Re-enable.
10462         (find_candidates_in_block): Re-enable slsr_process_phi.
10463         (create_phi_basis): Fix double counting of candidate adjustment.
10465 2013-05-06  Richard Biener  <rguenther@suse.de>
10467         PR middle-end/57147
10468         * tree-cfg.c (gimple_purge_dead_abnormal_call_edges): If
10469         the edge is also fallthru, preserve it and just clear the
10470         abnormal flag.
10471         * tree-cfgcleanup.c (remove_fallthru_edge): If the edge is
10472         also complex, preserve that and just clear the fallthru flag.
10473         * tree-inline.c (update_ssa_across_abnormal_edges): Also
10474         update virtual operands.
10476 2013-05-06  Alan Modra  <amodra@gmail.com>
10478         * config/rs6000/linux.h (DEFAULT_ASM_ENDIAN): Define.
10479         (LINK_OS_LINUX_EMUL): Use ENDIAN_SELECT.
10480         * config/rs6000/linux64.h (DEFAULT_ASM_ENDIAN): Define.
10481         * config/rs6000/sysv4le.h (DEFAULT_ASM_ENDIAN): Define.
10482         (LINK_TARGET_SPEC): Use ENDIAN_SELECT.
10483         * config/rs6000/sysv4.h (DEFAULT_ASM_ENDIAN): Define as -mbig.
10485 2013-05-06  Alan Modra  <amodra@gmail.com>
10487         * config/rs6000/sysv4.h (ENDIAN_SELECT): Define, extracted from
10488         (ASM_SPEC): ..here.  Emit DEFAULT_ASM_ENDIAN too.
10489         (DEFAULT_ASM_ENDIAN): Define.
10490         (CC1_SPEC, LINK_TARGET_SPEC): Use ENDIAN_SELECT.
10491         * config/rs6000/linux64.h (ASM_SPEC32): Remove endian options.
10492         Update -K PIC clause from sysv4.h.
10493         (ASM_SPEC_COMMON): Use ENDIAN_SELECT.
10494         (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Likewise.
10496 2013-05-06  Alan Modra  <amodra@gmail.com>
10498         * config/rs6000/rs6000.md (bswapdi 2nd splitter): Don't swap words
10499         twice for little-endian.
10500         (ashrdi3_no_power, ashrdi3): Support little-endian.
10502 2013-05-06  Oleg Endo  <olegendo@gcc.gnu.org>
10504         PR target/55303
10505         * config/sh/sh.c (sh_rtx_costs): Handle SMIN and SMAX cases.
10506         * config/sh/sh.md (*clips, uminsi3, *clipu, clipu_one): New insns and
10507         related expanders.
10508         * config/sh/iterators.md (SMIN_SMAX): New code iterator.
10509         * config/sh/predicates.md (arith_reg_or_0_or_1_operand,
10510         clips_min_const_int, clips_max_const_int, clipu_max_const_int):
10511         New predicates.
10513 2013-05-05  Steven Bosscher  <steven@gcc.gnu.org>
10514             John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
10516         * config.gcc (hppa*-*-*): Remove MASK_BIG_SWITCH from CPU default.
10517         * config/pa/pa.opt: Make mbig-switch a no-op.
10518         * config/pa/pa.h (TARGET_DEFAULT): Remove MASK_BIG_SWITCH.
10519         (CASE_VECTOR_MODE): Always return SImode.
10520         (ASM_OUTPUT_ADDR_VEC_ELT, ASM_OUTPUT_ADDR_DIFF_ELT): Remove code
10521         for the !TARGET_BIG_SWITCH case.
10522         * config/pa/pa-linux.h: Likewise.
10523         * config/pa/pa-openbsd.h: Likewise.
10524         * config/pa/pa-hpux.h: Define TARGET_DEFAULT to 0.
10525         * config/pa/pa.md (short_jump): Remove define_insn.
10526         (casesi): Remove code for the !TARGET_BIG_SWITCH case.
10527         (casesi0): Remove define_insn.
10528         (type): Remove btable_branch.
10529         (pa_combine_type): Likewise.
10530         (in_nullified_branch_delay): Likewise.
10531         (in_call_delay): Likewise.
10532         (define_delay): Likewise.
10533         (define_insn_reservation "Z3"): Likewise.
10534         (define_insn_reservation "Z4"): Likewise.
10535         * config/pa/pa.c  (pa_reorg): Remove code for !TARGET_BIG_SWITCH.
10536         (pa_adjust_insn_length): Remove adjustment for btable branches.
10537         * doc/invoke.texi (HPPA Options): Delete documentation for mbig-switch
10538         and mno-big-switch
10540 2013-05-05  Uros Bizjak  <ubizjak@gmail.com>
10542         * config/i386/sse.md (*vec_extract<ssevecmodelower>_0): Merge
10543         from sse2_stored and *sse2_storeq_rex64 using SWI48 mode iterator.
10544         Add m->r,x alternatives.
10545         (*vec_extract<ssevecmodelower>_0 splitters): Merge V2DI and V4SI
10546         splitters using SWI48x mode iterator.
10547         (*vec_extract_v2di_0_sse): Rename from *sse2_storeq.  Disable for
10548         TARGET_64BIT.  Add m->x alternative.
10549         (*vec_extractv4si_mem): Rename from *vec_ext_v4si_mem.
10550         Add o->x alternative.  Enable for TARGET_SSE.
10551         (sse_storeq): Remove expander.
10552         (*vec_extractv2di_1): Enable for TARGET_SSE. Split alternatives
10553         with memory input operand.
10554         (*vec_extractv2di_1 splitter): New.
10555         (*vec_extractv4sf_mem): Rename from *vec_extract_v4sf_mem.
10556         * config/i386/i386.md (ssevecmodelower): New mode attribute.
10558 2013-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
10560         * config/rs6000/rs6000.c (INT_P): Reformat.  Delete obsolete comment.
10561         (INT_LOWPART): Delete.
10562         (extract_MB): Adjust.
10563         (extract_ME): Adjust.
10564         (print_operand): Adjust.
10566 2013-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
10568         * config/rs6000/predicates.md (reg_or_add_cint_operand,
10569         reg_or_sub_cint_operand): Delete "HOST_BITS_PER_WIDE_INT == 32" case.
10570         (reg_or_logical_cint_operand, easy_fp_constant,
10571         logical_const_operand): Delete "CONST_DOUBLE" case.
10572         * config/rs6000/rs6000.c (num_insns_constant_wide): Delete
10573         "HOST_BITS_PER_WIDE_INT == 64" test.
10574         (num_insns_constant): Ditto.  Delete CONST_DOUBLE DImode/VOIDmode case.
10575         (build_mask64_2_operands): Delete "HOST_BITS_PER_WIDE_INT >= 64" test.
10576         (rs6000_emit_set_const): Delete CONST_DOUBLE case.
10577         (rs6000_emit_set_long_const): Delete "HOST_BITS_PER_WIDE_INT >= 64"
10578         test.
10579         (includes_rldic_lshift_p, includes_rldicr_lshift_p): Delete
10580         CONST_DOUBLE DImode/VOIDmode case.
10581         (INT_P, INT_LOWPART): Delete CONST_DOUBLE case.
10582         (print_operand): Delete "HOST_BITS_PER_WIDE_INT == 32" case.  Delete
10583         CONST_DOUBLE VOIDmode case.
10584         (output_toc): Delete "HOST_BITS_PER_WIDE_INT == 32" case.
10585         (rs6000_rtx_costs): Delete CONST_DOUBLE DImode/VOIDmode case.
10586         * config/rs6000/rs6000.md (iordi3, xordi3, splitter for these):
10587         Delete CONST_DOUBLE case.
10588         (splitters for mov FMOVE64 const_double): Delete
10589         "HOST_BITS_PER_WIDE_INT == 32" case.  Delete
10590         "HOST_BITS_PER_WIDE_INT >= 64" test.
10591         (splitter for mov DI const_int): Delete "HOST_BITS_PER_WIDE_INT == 32"
10592         case.
10593         (mov DI const_double): Delete.
10595 2013-05-04  Jakub Jelinek  <jakub@redhat.com>
10597         * combine.c (combine_simplify_rtx) <case SUBREG>: If nonzero_bits
10598         on op shows all bits zero in mode of a lowpart subreg, return zero.
10600 2013-05-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
10602         PR target/57150
10603         * config/rs6000/rs6000.h (HARD_REGNO_CALLER_SAVE_MODE): Use DFmode
10604         to save TFmode registers and DImode to save TImode registers for
10605         caller save operations.
10606         (HARD_REGNO_CALL_PART_CLOBBERED): TFmode and TDmode do not need to
10607         mark being partially clobbered since they only use the first
10608         double word.
10610         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): TFmode
10611         and TDmode only use the upper 64-bits of each VSX register.
10613 2013-05-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10615         * gimple-ssa-strength-reduction.c (slsr_process_phi): Disable.
10616         (find_candidates_in_block): Disable slsr_process_phi.
10618 2013-05-03  Guozhi Wei  <carrot@google.com>
10620         * coverage.c (coverage_obj_init): Move the construction of gcov
10621         constructor to ...
10622         (build_init_ctor): ... here.
10624 2013-05-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
10626         * gimple-ssa-strength-reduction.c (cand_kind): Add CAND_PHI.
10627         (slsr_cand_d): Redefine def_phi.
10628         (stride_status, phi_adjust_status, count_phis_status): New enums.
10629         (find_phi_def): New.
10630         (find_basis_for_base_expr): New.
10631         (find_basis_for_candidate): Handle hidden bases.
10632         (alloc_cand_and_find_basis): Handle phi candidates.
10633         (slsr_process_phi): New.
10634         (create_mul_ssa_cand): Exclude phi base candidates; use integer_onep.
10635         (create_mul_imm_cand): Likewise.
10636         (create_add_ssa_cand): Exclude phi base candidates.
10637         (create_add_imm_cand): Likewise.
10638         (slsr_process_cast): Likewise.
10639         (slsr_process_copy): Likewise.
10640         (find_candidates_in_block): Handle phi candidates.
10641         (dump_candidate): Likewise.
10642         (unconditional_cands): Delete.
10643         (unconditional_cands_with_known_stride_p): Delete.
10644         (phi_dependent_cand_p): New.
10645         (cand_increment): Handle phi-dependent candidates.
10646         (replace_dependent): Delete.
10647         (replace_mult_candidate): New.
10648         (replace_unconditional_candidate): New.
10649         (incr_vec_index): Move to avoid forward reference.
10650         (create_add_on_incoming_edge): New.
10651         (create_phi_basis): New.
10652         (replace_dependents): Delete.
10653         (replace_conditional_candidate): New.
10654         (phi_add_costs): New.
10655         (replace_uncond_cands_and_profitable_phis): New.
10656         (record_increment): Handle phi adjustments.
10657         (record_phi_increments): New.
10658         (record_increments): Handle phi adjustments.
10659         (phi_incr_cost): New.
10660         (lowest_cost_path): Handle phis.
10661         (total_savings): Likewise.
10662         (analyze_increments): Likewise.
10663         (ncd_with_phi): New.
10664         (ncd_of_cand_and_phis): New.
10665         (nearest_common_dominator_for_cands): Handle phi increments.
10666         (all_phi_incrs_profitable): New.
10667         (replace_profitable_candidates): Handle phi-dependent candidates.
10668         (analyze_candidates_and_replace): Likewise.
10670 2013-05-03  Teresa Johnson  <tejohnson@google.com>
10672         PR bootstrap/57154
10673         * sched-rgn.c (compute_dom_prob_ps): Ensure accumulated probabilities
10674         do not exceed REG_BR_PROB_BASE.
10676 2013-05-03  Jeff Law  <law@redhat.com>
10678         PR tree-optimization/57144
10679         * tree-vrp.c (simplify_cond_using_ranges): Verify the constant
10680         operand of the condition will bit into the new type when eliminating
10681         a cast feeding a condition.
10683 2013-05-03  Jakub Jelinek  <jakub@redhat.com>
10685         PR rtl-optimization/57130
10686         * combine.c (make_compound_operation) <case SUBREG>: Pass SET instead
10687         of COMPARE as in_code to the recursive call if needed.
10689 2013-05-03  Uros Bizjak  <ubizjak@gmail.com>
10691         * config/i386/i386.md (isa): Add x64_sse4_noavx and x64_avx members.
10692         (enabled): Handle new members.
10693         * config/i386/sse.md (*vec_concatv2si): Merge from
10694         *vec_concatv2si_sse2 and vec_concatv2si_sse.
10695         (vec_concatv2di): Merge with *vec_concatv2di_rex64.
10697 2013-05-03  Joern Rennecke  <joern.rennecke@embecosm.com>
10699         PR tree-optimization/57027
10700         * tree-ssa-math-opts.c (convert_mult_to_fma): When checking
10701         for fnms opportunity, check we got the prerequisite kind
10702         of tree / gimple before using accessor functions.
10704 2013-05-03  Richard Biener  <rguenther@suse.de>
10706         * double-int.h (lshift): New overload without precision
10707         and arith argument.
10708         (operator *=, operator +=, operator -=): Move ...
10709         * double-int.c (operator *=, operator +=, operator -=): ... here
10710         and implement more efficiently.
10711         (mul_double_with_sign): Remove.
10712         (lshift_double): Adjust to take unsinged shift argument, push
10713         dispatching code to callers.
10714         (mul_double_wide_with_sign): Add early out for callers that
10715         are not interested in high parts or overflow.
10716         (lshift): New function.
10717         (lshift, rshift, alshift, arshift, llshift, lrshift): Add
10718         dispatch code here.
10719         (lrotate, rrotate): Use logical shifts.
10720         * expr.c (get_inner_reference): Use lshift.
10721         * fixed-value.c (do_fixed_divide): Likewise.
10722         * tree-dfa.c (get_ref_base_and_extent): Likewise.
10723         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
10724         (indirect_refs_may_alias_p): Likewise.
10725         (stmt_kills_ref_p_1): Likewise.
10727 2013-05-03  Vidya Praveen  <vidyapraveen@arm.com>
10729         * config/aarch64/aarch64-simd.md (simd_fabd): Correct the description.
10731 2013-05-03  Vidya Praveen  <vidyapraveen@arm.com>
10733         * config/aarch64/aarch64-simd.md (*fabd_scalar<mode>3): Support
10734         scalar form of FABD instruction.
10736 2013-05-02  Vladimir Makarov  <vmakarov@redhat.com>
10738         * lra-constraints.c (process_alt_operands): Add checking alt
10739         number to choose the best alternative.
10741 2013-05-02  Richard Biener  <rguenther@suse.de>
10743         * tree-eh.c (cleanup_empty_eh_merge_phis): Remove rename_virts
10744         bitmap and its handling.
10745         (pass_cleanup_eh): Set todo_flags_finish to TODO_verify_ssa.
10747 2013-05-02  Richard Biener  <rguenther@suse.de>
10749         PR middle-end/57140
10750         * tree-inline.c (copy_loops): Properly handle removed loops.
10751         (copy_cfg_body): Mark destination loops for fixup if source
10752         loops needed fixup.
10754 2013-05-02  Greta Yorsh  <Greta.Yorsh@arm.com>
10756         PR target/56732
10757         * config/arm/arm.c (arm_expand_epilogue): Check really_return before
10758         generating simple_return for naked functions.
10760 2013-05-02  Martin Jambor  <mjambor@suse.cz>
10762         PR middle-end/56988
10763         * ipa-prop.h (ipa_agg_replacement_value): New flag by_ref.
10764         * ipa-cp.c (ipa_get_indirect_edge_target_1): Also check that by_ref
10765         flags match.
10766         (find_aggregate_values_for_callers_subset): Fill in the by_ref flag of
10767         ipa_agg_replacement_value structures.
10768         (known_aggs_to_agg_replacement_list): Likewise.
10769         * ipa-prop.c (write_agg_replacement_chain): Stream by_ref flag.
10770         (read_agg_replacement_chain): Likewise.
10771         (ipcp_transform_function): Also check that by_ref flags match.
10773 2013-05-02  Richard Biener  <rguenther@suse.de>
10775         * graphds.h (struct graph): Add obstack member.
10776         * graphds.c (new_graph): Initialize obstack and allocate
10777         vertices from it.
10778         (add_edge): Allocate edge from the obstack.
10779         (free_graph): Free the obstack instead of all edges and vertices.
10781 2013-05-02  Teresa Johnson  <tejohnson@google.com>
10783         * loop-unswitch.c (unswitch_loop): Use helper routines with rounding
10784         divides.
10785         * cfg.c (update_bb_profile_for_threading): Ditto.
10786         * tree-inline.c (copy_bb): Ditto.
10787         (copy_edges_for_bb): Ditto.
10788         (initialize_cfun): Ditto.
10789         (copy_cfg_body): Ditto.
10790         (expand_call_inline): Ditto.
10791         * ipa-inline-analysis.c (estimate_edge_size_and_time): Ditto.
10792         (estimate_node_size_and_time): Ditto.
10793         (inline_merge_summary): Ditto.
10794         * cgraphclones.c (cgraph_clone_edge): Ditto.
10795         (cgraph_clone_node): Ditto.
10796         * sched-rgn.c (compute_dom_prob_ps): Ditto.
10797         (compute_trg_info): Ditto.
10799 2013-05-02  Ian Bolton  <ian.bolton@arm.com>
10801         * config/aarch64/aarch64.md (movsi_aarch64): Only allow to/from
10802         S reg when fp attribute set.
10803         (movdi_aarch64): Only allow to/from D reg when fp attribute set.
10805 2013-05-02  Ian Bolton  <ian.bolton@arm.com>
10807         * config/aarch64/aarch64.md (*and_one_cmpl<mode>3_compare0):
10808         New pattern.
10809         (*and_one_cmplsi3_compare0_uxtw): Likewise.
10810         (*and_one_cmpl_<SHIFT:optab><mode>3_compare0): Likewise.
10811         (*and_one_cmpl_<SHIFT:optab>si3_compare0_uxtw): Likewise.
10813 2013-05-02  Richard Biener  <rguenther@suse.de>
10815         * tree-scalar-evolution.c (scev_info_hasher): Remove.
10816         (struct instantiate_cache_entry): New type.
10817         (struct instantiate_cache_entry_hasher): New hashtable descriptor.
10818         (struct instantiate_cache_type): New type.
10819         (set_instantiated_value, get_instantiated_value): Remove.
10820         (get_instantiated_value_entry): New function.
10821         (instantiate_scev_name): Use the new cache and adjust.
10822         (instantiate_scev_poly): Adjust.
10823         (instantiate_scev_binary): Likewise.
10824         (instantiate_array_ref): Likewise.
10825         (instantiate_scev_convert): Likewise.
10826         (instantiate_scev_not): Likewise.
10827         (instantiate_scev_3): Likewise.
10828         (instantiate_scev_2): Likewise.
10829         (instantiate_scev_r): Likewise.
10830         (instantiate_scev): Likewise.
10831         (resolve_mixers): Likewise.
10833 2013-05-01  Vladimir Makarov  <vmakarov@redhat.com>
10835         PR target/57091
10836         * lra-constraints.c (best_small_class_operands_num): Remove.
10837         (process_alt_operands): Remove small_class_operands_num.  Take
10838         small classes operands into losers and only if the operand is not
10839         matched.  Modify debugging output.
10840         (curr_insn_transform): Remove best_small_class_operands_num.
10841         Print insn name.
10843 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10845         * config/aarch64/aarch64-builtins.c
10846         (aarch64_gimple_fold_builtin.c): Fold more modes for reduc_splus_.
10847         * config/aarch64/aarch64-simd-builtins.def
10848         (reduc_splus_): Add new modes.
10849         (reduc_uplus_): New.
10850         * config/aarch64/aarch64-simd.md (aarch64_addvv4sf): Remove.
10851         (reduc_uplus_v4sf): Likewise.
10852         (reduc_splus_v4sf): Likewise.
10853         (aarch64_addv<mode>): Likewise.
10854         (reduc_uplus_<mode>): Likewise.
10855         (reduc_splus_<mode>): Likewise.
10856         (aarch64_addvv2di): Likewise.
10857         (reduc_uplus_v2di): Likewise.
10858         (reduc_splus_v2di): Likewise.
10859         (aarch64_addvv2si): Likewise.
10860         (reduc_uplus_v2si): Likewise.
10861         (reduc_splus_v2si): Likewise.
10862         (reduc_<sur>plus_<mode>): New.
10863         (reduc_<sur>plus_v2di): Likewise.
10864         (reduc_<sur>plus_v2si): Likewise.
10865         (reduc_<sur>plus_v4sf): Likewise.
10866         (aarch64_addpv4sf): Likewise.
10867         * config/aarch64/arm_neon.h
10868         (vaddv<q>_<s,u,f><8, 16, 32, 64): Rewrite using builtins.
10869         * config/aarch64/iterators.md (unspec): Remove UNSPEC_ADDV,
10870         add UNSPEC_SADDV, UNSPEC_UADDV.
10871         (SUADDV): New.
10872         (sur): Add UNSPEC_SADDV, UNSPEC_UADDV.
10874 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10876         * config/aarch64/arm_neon.h
10877         (v<max,min><nm><q><v>_<sfu><8, 16, 32, 64>): Rewrite using builtins.
10879 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10881         * config/aarch64/aarch64-builtins
10882         (aarch64_gimple_fold_builtin): Fold reduc_<su><maxmin>_ builtins.
10884 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10886         * config/aarch64/aarch64-simd-builtins.def
10887         (reduc_smax_): New.
10888         (reduc_smin_): Likewise.
10889         (reduc_umax_): Likewise.
10890         (reduc_umin_): Likewise.
10891         (reduc_smax_nan_): Likewise.
10892         (reduc_smin_nan_): Likewise.
10893         (fmax): Remove.
10894         (fmin): Likewise.
10895         (smax): Update for V2SF, V4SF and V2DF modes.
10896         (smin): Likewise.
10897         (smax_nan): New.
10898         (smin_nan): Likewise.
10899         * config/aarch64/aarch64-simd.md (<maxmin><mode>3): Rename to...
10900         (<su><maxmin><mode>3): ...This, refactor.
10901         (s<maxmin><mode>3): New.
10902         (<maxmin_uns><mode>3): Likewise.
10903         (reduc_<maxmin_uns>_<mode>): Refactor.
10904         (reduc_<maxmin_uns>_v4sf): Likewise.
10905         (reduc_<maxmin_uns>_v2si): Likewise.
10906         (aarch64_<fmaxmin><mode>: Remove.
10907         * config/aarch64/arm_neon.h (vmax<q>_f<32,64>): Rewrite to use
10908         new builtin names.
10909         (vmin<q>_f<32,64>): Likewise.
10910         * config/iterators.md (unspec): Add UNSPEC_FMAXNMV, UNSPEC_FMINNMV.
10911         (FMAXMIN): New.
10912         (su): Add mappings for smax, smin, umax, umin.
10913         (maxmin): New.
10914         (FMAXMINV): Add UNSPEC_FMAXNMV, UNSPEC_FMINNMV.
10915         (FMAXMIN): Rename as...
10916         (FMAXMIN_UNS): ...This.
10917         (maxminv): Remove.
10918         (fmaxminv): Likewise.
10919         (fmaxmin): Likewise.
10920         (maxmin_uns): New.
10921         (maxmin_uns_op): Likewise.
10923 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10925         * config/aarch64/arm_neon.h
10926         (vac<ge, gt><sd>_f<32, 64>): Rename to...
10927         (vca<ge, gt><sd>_f<32, 64>): ...this, reimpliment in C.
10928         (vca<ge, gt, lt, le><q>_f<32, 64>): Reimpliment in C.
10930 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10932         * config/aarch64/aarch64-simd.md (*aarch64_fac<optab><mode>): New.
10933         * config/aarch64/iterators.md (FAC_COMPARISONS): New.
10935 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10937         * config/aarch64/aarch64-simd.md
10938         (vcond<mode>_internal): Handle special cases for constant masks.
10939         (vcond<mode><mode>): Allow nonmemory_operands for outcome vectors.
10940         (vcondu<mode><mode>): Likewise.
10941         (vcond<v_cmp_result><mode>): New.
10943 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10945         * config/aarch64/aarch64-builtins.c (BUILTIN_VALLDI): Define.
10946         (aarch64_fold_builtin): Add folding for cm<eq,ge,gt,tst>.
10947         * config/aarch64/aarch64-simd-builtins.def
10948         (cmeq): Update to BUILTIN_VALLDI.
10949         (cmgt): Likewise.
10950         (cmge): Likewise.
10951         (cmle): Likewise.
10952         (cmlt): Likewise.
10953         * config/aarch64/arm_neon.h
10954         (vc<eq, lt, le, gt, ge, tst><z><qsd>_<fpsu><8,16,32,64>): Remap
10955         to builtins or C as appropriate.
10957 2013-05-01  James Greenhalgh  <james.greenhalgh@arm.com>
10959         * config/aarch64/aarch64-simd-builtins.def (cmhs): Rename to...
10960         (cmgeu): ...This.
10961         (cmhi): Rename to...
10962         (cmgtu): ...This.
10963         * config/aarch64/aarch64-simd.md
10964         (simd_mode): Add SF.
10965         (aarch64_vcond_internal): Use new names for unsigned comparison insns.
10966         (aarch64_cm<optab><mode>): Rewrite to not use UNSPECs.
10967         * config/aarch64/aarch64.md (*cstore<mode>_neg): Rename to...
10968         (cstore<mode>_neg): ...This.
10969         * config/aarch64/iterators.md
10970         (VALLF): new.
10971         (unspec): Remove UNSPEC_CM<EQ, LE, LT, GE, GT, HS, HI, TST>.
10972         (COMPARISONS): New.
10973         (UCOMPARISONS): Likewise.
10974         (optab): Add missing comparisons.
10975         (n_optab): New.
10976         (cmp_1): Likewise.
10977         (cmp_2): Likewise.
10978         (CMP): Likewise.
10979         (cmp): Remove.
10980         (VCMP_S): Likewise.
10981         (VCMP_U): Likewise.
10982         (V_cmp_result): Add DF, SF modes.
10983         (v_cmp_result): Likewise.
10984         (v): Likewise.
10985         (vmtype): Likewise.
10986         * config/aarch64/predicates.md (aarch64_reg_or_fp_zero): New.
10988 2013-05-01  Greta Yorsh  <Greta.Yorsh@arm.com>
10990         * config/arm/thumb2.md (thumb2_smaxsi3,thumb2_sminsi3): Convert
10991         define_insn to define_insn_and_split.
10992         (thumb32_umaxsi3,thumb2_uminsi3): Likewise.
10993         (thumb2_negdi2,thumb2_abssi2,thumb2_neg_abssi2): Likewise.
10994         (thumb2_mov_scc,thumb2_mov_negscc,thumb2_mov_notscc): Likewise.
10995         (thumb2_movsicc_insn,thumb2_and_scc,thumb2_ior_scc): Likewise.
10996         (thumb2_negscc): Likewise.
10998 2013-04-30  Greta Yorsh  <Greta.Yorsh@arm.com>
11000         * config/arm/thumb2.md (thumb2_incscc, thumb2_decscc): Delete.
11002 2013-04-30  Greta Yorsh  <Greta.Yorsh@arm.com>
11004         * config/arm/thumb2.md: Remove trailing whitespaces.
11006 2013-04-30  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
11008         * explow.c (plus_constant): Pass "mode" to immed_double_int_const.
11009         Use gen_int_mode rather than GEN_INT.
11011 2013-04-30  H.J. Lu  <hongjiu.lu@intel.com>
11013         * value-prof.c (stream_in_histogram_value): Remove the strayed
11014         debug_gimple_stmt.
11016 2013-04-30  Richard Biener  <rguenther@suse.de>
11018         PR middle-end/57122
11019         * cfghooks.c (split_edge): Properly check for the loop latch edge.
11021 2013-04-30  Richard Biener  <rguenther@suse.de>
11023         PR middle-end/57107
11024         * tree-eh.c (sink_clobbers): Preserve virtual SSA form.
11026 2013-04-30  Andrey Belevantsev  <abel@ispras.ru>
11028         PR rtl-optimization/56957
11029         PR rtl-optimization/57105
11030         * sel-sched.c (move_op_orig_expr_found): Remove insn_emitted
11031         variable.  Use just INSN_UID for determining whether an insn
11032         should be only disconnected from the insn stream.
11033         * sel-sched-ir.h (EXPR_WAS_CHANGED): Remove.
11035 2013-04-30  Jakub Jelinek  <jakub@redhat.com>
11037         PR tree-optimization/57104
11038         * tsan.c (instrument_expr): Don't instrument accesses to
11039         DECL_HARD_REGISTER VAR_DECLs.
11041 2013-04-30  Richard Biener  <rguenther@suse.de>
11043         * function.h (loops_for_fn): New inline function.
11044         (set_loops_for_fn): Likewise.
11045         * cfgloop.h (place_new_loop): Add struct function parameter.
11046         (get_loop): Likewise.
11047         (get_loops): Likewise.
11048         (number_of_loops): Likewise.
11049         (fel_next): Adjust.
11050         (fel_init): Likewise.
11051         * cfg.c (get_loop_copy): Adjust.
11052         * cfgloop.c (flow_loops_dump): Likewise.
11053         (record_loop_exits): Likewise.
11054         (verify_loop_structure): Likewise.
11055         * cfgloopanal.c (mark_irreducible_loops): Likewise.
11056         (estimate_reg_pressure_cost): Likewise.
11057         (mark_loop_exit_edges): Likewise.
11058         * cfgloopmanip.c (place_new_loop): Likewise.
11059         (add_loop): Likewise.
11060         (duplicate_loop): Likewise.
11061         * graph.c (draw_cfg_nodes): Likewise.
11062         * graphite-clast-to-gimple.c (translate_clast_user): Likewise.
11063         * graphite-sese-to-poly.c (build_scop_scattering): Likewise.
11064         (extract_affine_chrec): Likewise.
11065         (build_scop_iteration_domain): Likewise.
11066         * graphite.c (graphite_initialize): Likewise.
11067         * ira-build.c (create_loop_tree_nodes): Likewise.
11068         (more_one_region_p): Likewise.
11069         (rebuild_regno_allocno_maps): Likewise.
11070         (mark_loops_for_removal): Likewise.
11071         (mark_all_loops_for_removal): Likewise.
11072         (remove_unnecessary_regions): Likewise.
11073         (ira_build): Likewise.
11074         * ira-emit.c (setup_entered_from_non_parent_p): Likewise.
11075         * loop-init.c (fix_loop_structure): Likewise.
11076         (gate_rtl_move_loop_invariants): Likewise.
11077         (gate_rtl_unswitch): Likewise.
11078         (gate_rtl_unroll_and_peel_loops): Likewise.
11079         (rtl_doloop): Likewise.
11080         * lto-streamer-in.c (input_cfg): Likewise.
11081         * lto-streamer-out.c (output_cfg): Likewise.
11082         * modulo-sched.c (sms_schedule): Likewise.
11083         * predict.c (tree_estimate_probability): Likewise.
11084         (tree_estimate_probability_driver): Likewise.
11085         (estimate_loops): Likewise.
11086         * tree-cfg.c (fixup_loop_arrays_after_move): Likewise.
11087         (move_sese_region_to_fn): Likewise.
11088         (debug_loop_num): Likewise.
11089         * tree-chrec.c (chrec_evaluate): Likewise.
11090         (hide_evolution_in_other_loops_than_loop): Likewise.
11091         (chrec_component_in_loop_num): Likewise.
11092         (reset_evolution_in_loop): Likewise.
11093         (evolution_function_is_invariant_rec_p): Likewise.
11094         * tree-if-conv.c (main_tree_if_conversion): Likewise.
11095         * tree-inline.c (copy_loops): Likewise.
11096         (copy_cfg_body): Likewise.
11097         (tree_function_versioning): Likewise.
11098         * tree-loop-distribution.c (rdg_flag_loop_exits): Likewise.
11099         * tree-scalar-evolution.c (chrec_contains_symbols_defined_in_loop):
11100         Likewise.
11101         (add_to_evolution_1): Likewise.
11102         (scev_const_prop): Likewise.
11103         * tree-scalar-evolution.h (get_chrec_loop): Likewise.
11104         * tree-ssa-loop-ch.c (copy_loop_headers): Likewise.
11105         * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
11106         (tree_ssa_lim_initialize): Likewise.
11107         * tree-ssa-loop-manip.c (rewrite_into_loop_closed_ssa): Likewise.
11108         (verify_loop_closed_ssa): Likewise.
11109         * tree-ssa-loop.c  (tree_ssa_loop_init): Likewise.
11110         (tree_ssa_loop_im): Likewise.
11111         (tree_ssa_loop_unswitch): Likewise.
11112         (tree_vectorize): Likewise.
11113         (check_data_deps): Likewise.
11114         (tree_ssa_loop_ivcanon): Likewise.
11115         (tree_ssa_loop_bounds): Likewise.
11116         (tree_complete_unroll): Likewise.
11117         (tree_complete_unroll_inner): Likewise.
11118         (tree_parallelize_loops): Likewise.
11119         (tree_ssa_loop_prefetch): Likewise.
11120         (tree_ssa_loop_ivopts): Likewise.
11121         * tree-ssa.c (execute_update_addresses_taken): Liekwise.
11122         * tree-vectorizer.c (vectorize_loops): Likewise.
11124 2013-04-29  Mike Frysinger  <vapier@gentoo.org>
11126         * config/arm/bpabi.h (EABI_LINK_SPEC): Define.
11127         (BPABI_LINK_SPEC): Use new EABI_LINK_SPEC.
11128         * config/arm/linux-eabi.h (LINK_SPEC): Replace BE8_LINK_SPEC
11129         with EABI_LINK_SPEC.
11131 2013-04-29  Uros Bizjak  <ubizjak@gmail.com>
11133         PR target/44578
11134         * config/i386/i386.md (*zero_extendsidi2): Add "!" to m->?*y
11135         alternative.
11137 2013-04-29  Vladimir Makarov  <vmakarov@redhat.com>
11139         PR target/57097
11140         * lra-constraints.c (process_alt_operands): Discourage a bit more
11141         using memory for pseudos.  Print cost dump for alternatives.
11142         Modify cost values for conflicts with early clobbers.
11143         (curr_insn_transform): Spill pseudos reassigned to NO_REGS.
11145 2013-04-29  Uros Bizjak  <ubizjak@gmail.com>
11147         PR target/57098
11148         * config/i386/i386.c (ix86_expand_vec_perm): Validize constant memory.
11150 2013-04-29  Ian Bolton  <ian.bolton@arm.com>
11152         * config/aarch64/aarch64.md (movsi_aarch64): Support LDR/STR
11153         from/to S register.
11154         (movdi_aarch64): Support LDR/STR from/to D register.
11156 2013-04-29  Ian Bolton  <ian.bolton@arm.com>
11158         * common/config/aarch64/aarch64-common.c: Enable REE pass at O2
11159         or higher by default.
11161 2013-04-29  Richard Biener  <rguenther@suse.de>
11163         PR middle-end/57075
11164         * tree-inline.c (copy_edges_for_bb): Still split the bbs,
11165         even if not adding abnormal edges for calls that can make
11166         abnormal gotos.
11168 2013-04-29  Richard Biener  <rguenther@suse.de>
11170         PR middle-end/57103
11171         * tree-cfg.c (move_stmt_op): Fix condition under which to update
11172         TREE_BLOCK.
11173         (move_stmt_r): Remove redundant checking.
11175 2013-04-29  Teresa Johnson  <tejohnson@google.com>
11177         PR bootstrap/57077
11178         * basic-block.h (apply_scale): New function.
11179         (apply_probability): Use apply_scale.
11180         * gimple-streamer-in.c (input_bb): Ditto.
11181         * lto-streamer-in.c (input_cfg): Ditto.
11182         * lto-cgraph.c (merge_profile_summaries): Ditto.
11183         * tree-optimize.c (execute_fixup_cfg): Ditto.
11184         * tree-inline.c (copy_bb): Update comment to use apply_scale.
11185         (copy_edges_for_bb): Ditto.
11186         (copy_cfg_body): Ditto.
11188 2013-04-29  Tom de Vries  <tom@codesourcery.com>
11190         * tree-ssa-tail-merge.c (find_same_succ_bb): Skip loop latch bbs.
11191         (replace_block_by): Don't set LOOPS_NEED_FIXUP.
11192         (tail_merge_optimize): Handle current_loops == NULL.
11194 2013-04-26  Jeff Law  <law@redhat.com>
11196         * tree-vrp.c (range_fits_type_p): Move to earlier point in file.
11197         (simplify_cond_using_ranges): Generalize code to simplify
11198         COND_EXPRs where one argument is a constant and the other
11199         is an SSA_NAME created by an integral type conversion.
11201 2013-04-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11203         * config/arm/arm.md (store_minmaxsi): Use only when
11204         optimize_insn_for_size_p.
11206 2013-04-29  Christian Bruel  <christian.bruel@st.com>
11208         PR target/57108
11209         * sh.md (tstsi_t_zero_extract_eq): Set mode for operand 0.
11211 2013-04-29  Richard Biener  <rguenther@suse.de>
11213         PR middle-end/57089
11214         * omp-low.c (expand_omp_taskreg): If the parent function had a broken
11215         loop tree make sure to schedule a fixup for the child as well.
11216         (expand_omp_for_generic): Properly add loops.
11217         (expand_omp_for_static_nochunk): Likewise.
11218         (expand_omp_for_static_chunk): Likewise.
11219         (expand_omp_for): For the degenerate case fixup loops.
11220         (expand_omp_sections): Fix default bb placement in loops.
11221         (expand_omp_atomic_pipeline): Properly add loops.
11223 2013-04-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11225         * predict.c: Fix typo in comment above #define PROB_VERY_UNLIKELY.
11227 2013-04-29  Tom de Vries  <tom@codesourcery.com>
11229         * tree-ssa-tail-merge.c: Update header comment.
11231 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11233         * config/aarch64/arm_neon.h
11234         (vcvt<sd>_f<32,64>_s<32,64>): Rewrite in C.
11235         (vcvt<q>_f<32,64>_s<32,64>): Rewrite using builtins.
11236         (vcvt_<high_>_f<32,64>_f<32,64>): Likewise.
11237         (vcvt<qsd>_<su><32,64>_f<32,64>): Likewise.
11238         (vcvta<qsd>_<su><32,64>_f<32,64>): Likewise.
11239         (vcvtm<qsd>_<su><32,64>_f<32,64>): Likewise.
11240         (vcvtn<qsd>_<su><32,64>_f<32,64>): Likewise.
11241         (vcvtp<qsd>_<su><32,64>_f<32,64>): Likewise.
11243 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11245         * config/aarch64/aarch64-simd.md
11246         (<optab><VDQF:mode><fcvt_target>2): New, maps to fix, fixuns.
11247         (<fix_trunc_optab><VDQF:mode><fcvt_target>2): New, maps to
11248         fix_trunc, fixuns_trunc.
11249         (ftrunc<VDQF:mode>2): New.
11250         * config/aarch64/iterators.md (optab): Add fix, fixuns.
11251         (fix_trunc_optab): New.
11253 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11255         * config/aarch64/aarch64-builtins.c
11256         (aarch64_builtin_vectorized_function): Vectorize over ifloorf,
11257         iceilf, lround, iroundf.
11259 2013-04-29  Uros Bizjak  <ubizjak@gmail.com>
11261         PR target/54349
11262         * config/i386/i386.h (enum ix86_tune_indices)
11263         <X86_TUNE_INTER_UNIT_MOVES_TO_VEC, X86_TUNE_INTER_UNIT_MOVES_FROM_VEC>:
11264         New, split from X86_TUNE_INTER_UNIT_MOVES.
11265         <X86_TUNE_INTER_UNIT_MOVES>: Remove.
11266         (TARGET_INTER_UNIT_MOVES_TO_VEC): New define.
11267         (TARGET_INTER_UNIT_MOVES_FROM_VEC): Ditto.
11268         (TARGET_INTER_UNIT_MOVES): Remove.
11269         * config/i386/i386.c (initial_ix86_tune_features): Update.
11270         Disable X86_TUNE_INTER_UNIT_MOVES_FROM_VEC for m_ATHLON_K8 only.
11271         (ix86_expand_convert_uns_didf_sse): Use
11272         TARGET_INTER_UNIT_MOVES_TO_VEC instead of TARGET_INTER_UNIT_MOVES.
11273         (ix86_expand_vector_init_one_nonzero): Ditto.
11274         (ix86_expand_vector_init_interleave): Ditto.
11275         (inline_secondary_memory_needed): Return true for moves from SSE class
11276         registers for !TARGET_INTER_UNIT_MOVES_FROM_VEC targets and for moves
11277         to SSE class registers for !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
11278         * config/i386/constraints.md (Yi, Ym): Depend on
11279         TARGET_INTER_UNIT_MOVES_TO_VEC.
11280         (Yj, Yn): New constraints.
11281         * config/i386/i386.md (*movdi_internal): Change constraints of
11282         operand 1 from Yi to Yj and from Ym to Yn.
11283         (*movsi_internal): Ditto.
11284         (*movdf_internal): Ditto.
11285         (*movsf_internal): Ditto.
11286         (*float<SWI48x:mode><X87MODEF:mode>2_1): Use
11287         TARGET_INTER_UNIT_MOVES_TO_VEC instead of TARGET_INTER_UNIT_MOVES.
11288         (*float<SWI48x:mode><X87MODEF:mode>2_1 splitters): Ditto.
11289         (floatdi<X87MODEF:mode>2_i387_with_xmm): Ditto.
11290         (floatdi<X87MODEF:mode>2_i387_with_xmm splitters): Ditto.
11291         * config/i386/sse.md (movdi_to_sse): Ditto.
11292         (sse2_stored): Change constraint of operand 1 from Yi to Yj.
11293         Use TARGET_INTER_UNIT_MOVES_FROM_VEC instead of
11294         TARGET_INTER_UNIT_MOVES.
11295         (sse_storeq_rex64): Change constraint of operand 1 from Yi to Yj.
11296         (sse_storeq_rex64 splitter): Use TARGET_INTER_UNIT_MOVES_FROM_VEC
11297         instead of TARGET_INTER_UNIT_MOVES.
11298         * config/i386/mmx.md (*mov<mode>_internal): Change constraint of
11299         operand 1 from Yi to Yj and from Ym to Yn.
11301 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11303         * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi_): New.
11304         (float_truncate_hi_): Likewise.
11305         (float_extend_lo_): Likewise.
11306         (float_truncate_lo_): Likewise.
11307         * config/aarch64/aarch64-simd.md (vec_unpacks_lo_v4sf): New.
11308         (aarch64_float_extend_lo_v2df): Likewise.
11309         (vec_unpacks_hi_v4sf): Likewise.
11310         (aarch64_float_truncate_lo_v2sf): Likewise.
11311         (aarch64_float_truncate_hi_v4sf): Likewise.
11312         (vec_pack_trunc_v2df): Likewise.
11313         (vec_pack_trunc_df): Likewise.
11315 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11317         * config/aarch64/aarch64-builtins.c
11318         (aarch64_fold_builtin): Fold float conversions.
11319         * config/aarch64/aarch64-simd-builtins.def
11320         (floatv2si, floatv4si, floatv2di): New.
11321         (floatunsv2si, floatunsv4si, floatunsv2di): Likewise.
11322         * config/aarch64/aarch64-simd.md
11323         (<optab><fcvt_target><VDQF:mode>2): New, expands to float and floatuns.
11324         * config/aarch64/iterators.md (FLOATUORS): New.
11325         (optab): Add float, floatuns.
11326         (su_optab): Likewise.
11328 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11330         * config/aarch64/aarch64-builtins.c
11331         (aarch64_builtin_vectorized_function): Use new names for
11332         fcvt builtins.
11333         * config/aarch64/aarch64-simd-builtins.def (fcvtzs): Split as...
11334         (lbtruncv2sf, lbtruncv4sf, lbtruncv2df): ...This.
11335         (fcvtzu): Split as...
11336         (lbtruncuv2sf, lbtruncuv4sf, lbtruncuv2df): ...This.
11337         (fcvtas): Split as...
11338         (lroundv2sf, lroundv4sf, lroundv2df, lroundsf, lrounddf): ...This.
11339         (fcvtau): Split as...
11340         (lrounduv2sf, lrounduv4sf, lrounduv2df, lroundusf, lroundudf): ...This.
11341         (fcvtps): Split as...
11342         (lceilv2sf, lceilv4sf, lceilv2df): ...This.
11343         (fcvtpu): Split as...
11344         (lceiluv2sf, lceiluv4sf, lceiluv2df, lceilusf, lceiludf): ...This.
11345         (fcvtms): Split as...
11346         (lfloorv2sf, lfloorv4sf, lfloorv2df): ...This.
11347         (fcvtmu): Split as...
11348         (lflooruv2sf, lflooruv4sf, lflooruv2df, lfloorusf, lfloorudf): ...This.
11349         (lfrintnv2sf, lfrintnv4sf, lfrintnv2df, lfrintnsf, lfrintndf): New.
11350         (lfrintnuv2sf, lfrintnuv4sf, lfrintnuv2df): Likewise.
11351         (lfrintnusf, lfrintnudf): Likewise.
11352         * config/aarch64/aarch64-simd.md
11353         (l<fcvt_pattern><su_optab><fcvt_target><VDQF:mode>2): Convert to
11354         define_insn.
11355         (aarch64_fcvt<frint_suffix><su><mode>): Remove.
11356         * config/aarch64/iterators.md (FCVT): Include UNSPEC_FRINTN.
11357         (fcvt_pattern): Likewise.
11359 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11361         * config/aarch64/aarch64-simd.md
11362         (l<fcvt_pattern><su_optab><fcvt_target><VDQF:mode>2): Rename to...
11363         (l<fcvt_pattern><su_optab><VDQF:mode><fcvt_target>2): ... This.
11365 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11367         * config/aarch64/arm_neon.h (vrndq<a,m,n,p>_f<32, 64>): Rename to...
11368         (vrnd<a,m,n,p>q_f<32, 64>): ...This, implement using builtin.
11369         (vrnd<a,m,n,p>_f32): Implement using builtins.
11370         (vrnd<i,x><q>_f<32, 64>): New.
11372 2013-04-29  James Greenhalgh  <james.greenhalgh@arm.com>
11374         * config/aarch64/aarch64-builtins.c
11375         (aarch64_builtin_vectorized_function): Fold to standard pattern names.
11376         * config/aarch64/aarch64-simd-builtins.def (frintn): New.
11377         (frintz): Rename to...
11378         (btrunc): ...this.
11379         (frintp): Rename to...
11380         (ceil): ...this.
11381         (frintm): Rename to...
11382         (floor): ...this.
11383         (frinti): Rename to...
11384         (nearbyint): ...this.
11385         (frintx): Rename to...
11386         (rint): ...this.
11387         (frinta): Rename to...
11388         (round): ...this.
11389         * config/aarch64/aarch64-simd.md
11390         (aarch64_frint<frint_suffix><mode>): Delete.
11391         (<frint_pattern><mode>2): Convert to insn.
11392         * config/aarch64/aarch64.md (unspec): Add UNSPEC_FRINTN.
11393         * config/aarch64/iterators.md (FRINT): Add UNSPEC_FRINTN.
11394         (frint_pattern): Likewise.
11395         (frint_suffix): Likewise.
11397 2013-04-29  Richard Biener  <rguenther@suse.de>
11399         PR tree-optimization/57081
11400         * loop-init.c: Include tree-flow.h.
11401         (loop_optimizer_finalize): Free number of iteration estimates.
11402         * Makefile.in (loop-init.o): Add $(TREE_FLOW_H) dependency.
11404 2013-04-29  Jakub Jelinek  <jakub@redhat.com>
11406         PR tree-optimization/57083
11407         * tree-vrp.c (extract_range_from_binary_expr_1): For LSHIFT_EXPR with
11408         non-singleton shift count range, zero extend low_bound for uns case.
11410         * config/i386/predicates.md (general_vector_operand): New predicate.
11411         * config/i386/i386.c (const_vector_equal_evenodd_p): New function.
11412         (ix86_expand_mul_widen_evenodd): Force op1 resp. op2 into register
11413         if they aren't nonimmediate operands.  If their original values
11414         satisfy const_vector_equal_evenodd_p, don't shift them.
11415         * config/i386/sse.md (mul<mode>3): Use general_vector_operand
11416         predicates.  For the SSE4.1 case force operands[{1,2}] into registers
11417         if not nonimmediate_operand.
11418         (vec_widen_smult_even_v4si): Use nonimmediate_operand predicates
11419         instead of register_operand.
11420         (vec_widen_<s>mult_odd_<mode>): Use general_vector_operand predicates.
11422 2013-04-28  Eric Botcazou  <ebotcazou@adacore.com>
11424         * stor-layout.c (finalize_size_functions): Allocate a structure and
11425         reset cfun before dumping the functions.
11427 2013-04-27  Jakub Jelinek  <jakub@redhat.com>
11429         * config/i386/i386.c (ix86_expand_call): Make cregs_size unsigned.
11431         PR target/56866
11432         * config/i386/i386.c (ix86_expand_mul_widen_evenodd): Don't
11433         use xop_pmacsdqh if uns_p.
11434         * config/i386/sse.md (xop_rotr<mode>3): Fix up computation of
11435         the immediate rotate count.
11437 2013-04-26  Vladimir Makarov  <vmakarov@redhat.com>
11439         * rtl.h (struct rtx_def): Add comment for field jump.
11440         (LRA_SUBREG_P): New macro.
11441         * recog.c (register_operand): Check LRA_SUBREG_P.
11442         * lra.c (lra): Add note at the end of RTL code. Align non-empty
11443         stack frame.
11444         * lra-spills.c (lra_spill): Align stack after spilling pseudos.
11445         (lra_final_code_change): Skip subreg change for operators.
11446         * lra-eliminations.c (eliminate_regs_in_insn): Make return earlier
11447         if there are no operand changes.
11448         * lra-constraints.c (curr_insn_set): New.
11449         (match_reload): Set LRA_SUBREG_P.
11450         (emit_spill_move): Ditto.
11451         (check_and_process_move): Use curr_insn_set. Process only single
11452         set insns.  Don't initialize sec_mem_p and change_p.
11453         (simplify_operand_subreg): Use LRA_SUBREG_P.
11454         (reg_in_class_p): New function.
11455         (process_alt_operands): Use it.  Use #if HAVE_ATTR_enabled instead
11456         of #ifdef.  Add code to remove cycling.
11457         (process_address): Check EXTRA_CONSTRAINT_STR. Process even if
11458         non-null disp.  Reload inner instead of disp when base and index
11459         are null.  Try to put lo_sum into register.
11460         (EBB_PROBABILITY_CUTOFF): Redefine probability in percents.
11461         (check_and_process_move): Move code for move cost check to
11462         simple_move_p.  Remove equiv_substitution.
11463         (simple_move_p): New function.
11464         (curr_insn_transform): Initialize sec_mem_p and change_p.  Set up
11465         curr_insn_set.  Call check_and_process_move only for single set
11466         insns.  Use the new function.  Move call of check_and_process_move
11467         after operand equiv substitution and address process.
11469 2013-04-26  Jakub Jelinek  <jakub@redhat.com>
11471         PR go/57045
11472         * tree-ssa-uninit.c (compute_uninit_opnds_pos): In functions
11473         with nonlocal goto receivers or returns twice calls, ignore
11474         unininitialized values from abnormal edges to nl goto receiver
11475         or returns twice call.
11477 2013-04-26  Jakub Jelinek  <jakub@redhat.com>
11479         PR tree-optimization/57051
11480         * fold-const.c (const_binop): Handle VEC_LSHIFT_EXPR
11481         and VEC_RSHIFT_EXPR if shift count is a multiple of element
11482         bitsize.
11484 2013-04-26  Richard Biener  <rguenther@suse.de>
11486         * omp-low.c (finalize_task_copyfn): Do not drop PROP_loops.
11487         (expand_omp_taskreg): Likewise.  Mark loops for fixup.
11488         * tree-cfg.c (move_block_to_fn): Remap loop fathers.
11489         (fixup_loop_arrays_after_move): New function.
11490         (move_sese_region_to_fn): Properly outline the loop tree parts
11491         of the SESE region.
11493 2013-04-26  Uros Bizjak  <ubizjak@gmail.com>
11495         * config/i386/i386.md (type, unit): Fix long lines.
11497 2013-04-26  Richard Biener  <rguenther@suse.de>
11499         * Makefile.in (lto-streamer-in.o): Add $(CFGLOOP_H) dependency.
11500         (lto-streamer-out.o): Likewise.
11501         * cfgloop.c (init_loops_structure): Export, add struct function
11502         argument and adjust.
11503         (flow_loops_find): Adjust.
11504         * cfgloop.h (enum loop_estimation): Add EST_LAST.
11505         (init_loops_structure): Declare.
11506         * lto-streamer-in.c: Include cfgloop.h.
11507         (input_cfg): Input the loop tree.
11508         * lto-streamer-out.c: Include cfgloop.h.
11509         (output_cfg): Output the loop tree.
11510         (output_struct_function_base): Do not drop PROP_loops.
11512 2013-03-26  Richard Biener  <rguenther@suse.de>
11514         * tree-cfg.c (execute_build_cfg): Build the loop tree.
11515         (pass_build_cfg): Provide PROP_loops.
11516         (move_sese_region_to_fn): Remove loops that are outlined into fn
11517         for now.
11518         * tree-inline.c: Include cfgloop.h.
11519         (initialize_cfun): Do not drop PROP_loops.
11520         (copy_loops): New function.
11521         (copy_cfg_body): Copy loop structure.
11522         (tree_function_versioning): Initialize destination loop tree.
11523         * tree-ssa-loop.c (pass_tree_loop_init): Do not provide PROP_loops.
11524         (pass_parallelize_loops): Do IL verification.
11525         * loop-init.c (loop_optimizer_init): Fixup loops if required.
11526         * tree-optimize.c (execute_fixup_cfg): If we need to cleanup
11527         the CFG make sure we fixup loops as well.
11528         * tree-ssa-tail-merge.c: Include cfgloop.h.
11529         (replace_block_by): When merging loop latches mark loops for fixup.
11530         * lto-streamer-out.c (output_struct_function_base): Drop
11531         PROP_loops for now.
11532         * tree-ssa-phiopt.c: Include tree-scalar-evolution.h.
11533         (tree_ssa_cs_elim): Initialize the loop optimizer and SCEV.
11534         * ipa-split.c: Include cfgloop.h.
11535         (split_function): Add the new return block to the loop tree root.
11536         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Return
11537         whether we have removed the forwarder block.
11538         (merge_phi_nodes): If we removed a forwarder mark loops for fixup.
11539         * cfgloop.h (place_new_loop): Declare.
11540         * cfgloopmanip.c (place_new_loop): Export.
11541         * Makefile.in (asan.o): Add $(CFGLOOP_H) dependency.
11542         (tree-switch-conversion.o): Likewise.
11543         (tree-complex.o): Likewise.
11544         (tree-inline.o): Likewise.
11545         (tree-ssa-tailmerge.o): Likewise.
11546         (ipa-split.o): Likewise.
11547         (tree-ssa-phiopt.o): Add $(SCEV_H) dependency.
11548         (tree-ssa-copy.o): Likewise.
11549         * tree-switch-conversion.c: Include cfgloop.h
11550         (process_switch): If we emit a bit-test cascade, schedule loops
11551         for fixup.
11552         * tree-complex.c: Include cfgloop.h.
11553         (expand_complex_div_wide): Properly add new basic-blocks to loops.
11554         * asan.c: Include cfgloop.h.
11555         (create_cond_insert_point): Properly add new basic-blocks to
11556         loops, schedule loop fixup.
11557         * cfgloop.c (verify_loop_structure): Check that looks are not
11558         marked for fixup.
11559         * omp-low.c (expand_parallel_call): Properly add new basic-blocks
11560         to loops.
11561         (expand_omp_for_generic): Likewise.
11562         (expand_omp_sections): Likewise.
11563         (expand_omp_atomic_pipeline): Schedule loops for fixup.
11564         * tree-ssa-copy.c: Include tree-scalar-evolution.h.
11565         (fini_copy_prop): Disable DCE in substitute_and_fold if SCEV
11566         is initialized, not when loops are present.
11567         * tree-parloops.c (parallelize_loops): Remove checking here.
11568         * passes.c (init_optimization_passes): Schedule a copy-propagation
11569         pass before complete unrolling of inner loops.
11571 2013-04-26  Jakub Jelinek  <jakub@redhat.com>
11573         * Makefile.in (toplev.o): Depend on diagnostic-color.h.
11574         * diagnostic-color.c (should_colorize): Remove _WIN32 version.
11575         (colorize_init): Add argument to _WIN32 version.
11576         * toplev.c: Include diagnostic-color.h.
11577         (process_options): Default to -fdiagnostics-color=auto if
11578         GCC_COLORS env var is in the environment.
11579         * common.opt (fdiagnostics-color=): Add Var and Init.
11580         * doc/invoke.texi (-fdiagnostics-color=): Document that if GCC_COLORS
11581         env var is in the environment, the default is auto rather than never.
11583         * diagnostic.h (file_name_as_prefix): Add context argument.
11584         * diagnostic.c (file_name_as_prefix): Likewise.  Colorize
11585         the string as locus.
11586         * langhooks.c (lhd_print_error_function): Adjust caller.
11588 2013-04-25  Lawrence Crowl  <crowl@google.com>
11590         * var-tracking.c (shared_hash_def::htab):
11591         Change type to hash_table.  Update dependent calls and types.
11593 2013-04-25  Lawrence Crowl  <crowl@google.com>
11595         * Makefile.in: Update as needed below.
11597         * alloc-pool.c (static hash_table <alloc_pool_hasher> alloc_pool_hash):
11598         Move declaration to after the type's method definitons.
11600         * attribs.c (htab_t scoped_attributes::attribute_hash):
11601         Change type to hash_table.  Update dependent calls and types.
11603         * bitmap.c (htab_t bitmap_desc_hash):
11604         Change type to hash_table.  Update dependent calls and types.
11606         * cselib.c (htab_t cselib_hash_table):
11607         Change type to hash_table.  Update dependent calls and types.
11609         * data-streamer.h (struct string_slot): Move to lto-streamer.h.
11610         (hash_string_slot_node): Move implementation into lto-streamer.h
11611         struct string_slot_hasher.
11612         (eq_string_slot_node): Likewise.
11614         * data-streamer-out.c: Update output_block::string_hash_table
11615         dependent calls and types.
11617         * dwarf2cfi.c (htab_t trace_index):
11618         Change type to hash_table.  Update dependent calls and types.
11620         * dwarf2out.c (htab_t break_out_includes::cu_hash_table):
11621         Change type to hash_table.  Update dependent calls and types.
11622         (htab_t copy_decls_for_unworthy_types::decl_table): Likewise.
11623         (htab_t optimize_external_refs::map): Likewise.
11624         (htab_t output_comp_unit::extern_map): Likewise.
11625         (htab_t output_comdat_type_unit::extern_map): Likewise.
11626         (htab_t output_macinfo::macinfo_htab): Likewise.
11627         (htab_t optimize_location_lists::htab): Likewise.
11628         (htab_t dwarf2out_finish::comdat_type_table): Likewise.
11630         * except.c (htab_t ehspec_hash_type):
11631         Change type to hash_table.  Update dependent calls and types.
11632         (assign_filter_values::ttypes): Likewise.
11633         (assign_filter_values::ehspec): Likewise.
11634         (sjlj_assign_call_site_values::ar_hash): Likewise.
11635         (convert_to_eh_region_ranges::ar_hash): Likewise.
11637         * gcse.c (htab_t pre_ldst_table):
11638         Change type to hash_table.  Update dependent calls and types.
11640         * ggc-common.c (htab_t saving_htab):
11641         Change type to hash_table.  Update dependent calls and types.
11642         (htab_t loc_hash): Likewise.
11643         (htab_t ptr_hash): Likewise.
11644         (call_count): Rename ggc_call_count.
11645         (call_alloc): Rename ggc_call_alloc.
11646         (loc_descriptor): Rename make_loc_descriptor.
11647         (add_statistics): Rename ggc_add_statistics.
11649         * ggc-common.c (saving_htab):
11650         Change type to hash_table.  Update dependent calls and types.
11652         * gimple.h (struct gimplify_ctx): Move to gimplify-ctx.h.
11653         (push_gimplify_context): Likewise.
11654         (pop_gimplify_context): Likewise.
11655         (struct gimple_temp_hash_elt): Added.
11656         (struct gimplify_hasher): Likewise.
11657         (struct gimplify_ctx.temp_htab):
11658         Change type to hash_table.  Update dependent calls and types.
11660         * gimple-fold.c: Include gimplify-ctx.h.
11662         * gimple-ssa-strength-reduction.c (htab_t base_cand_map):
11663         Change type to hash_table.  Update dependent calls and types.
11664         (base_cand_dump_callback): Rename to ssa_base_cand_dump_callback to
11665         avoid potential global name collision.
11667         * gimplify.c: Include gimplify-ctx.h.
11668         (struct gimple_temp_hash_elt): Move to gimplify-ctx.h.
11669         (htab_t gimplify_ctx::temp_htab):
11670         Update dependent calls and types for new type hash_table.
11671         (gimple_tree_hash): Move into gimplify_hasher in gimplify-ctx.h.
11672         (gimple_tree_eq): Move into gimplify_hasher in gimplify-ctx.h.
11674         * gimplify-ctx.h: New.
11675         (struct gimple_temp_hash_elt): Move from gimplify.c.
11676         (class gimplify_hasher): New.
11677         (struct gimplify_ctx): Move from gimple.h.
11678         (htab_t gimplify_ctx::temp_htab):
11679         Change type to hash_table.  Update dependent calls and types.
11681         * graphite-clast-to-gimple.c: Include graphite-htab.h.
11682         (htab_t ivs_params::newivs_index):
11683         Change type to hash_table.  Update dependent calls and types.
11684         (htab_t ivs_params::params_index): Likewise.
11685         (htab_t print_generated_program::params_index): Likewise.
11686         (htab_t gloog::newivs_index): Likewise.
11687         (htab_t gloog::params_index): Likewise.
11689         * graphite.c: Include graphite-htab.h.
11690         4htab_t graphite_transform_loops::bb_pbb_mapping):
11691         Change type to hash_table.  Update dependent calls and types.
11693         * graphite-clast-to-gimple.h: (extern gloog) Move to graphite-htab.h.
11694         (bb_pbb_map_hash): Fold into bb_pbb_htab_type in graphite-htab.h.
11695         (eq_bb_pbb_map): Fold into bb_pbb_htab_type in graphite-htab.h.
11697         * graphite-dependences.c: Include graphite-htab.h.
11698         (loop_is_parallel_p): Change hash table type of parameter.
11700         * graphite-htab.h: New.
11701         (typedef hash_table <bb_pbb_hasher> bb_pbb_htab_type): New.
11702         (extern find_pbb_via_hash): Move from graphite-poly.h.
11703         (extern loop_is_parallel_p): Move from graphite-poly.h.
11704         (extern get_loop_body_pbbs): Move from graphite-poly.h.
11706         * graphite-poly.h (extern find_pbb_via_hash): Move to graphite-htab.h.
11707         (extern loop_is_parallel_p): Move to graphite-htab.h.
11708         (extern get_loop_body_pbbs): Move to graphite-htab.h.
11710         * haifa-sched.c (htab_t delay_htab):
11711         Change type to hash_table.  Update dependent calls and types.
11712         (htab_t delay_htab_i2): Likewise.
11714         * ira-color.c (htab_t allocno_hard_regs_htab):
11715         Change type to hash_table.  Update dependent calls and types.
11717         * ira-costs.c (htab_t cost_classes_htab):
11718         Change type to hash_table.  Update dependent calls and types.
11720         * loop-invariant.c (htab_t merge_identical_invariants::eq):
11721         Change type to hash_table.  Update dependent calls and types.
11723         * loop-iv.c (htab_t bivs):
11724         Change type to hash_table.  Update dependent calls and types.
11726         * loop-unroll.c (htab_t opt_info::insns_to_split):
11727         Change type to hash_table.  Update dependent calls and types.
11728         (htab_t opt_info::insns_with_var_to_expand): Likewise.
11730         * lto-streamer.h (struct string_slot): Move from data-streamer.h
11731         (struct string_slot_hasher): New.
11732         (htab_t output_block::string_hash_table):
11733         Change type to hash_table.  Update dependent calls and types.
11735         * lto-streamer-in.c (freeing_string_slot_hasher): New.
11736         (htab_t file_name_hash_table):
11737         Change type to hash_table.  Update dependent calls and types.
11739         * lto-streamer-out.c: Update output_block::string_hash_table dependent
11740         calls and types.
11742         * lto-streamer.c (htab_t tree_htab):
11743         Change type to hash_table.  Update dependent calls and types.
11745         * omp-low.c: Include gimplify-ctx.h.
11747         * passes.c (htab_t name_to_pass_map):
11748         Change type to hash_table.  Update dependent calls and types.
11749         (pass_traverse): Rename to passes_pass_traverse.
11751         * plugin.c (htab_t event_tab):
11752         Change type to hash_table.  Update dependent calls and types.
11754         * postreload-gcse.c (htab_t expr_table):
11755         Change type to hash_table.  Update dependent calls and types.
11756         (dump_hash_table_entry): Rename dump_expr_hash_table_entry.
11758         * sese.c (debug_rename_map_1): Make extern.
11759         (htab_t copy_bb_and_scalar_dependences::rename_map):
11760         Change type to hash_table.  Update dependent calls and types.
11762         * sese.h (extern debug_rename_map): Move to .c file.
11764         * store-motion.c (htab_t store_motion_mems_table):
11765         Change type to hash_table.  Update dependent calls and types.
11767         * trans-mem.c (htab_t tm_new_mem_hash):
11768         Change type to hash_table.  Update dependent calls and types.
11770         * tree-browser.c (htab_t TB_up_ht):
11771         Change type to hash_table.  Update dependent calls and types.
11773         * tree-cfg.c (htab_t discriminator_per_locus):
11774         Change type to hash_table.  Update dependent calls and types.
11776         * tree-complex.c: Include tree-hasher.h
11777         (htab_t complex_variable_components):
11778         Change type to hash_table.  Update dependent calls and types.
11780         * tree-eh.c (htab_t finally_tree):
11781         Change type to hash_table.  Update dependent calls and types.
11783         * tree-flow.h (extern int_tree_map_hash): Moved into tree-hasher
11784         struct int_tree_hasher.
11785         (extern int_tree_map_eq): Likewise.
11786         (uid_decl_map_hash): Removed.
11787         (extern decl_tree_map_eq): Likewise.
11789         * tree-hasher.h: New.
11790         (struct int_tree_hasher): New.
11791         (typedef int_tree_htab_type): New.
11793         * tree-inline.c: Include gimplify-ctx.h.
11795         * tree-mudflap.c: Include gimplify-ctx.h.
11797         * tree-parloops.c: Include tree-hasher.h.
11798         (htab_t eliminate_local_variables_stmt::decl_address):
11799         Change type to hash_table.  Update dependent calls and types.
11800         (htab_t separate_decls_in_region::decl_copies): Likewise.
11802         * tree-scalar-evolution.c (htab_t resolve_mixers::cache):
11803         Change type to hash_table.  Update dependent calls and types.
11805         * tree-sra.c (candidates):
11806         Change type to hash_table.  Update dependent calls and types.
11808         * tree-ssa.c (int_tree_map_eq): Moved into struct int_tree_hasher
11809         in tree-flow.h.
11810         (int_tree_map_hash): Likewise.
11812         * tree-ssa-dom.c (htab_t avail_exprs):
11813         Change type to hash_table.  Update dependent calls and types.
11815         * tree-ssa-live.c (var_map_base_init::tree_to_index):
11816         Change type to hash_table.  Update dependent calls and types.
11818         * tree-ssa-loop-ivopts.c (struct ivopts_data.inv_expr_tab):
11819         Change type to hash_table.  Update dependent calls and types.
11821         * tree-ssa-phiopt.c (seen_ssa_names):
11822         Change type to hash_table.  Update dependent calls and types.
11824         * tree-ssa-strlen.c (decl_to_stridxlist_htab):
11825         Change type to hash_table.  Update dependent calls and types.
11827         * tree-ssa-uncprop.c (equiv):
11828         Change type to hash_table.  Update dependent calls and types.
11830 2013-04-25  Jakub Jelinek  <jakub@redhat.com>
11832         PR rtl-optimization/57003
11833         * regcprop.c (copyprop_hardreg_forward_1): If ksvd.ignore_set_reg,
11834         call note_stores with kill_clobbered_value callback again after
11835         killing regs_invalidated_by_call.
11837 2013-04-25  James Greenhalgh  <james.greenhalgh@arm.com>
11839         * config/aarch64/aarch64-simd.md
11840         (aarch64_simd_bsl<mode>_internal): Rewrite RTL to not use UNSPEC_BSL.
11841         (aarch64_simd_bsl<mode>): Likewise.
11842         * config/aarch64/iterators.md (unspec): Remove UNSPEC_BSL.
11844 2013-04-25  Marek Polacek  <polacek@redhat.com>
11846         PR tree-optimization/57066
11847         * builtins.c (fold_builtin_logb): Return +Inf for -Inf.
11849 2013-04-25  James Greenhalgh  <jame.greenhalgh@arm.com>
11851         * config/aarch64/aarch64-simd.md (neg<mode>2): Use VDQ iterator.
11853 2013-04-25  James Greenhalgh  <james.greenhalgh@arm.com>
11855         * config/aarch64/aarch64-builtins.c
11856         (aarch64_fold_builtin): New.
11857         * config/aarch64/aarch64-protos.h (aarch64_fold_builtin): New.
11858         * config/aarch64/aarch64.c (TARGET_FOLD_BUILTIN): Define.
11859         * config/aarch64/aarch64-simd-builtins.def (abs): New.
11860         * config/aarch64/arm_neon.h
11861         (vabs<q>_<f32, 64>): Implement using __builtin_aarch64_fabs.
11863 2013-04-25  James Greenhalgh  <james.greenhalgh@arm.com>
11864             Tejas Belagod  <tejas.belagod@arm.com>
11866         * config/aarch64/aarch64-builtins.c
11867         (aarch64_gimple_fold_builtin): New.
11868         * config/aarch64/aarch64-protos.h (aarch64_gimple_fold_builtin): New.
11869         * config/aarch64/aarch64-simd-builtins.def (addv): New.
11870         * config/aarch64/aarch64-simd.md (addpv4sf): New.
11871         (addvv4sf): Update.
11872         * config/aarch64/aarch64.c (TARGET_GIMPLE_FOLD_BUILTIN): Define.
11874 2013-04-25  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
11876         * config/aarch64/aarch64.md
11877         (*cmp_swp_<optab><ALLX:mode>_shft_<GPI:mode>): New pattern.
11879 2013-04-25  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
11881         * config/aarch64/aarch64.md (*ngc<mode>): New pattern.
11882         (*ngcsi_uxtw): New pattern.
11884 2013-04-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11885             Julian Brown  <julian@codesourcery.com>
11887         * config/arm/arm.c (neon_builtin_type_mode): Add T_V4HF.
11888         (TB_DREG): Add T_V4HF.
11889         (v4hf_UP): New macro.
11890         (neon_itype): Add NEON_FLOAT_WIDEN, NEON_FLOAT_NARROW.
11891         (arm_init_neon_builtins): Handle NEON_FLOAT_WIDEN, NEON_FLOAT_NARROW.
11892         Handle initialisation of V4HF. Adjust initialisation of reinterpret
11893         built-ins.
11894         (arm_expand_neon_builtin): Handle NEON_FLOAT_WIDEN, NEON_FLOAT_NARROW.
11895         (arm_vector_mode_supported_p): Handle V4HF.
11896         (arm_mangle_map): Handle V4HFmode.
11897         * config/arm/arm.h (VALID_NEON_DREG_MODE): Add V4HF.
11898         * config/arm/arm_neon_builtins.def: Add entries for
11899         vcvtv4hfv4sf, vcvtv4sfv4hf.
11900         * config/arm/neon.md (neon_vcvtv4sfv4hf): New pattern.
11901         (neon_vcvtv4hfv4sf): Likewise.
11902         * config/arm/neon-gen.ml: Handle half-precision floating point
11903         features.
11904         * config/arm/neon-testgen.ml: Handle Requires_FP_bit feature.
11905         * config/arm/arm_neon.h: Regenerate.
11906         * config/arm/neon.ml (type elts): Add F16.
11907         (type vectype): Add T_float16x4, T_floatHF.
11908         (type vecmode): Add V4HF.
11909         (type features): Add Requires_FP_bit feature.
11910         (elt_width): Handle F16.
11911         (elt_class): Likewise.
11912         (elt_of_class_width): Likewise.
11913         (mode_of_elt): Refactor.
11914         (type_for_elt): Handle F16, fix error messages.
11915         (vectype_size): Handle T_float16x4.
11916         (vcvt_sh): New function.
11917         (ops): Add entries for vcvt_f16_f32, vcvt_f32_f16.
11918         (string_of_vectype): Handle T_floatHF, T_float16, T_float16x4.
11919         (string_of_mode): Handle V4HF.
11920         * doc/arm-neon-intrinsics.texi: Regenerate.
11922 2013-04-25  James Greenhalgh  <james.greenhalgh@arm.com>
11924         * config/aarch64/aarch64.c (aarch64_print_operand): Fix asm_fprintf
11925         format specifier in 'X' case.
11927 2013-04-25  Alan Modra  <amodra@gmail.com>
11929         PR target/57052
11930         * config/rs6000/rs6000.md (rotlsi3_internal7): Rename to
11931         rotlsi3_internal7le and condition on !BYTES_BIG_ENDIAN.
11932         (rotlsi3_internal8be): New BYTES_BIG_ENDIAN insn.
11933         Repeat for many other rotate/shift and mask patterns using subregs.
11934         Name lshiftrt insns.
11935         (ashrdisi3_noppc64): Rename to ashrdisi3_noppc64be and condition
11936         on WORDS_BIG_ENDIAN.
11938 2013-04-25  Alan Modra  <amodra@gmail.com>
11940         * config.gcc: Support little-endian powerpc-linux targets.
11941         * config/rs6000/linux.h (LINK_OS_LINUX_EMUL): Define.
11942         (LINK_OS_LINUX_SPEC): Define.
11943         * config/rs6000/linuxspe.h (TARGET_DEFAULT):
11944         Preserve MASK_LITTLE_ENDIAN.
11945         * config/rs6000/default64.h (TARGET_DEFAULT): Likewise.
11946         * config/rs6000/linuxaltivec.h (TARGET_DEFAULT): Likewise.
11947         * config/rs6000/linux64.h (OPTION_LITTLE_ENDIAN): Don't zero.
11948         (LINK_OS_LINUX_EMUL32, LINK_OS_LINUX_EMUL64): Define.
11949         (LINK_OS_LINUX_SPEC32, LINK_OS_LINUX_SPEC64): Use above.
11950         * config/rs6000/rs6000.c (output_toc): Don't use .tc for TARGET_ELF.
11951         Correct fp word order for little-endian.  Don't shift toc entries
11952         smaller than a word for little-endian.
11953         * config/rs6000/rs6000.md (bswaphi2, bswapsi2 split): Comment.
11954         (bswapdi2 splits): Correct low-part subreg for little-endian.
11955         Remove wrong BYTES_BIG_ENDIAN tests, and rename vars to remove
11956         low/high where such is correct only for be.
11957         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Allow
11958         little-endian for -mcall-aixdesc.
11960 2013-04-25  Alan Modra  <amodra@gmail.com>
11962         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
11963         replace_equiv_address_nv.
11965 2013-04-25  Alan Modra  <amodra@gmail.com>
11967         * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Tidy.
11969 2013-04-24  Vladimir Makarov  <vmakarov@redhat.com>
11971         Revert:
11972         2013-04-24  Vladimir Makarov  <vmakarov@redhat.com>
11973         * rtl.h (struct rtx_def): ...
11975 2013-04-24  Vladimir Makarov  <vmakarov@redhat.com>
11977         PR rtl-optimizations/57046
11978         * lra-constraints (split_reg): Set up lra_risky_transformations_p
11979         for multi-reg splits.
11981 2013-04-24  H.J. Lu  <hongjiu.lu@intel.com>
11983         * config/i386/x86-64.h (ASM_SPEC): Support -mx32.
11985 2013-04-24  Sterling Augustine  <saugustine@google.com>
11987         * dwarf2out.c (skeleton_debug_str_hash, add_skeleton_AT_string)
11988         (comp_dir_string, debug_str_dwo_section): New.
11989         (DEBUG_STR_DWO_SECTION): Rename to ...
11990         (DEBUG_DWO_STR_SECTION): ... this.
11991         (DEBUG_NORM_STR_SECTION): Delete.
11992         (DEBUG_STR_SECTION, DEBUG_STR_SECTION_FLAGS): Edit definitions.
11993         (DEBUG_STR_DWO_SECTION_FLAGS): New.
11994         (find_AT_string): Move most logic to ...
11995         (find_AT_string_in_table): ... here.  New.
11996         (add_top_level_skeleton_die_attrs): Call comp_dir_string and
11997         add_skeleton_AT_string.  Delete logic.
11998         (output_skeleton_debug_sections): Remove call to
11999         add_top_level_skeleton_die_attrs.
12000         (add_comp_dir_attribute): Move logic to comp_dir_string.
12001         (dwarf2out_init): Initialize debug_str_dwo_section.
12002         (output_indirect_string): Call find_string_form.
12003         (output_indirect_strings): Rewrite.
12004         (prune_unused_types): Empty skeleton_debug_str_hash.
12005         Call get_skeleton_type_unit and add_top_level_skeleton_die_attrs.
12006         (dwarf2out_finish):  Call output_indirect_strings.
12008 2013-04-24  Paolo Carlini  <paolo.carlini@oracle.com>
12010         * doc/cpp.texi: Remove __GXX_EXPERIMENTAL_CXX1Y__.
12012 2013-04-24  Vladimir Makarov  <vmakarov@redhat.com>
12014         * rtl.h (struct rtx_def): Add comment for field jump.
12015         (LRA_SUBREG_P): New macro.
12016         * recog.c (register_operand): Check LRA_SUBREG_P.
12017         * lra.c (lra): Add note at the end of RTL code. Align non-empty
12018         stack frame.
12019         * lra-spills.c (lra_spill): Align stack after spilling pseudos.
12020         (lra_final_code_change): Skip subreg change for operators.
12021         * lra-eliminations.c (eliminate_regs_in_insn): Make return earlier
12022         if there are no operand changes.
12023         * lra-constraints.c (curr_insn_set): New.
12024         (match_reload): Set LRA_SUBREG_P.
12025         (emit_spill_move): Ditto.
12026         (check_and_process_move): Use curr_insn_set. Process only single
12027         set insns.  Don't initialize sec_mem_p and change_p.
12028         (simplify_operand_subreg): Use LRA_SUBREG_P.
12029         (reg_in_class_p): New function.
12030         (process_alt_operands): Use it.  Use #if HAVE_ATTR_enabled instead
12031         of #ifdef.  Add code to remove cycling.
12032         (process_address): Check EXTRA_CONSTRAINT_STR. Process even if
12033         non-null disp.  Reload inner instead of disp when base and index
12034         are null.  Try to put lo_sum into register.
12035         (EBB_PROBABILITY_CUTOFF): Redefine probability in percents.
12036         (check_and_process_move): Move code for move cost check to
12037         simple_move_p.  Remove equiv_substitution.
12038         (simple_move_p): New function.
12039         (curr_insn_transform): Initialize sec_mem_p and change_p.  Set up
12040         curr_insn_set.  Call check_and_process_move only for single set
12041         insns.  Use the new function.  Move call of check_and_process_move
12042         after operand equiv substitution and address process.
12044 2013-04-24  James Greenhalgh  <james.greenhalgh@arm.com>
12046         * config/aarch64/arm_neon.h (vld1<q>_lane*): Fix constraints.
12047         (vld1<q>_dup_<sufp><8, 16, 32, 64>): Likewise.
12048         (vld1<q>_<sufp><8, 16, 32, 64>): Likewise.
12050 2013-04-24  Paolo Carlini  <paolo.carlini@oracle.com>
12052         * doc/cpp.texi: Document __GXX_EXPERIMENTAL_CXX1Y__.
12054 2013-04-24  Marek Polacek  <polacek@redhat.com>
12056         * tree-scalar-evolution.h (analyze_scalar_evolution): Remove.
12057         * tree-scalar-evolution.c (get_exit_conditions_rec): Likewise.
12058         (select_loops_exit_conditions): Likewise.
12059         (number_of_iterations_for_all_loops): Likewise.
12060         (analyze_scalar_evolution_for_all_loop_phi_nodes): Likewise.
12061         (scev_analysis): Likewise.
12063 2013-04-02  Catherine Moore  <clm@codesourcery.com>
12064             Chao-ying Fu  <fu@mips.com>
12066         * config/mips/micromips.md (jraddiusp): New pattern.
12067         * config/mips/mips.c (mips_expand_epilogue): Use the JRADDIUSP
12068         instruction if possible.
12070 2013-04-24  Alan Modra  <amodra@gmail.com>
12072         * config/rs6000/driver-rs6000.c (elf_dcachebsize): Fix comment pasto.
12074 2013-04-24  Julian Brown  <julian@codesourcery.com>
12075             Chung-Lin Tang  <cltang@codesourcery.com>
12077         * dwarf2out.c (gen_enumeration_type_die): Fix HOST_BITS_PER_WIDE_INT
12078         dependency behavior in enumeration type DIE generation. Add TODO note
12079         to comments about future DW_FORM_sdata/udata re-work of related code.
12081 2013-04-23  Lawrence Crowl  <crowl@google.com>
12083         * Makefile.in: Update as needed below.
12085         * hash-table.h (class hash_table):
12086         Correct many methods with parameter types compare_type to the correct
12087         value_type.  (Correct code was unlikely to notice the change.)
12088         (hash_table::elements_with_deleted) New.
12089         (class hashtable::iterator): New.
12090         (hashtable::begin()): New.
12091         (hashtable::end()): New.
12092         (FOR_EACH_HASH_TABLE_ELEMENT): New.
12094         * statistics.c (statistics_hashes):
12095         Change type to hash_table.  Update dependent calls and types.
12097         * tree-into-ssa.c (var_infos):
12098         Change type to hash_table.  Update dependent calls and types.
12100         * tree-ssa-coalesce.c (struct coalesce_list_d.list):
12101         Change type to hash_table.  Update dependent calls and types.
12103         * tree-ssa-loop-im.c (struct mem_ref.refs):
12104         Change type to hash_table.  Update dependent calls and types.
12106         * tree-ssa-reassoc.c (undistribute_ops_list::ctable):
12107         Change type to hash_table.  Update dependent calls and types.
12109         * tree-ssa-sccvn.c (vn_tables_s::nary):
12110         Change type to hash_table.  Update dependent calls and types.
12111         (vn_tables_s::phis): Likewise.
12112         (vn_tables_s::references): Likewise.
12114         * tree-ssa-sccvn.h (vn_nary_op_eq): Update parameter and return types.
12115         (vn_reference_eq): Update parameter and return types.
12117         * tree-ssa-structalias.c (pointer_equiv_class_table):
12118         Change type to hash_table.  Update dependent calls and types.
12119         (location_equiv_class_table): Likewise.
12121         * tree-vect-data-refs.c: Consequential changes for making
12122         peeling a hash_table.
12124         * tree-vect-loop.c (new_loop_vec_info): Dependent hash_table update.
12125         (destroy_loop_vec_info): Dependent hash_table update.
12127         * tree-vectorizer.h (peeling_htab):
12128         Change type to hash_table.  Update dependent calls and types.
12130 2013-04-23  Shiva Chen  <shiva0217@gmail.com>
12132         * lra-assigns.c (find_hard_regno_for): Use lra_reg_val_equal_p
12133         to check the register content is equal or not.
12134         * lra-constraints.c (match_reload): Use lra_assign_reg_val
12135         to assign register content record.
12136         * lra-eliminations.c (update_reg_eliminate): Use
12137         lra_update_reg_val_offset to update register content offset.
12138         * lra-int.h (struct lra_reg): Add offset member.
12139         (lra_reg_val_equal_p): New static inline function.
12140         (lra_update_reg_val_offset): New static inline function.
12141         (lra_assign_reg_val): New static inline function.
12142         * lra.c (lra_create_new_reg): Use lra_assign_reg_val
12143         to assign register content record.
12144         (initialize_lra_reg_info_element): Initial offset to zero.
12146 2013-04-23  Catherine Moore  <clm@codesourcery.com>
12148         * config/mips/mips.md (*movhi_internal, *movqi_internal): New
12149         operands.  Record compression.
12151 2013-04-23  Xinliang David Li  <davidxl@google.com>
12153         * cfghhooks.c (dump_bb_for_graph): Support 'slim' graph dump.
12155 2013-04-23  Richard Biener  <rguenther@suse.de>
12157         PR middle-end/57036
12158         * tree-inline.c (copy_edges_for_bb): Add can_make_abnormal_goto
12159         parameter, only add abnormal goto edges from the copied body
12160         if the call could perform abnormal gotos.
12161         (copy_cfg_body): Adjust.
12163 2013-04-23  Sofiane Naci  <sofiane.naci@arm.com>
12165         * config/aarch64/aarch64.md (*mov<mode>_aarch64): Add simd attribute.
12167 2013-04-23  Andreas Schwab  <schwab@linux-m68k.org>
12169         * coretypes.h (gimple_stmt_iterator): Add struct to make
12170         compatible with C.
12172 2013-04-23  Richard Biener  <rguenther@suse.de>
12174         PR tree-optimization/57026
12175         * tree-vrp.c (simplify_conversion_using_ranges): Do not propagate
12176         from SSA names occuring in abnormal PHI nodes.
12178 2013-04-22  Andi Kleen  <ak@linux.intel.com>
12180         * lto/lto.c (print_lto_report_1): Fix LTO report names.
12182 2013-04-22  Andi Kleen  <ak@linux.intel.com>
12184         * lto/lto.c (print_lto_report_1): Declare early.
12185         (read_cgraph_and_symbols): Call print_lto_report_1 early.
12187 2013-04-22  Andi Kleen  <ak@linux.intel.com>
12189         * common.opt (-flto-report-wpa): Add.
12190         * doc/invoke.texi (-flto-report-wpa): Add.
12191         * lto/lto.c (do_whole_program_analysis): Check for lto-report-wpa.
12192         (lto_main): dito.
12194 2013-04-22  Xinliang David Li  <davidxl@google.com>
12196         * graph.c (draw_cfg_node_succ_edges): Add branch probility as label.
12197         * cfghhooks.c (dump_bb_for_graph): Dump profile count and frquency.
12198         * Makefile.in: New dependency
12200             David Daney <ddaney.cavm@gmail.com>
12202         * configure.ac (gcc_cv_as_micromips_support):  Use the
12203         --fatal-warnings option.
12204         * configure: Regenerate.
12206 2013-04-22  Marek Polacek  <polacek@redhat.com>
12208         PR sanitizer/56990
12209         * tsan.c (instrument_expr): Don't instrument expression
12210         in case its size is zero.
12212 2013-04-22  Uros Bizjak  <ubizjak@gmail.com>
12214         PR target/57032
12215         Revert:
12216         2013-03-17  Uros Bizjak  <ubizjak@gmail.com>
12218         * config/alpha/alpha.c (TARGET_LRA_P): New define.
12220 2013-04-22  James Greenhalgh  <james.greenhalgh@arm.com>
12222         * coretypes.h (gimple_stmt_iterator_d): Forward declare.
12223         (gimple_stmt_iterator): New typedef.
12224         * gimple.h (gimple_stmt_iterator): Rename to...
12225         (gimple_stmt_iterator_d): ... This.
12226         * doc/tm.texi.in (TARGET_FOLD_BUILTIN): Detail restriction that
12227         trees be valid for GIMPLE and GENERIC.
12228         (TARGET_GIMPLE_FOLD_BUILTIN): New.
12229         * gimple-fold.c (gimple_fold_call): Call target hook
12230         gimple_fold_builtin.
12231         * hooks.c (hook_bool_gsiptr_false): New.
12232         * hooks.h (hook_bool_gsiptr_false): New.
12233         * target.def (fold_stmt): New.
12234         * doc/tm.texi: Regenerate.
12236 2013-04-22  Vladimir Makarov  <vmakarov@redhat.com>
12238         PR target/57018
12239         * lra-eliminations.c (mark_not_eliminable): Prevent elimination of
12240         a set sp if no stack realignment.
12242 2013-04-22  Nick Clifton  <nickc@redhat.com>
12244         * config.gcc (tilegx-linux): Extend extra_objs rather than
12245         overwriting it.
12246         (tilepro-linux): Likewise.
12248 2013-04-22  James Greenhalgh  <james.greenhalgh@arm.com>
12250         * config/aarch64/aarch64-builtins.c
12251         (CF): Remove.
12252         (CF0, CF1, CF2, CF3, CF4, CF10): New.
12253         (VAR<1-12>): Add MAP parameter.
12254         (BUILTIN_*): Likewise.
12255         * config/aarch64/aarch64-simd-builtins.def: Set MAP parameter.
12256         * config/aarch64/aarch64-simd.md (aarch64_sshl_n<mode>): Remove.
12257         (aarch64_ushl_n<mode>): Likewise.
12258         (aarch64_sshr_n<mode>): Likewise.
12259         (aarch64_ushr_n<mode>): Likewise.
12260         (aarch64_<maxmin><mode>): Likewise.
12261         (aarch64_sqrt<mode>): Likewise.
12262         * config/aarch64/arm_neon.h (vshl<q>_n_*): Use new builtin names.
12263         (vshr<q>_n_*): Likewise.
12265 2013-04-22  James Greenhalgh  <james.greenhalgh@arm.com>
12267         * config/aarch64/aarch64-builtins.c
12268         (aarch64_simd_builtin_type_mode): Handle SF types.
12269         (sf_UP): Define.
12270         (BUILTIN_GPF): Define.
12271         (aarch64_init_simd_builtins): Handle SF types.
12272         * config/aarch64/aarch64-simd-builtins.def (frecpe): Add support.
12273         (frecps): Likewise.
12274         (frecpx): Likewise.
12275         * config/aarch64/aarch64-simd.md
12276         (simd_types): Update simd_frcp<esx> to simd_frecp<esx>.
12277         (aarch64_frecpe<mode>): New.
12278         (aarch64_frecps<mode>): Likewise.
12279         * config/aarch64/aarch64.md (unspec): Add UNSPEC_FRECP<ESX>.
12280         (v8type): Add frecp<esx>.
12281         (aarch64_frecp<FRECP:frecp_suffix><mode>): New.
12282         (aarch64_frecps<mode>): Likewise.
12283         * config/aarch64/iterators.md (FRECP): New.
12284         (frecp_suffix): Likewise.
12285         * config/aarch64/arm_neon.h
12286         (vrecp<esx><qsd>_<fd><32, 64>): Convert to using builtins.
12288 2013-04-22  Christian Bruel  <christian.bruel@st.com>
12290         PR target/56995
12291         * config/sh/sh.h (enum reg_class): Remove DF_HI_REGS.
12292         (REG_CLASS_NAMES): Idem.
12293         (REG_CLASS_CONTENTS): Idem.
12294         (REGCLASS_HAS_FP_REG): Idem.
12295         * config/sh/sh.c (sh_cannot_change_mode_class): Idem.
12296         (sh_conditional_register_usage): Idem.
12298 2013-04-21  Jeff Law  <law@redhat.com>
12300         * tree-ssa-forwprop.c (simplify_conversion_from_bitmask): New function.
12301         (ssa_forward_propagate_and_combine): Use it.
12303 2013-04-19  Vladimir Makarov  <vmakarov@redhat.com>
12305         * lra.c: Update the flow chart diagram.
12307 2013-04-19  Vladimir Makarov  <vmakarov@redhat.com>
12309         PR rtl-optimization/56847
12310         * lra-constraints.c (process_alt_operands): Discourage alternative
12311         with non-matche doffsettable memory constraint fro memory with
12312         known offset.
12314 2013-04-19  Richard Biener  <rguenther@suse.de>
12316         PR tree-optimization/56982
12317         * builtins.def (BUILT_IN_LONGJMP): longjmp is not a leaf
12318         function.
12319         * gimplify.c (gimplify_call_expr): Notice special calls.
12320         (gimplify_modify_expr): Likewise.
12321         * tree-cfg.c (make_abnormal_goto_edges): Handle setjmp-like
12322         abnormal control flow receivers.
12323         (call_can_make_abnormal_goto): Handle cfun->calls_setjmp
12324         in the same way as cfun->has_nonlocal_labels.
12325         (gimple_purge_dead_abnormal_call_edges): Likewise.
12326         (stmt_starts_bb_p): Make setjmp-like abnormal control flow
12327         receivers start a basic-block.
12329 2013-04-19  Richard Biener  <rguenther@suse.de>
12331         * tree-vectorizer.h (struct _slp_instance): Move load_permutation
12332         member ...
12333         (struct _slp_tree): ... here.  Make it a vector of unsigned ints.
12334         (SLP_INSTANCE_LOAD_PERMUTATION): Remove.
12335         (SLP_TREE_LOAD_PERMUTATION): Add.
12336         (vect_transform_slp_perm_load): Adjust prototype.
12337         * tree-vect-slp.c (vect_free_slp_tree): Adjust.
12338         (vect_free_slp_instance): Likewise.
12339         (vect_create_new_slp_node): Likewise.
12340         (vect_supported_slp_permutation_p): Remove.
12341         (vect_slp_rearrange_stmts): Adjust.
12342         (vect_supported_load_permutation_p): Likewise.  Inline
12343         vect_supported_slp_permutation_p here.
12344         (vect_analyze_slp_instance): Compute load permutations per
12345         slp node instead of per instance.
12346         (vect_get_slp_defs): Adjust.
12347         (vect_transform_slp_perm_load): Likewise.
12348         (vect_schedule_slp_instance): Remove redundant code.
12349         (vect_schedule_slp): Remove hack for PR56270, add it ...
12350         * tree-vect-stmts.c (vectorizable_load): ... here, do not
12351         CSE loads for SLP.  Adjust.
12353 2013-04-19  Greta Yorsh  <Greta.Yorsh@arm.com>
12355         * config/arm/arm.c (load_multiple_sequence, ldm_stm_operation_p): Fix
12356         spelling in two comments.
12358 2013-04-19  Greta Yorsh  <Greta.Yorsh@arm.com>
12360         PR target/56797
12361         * config/arm/arm.c (load_multiple_sequence): Require SP
12362         as base register for loads if SP is in the register list.
12364 2013-04-19  Martin Jambor  <mjambor@suse.cz>
12366         PR tree-optimization/56718
12367         * ipa-cp.c (ipa_value_from_known_type_jfunc): Moved...
12368         * ipa-prop.c (ipa_binfo_from_known_type_jfunc): ...here, renamed
12369         and made public.  Adjusted all callers.
12370         (ipa_intraprocedural_devirtualization): New function.
12371         * ipa-prop.h (ipa_binfo_from_known_type_jfunc): Declare.
12372         (ipa_intraprocedural_devirtualization): Likewise.
12373         * Makefile.in (tree-ssa-pre.o): Add ipa-prop.h to dependencies.
12375 2013-04-19  Richard Biener  <rguenther@suse.de>
12377         PR tree-optimization/57000
12378         * tree-ssa-reassoc.c (pass_reassoc): Add TODO_update_ssa_only_virtuals.
12380 2013-04-19  Terry Guo  <terry.guo@arm.com>
12382         * config/arm/cortex-m4-fpu.md (cortex_m4_v): Delete cpu unit.
12383         Replace with ...
12384         (cortex_m4_v_a,  cortex_m4_v_b): ... new cpu units.
12385         (cortex_m4_v, cortex_m4_exa_va, cortex_m4_exb_vb): New reservations.
12386         (cortex_m4_fmacs): Use new reservations.
12387         (cortex_m4_f_load, cortex_m4_f_store): Likewise.
12389 2013-04-18  Vladimir Makarov  <vmakarov@redhat.com>
12391         PR rtl-optimization/56999
12392         * lra-coalesce.c (coalescable_pseudo_p): Remove 2nd parameter and
12393         related code.
12394         (lra_coalesce): Remove split_origin_bitmap and related code.
12395         * lra.c (lra): Coalesce after undoing inheritance.  Recreate live
12396         ranges if necessary.
12398 2013-04-18  Uros Bizjak  <ubizjak@gmail.com>
12400         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers):
12401         New array.
12402         (ix86_expand_call): Remove clobbered_registers array and use
12403         x86_64_ms_sysv_extra_clobbered_registers instead.
12404         * config/i386/i386.h (x86_64_ms_sysv_extra_clobbered_registers):
12405         Declare here.
12406         * config/i386/predicates.md (call_rex64_ms_sysv_operation): New
12407         predicate.
12408         * config/i386/i386.md (*call_rex64_ms_sysv): Use
12409         call_rex64_ms_sysv_operation predicate.  Remove explicit clobbers.
12410         (*call_value_rex64_ms_sysv): Ditto.
12412 2013-04-18  Cary Coutant  <ccoutant@google.com>
12414         * dwarf2out.c (output_pubnames): Check die_perennial_p of
12415         parent instead of die_mark.
12417 2013-04-18  Diego Novillo  <dnovillo@google.com>
12419         * gimple.c (create_gimple_tmp): New.
12420         (get_expr_type): New.
12421         (build_assign): New.
12422         (build_type_cast): New.
12423         * gimple.h (enum ssa_mode): Define.
12424         (gimple_seq_set_location): New.
12425         * asan.c (build_check_stmt): Change some gimple_build_* calls
12426         to use build_assign and build_type_cast.
12428 2013-04-18  Richard Biener  <rguenther@suse.de>
12430         * tree-vect-data-refs.c (vect_analyze_group_access): Properly
12431         handle negative step.  Remove redundant checks.
12432         (vect_create_data_ref_ptr): Avoid ICEs with non-constant steps.
12433         * tree-vect-stmts.c (vectorizable_load): Instead of asserting
12434         for negative step and grouped loads fail to vectorize.
12436 2013-04-18  Steven Bosscher  <steven@gcc.gnu.org>
12438         * emit-rtl.c (reset_insn_used_flags): New function.
12439         (reset_all_used_flags): Use it.
12440         (verify_insn_sharing): New function.
12441         (verify_rtl_sharing): Fix verification for SEQUENCEs.
12443 2013-04-18  Jakub Jelinek  <jakub@redhat.com>
12445         PR tree-optimization/56984
12446         * tree-vrp.c (register_edge_assert_for_2): For (x >> M) < N
12447         and (x >> M) >= N don't register any assertion if N << M is the
12448         minimum value.
12450 2013-04-18  Steven Bosscher  <steven@gcc.gnu.org>
12452         * lower-subreg.c (resolve_simple_move): If called self-recursive,
12453         do not delete_insn insns that have not yet been emitted, only
12454         unlink them with remove_insn.
12455         * df-scan.c (df_insn_delete): Revert r197492.
12457 2013-04-17  Steven Bosscher  <steven@gcc.gnu.org>
12459         * emit-rtl.c (link_insn_into_chain): Handle chaining of SEQUENCEs.
12460         * reorg.c (emit_delay_sequence): Simplify with emit-rtl API.
12462 2013-04-17  Greta Yorsh  <Greta.Yorsh@arm.com>
12464         * config/arm/arm.md (movsicc_insn): Convert define_insn into
12465         define_insn_and_split.
12466         (and_scc,ior_scc,negscc): Likewise.
12467         (cmpsi2_addneg, subsi3_compare): Convert to named patterns.
12469 2013-04-17  Greta Yorsh  <Greta.Yorsh@arm.com>
12471         * config/arm/arm.c (use_return_insn): Return 0 for targets that
12472         can benefit from using a sequence of LDRD instructions in epilogue
12473         instead of a single LDM instruction.
12475 2013-04-17  Manuel López-Ibáñez  <manu@gcc.gnu.org>
12477         PR 45688
12478         * doc/extend.texi: Fix typo.
12480 2013-04-17  Richard Biener  <rguenther@suse.de>
12482         * tree-vect-slp.c (vect_build_slp_tree_1): Split out from ...
12483         (vect_build_slp_tree): ... here.
12484         (vect_build_slp_tree_1): Compute which stmts of the SLP group
12485         match.  Remove special-casing of mismatched complex loads.
12486         (vect_build_slp_tree): Based on the result from vect_build_slp_tree_1
12487         re-try the match with swapped commutative operands.
12488         (vect_supported_load_permutation_p): Remove special-casing of
12489         mismatched complex loads.
12490         (vect_analyze_slp_instance): Adjust.
12492 2013-04-17  Richard Biener  <rguenther@suse.de>
12494         PR rtl-optimization/56921
12495         * cfgloop.h (struct loop): Add simple_loop_desc member.
12496         (struct niter_desc): Mark with GTY(()).
12497         (simple_loop_desc): Do not use aux field but simple_loop_desc.
12498         * loop-iv.c (get_simple_loop_desc): Likewise.
12499         (free_simple_loop_desc): Likewise.
12501         Revert
12502         2013-04-16  Richard Biener  <rguenther@suse.de>
12504         PR rtl-optimization/56921
12505         * loop-init.c (pass_rtl_move_loop_invariants): Add
12506         TODO_do_not_ggc_collect to todo_flags_finish.
12507         (pass_rtl_unswitch): Same.
12508         (pass_rtl_unroll_and_peel_loops): Same.
12509         (pass_rtl_doloop): Same.
12511 2013-04-17  Eric Botcazou  <ebotcazou@adacore.com>
12513         * tree-ssa-alias.c (nonoverlapping_component_refs_of_decl_p): New.
12514         (decl_refs_may_alias_p): Add REF1 and REF2 parameters.
12515         Use nonoverlapping_component_refs_of_decl_p to disambiguate component
12516         references.
12517         (refs_may_alias_p_1): Adjust call to decl_refs_may_alias_p.
12518         * tree-streamer.c (record_common_node): Adjust reference in comment.
12520 2013-04-17  Terry Guo  <terry.guo@arm.com>
12522         * config/arm/cortex-m4.md: Add a new bypass.
12524 2013-04-16  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
12526         * config/aarch64/aarch64.md (*adds_<optab><mode>_multp2):
12527         New pattern.
12528         (*subs_<optab><mode>_multp2): New pattern.
12529         (*adds_<optab><ALLX:mode>_<GPI:mode>): New pattern.
12530         (*subs_<optab><ALLX:mode>_<GPI:mode>): New pattern.
12532 2013-04-16  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
12534         * config/aarch64/aarch64.md (*adds_mul_imm_<mode>): New pattern.
12535         (*subs_mul_imm_<mode>): New pattern.
12537 2013-04-16  David Edelsohn  <dje.gcc@gmail.com>
12539         PR target/56948
12540         * config/rs6000/vsx.md (vsx_mov<mode>): Add j->r alternative.
12541         (vsx_movti_64bit): Change j->wa to O->wa.  Add n->r alternative.
12542         (vsx_movti_32bit): Change j->wa to O->wa.
12544 2013-04-16  Richard Biener  <rguenther@suse.de>
12546         PR rtl-optimization/56921
12547         * loop-init.c (pass_rtl_move_loop_invariants): Add
12548         TODO_do_not_ggc_collect to todo_flags_finish.
12549         (pass_rtl_unswitch): Same.
12550         (pass_rtl_unroll_and_peel_loops): Same.
12551         (pass_rtl_doloop): Same.
12553 2013-04-16  Greta Yorsh  <Greta.Yorsh@arm.com>
12555         * config/arm/arm.c (emit_multi_reg_push): New declaration
12556         for an existing function.
12557         (arm_emit_strd_push): New function.
12558         (arm_expand_prologue): Used here.
12559         (arm_emit_ldrd_pop): New function.
12560         (arm_expand_epilogue): Used here.
12561         (arm_get_frame_offsets): Update condition.
12562         (arm_emit_multi_reg_pop): Add a special case for load of a single
12563         register with writeback.
12565 2013-04-16  Uros Bizjak  <ubizjak@gmail.com>
12567         * doc/invoke.texi (i386 Option): Reword -mstack-protector-guard
12568         description.
12570 2013-04-16  Richard Biener  <rguenther@suse.de>
12572         PR tree-optimization/56756
12573         * tree-ssa-loop-im.c (struct first_mem_ref_loc_1): New functor.
12574         (first_mem_ref_loc): New.
12575         (execute_sm): Place the load temporarily before a previous
12576         access instead of in the latch edge to ensure its SSA dependencies
12577         are defined at points dominating the load.
12579 2013-04-16  Steven Bosscher  <steven@gcc.gnu.org>
12581         * cfgrtl.c (cfg_layout_merge_blocks): Revert r184005, implement
12582         correct fix by moving header and footer insn to the footer of
12583         the merged basic block.  Clear BB_END of the merged-away block.
12585         PR middle-end/43631
12586         * emit-rtl.c (make_note_raw): New function.
12587         (link_insn_into_chain): New static inline function.
12588         (add_insn): Use it.
12589         (add_insn_before, add_insn_after): Factor insn chain linking code...
12590         (add_insn_before_nobb, add_insn_after_nobb): ...here, new functions
12591         using link_insn_into_chain.
12592         (note_outside_basic_block_p): New helper function for emit_note_after
12593         and emit_note_before.
12594         (emit_note_after): Use nobb variant of add_insn_after if the note
12595         should not be contained in a basic block.
12596         (emit_note_before): Use nobb variant of add_insn_before if the note
12597         should not be contained in a basic block.
12598         (emit_note_copy): Use make_note_raw.
12599         (emit_note): Likewise.
12600         * bb-reorder.c (insert_section_boundary_note): Remove hack to set
12601         BLOCK_FOR_INSN to NULL manually for NOTE_INSN_SWITCH_TEXT_SECTIONS.
12602         * jump.c (cleanup_barriers): Use reorder_insns_nobb to avoid making
12603         the moved barrier the tail of the basic block it follows.
12604         * var-tracking.c (pass_variable_tracking): Add TODO_verify_flow.
12606 2013-04-15  Jakub Jelinek  <jakub@redhat.com>
12608         PR tree-optimization/56962
12609         * gimple-ssa-strength-reduction.c (record_increment): Only set
12610         initializer if gimple_assign_rhs_code is {,POINTER_}PLUS_EXPR and
12611         either rhs1 or rhs2 is equal to c->base_expr.
12613 2013-04-15  Richard Biener  <rguenther@suse.de>
12615         PR tree-optimization/56933
12616         * tree-vectorizer.h (struct _stmt_vec_info): Remove read_write_dep
12617         member.
12618         (GROUP_READ_WRITE_DEPENDENCE): Remove.
12619         (STMT_VINFO_GROUP_READ_WRITE_DEPENDENCE): Likewise.
12620         * tree-vect-data-refs.c (vect_analyze_group_access): Move
12621         dependence check ...
12622         vect_analyze_data_ref_dependence (vect_analyze_data_ref_dependence):
12623         ... here.
12624         * tree-vect-stmts.c (new_stmt_vec_info): Do not initialize
12625         GROUP_READ_WRITE_DEPENDENCE.
12627 2013-04-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
12629         * emit-rtl.c (reset_all_used_flags): New function.
12630         (verify_rtl_sharing): Call reset_all_used_flags before and after
12631         performing the checks.
12633 2013-04-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12635         * config/arm/arm.c (const_ok_for_dimode_op): Handle AND case.
12636         * config/arm/arm.md (*anddi3_insn): Change to insn_and_split.
12637         * config/arm/constraints.md (De): New constraint.
12638         * config/arm/neon.md (anddi3_neon): Delete.
12639         (neon_vand<mode>): Expand to standard anddi3 pattern.
12640         * config/arm/predicates.md (imm_for_neon_inv_logic_operand):
12641         Move earlier in the file.
12642         (neon_inv_logic_op2): Likewise.
12643         (arm_anddi_operand_neon): New predicate.
12645 2013-04-15  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12647         * configure.ac (gcc_cv_ld_as_needed): Set
12648         gcc_cv_ld_as_needed_option, gcc_cv_no_as_needed_option.
12649         Use -z ignore, -z record on *-*-solaris2*.
12650         (HAVE_LD_AS_NEEDED): Update comment.
12651         (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Define.
12652         * configure: Regenerate.
12653         * config.in: Regenerate.
12654         * gcc.c (init_gcc_specs) [USE_LD_AS_NEEDED]: Use
12655         LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION.
12656         * config/sol2.h [HAVE_LD_AS_NEEDED] (USE_LD_AS_NEEDED): Define.
12657         * doc/tm.texi.in (USE_LD_AS_NEEDED): Allow for --as-needed
12658         equivalents.  Fix markup.
12659         * doc/tm.texi: Regenerate.
12661 2013-04-15  Andrew Hsieh  <andrewhsieh.google.com>
12663         * config/i386/i386.opt: New option mstack-protector-guard=.
12664         * config/i386/i386-opts.h: Add enum stack_protector_guard.
12665         * config/i386/i386.h: Define TARGET_SSP_GLOBAL_GUARD and
12666         TARGET_SSP_TLS_GUARD.
12667         * config/i386/i386.c (ix86_option_override_internal): Set
12668         ix86_stack_protector_guard.
12669         * config/i386/i386.md (stack_protect_set): Enable for
12670         TARGET_SSP_TLS_GUARD only.
12671         (stack_protect_set_<mode>): Ditto.
12672         (stack_protect_test): Ditto.
12673         (stack_protect_test_<mode>): Ditto.
12674         * doc/invoke.texi (i386 Option): Document.
12676 2013-04-15  Eric Botcazou  <ebotcazou@adacore.com>
12678         PR target/56890
12679         * config/sparc/sparc.c (enum sparc_mode_class): Add H_MODE value.
12680         (S_MODES): Set H_MODE bit.
12681         (SF_MODES): Set only S_MODE and SF_MODE bits.
12682         (DF_MODES): Set SF_MODES and only D_MODE and DF_MODE bits.
12683         (sparc_init_modes) <MODE_INT>: Set H_MODE bit for sub-word modes.
12684         <MODE_VECTOR_INT>: Do not set SF_MODE for sub-word modes.
12685         <MODE_FLOAT>: Likewise.
12687 2013-04-15  Joey Ye  <joey.ye@arm.com>
12689         * config/arm/arm.c (thumb_far_jump_used_p): Fix typo in comments.
12691 2013-04-15  Joey Ye  <joey.ye@arm.com>
12693         * config/arm/arm.c (thumb1_final_prescan_insn): Assert lr save
12694         for real far jump.
12695         (thumb_far_jump_used_p): Count instruction size and set
12696         far_jump_used.
12698 2013-04-14  Eric Botcazou  <ebotcazou@adacore.com>
12700         * reorg.c (fill_simple_delay_slots): Reindent block of code.
12701         * resource.c (mark_target_live_regs): Reformat conditional block.
12703 2013-04-13  Steven Bosscher  <steven@gcc.gnu.org>
12705         * sched-deps.c (deps_analyze_insn): Do not check for EH_REGION insn
12706         notes, they are emitted only just before final.
12707         * sched-int.h: Include insn-attr.h before checking INSN_SCHEDULING.
12709 2013-04-13  Steven Bosscher  <steven@gcc.gnu.org>
12711         * emit-rtl.c (remove_insn): Do not call df_insn_delete here.
12712         * cfgrtl.c (delete_insn): Call it here instead.
12713         * lra-spills.c (lra_final_code_change): Use delete_insn.
12714         * haifa-sched.c (sched_remove_insn): Likewise.
12715         * sel-sched-ir.c (return_nop_to_pool): Clear INSN_DELETED_P for nops
12716         returning to the nop pool.
12717         (sel_remove_insn): Simplify the only_disconnect case via remove_insn,
12718         use delete_insn for definitive removal.  Clear BLOCK_FOR_INSN.
12720 2013-04-12  Steven Bosscher  <steven@gcc.gnu.org>
12722         * doc/tm.texi.in (LOOP_ALIGN): Remove loop note references.
12723         * doc/tm.texi: Regenerated.
12725 2013-04-12  Uros Bizjak  <ubizjak@gmail.com>
12727         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use ANY_QI_REGNO_P in
12728         QImode checks.
12730 2013-04-12  Steven Bosscher  <steven@gcc.gnu.org>
12732         * df-core.c (df_find_def): Compare register numbers.
12733         (df_find_use): Likewise.
12735 2013-04-12  Vladimir Makarov  <vmakarov@redhat.com>
12737         PR target/56903
12738         * config/i386/i386.c (ix86_hard_regno_mode_ok): Add
12739         lra_in_progress for return.
12741 2013-04-12  Greta Yorsh  <Greta.Yorsh@arm.com>
12743         * config/arm/arm.md (mov_scc,mov_negscc,mov_notscc): Convert
12744         define_insn into define_insn_and_split and emit movsicc patterns.
12746 2013-04-12  Greta Yorsh  <Greta.Yorsh@arm.com>
12748         * config/arm/arm.c (gen_operands_ldrd_strd): Initialize "base".
12750 2013-04-12  Richard Biener  <rguenther@suse.de>
12752         * tree-pass.h (TODO_do_not_ggc_collect): New.
12753         * passes.c (execute_one_ipa_transform_pass): Honor
12754         TODO_do_not_ggc_collect.
12755         (execute_one_pass): Likewise.
12757         Revert
12758         2013-04-10  Richard Biener  <rguenther@suse.de>
12760         * passes.c (init_optimization_passes): Remove reload pass.
12761         * ira.c (do_reload): Merge into ...
12762         (ira): ... this.
12763         (rest_of_handle_reload): Remove.
12764         (pass_reload): Likewise.
12765         * config/i386/i386.c (ix86_option_override): Refer to ira instead
12766         of reload for vzeroupper pass placement.
12768 2013-04-12  Jakub Jelinek  <jakub@redhat.com>
12770         PR tree-optimization/56918
12771         PR tree-optimization/56920
12772         * fold-const.c (int_const_binop_1): Use op1.mul_with_sign (op2, ...)
12773         instead of op1 - op2.  Pass 2 * TYPE_PRECISION (type) as second
12774         argument to rshift method.  For 2 * HOST_BITS_PER_WIDE_INT precision
12775         use wide_mul_with_sign method.
12777 2013-04-12  Richard Biener  <rguenther@suse.de>
12779         * gimple.c (is_gimple_constant): Vector CONSTRUCTORs should
12780         not be considered a gimple constant.
12782 2013-04-12  Marc Glisse  <marc.glisse@inria.fr>
12784         * fold-const.c (const_binop): Handle vector shifts by a scalar.
12785         (fold_binary_loc): Call const_binop also for mixed vector-scalar
12786         operations.
12788 2013-04-12  Manuel López-Ibáñez  <manu@gcc.gnu.org>
12789             Jakub Jelinek  <jakub@redhat.com>
12791         * opts.c: Include diagnostic-color.h.
12792         (common_handle_option): Handle OPT_fdiagnostics_color_.
12793         * Makefile.in (OBJS-libcommon): Add diagnostic-color.o.
12794         (diagnostic.o, opts.o, pretty-print.o): Depend on diagnostic-color.h.
12795         (diagnostic-color.o): New.
12796         * common.opt (fdiagnostics-color, fdiagnostics-color=): New options.
12797         (diagnostic_color_rule): New enum.
12798         * dwarf2out.c (gen_producer_string): Don't print -fdiagnostics-color*.
12799         * langhooks.c (lhd_print_error_function): Add %r "locus" and %R around
12800         the location string.
12801         * diagnostic.def: Add 3rd argument to DEFINE_DIAGNOSTIC_KIND macros,
12802         either NULL, or color kind.
12803         * diagnostic-color.c: New file.
12804         * diagnostic-color.h: New file.
12805         * diagnostic-core.h (DEFINE_DIAGNOSTIC_KIND): Adjust macro for 3
12806         arguments.
12807         * doc/invoke.texi (-fdiagnostics-color): Document.
12808         * pretty-print.h (pp_show_color): Define.
12809         (struct pretty_print_info): Add show_color field.
12810         * diagnostic.c: Include diagnostic-color.h.
12811         (diagnostic_build_prefix): Adjust for 3 argument DEFINE_DIAGNOSTIC_KIND
12812         macros.  Colorize error:, warning: etc. strings and also the location
12813         string.
12814         (diagnostic_show_locus): Colorize the caret line.
12815         * pretty-print.c: Include diagnostic-color.h.
12816         (pp_base_format): Handle %r and %R format specifiers.  Colorize strings
12817         inside of %< %> quotes or quoted through q format modifier.
12819 2013-04-12  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
12821         * ifcvt.c (end_ifcvt_sequence): Mark a and b for unsharing as well.
12823 2013-04-11  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
12825         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Allow NEG
12826         code in CC_NZ mode.
12827         * config/aarch64/aarch64.md (*neg_<shift><mode>3_compare0): New
12828         pattern.
12830 2013-04-11  Marek Polacek  <polacek@redhat.com>
12832         PR tree-optimization/48184
12833         * params.def (PARAM_ALIGN_THRESHOLD): Increase the minimum value to 1.
12835 2013-04-11  Eric Botcazou  <ebotcazou@adacore.com>
12837         * stor-layout.c (skip_simple_constant_arithmetic): Move to...
12838         * tree.c (skip_simple_constant_arithmetic): ...here and make public.
12839         (skip_simple_arithmetic): Tidy up.
12840         * tree.h (skip_simple_constant_arithmetic): Declare.
12842 2013-04-11  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
12844         * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Define.
12846 2013-04-11  Richard Biener  <rguenther@suse.de>
12848         * tree-vect-loop.c (get_initial_def_for_induction): Properly
12849         generate vector constants.
12851 2013-04-11  Richard Biener  <rguenther@suse.de>
12853         PR tree-optimization/56878
12854         * tree-flow.h (outermost_invariant_loop_for_expr): Declare.
12855         * tree-ssa-loop-ivopts.c (outermost_invariant_loop_for_expr):
12856         New function.
12857         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
12858         Prefer to align the DR with the most invariant base address.
12860 2013-04-11  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
12862         * opts.c (common_handle_option): Fix formatting and add FALLTHRU
12863         comment.
12865 2013-04-11  James Greenhalgh  <james.greenhalgh@arm.com>
12867         * config/aarch64/aarch64-simd.md (aarch64_vcond_internal): Fix
12868         floating-point vector comparisons against 0.
12870 2013-04-11  Jakub Jelinek  <jakub@redhat.com>
12872         PR tree-optimization/56899
12873         * fold-const.c (extract_muldiv_1): Apply distributive law
12874         only if TYPE_OVERFLOW_WRAPS (ctype).
12876 2013-04-11  Bin Cheng  <bin.cheng@arm.com>
12878         PR target/56124
12879         * ira-costs.c (scan_one_insn): Check whether the source rtx of
12880         loading has side effect.
12882 2013-04-10  Steven Bosscher  <steven@gcc.gnu.org>
12884         * config/sparc/sparc.c: Include tree-pass.h.
12885         (TARGET_MACHINE_DEPENDENT_REORG): Do not redefine.
12886         (sparc_reorg): Rename to sparc_do_work_around_errata.  Move to
12887         head of file.  Change return type.  Split off gate function.
12888         (sparc_gate_work_around_errata): New function.
12889         (pass_work_around_errata): New pass definition.
12890         (insert_pass_work_around_errata) New pass insert definition to
12891         insert pass_work_around_errata just after delayed-branch scheduling.
12892         (sparc_option_override): Insert the pass.
12893         * config/sparc/t-sparc (sparc.o): Add TREE_PASS_H dependence.
12895 2013-04-10  David S. Miller  <davem@davemloft.net>
12897         * config/sparc/sparc.h (ASM_CPU_SPEC): Pass -Av8 if -mcpu=supersparc
12898         or -mcpu=hypersparc.
12900         * target.def (cstore_mode): New hook.
12901         * target.h: Include insn-codes.h
12902         * targhooks.c: Likewise.
12903         (default_cstore_mode): New function.
12904         * targhooks.h: Declare it.
12905         * doc/tm.texi.in: New hook slot for TARGET_CSTORE_MODE.
12906         * doc/tm.texi: Rebuild.
12907         * expmed.c (emit_cstore): Obtain cstore boolean result mode using
12908         target hook, rather than inspecting the insn_data.
12909         * config/sparc/sparc.c (sparc_cstore_mode): New function.
12910         (TARGET_CSTORE_MODE): Redefine.
12911         (emit_scc_insn): When TARGET_ARCH64, emit new 64-bit boolean
12912         result patterns.
12913         * config/sparc/predicates.md (cstore_result_operand): New special
12914         predicate.
12915         * config/sparc/sparc.md (cstoresi4, cstoredi4, cstore<F:mode>4):
12916         Use it for operand 0.
12917         (*seqsi_special): Rewrite using 'P' mode iterator on operand 0.
12918         (*snesi_special): Likewise.
12919         (*snesi_zero): Likewise.
12920         (*seqsi_zero): Likewise.
12921         (*sltu_insn): Likewise.
12922         (*sgeu_insn): Likewise.
12923         (*seqdi_special): Make operand 0 and comparison operation be of
12924         DImode.
12925         (*snedi_special): Likewise.
12926         (*snedi_special_vis3): Likewise.
12927         (*neg_snesi_zero): Rename to *neg_snesisi_zero.
12928         (*neg_snesi_sign_extend): Rename to *neg_snesidi_zero.
12929         (*snesi_zero_extend): Delete, covered by 'P' mode iterator.
12930         (*neg_seqsi_zero): Rename to *neg_seqsisi_zero.
12931         (*neg_seqsi_sign_extend): Rename to *neg_seqsidi_zero.
12932         (*seqsi_zero_extend): Delete, covered by 'P' mode iterator.
12933         (*sltu_extend_sp64): Likewise.
12934         (*neg_sltu_insn): Rename to *neg_sltusi_insn.
12935         (*neg_sltu_extend_sp64): Rename to *neg_sltudi_insn.
12936         (*sgeu_extend_sp64): Delete, covered by 'P' mode iterator.
12937         (*neg_sgeu_insn): Rename to *neg_sgeusi_insn.
12938         (*neg_sgeu_extend_sp64): Rename to *neg_sgeudi_insn.
12940 2013-04-10  Yufeng Zhang  <yufeng.zhang@arm.com>
12942         * config/aarch64/aarch64.c (aarch64_print_extension): New function.
12943         (aarch64_start_file): Use the new function.
12945 2013-04-10  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
12946             Jason Merrill  <jason@redhat.com>
12948         * common.opt: Add -gdwarf.
12949         * opts.c (common_handle_option): Handle it.
12950         * gcc.c (ASM_DEBUG_SPEC): Don't expect "-2" for DWARF.
12952 2013-04-10  Richard Biener  <rguenther@suse.de>
12954         * passes.c (execute_todo): Do not call ggc_collect conditional here.
12955         (execute_one_ipa_transform_pass): But unconditionally here.
12956         (execute_one_pass): And here.
12957         (init_optimization_passes): Remove reload pass.
12958         * tree-pass.h (TODO_ggc_collect): Remove.
12959         (pass_reload): Likewise.
12960         * ira.c (do_reload): Merge into ...
12961         (ira): ... this.
12962         (rest_of_handle_reload): Remove.
12963         (pass_reload): Likewise.
12964         * config/i386/i386.c (ix86_option_override): Refer to ira instead
12965         of reload for vzeroupper pass placement.
12966         * <everywhere>: Remove TODO_ggc_collect from todo_flags_start
12967         and todo_flags_finish of all passes.
12969 2013-04-10  Richard Biener  <rguenther@suse.de>
12971         * tree-vectorizer.h (struct _slp_oprnd_info): Remove
12972         first_const_oprnd field, rename first_def_type to first_op_type.
12973         * tree-vect-slp.c (vect_create_oprnd_info): Adjust.
12974         (vect_get_and_check_slp_defs): Always use the type of the
12975         operand.  Allow mixed vect_external_def, vect_constant_def types.
12976         (vect_get_constant_vectors): Handle mixed vect_external_def,
12977         vect_constant_def types.
12979 2013-04-10  Joern Rennecke  <joern.rennecke@embecosm.com>
12981         PR tree-optimization/55524
12982         * tree-ssa-math-opts.c
12983         (convert_mult_to_fma): Don't use an fms construct
12984         when we don't have an fms operation, but fnma, and it looks
12985         likely that we'll be able to use the latter.
12987 2013-04-10  Zhouyi Zhou  <yizhouzhou@ict.ac.cn>
12989         * cif-code.def (OVERWRITABLE): Correct the comment for overwritable
12990         function.
12991         * ipa-inline.c (can_inline_edge_p): Let dump mechanism report the
12992         inline fail caused by overwritable functions.
12994 2013-04-10  Chung-Ju Wu  <jasonwucj@gmail.com>
12996         * combine.c (simplify_compare_const): Use GET_MODE_MASK to filter out
12997         unnecessary bits in the constant power of two case.
12999 2013-04-10  Richard Biener  <rguenther@suse.de>
13001         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove
13002         broken code swapping operands.
13003         (vect_build_slp_tree): Do not compute load permutations here.
13004         (vect_analyze_slp_instance): Compute load permutations here,
13005         after building the SLP tree.
13007 2013-04-09  Christian Bruel  <christian.bruel@st.com>
13009         * config/sh/sh.md (barrier_align): Use next/prev_active_insn instead
13010         of next/prev_real_insn.
13012 2013-04-09  Jan Hubicka  <jh@suse.cz>
13014         * ipa.c (cgraph_externally_visible_p, varpool_externally_visible_p):
13015         Drop aliased parameter.
13016         (function_and_variable_visibility): Do not handle alias pairs.
13017         * cgraph.c (varpool_externally_visible_p): Update prototype.
13018         * varpool.c (varpool_add_new_variable): Update.
13020 2013-04-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13022         * config/arm/arm.md (minmax_arithsi_non_canon): New pattern.
13024 2013-04-09  Steven Bosscher  <steven@gcc.gnu.org>
13026         * sched-vis.c (print_pattern): Print SEQUENCE of insns as insns.
13028         * config/sparc/sparc.md: Use define_c_enum for "unspec" and "unspecv".
13030 2013-04-09  Marek Polacek  <polacek@redhat.com>
13032         PR tree-optimization/48762
13033         * params.def (PARAM_MAX_CSE_INSNS): Increase the minimum value to 1.
13035 2013-04-09  Richard Biener  <rguenther@suse.de>
13037         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove code
13038         dealing with cost.
13039         (vect_build_slp_tree): Likewise.
13040         (vect_analyze_slp_cost_1, vect_analyze_slp_cost): New functions
13041         calculating the cost of a SLP instance.
13042         (vect_analyze_slp_instance): Use it from here, after building
13043         the SLP tree.
13045 2013-04-09  Jakub Jelinek  <jakub@redhat.com>
13047         PR middle-end/56883
13048         * omp-low.c (expand_omp_for_generic, expand_omp_for_static_nochunk,
13049         expand_omp_for_static_chunk): Use simple_p = true in
13050         force_gimple_operand_gsi calls when assigning to addressable decls.
13052 2013-04-09  Jeff Law  <law@redhat.com>
13054         * tree-vrp.c (simplify_cond_using_ranges): Simplify test of boolean
13055         when the boolean was created by converting a wider object which
13056         had a boolean range.
13058 2013-04-09  Richard Biener  <rguenther@suse.de>
13060         * tree-vectorizer.h (slp_void_p): Remove.
13061         (slp_tree): Typedef before _slp_tree declaration.
13062         (struct _slp_tree): Use a vector of slp_tree as children.
13063         (vect_get_place_in_interleaving_chain): Remove.
13064         * tree-vect-data-refs.c (vect_get_place_in_interleaving_chain):
13065         Move ...
13066         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): ... here
13067         and make static.
13068         (vect_free_slp_tree, vect_print_slp_tree, vect_mark_slp_stmts,
13069         vect_mark_slp_stmts_relevant, vect_slp_rearrange_stmts,
13070         vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations,
13071         vect_schedule_slp_instance, vect_remove_slp_scalar_calls):
13072         Use slp_node instead of slp_void_p and adjust.
13074 2013-04-09  Richard Biener  <rguenther@suse.de>
13076         * tree-ssa-loop-manip.c (rewrite_into_loop_closed_ssa): Avoid
13077         work that is not necessary.
13079 2013-04-09  Jakub Jelinek  <jakub@redhat.com>
13081         PR tree-optimization/56854
13082         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Don't
13083         forward into clobber stmts if it would change MEM_REF lhs into
13084         non-MEM_REF.
13086 2013-04-09  Maxim Kuvyrkov  <maxim@kugelworks.com>
13088         * tree.c (type_hash_lookup, type_hash_add): Make static.
13089         * tree.h (type_hash_lookup, type_hash_add): Remove global declarations.
13091 2013-04-09  Richard Biener  <rguenther@suse.de>
13093         * tree.h (unsave_expr_now): Remove.
13094         * tree-inline.c (mark_local_for_remap_r): Remove.
13095         (unsave_expr_1): Likewise.
13096         (unsave_r): Likewise.
13097         (unsave_expr_now): Likewise.
13098         * tree-ssa-copy.c (replace_exp_1): Use unshare_expr.
13099         (propagate_tree_value): Likewise.
13101 2013-04-08  Steven Bosscher  <steven@gcc.gnu.org>
13103         * doc/rtl.texi (sequence): Rewrite documentation to match the
13104         current use of SEQUENCE rtl objects.
13105         * rtl.def (SEQUENCE): Likewise.
13107         * doc/rtl.texi (NOTE_INSN_EH_REGION_BEG, NOTE_INSN_EH_REGION_END):
13108         Update documentation.
13109         (NOTE_INSN_LOOP_BEG, NOTE_INSN_LOOP_END, NOTE_INSN_LOOP_CONT,
13110         NOTE_INSN_LOOP_VTOP): Remove documentation for non-existing notes.
13112         * reg-notes.def (REG_EH_CONTEXT): Remove unused note.
13114 2013-04-08  Teresa Johnson  <tejohnson@google.com>
13116         * basic-block.h (GCOV_COMPUTE_SCALE): Define.
13117         * ipa-inline-analysis.c (param_change_prob): Use helper rounding divide
13118         methods.
13119         (estimate_edge_size_and_time): Add comment to suggest using rounding
13120         methods.
13121         (estimate_node_size_and_time): Ditto.
13122         (remap_edge_change_prob): Use helper rounding divide methods.
13123         * value-prof.c (gimple_divmod_fixed_value_transform): Ditto.
13124         (gimple_mod_pow2_value_transform): Ditto.
13125         (gimple_mod_subtract_transform): Ditto.
13126         (gimple_ic_transform): Ditto.
13127         (gimple_stringops_transform): Ditto.
13128         * stmt.c (conditional_probability): Ditto.
13129         (emit_case_dispatch_table): Ditto.
13130         * lto-cgraph.c (merge_profile_summaries): Ditto.
13131         * tree-optimize.c (execute_fixup_cfg): Ditto.
13132         * cfgcleanup.c (try_forward_edges): Ditto.
13133         * cfgloopmanip.c (scale_loop_profile): Ditto.
13134         (loopify): Ditto.
13135         (duplicate_loop_to_header_edge): Ditto.
13136         (lv_adjust_loop_entry_edge): Ditto.
13137         * tree-vect-loop.c (vect_transform_loop): Ditto.
13138         * profile.c (compute_branch_probabilities): Ditto.
13139         * cfgbuild.c (compute_outgoing_frequencies): Ditto.
13140         * lto-streamer-in.c (input_cfg): Ditto.
13141         * gimple-streamer-in.c (input_bb): Ditto.
13142         * ipa-cp.c (update_profiling_info): Ditto.
13143         (update_specialized_profile): Ditto.
13144         * tree-vect-loop-manip.c (slpeel_tree_peel_loop_to_edge): Ditto.
13145         * cfg.c (update_bb_profile_for_threading): Add comment to suggest using
13146         rounding methods.
13147         * sched-rgn.c (compute_dom_prob_ps): Ditto.
13148         (compute_trg_info): Ditto.
13149         * cfgrtl.c (force_nonfallthru_and_redirect): Ditto.
13150         (purge_dead_edges): Ditto.
13151         * loop-unswitch.c (unswitch_loop): Ditto.
13152         * cgraphclones.c (cgraph_clone_edge): Ditto.
13153         (cgraph_clone_node): Ditto.
13154         * tree-inline.c (copy_bb): Ditto.
13155         (copy_edges_for_bb): Ditto.
13156         (initialize_cfun): Ditto.
13157         (copy_cfg_body): Ditto.
13158         (expand_call_inline): Ditto.
13160 2013-04-08  Kai Tietz  <ktietz@redhat.com>
13162         * config/i386/cygwin.h (EXTRA_OS_CPP_BUILTINS): Replaced
13163         TARGET_CYGWIN64 by TARGET_64BIT.
13165 2013-04-08  Joern Rennecke  <joern.rennecke@embecosm.com>
13167         * config/epiphany/epiphany.md (GPR_1): New constant.
13168         (define_expand "mov<mode>cc): FAIL if gen_compare_reg returned 0.
13169         * config/epiphany/epiphany.c (gen_compare_reg):
13170         For flag_finite_math_only, avoid swapping operands when r0 and/or r1
13171         is already in place.
13172         Use GPR_0 / GPR_1 instead of 0/1 for r0/r1 register numbers.
13173         Don't require being called during rtl expansion; If y operlaps r0,
13174         return 0.
13175         (epiphany_compute_frame_size, epiphany_expand_prologue): Use GPR_1.
13176         (epiphany_expand_epilogue): Likewise.
13178         * config/epiphany/epiphany.c (epiphany_select_cc_mode):
13179         Don't use CC_FPmode for ORDERED / UNORDERED.
13180         * config/epiphany/epiphany.md (cmpsf_ord): Make pattern unconditional.
13182         * config/epiphany/constraints.md (CnL): New constraint.
13183         * config/epiphany/epiphany.md (addsi3_i): Add r/r/CnL alternative.
13184         * config/epiphany/predicates.md (add_operand): Allow 1024.
13186         * config/epiphany/epiphany.md (logical_op): New code iterator.
13187         (op_mnc): New code attribute.
13188         (<op_mnc>_f, mov_f, cstoresi4): New patterns.
13189         (mov_f+1, mov_f+2): New peephole2 patterns.
13191         * config/epiphany/epiphany.md (mov_f+2): New peephole2 pattern.
13192         (cstoresi4): Also allow re-use of zero result when doing a NE
13193         comparison to a non-zero operand.
13194         Use (clobber (scratch)) for first insn if the gpr output is not needed.
13196         * config/epiphany/epiphany.md (<insn_opname>v2si3):
13197         Use gen_addsi3_i / gen_subsi3_i.
13199 2013-04-08  Jakub Jelinek  <jakub@redhat.com>
13201         PR c++/34949
13202         PR c++/50243
13203         * tree-eh.c (optimize_clobbers): Only remove clobbers if bb doesn't
13204         contain anything but clobbers, at most one __builtin_stack_restore,
13205         optionally debug stmts and final resx, and if it has at least one
13206         incoming EH edge.  Don't check for SSA_NAME on LHS of a clobber.
13207         (sink_clobbers): Don't check for SSA_NAME on LHS of a clobber.
13208         Instead of moving clobbers with MEM_REF LHS with SSA_NAME address
13209         which isn't defaut definition, remove them.
13210         (unsplit_eh, cleanup_empty_eh): Use single_{pred,succ}_{p,edge}
13211         instead of EDGE_COUNT comparisons or EDGE_{PRED,SUCC}.
13212         * tree-ssa-ccp.c (execute_fold_all_builtins): Remove clobbers
13213         with MEM_REF LHS with SSA_NAME address.
13215 2013-04-08  Jeff Law  <law@redhat.com>
13217         * gimple.c (canonicalize_cond_expr_cond): Rewrite x ^ y into x != y.
13219 2013-04-08  Richard Biener  <rguenther@suse.de>
13221         * gimple-pretty-print.c (debug_gimple_stmt): Do not print
13222         extra newline.
13223         * tree-vect-loop.c (vect_determine_vectorization_factor): Dump
13224         determined vector type.
13225         (vect_analyze_data_refs): Likewise.
13226         (vect_get_new_vect_var): Adjust.
13227         (vect_create_destination_var): Preserve SSA name versions.
13228         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Do
13229         not dump anything here.
13231 2013-04-08  Joern Rennecke  <joern.rennecke@embecosm.com>
13233         * config/epiphany/epiphany.h (struct GTY (()) machine_function):
13234         Add member lr_slot_known.
13235         * config/epiphany/epiphany.md (reload_insi_ra): Compute lr_slot_offs
13236         if necessary.
13237         * config/epiphany/epiphany.c (epiphany_compute_frame_size):
13238         Remove code that sets lr_slot_offset according to what a previous
13239         version of epiphany_emit_save_restore used to do.
13240         (epiphany_emit_save_restore): When doing an lr save or restore,
13241         set/verify lr_slot_known and lr_slot_offset.
13243 2013-04-08  Xinyu Qi  <xyqi@marvell.com>
13245         PR target/54338
13246         * config/arm/arm.h (REG_CLASS_CONTENTS): Include IWMMXT_GR_REGS
13247         in ALL_REGS.
13249 2013-04-08  Richard Biener  <rguenther@suse.de>
13251         * alias.c (find_base_term): Fix thinko in previous change.
13253 2013-04-08  Jakub Jelinek  <jakub@redhat.com>
13255         * tree-loop-distribution.c (const_with_all_bytes_same): New function.
13256         (generate_memset_builtin): Only handle integer_all_onesp as -1 val if
13257         TYPE_PRECISION is equal to mode bitsize.  Use const_with_all_bytes_same
13258         if possible to compute val.
13259         (classify_partition): Verify CONSTRUCTOR doesn't have any elts.
13260         For QImode integers don't require anything about precision.  Use
13261         const_with_all_bytes_same to find out if the constant doesn't have
13262         repeated bytes in it.
13264 2013-04-08  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
13266         * config/s390/s390.c (s390_expand_insv): Only accept insertions
13267         within mode size.
13269 2013-04-08  Marek Polacek  <polacek@redhat.com>
13271         PR rtl-optimization/48182
13272         * params.def (PARAM_MIN_CROSSJUMP_INSNS): Increase the minimum
13273         value to 1.
13275 2013-04-06  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
13277         PR target/55487
13278         * config/pa/pa.c (legitimize_pic_address): Before incrementing label
13279         nuses, make sure we have a label.
13281 2013-04-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13283         PR target/56843
13284         * config/rs6000/rs6000.c (rs6000_emit_swdiv_high_precision): Remove.
13285         (rs6000_emit_swdiv_low_precision): Remove.
13286         (rs6000_emit_swdiv): Rewrite to handle between one and four
13287         iterations of Newton-Raphson generally; modify required number of
13288         iterations for some cases.
13289         * config/rs6000/rs6000.h (RS6000_RECIP_HIGH_PRECISION_P): Remove.
13291 2013-04-05  Steven Bosscher  <steven@gcc.gnu.org>
13293         * bb-reorder.c (fix_crossing_unconditional_branches): Remove a
13294         set-but-unused variable.
13296         * cgraph.c (cgraph_release_function_body): Clear cfun->cfg to make
13297         basic blocks of released function bodies garbage-collectable.
13299         * ree.c (find_and_remove_re): Do not call df_finish_pass here.
13300         (struct rtl_opt_pass): Add TODO_df_finish.
13302         * rtl.def (DEFINE_SUBST, DEFINE_SUBST_ATTR): Add documentation.
13304 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13306         * config/arm/constraints.md (q): New constraint.
13307         * config/arm/ldrdstrd.md: New file.
13308         * config/arm/arm.md (ldrdstrd.md) New include.
13309         (arm_movdi): Use "q" instead of "r" constraint
13310         for double-word memory access.
13311         (movdf_soft_insn): Likewise.
13312         * config/arm/vfp.md (movdi_vfp): Likewise.
13313         * config/arm/t-arm (MD_INCLUDES): Add ldrdstrd.md.
13314         * config/arm/arm-protos.h (gen_operands_ldrd_strd): New declaration.
13315         * config/arm/arm.c (gen_operands_ldrd_strd): New function.
13316         (mem_ok_for_ldrd_strd): Likewise.
13317         (output_move_double): Update assertion.
13319 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13321         * config/arm/arm.md: Comment on splitting Thumb1 patterns.
13323 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13325         * config/arm/arm.md (arm_smax_insn): Convert define_insn into
13326         define_insn_and_split.
13327         (arm_smin_insn,arm_umaxsi3,arm_uminsi3): Likewise.
13329 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13331         * config/arm/arm.md (arm_ashldi3_1bit): Convert define_insn into
13332         define_insn_and_split.
13333         (arm_ashrdi3_1bit,arm_lshrdi3_1bit): Likewise.
13334         (shiftsi3_compare): New pattern.
13335         (rrx): New pattern.
13336         * config/arm/unspecs.md (UNSPEC_RRX): New.
13338 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13340         * config/arm/arm.md (negdi_extendsidi): New pattern.
13341         (negdi_zero_extendsidi): Likewise.
13343 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13345         * config/arm/arm.md (andsi_iorsi3_notsi): Convert define_insn into
13346         define_insn_and_split.
13347         (arm_negdi2,arm_abssi2,arm_neg_abssi2): Likewise.
13348         (arm_cmpdi_insn,arm_cmpdi_unsigned): Likewise.
13350 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13352         * config/arm/arm.md (arm_subdi3): Convert define_insn into
13353         define_insn_and_split.
13354         (subdi_di_zesidi,subdi_di_sesidi): Likewise.
13355         (subdi_zesidi_di,subdi_sesidi_di,subdi_zesidi_zesidi): Likewise.
13357 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13359         * config/arm/arm.md (subsi3_carryin): New pattern.
13360         (subsi3_carryin_const): Likewise.
13361         (subsi3_carryin_compare,subsi3_carryin_compare_const): Likewise.
13362         (subsi3_carryin_shift,rsbsi3_carryin_shift): Likewise.
13364 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13366         * config/arm/arm.md (incscc,arm_incscc,decscc,arm_decscc): Delete.
13368 2013-04-05  Greta Yorsh  <Greta.Yorsh@arm.com>
13370         * config/arm/arm.md (addsi3_carryin_<optab>): Set attribute predicable.
13371         (addsi3_carryin_alt2_<optab>,addsi3_carryin_shift_<optab>): Likewise.
13373 2013-04-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13375         * config/arm/arm.c (arm_expand_builtin): Change fcode
13376         type to unsigned int.
13378 2013-04-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13380         * doc/invoke.texi (ARM Options): Document cortex-a53 support.
13382 2013-04-04  Ian Lance Taylor  <iant@google.com>
13384         * doc/standards.texi (Standards): The Go frontend supports the Go 1
13385         language standard.
13387 2013-04-04  Steven Bosscher  <steven@gcc.gnu.org>
13389         PR middle-end/56729
13390         * df-scan.c (df_insn_delete): Disable failing assert.
13392 2013-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13394         * config/arm/arm-protos.h (arm_builtin_vectorized_function):
13395         New function prototype.
13396         * config/arm/arm.c (TARGET_VECTORIZE_BUILTINS): Define.
13397         (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Likewise.
13398         (arm_builtin_vectorized_function): New function.
13400 2013-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13402         * config/arm/arm_neon_builtins.def: New file.
13403         * config/arm/arm.c (neon_builtin_data): Move contents to
13404         arm_neon_builtins.def.
13405         (enum arm_builtins): Include neon builtin definitions.
13406         (ARM_BUILTIN_NEON_BASE): Move from enum to macro.
13407         * config/arm/t-arm (arm.o): Add dependency on arm_neon_builtins.def.
13409 2013-04-04  Marek Polacek  <polacek@redhat.com>
13411         PR tree-optimization/48186
13412         * predict.c (maybe_hot_frequency_p): Return false if
13413         HOT_BB_FREQUENCY_FRACTION is 0.
13414         (cgraph_maybe_hot_edge_p): Likewise.
13416 2013-04-04  Richard Biener  <rguenther@suse.de>
13418         PR tree-optimization/56826
13419         * tree-vect-slp.c (vect_build_slp_tree): Compute ncopies
13420         more accurately.
13422 2013-04-04  Richard Biener  <rguenther@suse.de>
13424         PR tree-optimization/56213
13425         * tree-vect-data-refs.c (vect_check_strided_load): Remove.
13426         (vect_analyze_data_refs): Allow all non-nested loads as strided loads.
13428 2013-04-04  Richard Biener  <rguenther@suse.de>
13430         PR tree-optimization/56837
13431         * tree-loop-distribution.c (classify_partition): For non-zero
13432         values require that the value has the same precision as its
13433         mode to be useful as memset value.
13435 2013-04-03  Nick Clifton  <nickc@redhat.com>
13437         * config/v850/v850e3v5.md (fmasf4): Use fmaf.s on E3V5 architectures.
13438         (fmssf4): Use fmsf.s on E3V5 architectures.
13439         (fnmasf4): Use fnmaf.s on E3V5 architectures.
13440         (fnmssf4): Use fnmsf.s on E3V5 architectures.
13442 2013-04-03  Jeff Law  <law@redhat.com>
13444         * Makefile.in (lra-constraints.o): Depend on $(OPTABS_H).
13445         (lra-eliminations.o): Likewise.
13447 2013-04-03  Teresa Johnson  <tejohnson@google.com>
13449         * gcov-io.c (compute_working_sets): Moved most of body of old
13450         compute_working_sets here from profile.c.
13451         * gcov-io.h (NUM_GCOV_WORKING_SETS): Moved here from profile.c.
13452         (gcov_working_set_t): Moved typedef here from basic-block.h
13453         (compute_working_set): Declare.
13454         * profile.c (NUM_GCOV_WORKING_SETS): Moved to gcov-io.h.
13455         (get_working_sets): Renamed from compute_working_set,
13456         replace most of body with call to new compute_working_sets.
13457         (get_exec_counts): Replace call to compute_working_sets
13458         to get_working_sets.
13459         * profile.h (get_working_sets): Renamed from compute_working_set.
13460         * lto-cgraph.c (input_symtab): Replace call to compute_working_sets
13461         to get_working_sets.
13462         * basic-block.h (gcov_working_set_t): Moved to gcov-io.h.
13463         * gcov-dump.c (dump_working_sets): New function.
13465 2013-04-03  Kenneth Zadeck  <zadeck@naturalbridge.com>
13467         * hwint.c (sext_hwi, zext_hwi): New functions.
13468         * hwint.h (HOST_BITS_PER_HALF_WIDE_INT, HOST_HALF_WIDE_INT,
13469         HOST_HALF_WIDE_INT_PRINT, HOST_HALF_WIDE_INT_PRINT_C,
13470         HOST_HALF_WIDE_INT_PRINT_DEC, HOST_HALF_WIDE_INT_PRINT_DEC_C,
13471         HOST_HALF_WIDE_INT_PRINT_UNSIGNED, HOST_HALF_WIDE_INT_PRINT_HEX,
13472         HOST_HALF_WIDE_INT_PRINT_HEX_PURE): New symbols.
13473         (sext_hwi, zext_hwi): New functions.
13475 2013-04-03  Jeff Law  <law@redhat.com>
13477         PR tree-optimization/56799
13478         * tree-ssa-dom.c (record_equivalences_from_incoming_edge): Bring
13479         back test for widening conversion erroneously dropped in prior change.
13481 2013-04-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13483         PR target/56809
13484         * config/aarch64/aarch64.c (is_jump_table): Use next_active_insn
13485         instead of next_real_insn.
13487 2013-04-03  Marek Polacek  <polacek@redhat.com>
13489         PR sanitizer/55702
13490         * tsan.c (instrument_func_exit): Allow BUILT_IN_RETURN functions.
13492 2013-04-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13494         PR target/56809
13495         * config/arm/arm.c (is_jump_table): Use next_active_insn instead of
13496         next_real_insn.
13497         (thumb1_output_casesi): Likewise.
13498         (thumb2_output_casesi): Likewise.
13500 2013-04-03  Richard Biener  <rguenther@suse.de>
13502         PR tree-optimization/56817
13503         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely):
13504         Split out ...
13505         (tree_unroll_loops_completely_1): ... new function to manually
13506         walk the loop tree, properly defering outer loops of unrolled
13507         loops to later iterations.
13509 2013-04-03  Marc Glisse  <marc.glisse@inria.fr>
13511         * tree-vect-stmts.c (vectorizable_store): Accept BIT_FIELD_REF.
13512         (vectorizable_load): Likewise.
13513         * tree-vect-slp.c (vect_build_slp_tree): Likewise.
13514         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Handle VECTOR_TYPE.
13516 2013-04-03  Marc Glisse  <marc.glisse@inria.fr>
13518         * tree-flow-inline.h (get_addr_base_and_unit_offset_1): Handle
13519         BIT_FIELD_REF.
13521 2013-04-03  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
13523         * config/spu/spu.c (emit_nop_for_insn): Handle JUMP_TABLE_DATA.
13525 2013-04-03  Bin Cheng  <bin.cheng@arm.com>
13527         * rtl.h (AUTO_INC_DEC): Fix typo of HAVE_POST_MODIFY_DISP.
13529 2013-04-03  Marc Glisse  <marc.glisse@inria.fr>
13531         PR tree-optimization/56790
13532         * fold-const.c (fold_ternary_loc) <VEC_COND_EXPR>: Add constant
13533         folding.
13535 2013-04-03  Marc Glisse  <marc.glisse@inria.fr>
13537         * simplify-rtx.c (simplify_binary_operation_1) <VEC_SELECT>:
13538         Handle VEC_MERGE.
13539         (simplify_ternary_operation) <VEC_MERGE>: Use unsigned HOST_WIDE_INT
13540         for masks. Test for side effects. Handle nested VEC_MERGE. Handle
13541         equal arguments.
13543 2013-04-03  Jakub Jelinek  <jakub@redhat.com>
13545         PR c/19449
13546         * tree.h (force_folding_builtin_constant_p): New decl.
13547         * builtins.c (force_folding_builtin_constant_p): New variable.
13548         (fold_builtin_constant_p): Fold immediately also if
13549         force_folding_builtin_constant_p.
13551 2013-04-03  Richard Biener  <rguenther@suse.de>
13553         PR tree-optimization/56812
13554         * tree-vect-data-refs.c (vect_slp_analyze_data_ref_dependence):
13555         DRs of the same interleaving chain are independent.
13557 2013-04-02  Jason Merrill  <jason@redhat.com>
13559         * gdbinit.in (pbb): Use debug fn.
13561 2013-04-02  Lawrence Crowl  <crowl@google.com>
13563         * sese.h (struct ivtype_map_elt_s): Remove unused.
13564         (extern debug_ivtype_map): Remove unused.
13565         (extern eq_ivtype_map_elts): Remove unused.
13566         * sese.c (debug_ivtype_map): Removed unused.
13567         (debug_ivtype_map_1): Removed unused.
13568         (debug_ivtype_elt): Remove unused.
13569         (eq_ivtype_map_elts): Remove unused.
13572 2013-04-02  Kai Tietz  <ktietz@redhat.com>
13574         PR target/52790
13575         * config/i386/cygming.h (SUB_TARGET_RECORD_STUB): New sub-target macro.
13576         * config/i386/i386-protos.h (i386_pe_record_stub): Add new prototype.
13577         * config/i386/i386.c (legitimize_pe_coff_extern_decl): New static
13578         function.
13579         (legitimize_pe_coff_symbol): Likewise.
13580         (is_imported_p): New helper-function.
13581         (ix86_option_override_internal): Make MEDIUM_PIC the default code-model
13582         for Windows x64 targets.
13583         (ix86_expand_prologue): Optimize for pe-coff targets.
13584         (ix86_expand_split_stack_prologue): Adjust for pe-coff targets.
13585         (legitimate_pic_address_disp_p): Adjust for x64 pe-coff to support
13586         medium/large code-model.
13587         (legitimize_pic_address): Likewise.
13588         (legitimize_tls_address): Likewise.
13589         (ix86_expand_call): Likewise.
13590         (x86_output_mi_thunk): Likewise.
13591         (get_dllimport_decl): Add new beimport argument.
13592         (construct_plt_address): Don't assert for x64 pe-coff targets.
13593         * config/i386/i386.h (PIC_OFFSET_TABLE_REGNUM): Adjust for x64 pe-coff
13594         targets.
13595         (SYMBOL_FLAG_STUBVAR): New macro.
13596         (SYMBOL_REF_STUBVAR_P): Likewise.
13597         * config/i386/winnt.c (stub_list): New structure.
13598         (stub_head): New local variable.
13599         (i386_pe_record_stub): New function.
13600         (i386_pe_file_end): Emit refptr-stubs.
13602 2013-04-02  Jakub Jelinek  <jakub@redhat.com>
13604         PR rtl-optimization/56745
13605         * ifcvt.c (cond_exec_find_if_block): Don't try to optimize
13606         if then_bb has no successors and else_bb is EXIT_BLOCK_PTR.
13608         PR c++/34949
13609         * tree-ssa-alias.c (stmt_kills_ref_p_1): If base != ref->base
13610         and both of them are MEM_REFs, just compare first argument for
13611         equality and attempt to deal even with differing offsets.
13613         PR c++/34949
13614         * tree-cfg.c (verify_gimple_assign_single): Allow lhs
13615         of gimple_clobber_p to be MEM_REF.
13616         * gimplify.c (gimplify_modify_expr): Gimplify *to_p of
13617         an assignment from TREE_CLOBBER_P.  Allow it to be MEM_REF
13618         after gimplification.
13619         * asan.c (get_mem_ref_of_assignment): Don't instrument
13620         gimple_clobber_p stmts.
13621         * tree-ssa-dse.c (dse_optimize_stmt): Allow DSE of
13622         gimple_clobber_p stmt if they have MEM_REF lhs and
13623         are dead because of another gimple_clobber_p stmt.
13624         * tree-ssa-live.c (clear_unused_block_pointer): Treat
13625         gimple_clobber_p stmts like debug stmts.
13626         (remove_unused_locals): Remove clobbers with MEM_REF lhs
13627         that refer to unused VAR_DECLs or uninitialized values.
13628         * tree-sra.c (sra_ipa_reset_debug_stmts): Also remove
13629         gimple_clobber_p stmts if they refer to removed parameters.
13630         (get_repl_default_def_ssa_name, sra_ipa_modify_expr): Fix up
13631         formatting.
13633 2013-04-02  Uros Bizjak  <ubizjak@gmail.com>
13635         * config/i386/i386.md (*testqi_ext_3): Merge with *testqi_ext_3_rex64
13636         using SWI48 mode attribute.
13638 2013-04-02  Wei Mi  <wmi@google.com>
13640         * config/i386/i386.c (ix86_rtx_costs): Set proper rtx cost for
13641         ashl<mode>3_mask, *<shift_insn><mode>3_mask and
13642         *<rotate_insn><mode>3_mask in i386.md.
13644 2013-04-02  Alexander Ivchenko  <alexander.ivchenko@intel.com>
13646         * config.gcc (arm*-*-linux-*): Remove duplicate t-linux-android.
13648 2013-04-02  Richard Biener  <rguenther@suse.de>
13650         PR tree-optimization/56778
13651         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13652         Runtime alias tests are not supported for gather loads.
13653         * tree-vect-loop-manip.c (vect_loop_versioning): Insert
13654         stmts referenced from SSA operands before updating SSA form.
13656 2013-04-02  Ian Caulfield  <ian.caulfield@arm.com>
13657             Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13659         * config/arm/arm-arches.def (armv8-a): Default to cortex-a53.
13660         * config/arm/t-arm (MD_INCLUDES): Depend on cortex-a53.md.
13661         * config/arm/cortex-a53.md: New file.
13662         * config/arm/bpabi.h (BE8_LINK_SPEC): Handle cortex-a53.
13663         * config/arm/arm.md (generic_sched, generic_vfp): Handle cortex-a53.
13664         * config/arm/arm.c (arm_issue_rate): Likewise.
13665         * config/arm/arm-tune.md: Regenerate
13666         * config/arm/arm-tables.opt: Regenerate.
13667         * config/arm/arm-cores.def: Add cortex-a53.
13669 2013-04-02  Zhenqiang Chen  <zhenqiang.chen@arm.com>
13671         * config/arm/uclinux-elf.h: Add %L to LINK_GCC_C_SEQUENCE_SPEC for
13672         non-static link.
13674 2013-04-02  Sofiane Naci  <sofiane.naci@arm.com>
13676         * config/aarch64/aarch64.md (*mov<mode>_aarch64): Add variants for
13677         scalar load/store operations using B/H registers.
13678         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64): Likewise.
13680 2013-04-02  Sofiane Naci  <sofiane.naci@arm.com>
13682         * config/aarch64/aarch64.md (*mov<mode>_aarch64): Add alternatives for
13683         scalar move.
13684         * config/aarch64/aarch64.c
13685         (aarch64_simd_scalar_immediate_valid_for_move): New.
13686         * config/aarch64/aarch64-protos.h
13687         (aarch64_simd_scalar_immediate_valid_for_move): New.
13688         * config/aarch64/constraints.md (Dh, Dq): New.
13689         * config/aarch64/iterators.md (hq): New.
13691 2013-04-02  Eric Botcazou  <ebotcazou@adacore.com>
13693         * reorg.c (get_branch_condition): Deal with conditional returns.
13694         (fill_simple_delay_slots): Remove dead code dealing with jumps.
13696 2013-04-01  Wei Mi  <wmi@google.com>
13698         * config/i386/i386.md (*ashl<mode>3_mask): Rewrite as define_insn.
13699         Truncate operand 2 using %b asm operand modifier.
13700         (*<shift_insn><mode>3_mask): Ditto.
13701         (*<rotate_insn><mode>3_mask): Ditto.
13703 2013-04-01  Steven Bosscher  <steven@gcc.gnu.org>
13705         PR middle-end/56798
13706         * cfgbuild.c (inside_basic_block_p): Restore check broken at r197234.
13708 2013-03-31  Kaz Kojima  <kkojima@gcc.gnu.org>
13710         * config/sh/sh.md (casesi_worker_1): Use next_active_insn instead
13711         of next_real_insn.
13712         (casesi_worker_2, casesi_shift_media, casesi_load_media): Likewise.
13714 2013-03-30  Lawrence Crowl  <crowl@google.com>
13716         * dse.c (clear_alias_sets): Remove never set.
13717         (disqualified_clear_alias_sets): Remove never set.
13718         (clear_alias_mode_pool): Remove never set.
13719         (dse_step0): Remove condition that is never true.
13720         (canon_address): Remove condition that is never true.
13721         (dse_step7): Remove condition that is never true.
13722         (rest_of_handle_dse): Remove condition that is never true.
13723         (rest_of_handle_dse::did_global): Remove never read from above.
13724         (dse_step2_spill): Remove never called from above.
13725         (dse_step5_spill): Remove never called from above.
13727 2013-03-30  Steven Bosscher  <steven@gcc.gnu.org>
13729         * doc/md.texi (Standard Names) <casesi>: Update documentation for
13730         JUMP_TABLE_DATA changes.
13731         * doc/tm.texi.in (Dispatch Tables) <ASM_OUTPUT_CASE_LABEL>: Likewise.
13732         * doc/rtl.texi (Flags) <INSN_DELETED_P, SCHED_GROUP_P>: Likewise.
13733         (Insns) <jump_table_data>: New entry.
13734         * doc/tm.texi: Regenerate.
13736         * cfgrtl.c (fixup_reorder_chain): Do not emit barriers to BB_FOOTER.
13738         * postreload-gcse.c (bb_has_well_behaved_predecessors): Correct test
13739         for table jump at the end of a basic block using tablejump_p.
13740         * targhooks.c (default_invalid_within_doloop): Likewise.
13741         * config/rs6000/rs6000.c (TARGET_INVALID_WITHIN_DOLOOP): Remove
13742         target hook implementation that is identical to the default hook.
13743         (rs6000_invalid_within_doloop): Remove.
13745         * bb-reorder.c (fix_crossing_unconditional_branches): Remove set but
13746         unused variable from tablejump_p call.
13748         * rtl.def (JUMP_TABLE_DATA): New RTX_INSN object.
13749         * rtl.h (RTX_PREV, RTX_NEXT): Adjust for new JUMP_TABLE_DATA.
13750         (INSN_DELETED_P): Likewise.
13751         (emit_jump_table_data): New prototype.
13752         * gengtype.c (adjust_field_rtx_def): Handle JUMP_TABLE_DATA fields
13753         after 4th as unused.
13754         * print-rtl.c (print_rtl): Handle JUMP_TABLE_DATA.
13755         * sched-vis.c (print_insn): Likewise.
13756         * emit-rtl.c (active_insn_p): Consider JUMP_TABLE_DATA an active
13757         insn for compatibility with back ends that use next_active_insn to
13758         identify jump table data.
13759         (set_insn_deleted): Remove no longer useful JUMP_TABLE_DATA_P check.
13760         (remove_insn): Likewise.
13761         (emit_insn): Do not accept JUMP_TABLE_DATA objects in insn chains
13762         to be emitted.
13763         (emit_debug_insn, emit_jump_insn, emit_call_insn, emit_label): Idem.
13764         (emit_jump_table_data): New function.
13766         * cfgbuild.c (inside_basic_block_p): A JUMP_INSN is always inside a
13767         basic block, a JUMP_TABLE_DATA never is.
13768         (control_flow_insn_p): JUMP_TABLE_DATA is not a control flow insn.
13769         * cfgrtl.c (duplicate_insn_chain): Split handling of JUMP_TABLE_DATA
13770         off from code handling real insns.
13771         * final.c (get_attr_length_1): Simplify for JUMP_INSNs.
13772         * function.c (instantiate_virtual_regs): Remove JUMP_TABLE_DATA_P
13773         test, now redundant because JUMP_TABLE_DATA is not an INSN_P insn.
13774         * gcse.c (insert_insn_end_basic_block): Likewise, JUMP_TABLE_DATA_P
13775         is not a NONDEBUG_INSN_P.
13776         * ira-costs.c (scan_one_insn): Likewise.
13777         * jump.c (mark_all_labels): Likewise.
13778         (mark_jump_label_1): Likewise.
13779         * lra-eliminations.c (eliminate_regs_in_insn): Likewise.
13780         * lra.c (get_insn_freq): Expect all insns reaching here to be in
13781         a basic block.
13782         (check_rtl): Remove JUMP_TABLE_DATA_P test, not a NONDEBUG_INSN_P insn.
13783         * predict.c (expensive_function_p): Use FOR_BB_INSNS.
13784         * reload1.c (calculate_needs_all_insns): Call set_label_offsets for
13785         JUMP_TABLE_DATA_P insns.
13786         (calculate_elim_costs_all_insns): Likewise.
13787         (set_label_offsets): Recurse on the PATTERN of JUMP_TABLE_DATA insns.
13788         (elimination_costs_in_insn): Remove redundant JUMP_TABLE_DATA_P test.
13789         (delete_output_reload): Code style fixups.
13790         * reorg.c (dbr_schedule): Move JUMP_TABLE_DATA_P up to avoid setting
13791         insn flags on this non-insn.
13792         * sched-rgn.c (add_branch_dependences): Treat JUMP_TABLE_DATA insns
13793         as scheduling barriers, for pre-change compatibility.
13794         * stmt.c (emit_case_dispatch_table): Emit jump table data not as
13795         JUMP_INSN objects but instead as JUMP_TABLE_DATA objects.
13797         * config/alpha/alpha.c (alpha_does_function_need_gp): Remove
13798         redundant JUMP_TABLE_DATA_P test.
13799         * config/arm/arm.c (thumb_far_jump_used_p): Likewise.
13800         * config/frv/frv.c (frv_function_contains_far_jump): Likewise.
13801         (frv_for_each_packet): Likewise.
13802         * config/i386/i386.c (min_insn_size): Likewise.
13803         (ix86_avoid_jump_mispredicts): Likewise.
13804         * config/m32r/m32r.c (m32r_is_insn): Likewise.
13805         * config/mep/mep.c (mep_reorg_erepeat): Likewise.
13806         * config/mips/mips.c (USEFUL_INSN_P): Likewise.
13807         (mips16_insn_length): Robustify.
13808         (mips_has_long_branch_p): Remove redundant JUMP_TABLE_DATA_P test.
13809         (mips16_split_long_branches): Likewise.
13810         * config/pa/pa.c (pa_combine_instructions): Likewise.
13811         * config/rs6000/rs6000.c (get_next_active_insn): Treat
13812         JUMP_TABLE_DATA objects as active insns, like in active_insn_p.
13813         * config/s390/s390.c (s390_chunkify_start): Treat JUMP_TABLE_DATA
13814         as contributing to pool range lengths.
13815         * config/sh/sh.c (find_barrier): Restore check for ADDR_DIFF_VEC.
13816         Remove redundant JUMP_TABLE_DATA_P test.
13817         (sh_loop_align): Likewise.
13818         (split_branches): Likewise.
13819         (sh_insn_length_adjustment): Likewise.
13820         * config/spu/spu.c (get_branch_target): Likewise.
13822 2013-03-29  Jan Hubicka  <jh@suse.cz>
13824         * lto-cgraph.c (output_profile_summary, input_profile_summary): Use
13825         gcov streaming; stream hot bb threshold to ltrans.
13826         * predict.c (get_hot_bb_threshold): Break out from ....
13827         (maybe_hot_count_p): ... here.
13828         (set_hot_bb_threshold): New function.
13829         * lto-section-in.c (lto_section_name): Add profile.
13830         * profile.h (get_hot_bb_threshold, set_hot_bb_threshold): Declare.
13831         * ipa.c: Include hash-table.h, tree-inline.h, profile.h, lto-streamer.h
13832         and data-streamer.h
13833         (histogram_entry): New structure.
13834         (histogram, histogram_pool): New global vars.
13835         (histogram_hash): New structure.
13836         (histogram_hash::hash): New method.
13837         (histogram_hash::equal): Likewise.
13838         (account_time_size): New function.
13839         (cmp_counts): New function.
13840         (dump_histogram): New function.
13841         (ipa_profile_generate_summary): New function.
13842         (ipa_profile_write_summary): New function.
13843         (ipa_profile_read_summary): New function.
13844         (ipa_profile): Decide on threshold.
13845         (pass_ipa_profile): Add ipa_profile_write_summary and
13846         ipa_profile_read_summary.
13847         * Makefile.in (ipa.o): Update dependencies.
13848         * lto-streamer.h (LTO_section_ipa_profile): New section.
13850 2013-03-29  Gabriel Dos Reis  <gdr@integrable-solutions.net>
13852         * tree.h (VAR_P): New.
13854 2013-03-29  Paolo Carlini  <paolo.carlini@oracle.com>
13856         PR lto/56777
13857         * doc/invoke.texi ([-fwhole-program]): Fix typo.
13859 2013-03-29  Steven Bosscher  <steven@gcc.gnu.org>
13861         * cfgbuild.c (inside_basic_block_p): Use JUMP_TABLE_DATA_P in lieu
13862         of tests for JUMP_P and a ADDR_DIFF_VEC or ADDR_VEC pattern.
13863         (control_flow_insn_p): Likewise.
13864         * cfgrtl.c (duplicate_insn_chain): Likewise.
13865         * final.c (get_attr_length_1): Likewise.
13866         (shorten_branches): Likewise.
13867         (final_scan_insn): Likewise.
13868         * function.c (instantiate_virtual_regs): Likewise.
13869         * gcse.c (insert_insn_end_basic_block): Likewise.
13870         * ira-costs.c (scan_one_insn): Likewise.
13871         * lra-eliminations.c (eliminate_regs_in_insn): Likewise.
13872         * lra.c (check_rtl): Likewise.
13873         * reload1.c (elimination_costs_in_insn): Likewise.
13874         * reorg.c (follow_jumps): Likewise.
13876         * config/arm/arm.c (is_jump_table): Use JUMP_TABLE_DATA_P in lieu
13877         of tests for JUMP_P and a ADDR_DIFF_VEC or ADDR_VEC pattern.
13878         (thumb_far_jump_used_p): Likewise.
13879         * config/bfin/bfin.c (workaround_rts_anomaly): Likewise.
13880         (workaround_speculation): Likewise.
13881         (add_sched_insns_for_speculation): Likewise.
13882         * config/c6x/c6x.c (reorg_emit_nops): Likewise.
13883         * config/frv/frv.c (frv_function_contains_far_jump): Likewise.
13884         (frv_for_each_packet): Likewise.
13885         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Likewise.
13886         * config/ia64/ia64.c (emit_all_insn_group_barriers): Likewise.
13887         (final_emit_insn_group_barriers): Likewise.
13888         * config/m32r/m32r.c (m32r_is_insn): Likewise.
13889         * config/mips/mips.c (USEFUL_INSN_P): Likewise.
13890         (mips16_insn_length): Likewise.
13891         * config/pa/pa.c (pa_reorg): Likewise.
13892         (pa_combine_instructions): Likewise.
13893         * config/rs6000/rs6000.c (rs6000_invalid_within_doloop): Likewise.
13894         * config/sh/sh.c (fixup_addr_diff_vecs): Likewise.
13895         (sh_reorg): Likewise.
13896         (split_branches): Likewise.
13897         * config/spu/spu.c (get_branch_target): Likewise.
13899         * config/s390/s390.c (s390_chunkify_start): Simplify logic using
13900         JUMP_TABLE_DATA_P.
13902 2013-03-29  Kirill Yukhin  <kirill.yukhin@intel.com>
13904         * config/i386/avx2intrin.h (_mm256_broadcastsi128_si256):
13905         Fix declaration name.
13907 2013-03-28  Lawrence Crowl  <crowl@google.com>
13909         * graphds.h (struct graph.indicies): Remove unused.
13910         * graphite-poly.h (struct graph.original_pddrs): Remove unused.
13911         (SCOP_ORIGINAL_PDDRS): Remove unused.
13912         * sese.h (extern insert_loop_close_phis): Removed unused.
13913         (extern insert_guard_phis): Removed unused.
13914         (extern ivtype_map_elt_info): Removed unused.
13915         (new_ivtype_map_elt): Removed unused.
13916         * sese.c (ivtype_map_elt_info): Removed unused.
13918 2013-03-28  Lawrence Crowl  <crowl@google.com>
13920         * Makefile.in: Add several missing include dependences.
13921         (DUMPFILE_H): New.
13922         (test-dump.o): New.  This object is not added to any executable,
13923         but is present for ad-hoc testing.
13924         * bitmap.c
13925         (debug (const bitmap_head_def &)): New.
13926         (debug (const bitmap_head_def *)): New.
13927         * bitmap.h
13928         (extern debug (const bitmap_head_def &)): New.
13929         (extern debug (const bitmap_head_def *)): New.
13930         * cfg.c
13931         (debug (edge_def &)): New.
13932         (debug (edge_def *)): New.
13933         * cfghooks.c
13934         (debug (basic_block_def &)): New.
13935         (debug (basic_block_def *)): New.
13936         * dumpfile.h
13937         (dump_node (const_tree, int, FILE *)): Correct source file.
13938         * dwarf2out.c
13939         (debug (die_struct &)): New.
13940         (debug (die_struct *)): New.
13941         * dwarf2out.h
13942         (extern debug (die_struct &)): New.
13943         (extern debug (die_struct *)): New.
13944         * gimple-pretty-print.c
13945         (debug (gimple_statement_d &)): New.
13946         (debug (gimple_statement_d *)): New.
13947         * gimple-pretty-print.h
13948         (extern debug (gimple_statement_d &)): New.
13949         (extern debug (gimple_statement_d *)): New.
13950         * ira-build.c
13951         (debug (ira_allocno_copy &)): New.
13952         (debug (ira_allocno_copy *)): New.
13953         (debug (ira_allocno &)): New.
13954         (debug (ira_allocno *)): New.
13955         * ira-int.h
13956         (extern debug (ira_allocno_copy &)): New.
13957         (extern debug (ira_allocno_copy *)): New.
13958         (extern debug (ira_allocno &)): New.
13959         (extern debug (ira_allocno *)): New.
13960         * ira-lives.c
13961         (debug (live_range &)): New.
13962         (debug (live_range *)): New.
13963         * lra-int.h
13964         (debug (lra_live_range &)): New.
13965         (debug (lra_live_range *)): New.
13966         * lra-lives.c
13967         (debug (lra_live_range &)): New.
13968         (debug (lra_live_range *)): New.
13969         * omega.c
13970         (debug (omega_pb_d &)): New.
13971         (debug (omega_pb_d *)): New.
13972         * omega.h
13973         (extern debug (omega_pb_d &)): New.
13974         (extern debug (omega_pb_d *)): New.
13975         * print-rtl.c
13976         (debug (const rtx_def &)): New.
13977         (debug (const rtx_def *)): New.
13978         * print-tree.c
13979         (debug_tree (tree): Move within file.
13980         (debug_raw (const tree_node &)): New.
13981         (debug_raw (const tree_node *)): New.
13982         (dump_tree_via_hooks (const tree_node *, int)): New.
13983         (debug (const tree_node &)): New.
13984         (debug (const tree_node *)): New.
13985         (debug_verbose (const tree_node &)): New.
13986         (debug_verbose (const tree_node *)): New.
13987         (debug_head (const tree_node &)): New.
13988         (debug_head (const tree_node *)): New.
13989         (debug_body (const tree_node &)): New.
13990         (debug_body (const tree_node *)): New.
13991         (debug_vec_tree (tree): Move and reimplement in terms of dump.
13992         (debug (vec<tree, va_gc> &)): New.
13993         (debug (vec<tree, va_gc> *)): New.
13994         * rtl.h
13995         (extern debug (const rtx_def &)): New.
13996         (extern debug (const rtx_def *)): New.
13997         * sbitmap.c
13998         (debug_raw (simple_bitmap_def &)): New.
13999         (debug_raw (simple_bitmap_def *)): New.
14000         (debug (simple_bitmap_def &)): New.
14001         (debug (simple_bitmap_def *)): New.
14002         * sbitmap.h
14003         (extern debug (simple_bitmap_def &)): New.
14004         (extern debug (simple_bitmap_def *)): New.
14005         (extern debug_raw (simple_bitmap_def &)): New.
14006         (extern debug_raw (simple_bitmap_def *)): New.
14007         * sel-sched-dump.c
14008         (debug (vinsn_def &)): New.
14009         (debug (vinsn_def *)): New.
14010         (debug_verbose (vinsn_def &)): New.
14011         (debug_verbose (vinsn_def *)): New.
14012         (debug (expr_def &)): New.
14013         (debug (expr_def *)): New.
14014         (debug_verbose (expr_def &)): New.
14015         (debug_verbose (expr_def *)): New.
14016         (debug (vec<rtx> &)): New.
14017         (debug (vec<rtx> *)): New.
14018         * sel-sched-dump.h
14019         (extern debug (vinsn_def &)): New.
14020         (extern debug (vinsn_def *)): New.
14021         (extern debug_verbose (vinsn_def &)): New.
14022         (extern debug_verbose (vinsn_def *)): New.
14023         (extern debug (expr_def &)): New.
14024         (extern debug (expr_def *)): New.
14025         (extern debug_verbose (expr_def &)): New.
14026         (extern debug_verbose (expr_def *)): New.
14027         (extern debug (vec<rtx> &)): New.
14028         (extern debug (vec<rtx> *)): New.
14029         * sel-sched-ir.h
14030         (_list_iter_cond_expr): Make inline instead of static.
14031         * sreal.c
14032         (debug (sreal &)): New.
14033         (debug (sreal *)): New.
14034         * sreal.h
14035         (extern debug (sreal &)): New.
14036         (extern debug (sreal *)): New.
14037         * tree.h
14038         (extern debug_raw (const tree_node &)): New.
14039         (extern debug_raw (const tree_node *)): New.
14040         (extern debug (const tree_node &)): New.
14041         (extern debug (const tree_node *)): New.
14042         (extern debug_verbose (const tree_node &)): New.
14043         (extern debug_verbose (const tree_node *)): New.
14044         (extern debug_head (const tree_node &)): New.
14045         (extern debug_head (const tree_node *)): New.
14046         (extern debug_body (const tree_node &)): New.
14047         (extern debug_body (const tree_node *)): New.
14048         (extern debug (vec<tree, va_gc> &)): New.
14049         (extern debug (vec<tree, va_gc> *)): New.
14050         * tree-cfg.c
14051         (debug (struct loop &)): New.
14052         (debug (struct loop *)): New.
14053         (debug_verbose (struct loop &)): New.
14054         (debug_verbose (struct loop *)): New.
14055         * tree-dump.c: Add header dependence.
14056         * tree-flow.h
14057         (extern debug (struct loop &)): New.
14058         (extern debug (struct loop *)): New.
14059         (extern debug_verbose (struct loop &)): New.
14060         (extern debug_verbose (struct loop *)): New.
14061         * tree-data-ref.c
14062         (debug (data_reference &)): New.
14063         (debug (data_reference *)): New.
14064         (debug (vec<data_reference_p> &)): New.
14065         (debug (vec<data_reference_p> *)): New.
14066         (debug (vec<ddr_p> &)): New.
14067         (debug (vec<ddr_p> *)): New.
14068         * tree-data-ref.h
14069         (extern debug (data_reference &)): New.
14070         (extern debug (data_reference *)): New.
14071         (extern debug (vec<data_reference_p> &)): New.
14072         (extern debug (vec<data_reference_p> *)): New.
14073         (extern debug (vec<ddr_p> &)): New.
14074         (extern debug (vec<ddr_p> *)): New.
14075         * tree-ssa-alias.c
14076         (debug (pt_solution &)): New.
14077         (debug (pt_solution *)): New.
14078         * tree-ssa-alias.h
14079         (extern debug (pt_solution &)): New.
14080         (extern debug (pt_solution *)): New.
14081         * tree-ssa-alias.c
14082         (debug (_var_map &)): New.
14083         (debug (_var_map *)): New.
14084         (debug (tree_live_info_d &)): New.
14085         (debug (tree_live_info_d *)): New.
14086         * tree-ssa-alias.h
14087         (extern debug (_var_map &)): New.
14088         (extern debug (_var_map *)): New.
14089         (extern debug (tree_live_info_d &)): New.
14090         (extern debug (tree_live_info_d *)): New.
14092 2013-03-28  Jan Hubicka  <jh@suse.cz>
14094         * lto-cgraph.c (merge_profile_summaries): Fix overflows.
14096 2013-03-28  Ian Bolton  <ian.bolton@arm.com>
14098         * config/aarch64/aarch64.md (aarch64_can_eliminate): Keep frame
14099         record only when desired or required.
14101 2013-03-28  Uros Bizjak  <ubizjak@gmail.com>
14103         * config/i386/i386.md (*vec_extract2vdi_1): Merge with
14104         *vec_extractv2di_1_rex64.  Use x64 isa attribute.
14106 2013-03-28  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
14108         * config/aarch64/aarch64.md (*and<mode>3_compare0): New pattern.
14109         (*andsi3_compare0_uxtw): New pattern.
14110         (*and_<SHIFT:optab><mode>3_compare0): New pattern.
14111         (*and_<SHIFT:optab>si3_compare0_uxtw): New pattern.
14113 2013-03-28  Jan Hubicka  <jh@suse.cz>
14115         * data-streamer-in.c (streamer_read_gcov_count): New function.
14116         * gimple-streamer-out.c: Include value-prof.h.
14117         (output_gimple_stmt): Output histogram.
14118         (output_bb): Use streamer_write_gcov_count.
14119         * value-prof.c: Include data-streamer.h
14120         (dump_histogram_value): Add HIST_TYPE_MAX.
14121         (stream_out_histogram_value): New function.
14122         (stream_in_histogram_value): New function.
14123         * value-prof.h (enum hist_type): Add HIST_TYPE_MAX.
14124         (stream_out_histogram_value, stream_in_histogram_value): Declare.
14125         * data-streamer-out.c (streamer_write_gcov_count): New function.
14126         (streamer_write_gcov_count_stream): New function.
14127         * lto-cgraph.c (lto_output_edge): Update counter streaming.
14128         (lto_output_node): Likewise.
14129         (input_node, input_edge): Likewise.
14130         * lto-streamer-out.c (output_cfg): Update streaming.
14131         * lto-streamer-in.c (input_cfg): Likewise.
14132         * data-streamer.h (streamer_write_gcov_count,
14133         streamer_write_gcov_count_stream, streamer_read_gcov_count): Declare.
14134         * gimple-streamer-in.c: Include value-prof.h
14135         (input_gimple_stmt): Input histograms.
14136         (input_bb): Update profile streaming.
14138 2013-03-28  Kenneth Zadeck  <zadeck@naturalbridge.com>
14140         * genmodes.c (emit_max_int): New function.
14141         (emit_insn_modes_h): Added call to emit_max_function.
14142         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_INT, MAX_BITSIZE_MODE_ANY_MODE):
14143         Added doc.
14144         * machmode.def: Fixed comment.
14146 2013-03-28  Kenneth Zadeck  <zadeck@naturalbridge.com>
14148         * combine.c (try_combine): Removed useless assert.
14149         * cselib.c (rtx_equal_for_cselib_1): Removed unnecessary parens.
14151 2013-03-28  Marek Polacek  <polacek@redhat.com>
14152             Richard Biener  <rguenther@suse.de>
14154         PR tree-optimization/56695
14155         * tree-vect-stmts.c (vectorizable_condition): Unconditionally
14156         build signed result of a vector comparison.
14157         * tree-cfg.c (verify_gimple_comparison): Check that a result
14158         of a vector comparison has signed type.
14160 2013-03-28  Richard Biener  <rguenther@suse.de>
14162         PR tree-optimization/37021
14163         * tree-vect-slp.c (vect_build_slp_tree): When not unrolling
14164         do not restrict gaps between groups.
14165         * tree-vect-stmts.c (vectorizable_load): Properly account for
14166         a gap between groups.
14168 2013-03-28  Eric Botcazou  <ebotcazou@adacore.com>
14170         * toplev.c (process_options): Do not disable -fomit-frame-pointer on a
14171         general basis if unwind info is requested and ACCUMULATE_OUTGOING_ARGS
14172         is not enabled.
14174 2013-03-27  Gerald Pfeifer  <gerald@pfeifer.com>
14176         * doc/invoke.texi (AVR Options): Tweak link for AVR-LibC user manual.
14177         * doc/extend.texi (Named Address Spaces): Ditto.
14178         (Variable Attributes): Ditto.
14180 2013-03-27  Kai Tietz  <ktietz@redhat.com>
14182         * config.build: Add support for cygwin x64 target.
14183         * config.gcc: Likewise.
14184         * config.host: Likewise.
14185         * configure.ac: Likewise
14186         * configure: Regenerated.
14188 2013-03-27  Kai Tietz  <ktietz@redhat.com>
14190         * config/i386/cygwin-stdint.h: Add support for cygwin x64 target.
14191         * config/i386/t-cygwin-w64: New file.
14192         * config/i386/cygwin-w64.h: New file.
14193         * config/i386/cygwin.h (EXTRA_OS_CPP_BUILTINS): Extend
14194         and add support for x64-cygwin target.
14195         (CPP_SPEC): Likewise.
14196         (CXX_WRAP_SPEC_LIST): Undefine before define.
14197         (LIBGCJ_SONAME): Use 15 as version.
14199 2013-03-27  Richard Biener  <rguenther@suse.de>
14201         PR tree-optimization/56716
14202         * tree-ssa-structalias.c (perform_var_substitution): Adjust
14203         dumping for ref nodes.
14205 2013-03-27  Martin Jambor  <mjambor@suse.cz>
14207         PR tree-optimization/55334
14208         * ipa-cp.c (initialize_node_lattices): Allow IPA-CP through and to
14209         restricted pointers to arrays.
14211 2013-03-27  Gabriel Dos Reis  <gdr@integrable-solutions.net>
14213         * Makefile.in (.SUFFIXES): Add .cc.
14214         (.c.o): Apply same recipe for implicit rule .cc.o.
14216 2013-03-27  Richard Biener  <rguenther@suse.de>
14218         PR tree-optimization/37021
14219         * tree-vect-data-refs.c (vect_check_strided_load): Allow
14220         REALPART/IMAGPART_EXPRs around the supported refs.
14221         * tree-ssa-structalias.c (find_func_aliases): Assume that
14222         floating-point values are not used to transfer pointers.
14224 2013-03-27  Alexander Ivchenko  <alexander.ivchenko@intel.com>
14226         * gcc/target.def (TARGET_HAS_IFUNC_P): New target hook.
14227         * gcc/doc/tm.texi.in (TARGET_HAS_IFUNC_P): New.
14228         * gcc/doc/tm.texi: Regenerate.
14229         * gcc/targhooks.h (default_has_ifunc_p): New.
14230         * gcc/targhooks.c (default_has_ifunc_p): Ditto.
14231         * gcc/config/linux-protos.h: New file.
14232         * gcc/config/linux-android.h (TARGET_HAS_IFUNC_P): Using version of
14233         this hook for linux which disables support of indirect functions in
14234         android.
14235         * gcc/config/linux-android.c: New file.
14236         * gcc/config/t-linux-android.c: Ditto.
14237         * gcc/config.gcc: Added new object file linux-android.o.
14238         * gcc/config/i386/i386.c (ix86_get_function_versions_dispatcher):
14239         Using TARGET_HAS_IFUNC hook instead of HAVE_GNU_INDIRECT_FUNCTION.
14240         * gcc/varasm.c (do_assemble_alias): Likewise.
14241         * configure.ac: Define HAVE_GNU_INDIRECT_FUNCTION as zero if the target
14242         doesn't support indirect functions.
14243         * configure: Regenerate.
14245 2013-03-27  Bin Cheng  <bin.cheng@arm.com>
14247         PR target/56102
14248         * config/arm/arm.c (thumb1_rtx_costs, thumb1_size_rtx_costs): Fix
14249         rtx costs for SET/ASHIFT/ASHIFTRT/LSHIFTRT/ROTATERT patterns with
14250         mult-word mode.
14252 2013-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
14254         * config/s390/s390.h (TARGET_FLT_EVAL_METHOD): Define.
14256 2013-03-27  Terry Guo  <terry.guo@arm.com>
14258         * config/arm/arm-cores.def: Added core cortex-r7.
14259         * config/arm/arm-tune.md: Regenerated.
14260         * config/arm/arm-tables.opt: Regenerated.
14261         * doc/invoke.texi: Added entry for core cortex-r7.
14263 2013-03-27  Walter Lee  <walt@tilera.com>
14265         * config/tilegx/tilegx.c (tilegx_expand_prologue): Avoid
14266         double-decrement of next_scratch_regno.
14268 2013-03-27  Walter Lee  <walt@tilera.com>
14270         * config/tilegx/tilegx.md (insn_v1mulu): Fix predicates on
14271         input operands.
14272         (insn_v1mulus): Ditto.
14273         (insn_v2muls): Ditto.
14275 2013-03-27  Walter Lee  <walt@tilera.com>
14277         * config/tilegx/tilegx.h (ASM_OUTPUT_ADDR_VEC_ELT): Delete extra tab.
14278         (ASM_OUTPUT_ADDR_DIFF_ELT): Ditto.
14280 2013-03-27  Walter Lee  <walt@tilera.com>
14282         * config/tilegx/tilegx.md (*sibcall_insn): Fix type atribute for jr.
14283         (*sibcall_value): Ditto.
14285 2013-03-27  Walter Lee  <walt@tilera.com>
14287         * config/tilegx/tilegx.md (insn_mnz_<mode>): Replaced by ...
14288         (insn_mnz_v8qi): ... this ...
14289         (insn_mnz_v4hi): ... and this.  Replace (const_int 0) with the
14290         vector equivalent.
14291         (insn_v<n>mnz): Replaced by ...
14292         (insn_v1mnz): ... this ...
14293         (insn_v2mnz): ... and this.  Replace (const_int 0) with the vector
14294         equivalent.
14295         (insn_mz_<mode>): Replaced by ...
14296         (insn_mz_v8qi): ... this ...
14297         (insn_mz_v4hi): ... and this.  Replace (const_int 0) with the
14298         vector equivalent.
14299         (insn_v<n>mz): Replaced by ...
14300         (insn_v1mz): ... this ...
14301         (insn_v2mz): ... and this.  Replace (const_int 0) with the vector
14302         equivalent.
14304 2013-03-26  Eric Botcazou  <ebotcazou@adacore.com>
14306         * doc/invoke.texi (SPARC options): Remove -mlittle-endian.
14308 2013-03-26  Roland McGrath  <mcgrathr@google.com>
14310         * config/arm/arm.c (arm_print_operand: case 'w'): Use fputs rather
14311         than fprintf with a non-constant, non-format string.
14313 2013-03-26  Uros Bizjak  <ubizjak@gmail.com>
14315         * config/i386/i386.md (*cmpqi_ext_1): Merge with *cmpqi_ext_1_rex64
14316         using nox64 isa attribute.  Use nonimmediate_x86nomem_operand as
14317         operand 0 predicate.
14318         (*cmpqi_ext_3): Merge with *cmpqi_ext_3_rex64 using nox64 isa
14319         attribute.  Use general_x64nomem_operand as operand 1 predicate.
14320         (*movqi_extv_1): Merge with *movqi_extv_1_rex64 using nox64 isa
14321         attribute.  Use nonimmediate_x64nomem_operand as operand 0 predicate.
14322         (*movqi_extzv_2): Merge with *movqi_extzv_2_rex64 using nox64 isa
14323         attribute.  Use nonimmediate_x64nomem_operand as operand 0 predicate.
14324         (mov<mode>_insv_1): Remove expander.  Merge insn with
14325         movsi_insv_1 using SWI48 mode iterator and nox64 isa attribute.
14326         Use general_x64nomem_operand as operand 1 predicate.
14327         (addqi_ext_1): Merge with *addqi_ext_1_rex64 using nox64 isa attribute.
14328         (*testqi_ext_1): Merge with *testqi_ext_1_rex64 using nox64 isa
14329         attribute.  Use nonimmediate_x64nomem_operand as operand 1 predicate.
14330         (*andqi_ext_1): Merge with *andqi_ext_1_rex64 using nox64 isa
14331         attribute.  Use nonimmediate_x64nomem_operand as operand 2 predicate.
14332         (*<code>qi_ext_1): Merge with *<code>qi_ext_1_rex64 using nox64 isa
14333         attribute.  Use nonimmediate_x64nomem_operand as operand 1 predicate.
14334         (*xorqi_cc_ext_1): Merge with *xorqi_cc_ext_1_rex64 using nox64
14335         isa attribute.  Use general_x64nomem_operand as operand 2 predicate.
14336         * config/i386/predicates.md (nonimmediate_x64nomem_operand): New.
14337         (general_x64nomem_operand): Ditto.
14339 2013-03-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
14341         * config/rtems.opt: Add -pthread option.
14343 2013-03-26  Richard Biener  <rguenther@suse.de>
14345         * alias.c (find_base_term): Avoid redundant and not used recursion.
14346         (base_alias_check): Get the initial base term from the caller.
14347         (true_dependence_1): Compute and pass base terms to base_alias_check.
14348         (write_dependence_p): Likewise.
14349         (may_alias_p): Likewise.
14351 2013-03-26  Sofiane Naci  <sofiane.naci@arm.com>
14353         * config/aarch64/aarch64.c (aarch64_classify_address): Support
14354         PC-relative load in SI modes and above only.
14356 2013-03-26  Xinyu Qi  <xyqi@marvell.com>
14358         * config/arm/arm.h (FIRST_IWMMXT_GR_REGNUM): Add comment.
14359         * config/arm/iwmmxt.md (WCGR0): Update.
14360         (WCGR1, WCGR2, WCGR3): Likewise.
14362 2013-03-26  Uros Bizjak  <ubizjak@gmail.com>
14364         * config/i386/i386.md (*movdfcc_1): Merge with *movdfcc_1_rex64.
14365         Use x64 and nox64 isa attributes.
14367 2013-03-26  Richard Biener  <rguenther@suse.de>
14369         * emit-rtl.c (set_mem_attributes_minus_bitpos): Remove
14370         alignment computations and rely on get_object_alignment_1
14371         for the !TYPE_P case.
14372         Commonize DECL/COMPONENT_REF handling in the ARRAY_REF path.
14374 2013-03-26  Walter Lee  <walt@tilera.com>
14376         * config/tilegx/tilegx.h (PROFILE_BEFORE_PROLOGUE): Define.
14377         * config/tilegx/tilepro.h (PROFILE_BEFORE_PROLOGUE): Define.
14379 2013-03-25  Jeff Law  <law@redhat.com>
14381         * tree-ssa-dom.c (record_equivalences_from_incoming_edge): Add missing
14382         check for INTEGRAL_TYPE_P that was missing due to checking in
14383         wrong version of prior patch.
14385 2013-03-25  Walter Lee  <walt@tilera.com>
14387         * config/tilegx/tilegx-builtins.h (enum tilegx_builtin): Add
14388         TILEGX_INSN_SHUFFLEBYTES1.
14389         * config/tilegx/tilegx.c (tilegx_builtin_info): Add entry for
14390         shufflebytes1.
14391         (tilegx_builtins): Ditto.
14392         * config/tilegx/tilegx.md (insn_shufflebytes1): New pattern.
14394 2013-03-25  Walter Lee  <walt@tilera.com>
14396         * config/tilegx/tilegx.md (floatsisf2): New pattern.
14397         (floatunssisf2): New pattern.
14398         (floatsidf2): New pattern.
14399         (floatunssidf2): New pattern.
14401 2013-03-25  Walter Lee  <walt@tilera.com>
14403         * config/tilegx/tilegx.c (expand_set_cint64_one_inst): Inline
14404         tests for constraint J, K, N, P.
14406 2013-03-25  Walter Lee  <walt@tilera.com>
14408         * config/tilegx/tilegx.c (tilegx_asm_preferred_eh_data_format):
14409         Use indirect/pcrel encoding.
14410         * config/tilepro/tilepro.c (tilepro_asm_preferred_eh_data_format):
14411         Ditto.
14413 2013-03-25  Steve Ellcey  <sellcey@mips.com>
14415         * config/mips/mmips-cpus.def (74kc, 74kf2_1, 74kf, 74kf, 74kf1_1,
14416         74kfx, 74kx, 74kf3_2): Add PTF_AVOID_IMADD.
14417         * config/mips/mips.c (mips_option_override): Set IMADD default.
14418         * config/mips/mips.h (PTF_AVOID_IMADD): New.
14419         (ISA_HAS_MADD_MSUB): Remove MIPS16 check.
14420         (GENERATE_MADD_MSUB): Remove TUNE_74K check, add MIPS16 check.
14421         * config/mips/mips.md (mimadd): New flag for integer madd/msub.
14422         * doc/invoke.texi (-mimadd/-mno-imadd): New.
14424 2013-03-25  Jeff Law  <law@redhat.com>
14426         * tree-ssa-dom.c (record_equivalences_from_incoming_edge): Rework
14427         slightly to avoid creating and folding useless trees.  Simplify
14428         slightly by restricting to INTEGER_CSTs and using int_fits_type_p.
14430 2013-03-25  Uros Bizjak  <ubizjak@gmail.com>
14432         * config/i386/i386.md (*zero_extendsidi2): Merge with
14433         *zero_extendsidi2_rex64.  Use x64 and nox64 isa attributes.
14434         * config/i386/predicates.md (x86_64_zext_operand): Rename from
14435         x86_64_zext_general_operand.  Use nonimmediate_operand on 32bit
14436         targets.  Clarify comment.
14438 2013-03-25  Martin Jambor  <mjambor@suse.cz>
14440         * ipa-prop.c (ipa_write_jump_function): Stream simple and aritmetic
14441         pass-through jump functions differently.
14442         (ipa_read_jump_function): Likewise.  Also use setter functions to set
14443         up jump functions.
14445 2013-03-25  Martin Jambor  <mjambor@suse.cz>
14447         * ipa-cp.c (ipa_get_indirect_edge_target): Renamed to
14448         ipa_get_indirect_edge_target_1, added parameter agg_reps and ability to
14449         process it.
14450         (ipa_get_indirect_edge_target): New function.
14451         (devirtualization_time_bonus): New parameter known_aggs, pass it to
14452         ipa_get_indirect_edge_target.  Update all callers.
14453         (ipcp_discover_new_direct_edges): New parameter aggvals.  Pass it to
14454         ipa_get_indirect_edge_target_1 instead of calling
14455         ipa_get_indirect_edge_target.
14456         (create_specialized_node): Pass aggvlas to
14457         ipcp_discover_new_direct_edges.
14459 2013-03-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14461         * config/arm/arm.md (f_sels, f_seld): New types.
14462         (*cmov<mode>): New pattern.
14463         * config/arm/predicates.md (arm_vsel_comparison_operator): New
14464         predicate.
14466 2013-03-25  Kai Tietz  <ktietz@redhat.com>
14468         * config/i386/xm-mingw32.h (__USE_MINGW_ANSI_STDIO): Enable
14469         POSIX-printf for mingw-hosted  builds.
14471 2013-03-25  Richard Biener  <rguenther@suse.de>
14473         PR middle-end/56694
14474         * tree-eh.c (lower_eh_must_not_throw): Strip BLOCKs from the
14475         must-not-throw stmt location.
14477 2013-03-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14479         * config/arm/arm.c (arm_emit_load_exclusive): Add acq parameter.
14480         Emit load-acquire versions when acq is true.
14481         (arm_emit_store_exclusive): Add rel parameter.
14482         Emit store-release versions when rel is true.
14483         (arm_split_compare_and_swap): Use acquire-release instructions
14484         instead.
14485         of barriers when appropriate.
14486         (arm_split_atomic_op): Likewise.
14487         * config/arm/arm.h (TARGET_HAVE_LDACQ): New macro.
14488         * config/arm/unspecs.md (VUNSPEC_LAX): New unspec.
14489         (VUNSPEC_SLX): Likewise.
14490         (VUNSPEC_LDA): Likewise.
14491         (VUNSPEC_STL): Likewise.
14492         * config/arm/sync.md (atomic_load<mode>): New pattern.
14493         (atomic_store<mode>): Likewise.
14494         (arm_load_acquire_exclusive<mode>): Likewise.
14495         (arm_load_acquire_exclusivesi): Likewise.
14496         (arm_load_acquire_exclusivedi): Likewise.
14497         (arm_store_release_exclusive<mode>): Likewise.
14499 2013-03-25  Catherine Moore  <clm@codesourcery.com>
14501         * config/mips/constraints.md (u, Udb7 Uead, Uean, Uesp, Uib3,
14502         Uuw6, Usb4, ZS, ZT, ZU, ZV, ZW): New constraints.
14503         * config/mip/predicates.md (lwsp_swsp_operand,
14504         lw16_sw16_operand, lhu16_sh16_operand, lbu16_operand,
14505         sb16_operand, db4_operand, db7_operand, ib3_operand,
14506         sb4_operand, ub4_operand, uh4_operand, uw4_operand,
14507         uw5_operand, uw6_operand, addiur2_operand, addiusp_operand,
14508         andi16_operand): New predicates.
14509         * config/mips/mips.md (compression): New attribute.
14510         (enabled): New attribute.
14511         (length): Consider compression in computing length.
14512         (shift_compression): New code attribute.
14513         (*add<mode>3): New operands. Record compression.
14514         (sub<mode>3): Likewise.
14515         (one_cmpl<mode>2): Likewise.
14516         (*and<mode>3): Likewise.
14517         (*ior<mode>3): Likewise.
14518         (unnamed pattern for xor): Likewise.
14519         (*zero_extend<SHORT:mode><GPR:mode>2): Likewise.
14520         (*<optab><mode>3): Likewise.
14521         (*mov<mode>_internal: Likewise.
14522         * config/mips/mips-protos.h (mips_signed_immediate_p): New.
14523         (mips_unsigned_immediate_p): New.
14524         (umips_lwsp_swsp_address_p): New.
14525         (m16_based_address_p): New.
14526         * config/mips/mips-protos.h (mips_signed_immediate_p): New prototype.
14527         (mips_unsigned_immediate_p): New prototype.
14528         (lwsp_swsp_address_p): New prototype.
14529         (m16_based_address_p): New prototype.
14530         * config/mips/mips.c (mips_unsigned_immediate_p): New function.
14531         (mips_signed_immediate_p): New function.
14532         (m16_based_address_p): New function.
14533         (lwsp_swsp_address_p): New function.
14534         (mips_print_operand_punctuation): Recognize short delay slot insns
14535         for microMIPS.add<mode>3"
14537 2013-03-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14539         PR target/56720
14540         * config/arm/iterators.md (v_cmp_result): New mode attribute.
14541         * config/arm/neon.md (vcond<mode><mode>): Handle unordered cases.
14543 2013-03-25  Richard Biener  <rguenther@suse.de>
14545         PR tree-optimization/56689
14546         * tree-vrp.c (execute_vrp): Mark loops for fixup if we removed
14547         any edge.
14549 2013-03-25  Richard Biener  <rguenther@suse.de>
14551         * tree-ssa-loop-im.c (struct mem_ref): Use bitmap_head instead
14552         of bitmap.
14553         (memory_references): Likewise.
14554         (outermost_indep_loop, mem_ref_alloc, mark_ref_stored,
14555         gather_mem_refs_stmt, record_dep_loop, ref_indep_loop_p_1,
14556         ref_indep_loop_p_2, find_refs_for_sm): Adjust.
14557         (gather_mem_refs_in_loops): Fold into ...
14558         (analyze_memory_references): ... this.  Move initialization
14559         to tree_ssa_lim_initialize.
14560         (fill_always_executed_in): Rename to ...
14561         (fill_always_executed_in_1): ... this.
14562         (fill_always_executed_in): Move contains_call computation to
14563         this new function from ...
14564         (tree_ssa_lim_initialize): ... here.
14565         (tree_ssa_lim): Call fill_always_executed_in.
14567 2013-03-25  Eric Botcazou  <ebotcazou@adacore.com>
14569         * postreload.c (reload_combine): Fix code detecting returns.
14571 2013-03-25  Eric Botcazou  <ebotcazou@adacore.com>
14573         * function.c (emit_use_return_register_into_block): On cc0 targets,
14574         do not emit the sequence between cc0 setter and user.
14576 2013-03-25  Kai Tietz  <ktietz@redhat.com>
14578         * config/i386/predicates.md (local_symbolic_operand): Interpret
14579         dll-imported symbols as none-local.
14581 2013-03-25  Richard Biener  <rguenther@suse.de>
14583         * tree-ssa-loop-im.c (struct depend): Remove.
14584         (struct lim_aux_data): Make depends a vec of gimples.
14585         (free_lim_aux_data): Adjust.
14586         (add_dependency): Likewise.
14587         (set_level): Likewise.
14589 2013-03-25  Richard Biener  <rguenther@suse.de>
14591         PR middle-end/56434
14592         * calls.c (expand_call): Use MALLOC_ABI_ALIGNMENT to annotate
14593         the pointer returned by calls with ECF_MALLOC set.
14595 2013-03-24  Uros Bizjak  <ubizjak@gmail.com>
14597         * config/i386/mmx.md (mov<mode>): Add ?!Ym,r and r,?!Ym alternatives.
14599 2013-03-24  Uros Bizjak  <ubizjak@gmail.com>
14601         * config/i386/mmx.md (mov<mode>): Merge with movv2sf expander
14602         using MMXMODE mode iterator.
14603         (*move<mode>_internal): Merge with *movv2sf_internal and
14604         *movv2sf_internal_rex64 using MMXMODE mode iterator.
14606 2013-03-23  Steven Bosscher  <steven@gcc.gnu.org>
14608         * gcse.c (oprs_unchanged_p): Respect flag_gcse_lm.
14609         (record_last_mem_set_info): Likewise.
14611         * df-core.c (rest_of_handle_df_initialize): Use XCNEWVEC instead
14612         of XNEWVEC followed by memset.
14613         (df_worklist_dataflow): Use XNEWVEC instead of xmalloc with a cast.
14615 2013-03-23  Steven Bosscher  <steven@gcc.gnu.org>
14617         * config/avr/avr.c, config/bfin/bfin.c, config/c6x/c6x.c,
14618         config/epiphany/epiphany.c, config/frv/frv.c, config/ia64/ia64.c,
14619         config/iq2000/iq2000.c, config/mcore/mcore.c, config/mep/mep.c,
14620         config/mmix/mmix.c, config/pa/pa.c, config/rs6000/rs6000.c,
14621         config/s390/s390.c, config/sparc/sparc.c, config/spu/spu.c,
14622         config/stormy16/stormy16.c, config/v850/v850.c, config/xtensa/xtensa.c,
14623         dwarf2out.c, hw-doloop.c, resource.c, rtl.h : Where applicable, use
14624         the predicates NOTE_P, NONJUMP_INSN_P, JUMP_P, CALL_P, LABEL_P, and
14625         BARRIER_P instead of GET_CODE.
14627 2013-03-23  Eric Botcazou  <ebotcazou@adacore.com>
14629         * config/sparc/sparc.c (sparc_emit_probe_stack_range): Fix small
14630         inaccuracy in the probing code.
14632         * config/sparc/sparc.md (ctrapsi4): Add predicate for operand #3.
14633         (ctrapdi4): Likewise.
14635 2013-03-23  Eric Botcazou  <ebotcazou@adacore.com>
14637         * calls.c (expand_call): Add missing guard to code handling return
14638         of non-BLKmode structures in MSB.
14639         * function.c (expand_function_end): Likewise.
14641 2013-03-23  Eric Botcazou  <ebotcazou@adacore.com>
14643         * combine.c (try_combine): Adjust comment.  Do not add the set of
14644         insn #0 if the destination indirectly is set or dies in insn #2.
14645         Tidy up code to distribute a new note.
14647 2013-03-22  Uros Bizjak  <ubizjak@gmail.com>
14649         * config/i386/i386.md (*movdi_internal): Set prefix_rex attribute
14650         also for alternatives 16 and 17.
14652 2013-03-22  Uros Bizjak  <ubizjak@gmail.com>
14654         * config/i386/sse.md (*mov<mode>_internal): Merge with
14655         *mov<mode>_internal_rex64.  Use x64 and nox64 isa attributes.
14656         Emit insn template depending on type attribute.  Use
14657         HAVE_AS_IX86_INTERUNIT_MOVQ to handle broken assemblers that require
14658         movd instead of movq mnemonic for interunit moves.  Rewrite mode
14659         attribute calculation.  Remove unit attribute calculation.
14660         Set prefix attribute to maybe_vex for sselog1 and ssemov types.
14661         Set prefix_data16 attribute for DImode ssemov types.
14662         Use Ym instead of y for SSE-MMX conversion alternatives.
14663         Reorder operand constraints.
14665 2013-03-22  Steven Bosscher  <steven@gcc.gnu.org>
14667         * df.h (df_insn_delete): Adjust prototype.
14668         * emit-rtl.c (remove_insn): Pass a basic block to df_insn_delete
14669         and let it decide whether mark the basic block dirty.
14670         (set_insn_deleted): Only pass INSN_P insns to df_insn_delete.
14671         * df-scan.c (df_insn_info_delete): New helper function, split
14672         off from df_insn_delete.
14673         (df_scan_free_bb_info): Use it.
14674         (df_insn_rescan, df_insn_rescan_all, df_process_deferred_rescans):
14675         Likewise.
14676         (df_insn_delete): Likewise.  Take insn rtx as argument.  Verify
14677         that the insn is actually an insn and it has a non-NULL basic block.
14678         Do not mark basic block dirty if only deleting a DEBUG_INSN.
14680 2013-03-22  Richard Biener  <rguenther@suse.de>
14682         * tree-ssa-loop-im.c (struct mem_ref): Remove indep_ref and
14683         dep_ref members.
14684         (mem_ref_alloc): Do not allocate them.
14685         (refs_independent_p): Do not query or maintain a cache.
14687 2013-03-22  Richard Biener  <rguenther@suse.de>
14689         * tree-ssa-loop-im.c (memory_references): Drop all_refs_in_loop.
14690         (gather_mem_refs_in_loops): Do not compute it.
14691         (analyze_memory_references): Do not allocate it.
14692         (tree_ssa_lim_finalize): Do not free it.
14693         (for_all_locs_in_loop): Do not query all_refs_in_loop.
14695 2013-03-22  Richard Biener  <rguenther@suse.de>
14697         * is-a.h (as_a): Use gcc_checking_assert.
14699 2013-03-22  Ian Bolton  <ian.bolton@arm.com>
14701         * config/aarch64/aarch64.c (aarch64_print_operand): New
14702         format specifier for printing a constant in hex.
14703         * config/aarch64/aarch64.md (insv_imm<mode>): Use the X
14704         format specifier for printing second operand.
14706 2013-03-22  Richard Biener  <rguenther@suse.de>
14708         * tree-ssa-loop-im.c (memory_references): Add refs_stored_in_loop
14709         bitmaps.
14710         (gather_mem_refs_in_loops): Perform store accumulation here.
14711         (create_vop_ref_mapping_loop): Remove.
14712         (create_vop_ref_mapping): Likewise.
14713         (analyze_memory_references): Initialize refs_stored_in_loop.
14714         (LOOP_DEP_BIT): New define to map to bits in (in)dep_loop bitmaps.
14715         (record_indep_loop): Remove.
14716         (record_dep_loop): New function.
14717         (ref_indep_loop_p_1): Adjust to only walk over references
14718         in the loop, not its subloops.
14719         (ref_indep_loop_p): Rename to ...
14720         (ref_indep_loop_p_2): ... this and recurse over the loop tree,
14721         maintaining a more fine-grained cache.
14722         (ref_indep_loop_p): Wrap ref_indep_loop_p_2.
14723         (tree_ssa_lim_finalize): Free refs_stored_in_loop.
14725 2013-03-22  Richard Biener  <rguenther@suse.de>
14727         * tree-ssa-loop-im.c (struct mem_ref_locs): Remove.
14728         (struct mem_ref): Make accesses_in_loop a vec of a vec of
14729         aggregate mem_ref_loc.
14730         (free_mem_ref_locs): Inline into ...
14731         (memref_free): ... this and adjust.
14732         (mem_ref_alloc): Adjust.
14733         (mem_ref_locs_alloc): Remove.
14734         (record_mem_ref_loc): Adjust.
14735         (get_all_locs_in_loop): Rewrite into ...
14736         (for_all_locs_in_loop): ... this iterator.
14737         (rewrite_mem_ref_loc): New functor.
14738         (rewrite_mem_refs): Use for_all_locs_in_loop.
14739         (sm_set_flag_if_changed): New functor.
14740         (execute_sm_if_changed_flag_set): Use for_all_locs_in_loop.
14741         (ref_always_accessed): New functor.
14742         (ref_always_accessed_p): Use for_all_locs_in_loop.
14744 2013-03-21  Marc Glisse  <marc.glisse@inria.fr>
14746         * tree-pass.h (PROP_gimple_lvec): New.
14747         * passes.c (dump_properties): Handle PROP_gimple_lvec.
14748         (init_optimization_passes): Move pass_lower_vector.
14749         * tree-vect-generic.c (gate_expand_vector_operations_ssa): Test
14750         PROP_gimple_lvec.
14751         (pass_lower_vector): Provide PROP_gimple_lvec.
14752         (pass_lower_vector_ssa): Likewise.
14753         * cfgexpand.c (pass_expand): Require PROP_gimple_lvec.
14755 2013-03-21  Mark Wielaard  <mjw@redhat.com>
14757         * dwarf2out.c (size_of_aranges): Skip DECL_IGNORED_P functions.
14759 2013-03-21  Uros Bizjak  <ubizjak@gmail.com>
14761         * config/i386/i386.md (*movdi_internal): Disparage slightly
14762         all MMX moves to/from memory.  Use Yi instead of x for SSE-MMX
14763         conversion alternatives.
14765 2013-03-21  Jakub Jelinek  <jakub@redhat.com>
14767         PR middle-end/48087
14768         * diagnostic.def (DK_WERROR): New kind.
14769         * diagnostic.h (werrorcount): Define.
14770         * diagnostic.c (diagnostic_report_diagnostic): For DK_WARNING
14771         promoted to DK_ERROR, increment DK_WERROR counter instead of
14772         DK_ERROR counter.
14773         * toplev.c (toplev_main): Call print_ignored_options even if
14774         just werrorcount is non-zero.  Exit with FATAL_EXIT_CODE
14775         even if just werrorcount is non-zero.
14777         PR debug/55608
14778         * dwarf2out.c (tree_add_const_value_attribute): Call ggc_free (array)
14779         on failure.
14780         (resolve_one_addr): Fail if referenced STRING_CST hasn't been written.
14781         (string_cst_pool_decl): New function.
14782         (optimize_one_addr_into_implicit_ptr): New function.
14783         (resolve_addr_in_expr): Optimize DWARF location expression
14784         DW_OP_addr DW_OP_stack_value where DW_OP_addr refers to some variable
14785         which doesn't live in memory, but has DW_AT_location or
14786         DW_AT_const_value, or refers to a string literal, into
14787         DW_OP_GNU_implicit_pointer.
14788         (optimize_location_into_implicit_ptr): New function.
14789         (resolve_addr): If removing DW_AT_location of a variable because
14790         it was DW_OP_addr of address of the variable, but the variable doesn't
14791         live in memory, try to emit const value attribute for the initializer.
14793 2013-03-21  Marc Glisse  <marc.glisse@inria.fr>
14795         * tree.h (VECTOR_TYPE_P): New macro.
14796         (VECTOR_INTEGER_TYPE_P, VECTOR_FLOAT_TYPE_P, FLOAT_TYPE_P,
14797         TYPE_MODE): Use it.
14798         * fold-const.c (fold_cond_expr_with_comparison): Use build_zero_cst.
14799         VEC_COND_EXPR cannot be lvalues.
14800         (fold_ternary_loc) <VEC_COND_EXPR>: Merge with the COND_EXPR case.
14802 2013-03-21  Marc Glisse  <marc.glisse@inria.fr>
14804         * simplify-rtx.c (simplify_binary_operation_1) <VEC_CONCAT>:
14805         Restrict the transformation to equal modes.
14807 2013-03-21  Richard Biener  <rguenther@suse.de>
14809         PR tree-optimization/39326
14810         * tree-ssa-loop-im.c (UNANALYZABLE_MEM_ID): New define.
14811         (MEM_ANALYZABLE): Adjust.
14812         (record_mem_ref_loc): Move bitmap ops ...
14813         (gather_mem_refs_stmt): ... here.  Use the shared mem-ref for
14814         unanalyzable refs, do not record locations for it.
14815         (analyze_memory_references): Allocate ref zero as shared
14816         unanalyzable ref.
14817         (refs_independent_p): Do not test for unanalyzed mems here.
14818         (ref_indep_loop_p_1): Special-case disambiguation against
14819         the unanalyzed ref.
14820         (ref_indep_loop_p): Assert we are not queried for the unanalyzed mem.
14822 2013-03-21  Christophe Lyon  <christophe.lyon@linaro.org>
14824         * config/arm/arm-protos.h (tune_params): Add
14825         prefer_neon_for_64bits field.
14826         * config/arm/arm.c (prefer_neon_for_64bits): New variable.
14827         (arm_slowmul_tune): Default prefer_neon_for_64bits to false.
14828         (arm_fastmul_tune, arm_strongarm_tune, arm_xscale_tune): Ditto.
14829         (arm_9e_tune, arm_v6t2_tune, arm_cortex_tune): Ditto.
14830         (arm_cortex_a15_tune, arm_cortex_a5_tune): Ditto.
14831         (arm_cortex_a9_tune, arm_v6m_tune, arm_fa726te_tune): Ditto.
14832         (arm_option_override): Handle -mneon-for-64bits new option.
14833         * config/arm/arm.h (TARGET_PREFER_NEON_64BITS): New macro.
14834         (prefer_neon_for_64bits): Declare new variable.
14835         * config/arm/arm.md (arch): Rename neon_onlya8 and neon_nota8 to
14836         avoid_neon_for_64bits and neon_for_64bits. Remove onlya8 and nota8.
14837         (arch_enabled): Handle new arch types. Remove support for onlya8
14838         and nota8.
14839         (one_cmpldi2): Use new arch names.
14840         (zero_extend<mode>di2, extend<mode>di2): Ditto.
14841         * config/arm/arm.opt (mneon-for-64bits): Add option.
14842         * config/arm/neon.md (adddi3_neon, subdi3_neon, iordi3_neon)
14843         (anddi3_neon, xordi3_neon, ashldi3_neon, <shift>di3_neon): Use
14844         neon_for_64bits instead of nota8 and avoid_neon_for_64bits instead
14845         of onlya8.
14846         * doc/invoke.texi (-mneon-for-64bits): Document.
14848 2013-03-21  Richard Biener  <rguenther@suse.de>
14850         PR tree-optimization/39326
14851         * tree-ssa-loop-im.c (bb_loop_postorder): New global static.
14852         (sort_bbs_in_loop_postorder_cmp): New function.
14853         (gather_mem_refs_in_loops): Assign mem-ref IDs in loop postorder.
14855 2013-03-21  Richard Biener  <rguenther@suse.de>
14857         * tree-vect-data-refs.c (vect_update_interleaving_chain): Remove.
14858         (vect_insert_into_interleaving_chain): Likewise.
14859         (vect_drs_dependent_in_basic_block): Inline ...
14860         (vect_slp_analyze_data_ref_dependence): ... here.  New function,
14861         split out from ...
14862         (vect_analyze_data_ref_dependence): ... here.  Simplify.
14863         (vect_check_interleaving): Simplify.
14864         (vect_analyze_data_ref_dependences): Likewise.  Split out ...
14865         (vect_slp_analyze_data_ref_dependences): ... this new function.
14866         (dr_group_sort_cmp): New function.
14867         (vect_analyze_data_ref_accesses): Compute data-reference groups
14868         here instead of in vect_analyze_data_ref_dependence.  Use
14869         a more efficient algorithm.
14870         * tree-vect-slp.c (vect_slp_analyze_bb_1): Use
14871         vect_slp_analyze_data_ref_dependences.  Call
14872         vect_analyze_data_ref_accesses earlier.
14873         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14874         * tree-vectorizer.h (vect_analyze_data_ref_dependences): Adjust.
14875         (vect_slp_analyze_data_ref_dependences): New prototype.
14877 2013-03-21  Richard Biener  <rguenther@suse.de>
14879         * tree-ssa-loop-im.c (can_sm_ref_p): Do not test whether
14880         ref is stored in the loop.
14881         (find_refs_for_sm): Walk only over all stores.
14882         (store_motion_loop): Allocate from lim_bitmap_obstack.
14883         (store_motion): Likewise.
14885 2013-03-21  Richard Biener  <rguenther@suse.de>
14887         * tree-vect-loop-manip.c (slpeel_tree_peel_loop_to_edge):
14888         Update virtual SSA form.
14890 2013-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14892         * configure.ac (gcc_cv_ld_eh_frame_ciev3): New test.
14893         * configure: Regenerate.
14894         * config.in: Regenerate.
14895         * config/sol2.c (solaris_override_options): Only enforce DWARF 2
14896         if !HAVE_LD_EH_FRAME_CIEV3.
14898 2013-03-21  Richard Biener  <rguenther@suse.de>
14900         * tree-cfg.c (verify_expr_no_block): New function.
14901         (verify_expr_location_1): Verify that neither DECL_DEBUG_EXPR
14902         nor DECL_VALUE_EXPR have locations with associated blocks.
14903         * tree-ssa-live.c (clear_unused_block_pointer_1): Remove.
14904         (clear_unused_block_pointer): Remove code dealing with
14905         blocks in DECL_DEBUG_EXPR locations.
14907 2013-03-21  Richard Biener  <rguenther@suse.de>
14909         * tree.h (DECL_DEBUG_EXPR_IS_FROM): Rename to ...
14910         (DECL_HAS_DEBUG_EXPR_P): ... this.  Guard properly.
14911         * tree.c (copy_node_stat): Do not copy DECL_HAS_DEBUG_EXPR_P.
14912         * dwarf2out.c (add_var_loc_to_decl): Use DECL_HAS_DEBUG_EXPR_P
14913         instead of DECL_DEBUG_EXPR_IS_FROM.
14914         * gimplify.c (gimplify_modify_expr): Likewise.
14915         * tree-cfg.c (verify_expr_location_1): Likewise.
14916         * tree-complex.c (create_one_component_var): Likewise.
14917         * tree-sra.c (create_access_replacement): Likewise.
14918         * tree-ssa-live.c (clear_unused_block_pointer_1): Likewise.
14919         (clear_unused_block_pointer): Likewise.
14920         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Likewise.
14921         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Likewise.
14922         * var-tracking.c (var_debug_decl): Likewise.
14923         (track_expr_p): Likewise.
14924         * tree-inline.c (add_local_variables): Likewise.  Set
14925         DECL_HAS_DEBUG_EXPR_P after copying it.
14926         * tree-diagnostic.c (default_tree_printer): Use DECL_HAS_DEBUG_EXPR_P
14927         instead of DECL_DEBUG_EXPR_IS_FROM.  Guard properly.
14929 2013-03-21  Uros Bizjak  <ubizjak@gmail.com>
14931         PR bootstrap/56656
14932         * configure.ac (HAVE_AS_IX86_INTERUNIT_MOVQ): New test.
14933         * configure: Regenerate.
14934         * config.in: Regenerate.
14935         * config/i386/i386.md (*movdf_internal): Use
14936         HAVE_AS_IX86_INTERUNIT_MOVQ to handle broken assemblers that require
14937         movd instead of movq mnemonic for interunit moves.
14938         (*movdi_internal): Ditto.
14940 2013-03-21  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
14942         * config/aarch64/aarch64-simd.md (simd_fabd): New Attribute.
14943         (abd<mode>_3): New pattern.
14944         (aba<mode>_3): New pattern.
14945         (fabd<mode>_3): New pattern.
14947 2013-03-21  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
14949         * config/aarch64/aarch64-elf.h (REGISTER_PREFIX): Remove.
14950         * config/aarch64/aarch64.c (aarch64_print_operand): Remove all
14951         occurrence of REGISTER_PREFIX as its empty string.
14953 2013-03-20  Jeff Law  <law@redhat.com>
14955         * tree-ssa-dom.c (record_equivalences_from_incoming_edge): Record
14956         addititional equivalences for equality comparisons between an SSA_NAME
14957         and a constant where the SSA_NAME was set from a widening conversion.
14959 2013-03-20  Walter Lee  <walt@tilera.com>
14961         * config/tilegx/sync.md (atomic_test_and_set): New pattern.
14963 2013-03-20  Uros Bizjak  <ubizjak@gmail.com>
14965         * config/i386/i386.md (*movoi_internal_avx): Emit insn template
14966         depending on type attribute.
14967         (*movti_internal): Ditto.
14968         (*movtf_internal): Ditto.
14969         (*movxf_internal): Ditto.
14970         (*movdf_internal): Ditto.
14971         (*movsf_internal): Ditto.
14973 2013-03-20  Uros Bizjak  <ubizjak@gmail.com>
14975         * config/i386/i386.md (*movti_internal): Set prefix attribute to
14976         maybe_vex for sselog1 and ssemov types.
14977         (*movdi_internal): Reorder operand constraints.
14978         (*movsi_internal): Ditto.  Set prefix attribute to
14979         maybe_vex for sselog1 and ssemov types.
14980         (*movtf_internal): Set prefix attribute to maybe_vex
14981         for sselog1 and ssemov types.
14982         (*movdf_internal): Ditto.  Set prefix_data16 attribute for
14983         DImode ssemov types.  Reorder operand constraints.
14984         (*movsf_internal): Set type of alternatives 3,4 to imov.  Set prefix
14985         attribute to maybe_vex for sselog1 and ssemov types.  Set prefix_data16
14986         attribute for SImode ssemov types.  Reorder operand constraints.
14988 2013-03-20  Martin Jambor  <mjambor@suse.cz>
14990         * params.def (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): New parameter.
14991         * ipa-cp.c (hint_time_bonus): Add abonus for known array indices.
14993 2013-03-20  Pat Haugen  <pthaugen@us.ibm.com>
14995         * config/rs6000/predicates.md (indexed_address, update_address_mem
14996         update_indexed_address_mem): New predicates.
14997         * config/rs6000/vsx.md (vsx_extract_<mode>_zero): Set correct "type"
14998         attribute for load/store instructions.
14999         * config/rs6000/dfp.md (movsd_store): Likewise.
15000         (movsd_load): Likewise.
15001         * config/rs6000/rs6000.md (zero_extend<mode>di2_internal1): Likewise.
15002         (unnamed HI->DI extend define_insn): Likewise.
15003         (unnamed SI->DI extend define_insn): Likewise.
15004         (unnamed QI->SI extend define_insn): Likewise.
15005         (unnamed QI->HI extend define_insn): Likewise.
15006         (unnamed HI->SI extend define_insn): Likewise.
15007         (unnamed HI->SI extend define_insn): Likewise.
15008         (extendsfdf2_fpr): Likewise.
15009         (movsi_internal1): Likewise.
15010         (movsi_internal1_single): Likewise.
15011         (movhi_internal): Likewise.
15012         (movqi_internal): Likewise.
15013         (movcc_internal1): Correct mnemonic for stw insn. Set correct "type"
15014         attribute for load/store instructions.
15015         (mov<mode>_hardfloat): Set correct "type" attribute for load/store
15016         instructions.
15017         (mov<mode>_softfloat): Likewise.
15018         (mov<mode>_hardfloat32): Likewise.
15019         (mov<mode>_hardfloat64): Likewise.
15020         (mov<mode>_softfloat64): Likewise.
15021         (movdi_internal32): Likewise.
15022         (movdi_internal64): Likewise.
15023         (probe_stack_<mode>): Likewise.
15025 2013-03-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
15027         * config/rs6000/vector.md (VEC_R): Add 32-bit integer, binary
15028         floating point, and decimal floating point to reload iterator.
15030         * config/rs6000/constraints.md (wl constraint): New constraints to
15031         return FLOAT_REGS if certain options are used to reduce the number
15032         of separate patterns that exist in the file.
15033         (wx constraint): Likewise.
15034         (wz constraint): Likewise.
15036         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
15037         -mdebug=reg, print wg, wl, wx, and wz constraints.
15038         (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
15039         Initialize the reload functions for 64-bit binary/decimal floating
15040         point types.
15041         (reg_offset_addressing_ok_p): If we are on a power7 or later, use
15042         LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
15043         create the buffer on the stack to overcome not having a 32-bit
15044         load and store.
15045         (rs6000_emit_move): Likewise.
15046         (rs6000_secondary_memory_needed_rtx): Likewise.
15047         (rs6000_alloc_sdmode_stack_slot): Likewise.
15048         (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
15049         via xxlxor, just like DFmode 0.0.
15051         * config/rs6000/rs6000.h (TARGET_NO_SDMODE_STACK): New macro,
15052         define as 1 if we are running on a power7 or newer.
15053         (enum r6000_reg_class_enum): Add new constraints.
15055         * config/rs6000/dfp.md (movsd): Delete, combine with binary
15056         floating point moves in rs6000.md.  Combine power6x (mfpgpr) moves
15057         with other moves by using conditional constraits (wg).  Use LFIWZX
15058         and STFIWX for loading SDmode on power7.  Use xxlxor to create 0.0f.
15059         (movsd splitter): Likewise.
15060         (movsd_hardfloat): Likewise.
15061         (movsd_softfloat): Likewise.
15063         * config/rs6000/rs6000.md (FMOVE32): New iterators to combine
15064         binary and decimal floating point moves.
15065         (fmove_ok): New attributes to combine binary and decimal floating
15066         point moves, and to combine power6x (mfpgpr) moves along normal
15067         floating moves.
15068         (real_value_to_target): Likewise.
15069         (f32_lr): Likewise.
15070         (f32_lm): Likewise.
15071         (f32_li): Likewise.
15072         (f32_sr): Likewise.
15073         (f32_sm): Likewise.
15074         (f32_si): Likewise.
15075         (movsf): Combine binary and decimal floating point moves.  Combine
15076         power6x (mfpgpr) moves with other moves by using conditional
15077         constraits (wg).  Use LFIWZX and STFIWX for loading SDmode on power7.
15078         (mov<mode> for SFmode/SDmode); Likewise.
15079         (SFmode/SDmode splitters): Likewise.
15080         (movsf_hardfloat): Likewise.
15081         (mov<mode>_hardfloat for SFmode/SDmode): Likewise.
15082         (movsf_softfloat): Likewise.
15083         (mov<mode>_softfloat for SFmode/SDmode): Likewise.
15085         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wl,
15086         wx and wz constraints.
15088         * config/rs6000/constraints.md (wg constraint): New constraint to
15089         return FLOAT_REGS if -mmfpgpr (power6x) was used.
15091         * config/rs6000/rs6000.h (enum r6000_reg_class_enum): Add wg
15092         constraint.
15094         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
15095         -mdebug=reg, print wg, wl, wx, and wz constraints.
15096         (rs6000_init_hard_regno_mode_ok): Initialize new constraints.
15097         Initialize the reload functions for 64-bit binary/decimal floating
15098         point types.
15099         (reg_offset_addressing_ok_p): If we are on a power7 or later, use
15100         LFIWZX and STFIWX to load/store 32-bit decimal types, and don't
15101         create the buffer on the stack to overcome not having a 32-bit
15102         load and store.
15103         (rs6000_emit_move): Likewise.
15104         (rs6000_secondary_memory_needed_rtx): Likewise.
15105         (rs6000_alloc_sdmode_stack_slot): Likewise.
15106         (rs6000_preferred_reload_class): On VSX, we can create SFmode 0.0f
15107         via xxlxor, just like DFmode 0.0.
15109         * config/rs6000/dfp.md (movdd): Delete, combine with binary
15110         floating point moves in rs6000.md.  Combine power6x (mfpgpr) moves
15111         with other moves by using conditional constraits (wg).  Use LFIWZX
15112         and STFIWX for loading SDmode on power7.
15113         (movdd splitters): Likewise.
15114         (movdd_hardfloat32): Likewise.
15115         (movdd_softfloat32): Likewise.
15116         (movdd_hardfloat64_mfpgpr): Likewise.
15117         (movdd_hardfloat64): Likewise.
15118         (movdd_softfloat64): Likewise.
15120         * config/rs6000/rs6000.md (FMOVE64): New iterators to combine
15121         64-bit binary and decimal floating point moves.
15122         (FMOVE64X): Likewise.
15123         (movdf): Combine 64-bit binary and decimal floating point moves.
15124         Combine power6x (mfpgpr) moves with other moves by using
15125         conditional constraits (wg).
15126         (mov<mode> for DFmode/DDmode): Likewise.
15127         (DFmode/DDmode splitters): Likewise.
15128         (movdf_hardfloat32): Likewise.
15129         (mov<mode>_hardfloat32 for DFmode/DDmode): Likewise.
15130         (movdf_softfloat32): Likewise.
15131         (movdf_hardfloat64_mfpgpr): Likewise.
15132         (movdf_hardfloat64): Likewise.
15133         (mov<mode>_hardfloat64 for DFmode/DDmode): Likewise.
15134         (movdf_softfloat64): Likewise.
15135         (mov<mode>_softfloat64 for DFmode/DDmode): Likewise.
15136         (reload_<mode>_load): Move to later in the file so they aren't in
15137         the middle of the floating point move insns.
15138         (reload_<mode>_store): Likewise.
15140         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wg
15141         constraint.
15143         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print out wg
15144         constraint if -mdebug=reg.
15145         (rs6000_initi_hard_regno_mode_ok): Enable wg constraint if -mfpgpr.
15146         Enable using dd reload support if needed.
15148         * config/rs6000/dfp.md (movtd): Delete, combine with 128-bit
15149         binary and decimal floating point moves in rs6000.md.
15150         (movtd_internal): Likewise.
15152         * config/rs6000/rs6000.md (FMOVE128): Combine 128-bit binary and
15153         decimal floating point moves.
15154         (movtf): Likewise.
15155         (movtf_internal): Likewise.
15156         (mov<mode>_internal, TDmode/TFmode): Likewise.
15157         (movtf_softfloat): Likewise.
15158         (mov<mode>_softfloat, TDmode/TFmode): Likewise.
15160         * config/rs6000/rs6000.md (movdi_mfpgpr): Delete, combine with
15161         movdi_internal64, using wg constraint for move direct operations.
15162         (movdi_internal64): Likewise.
15164         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print
15165         MODES_TIEABLE_P for selected modes.  Print the numerical value of
15166         the various virtual registers. Use GPR/FPR first/last values,
15167         instead of hard coding the register numbers.  Print which modes
15168         have reload functions registered.
15169         (rs6000_option_override_internal): If -mdebug=reg, trace the options
15170         settings before/after setting cpu, target and subtarget settings.
15171         (rs6000_secondary_reload_trace): Improve the RTL dump for -mdebug=addr
15172         and for secondary reload failures in rs6000_secondary_reload_inner.
15173         (rs6000_secondary_reload_fail): Likewise.
15174         (rs6000_secondary_reload_inner): Likewise.
15176         * config/rs6000/rs6000.md (FIRST_GPR_REGNO): Add convenience
15177         macros for first/last GPR and FPR registers.
15178         (LAST_GPR_REGNO): Likewise.
15179         (FIRST_FPR_REGNO): Likewise.
15180         (LAST_FPR_REGNO): Likewise.
15182         * config/rs6000/vector.md (mul<mode>3): Use the combined macro
15183         VECTOR_UNIT_ALTIVEC_OR_VSX_P instead of separate calls to
15184         VECTOR_UNIT_ALTIVEC_P and VECTOR_UNIT_VSX_P.
15185         (vcond<mode><mode>): Likewise.
15186         (vcondu<mode><mode>): Likewise.
15187         (vector_gtu<mode>): Likewise.
15188         (vector_gte<mode>): Likewise.
15189         (xor<mode>3): Don't allow logical operations on TImode in 32-bit
15190         to prevent the compiler from converting DImode operations to TImode.
15191         (ior<mode>3): Likewise.
15192         (and<mode>3): Likewise.
15193         (one_cmpl<mode>2): Likewise.
15194         (nor<mode>3): Likewise.
15195         (andc<mode>3): Likewise.
15197         * config/rs6000/constraints.md (wt constraint): New constraint
15198         that returns VSX_REGS if TImode is allowed in VSX registers.
15200         * config/rs6000/predicates.md (easy_fp_constant): 0.0f is an easy
15201         constant under VSX.
15203         * config/rs6000/rs6000-modes.def (PTImode): Define, PTImode is
15204         similar to TImode, but it is restricted to being in the GPRs.
15206         * config/rs6000/rs6000.opt (-mvsx-timode): New switch to allow
15207         TImode to occupy a single VSX register.
15209         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Default to
15210         -mvsx-timode for power7/power8.
15211         (power7 cpu): Likewise.
15212         (power8 cpu): Likewise.
15214         * config/rs6000/rs6000.c (rs6000_hard_regno_nregs_internal): Make
15215         sure that TFmode/TDmode take up two registers if they are ever
15216         allowed in the upper VSX registers.
15217         (rs6000_hard_regno_mode_ok): If -mvsx-timode, allow TImode in VSX
15218         registers.
15219         (rs6000_init_hard_regno_mode_ok): Likewise.
15220         (rs6000_debug_reg_global): Add debugging for PTImode and wt
15221         constraint.  Print if LRA is turned on.
15222         (rs6000_option_override_internal): Give an error if -mvsx-timode
15223         and VSX is not enabled.
15224         (invalid_e500_subreg): Handle PTImode, restricting it to GPRs.  If
15225         -mvsx-timode, restrict TImode to reg+reg addressing, and PTImode
15226         to reg+offset addressing.  Use PTImode when checking offset
15227         addresses for validity.
15228         (reg_offset_addressing_ok_p): Likewise.
15229         (rs6000_legitimate_offset_address_p): Likewise.
15230         (rs6000_legitimize_address): Likewise.
15231         (rs6000_legitimize_reload_address): Likewise.
15232         (rs6000_legitimate_address_p): Likewise.
15233         (rs6000_eliminate_indexed_memrefs): Likewise.
15234         (rs6000_emit_move): Likewise.
15235         (rs6000_secondary_reload): Likewise.
15236         (rs6000_secondary_reload_inner): Handle PTImode.  Allow 64-bit
15237         reloads to fpr registers to continue to use reg+offset addressing,
15238         but 64-bit reloads to altivec registers need reg+reg addressing.
15239         Drop test for PRE_MODIFY, since VSX loads/stores no longer support
15240         it.  Treat LO_SUM like a PLUS operation.
15241         (rs6000_secondary_reload_class): If type is 64-bit, prefer to use
15242         FLOAT_REGS instead of VSX_RGS to allow use of reg+offset addressing.
15243         (rs6000_cannot_change_mode_class): Do not allow TImode in VSX
15244         registers to share a register with a smaller sized type, since VSX
15245         puts scalars in the upper 64-bits.
15246         (print_operand): Add support for PTImode.
15247         (rs6000_register_move_cost): Use VECTOR_MEM_VSX_P instead of
15248         VECTOR_UNIT_VSX_P to catch types that can be loaded in VSX
15249         registers, but don't have arithmetic support.
15250         (rs6000_memory_move_cost): Add test for VSX.
15251         (rs6000_opt_masks): Add -mvsx-timode.
15253         * config/rs6000/vsx.md (VSm): Change to use 64-bit aligned moves
15254         for TImode.
15255         (VSs): Likewise.
15256         (VSr): Use wt constraint for TImode.
15257         (VSv): Drop TImode support.
15258         (vsx_movti): Delete, replace with versions for 32-bit and 64-bit.
15259         (vsx_movti_64bit): Likewise.
15260         (vsx_movti_32bit): Likewise.
15261         (vec_store_<mode>): Use VSX iterator instead of vector iterator.
15262         (vsx_and<mode>3): Delete use of '?' constraint on inputs, just put
15263         one '?' on the appropriate output constraint.  Do not allow TImode
15264         logical operations on 32-bit systems.
15265         (vsx_ior<mode>3): Likewise.
15266         (vsx_xor<mode>3): Likewise.
15267         (vsx_one_cmpl<mode>2): Likewise.
15268         (vsx_nor<mode>3): Likewise.
15269         (vsx_andc<mode>3): Likewise.
15270         (vsx_concat_<mode>): Likewise.
15271         (vsx_xxpermdi_<mode>): Fix thinko for non V2DF/V2DI modes.
15273         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Map from
15274         OPTION_MASK_VSX_TIMODE.
15275         (enum rs6000_reg_class_enum): Add RS6000_CONSTRAINT_wt.
15276         (STACK_SAVEAREA_MODE): Use PTImode instead of TImode.
15278         * config/rs6000/rs6000.md (INT mode attribute): Add PTImode.
15279         (TI2 iterator): New iterator for TImode, PTImode.
15280         (wd mode attribute): Add values for vector types.
15281         (movti_string): Replace TI move operations with operations for TImode
15282         and PTImode.  Add support for TImode being allowed in VSX registers.
15283         (mov<mode>_string, TImode/PTImode): Likewise.
15284         (movti_ppc64): Likewise.
15285         (mov<mode>_ppc64, TImode/PTImode): Likewise.
15286         (TI mode splitters): Likewise.
15288         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wt
15289         constraint.
15291 2013-03-20  Marc Glisse  <marc.glisse@inria.fr>
15293         PR tree-optimization/56355
15294         * fold-const.c (tree_binary_nonnegative_warnv_p) <MULT_EXPR>:
15295         Also handle integers with undefined overflow.
15297 2013-03-20  Catherine Moore  <clm@codesourcery.com>
15298             Maciej W. Rozycki  <macro@codesourcery.com>
15299             Tom de Vries  <tom@codesourcery.com>
15300             Nathan Sidwell  <nathan@codesourcery.com>
15301             Iain Sandoe  <iain@codesourcery.com>
15302             Nathan Froyd  <froydnj@codesourcery.com>
15303             Chao-ying Fu  <fu@mips.com>
15305         * doc/extend.texi: (micromips, nomicromips, nocompression):
15306         Document new function attributes.
15307         * doc/invoke.texi (minterlink-compressed, mmicromips,
15308         m14k, m14ke, m14kec): Document new options.
15309         (minterlink-mips16): Update documentation.
15310         * doc/md.texi (ZC, ZD): Document new constraints.
15311         * configure.ac (gcc_cv_as_micromips): Check if linker
15312         supports the .set micromips directive.
15313         * configure: Regenerate.
15314         * config.in: Regenerate.
15315         * config/mips/mips-tables.opt: Regenerate.
15316         * config/mips/micromips.md: New file.
15317         * constraints.md (ZC, ZD): New constraints.
15318         * config/mips/predicates.md (movep_src_register): New predicate.
15319         (movep_src_operand): New predicate.
15320         (non_volatile_mem_operand): New predicate.
15321         * config/mips/mips.md (multimem): New type.
15322         (length): Differentiate between 17-bit and 18-bit branch offsets.
15323         (MOVEP1, MOVEP2): New mode iterator.
15324         (mov_<load>l): Use ZC constraint.
15325         (mov_<load>r): Likewise.
15326         (mov_<store>l): Likewise.
15327         (mov_<store>r): Likewise.
15328         (*branch_equality<mode>_inverted): Add microMIPS support.
15329         (*branch_equality<mode>): Likewise.
15330         (*jump_absolute): Likewise.
15331         (indirect_jump_<mode>): Likewise.
15332         (tablejump_<mode>): Likewise.
15333         (<optab>_internal): Likewise.
15334         (sibcall_internal): Likewise.
15335         (sibcall_value_internal): Likewise.
15336         (prefetch): Use constraint ZD.
15337         * config/mips/mips.opt (minterlink-compressed): New option.
15338         (minterlink-mips16): Now an alias for minterlink-compressed.
15339         (mmicromips): New option.
15340         * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
15341         (compare_and_swap_12): Likewise.
15342         (sync_add<mode>): Likewise.
15343         (sync_<optab>_12): Likewise.
15344         (sync_old_<optab>_12): Likewise.
15345         (sync_new_<optab>_12): Likewise.
15346         (sync_nand_12): Likewise.
15347         (sync_old_nand_12): Likewise.
15348         (sync_new_nand_12): Likewise.
15349         (sync_sub<mode>): Likewise.
15350         (sync_old_add<mode>): Likewise.
15351         (sync_old_sub<mode>): Likewise.
15352         (sync_new_add<mode>): Likewise.
15353         (sync_new_sub<mode>): Likewise.
15354         (sync_<optab><mode>): Likewise.
15355         (sync_old_<optab><mode>): Likewise.
15356         (sync_new_<optab><mode>): Likewise.
15357         (sync_nand<mode>): Likewise.
15358         (sync_old_nand<mode>): Likewise.
15359         (sync_new_nand<mode>): Likewise.
15360         (sync_lock_test_and_set<mode>): Likewise.
15361         (test_and_set_12): Likewise.
15362         (atomic_compare_and_swap<mode>): Likewise.
15363         (atomic_exchange<mode>_llsc): Likewise.
15364         (atomic_fetch_add<mode>_llsc): Likewise.
15365         * config/mips/mips-cpus.def (m14kc, m14k): New processors.
15366         * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
15367         (umips_save_restore_pattern_p): Likewise.
15368         (umips_load_store_pair_p): Likewise.
15369         (umips_output_load_store_pair): Likewise.
15370         (umips_movep_target_p): Likewise.
15371         (umips_12bit_offset_address_p): Likewise.
15372         * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
15373         (mips_base_mips16): Rename this...
15374         (mips_base_compression_flags): ...to this. Update all uses.
15375         (mips_attribute_table): Add micromips, nomicromips and nocompression.
15376         (mips_mips16_decl_p): Delete.
15377         (mips_nomips16_decl_p): Delete.
15378         (mips_get_compress_on_flags): New function.
15379         (mips_get_compress_off_flags): New function.
15380         (mips_get_compress_mode): New function.
15381         (mips_get_compress_on_name): New function.
15382         (mips_get_compress_off_name): New function.
15383         (mips_insert_attributes): Support multiple compression types.
15384         (mips_merge_decl_attributes): Likewise.
15385         (umips_12bit_offset_address_p): New function.
15386         (mips_start_function_definition): Emit .set micromips directive.
15387         (mips_call_may_need_jalx_p): New function.
15388         (mips_function_ok_for_sibcall): Add microMIPS support.
15389         (mips_print_operand_punctuation): Support short delay slots and
15390         compact jumps.
15391         (umips_swm_mask, umips_swm_encoding): New.
15392         (umips_build_save_restore): New function.
15393         (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
15394         (was_mips16_p): Remove.
15395         (old_compression_mode): New.
15396         (mips_set_compression_mode): New function.
15397         (mips_set_current_function): Add microMIPS support.
15398         (mips_option_override): Likewise.
15399         (umips_save_restore_pattern_p): New function.
15400         (umips_output_save_restore): New function.
15401         (umips_load_store_pair_p_1): New function.
15402         (umips_load_store_pair_p): New function.
15403         (umips_output_load_store_pair_1): New function.
15404         (umips_output_load_store_pair): New function.
15405         (umips_movep_target_p) New function.
15406         (mips_prepare_pch_save): Add microMIPS support.
15407         * config/mips/mips.h (TARGET_COMPRESSION): New.
15408         (TARGET_CPU_CPP_BUILTINS): Update macro
15409         to use new compression flags and to support microMIPS.
15410         (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
15411         (MIPS_ARCH_FLOAT_SPEC): Likewise.
15412         (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
15413         (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
15414         (ASM_SPEC): Support mmicromips and mno-micromips.
15415         (M16STORE_REG_P): New macro.
15416         (MIPS_CALL): Support TARGET_MICROMIPS.
15417         (MICROMIPS_J): New macro.
15418         (mips_base_mips16): Rename this...
15419         (mips_base_compression_flags): ...to this.
15420         (UMIPS_12BIT_OFFSET_P): New macro.
15421         * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
15422         (MULTILIB_DIRNAMES): Likewise.
15423 2013-03-20  Richard Biener  <rguenther@suse.de>
15425         PR tree-optimization/56661
15426         * tree-ssa-sccvn.c (visit_use): Only value-number calls if
15427         the result does not have to be distinct.
15429 2013-03-20  Richard Biener  <rguenther@suse.de>
15431         * tree-inline.c (copy_tree_body_r): Sync MEM_REF code with
15432         remap_gimple_op_r.
15434 2013-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
15435             Steven Bosscher  <steven@gcc.gnu.org>
15437         PR rtl-optimization/56605
15438         * loop-iv.c (implies_p): Handle equal RTXs and subregs.
15440 2013-03-20  Uros Bizjak  <ubizjak@gmail.com>
15442         PR bootstrap/56656
15443         * config/i386/i386.md (*movdi_internal): Handle broken assemblers
15444         that require movd instead of movq.
15446 2013-03-20  Richard Biener  <rguenther@suse.de>
15448         * tree-ssa-structalias.c (struct variable_info): Add pointer
15449         to the first field of an aggregate with sub-vars.  Make
15450         this and the pointer to the next subfield its ID.
15451         (vi_next): New function.
15452         (nothing_id, anything_id, readonly_id, escaped_id, nonlocal_id,
15453         storedanything_id, integer_id): Increment by one.
15454         (new_var_info, get_call_vi, lookup_call_clobber_vi,
15455         get_call_clobber_vi): Adjust.
15456         (solution_set_expand): Simplify and speedup.
15457         (solution_set_add): Inline into ...
15458         (set_union_with_increment): ... this.  Adjust accordingly.
15459         (do_sd_constraint): Likewise.
15460         (do_ds_constraint): Likewise.
15461         (do_complex_constraint): Simplify.
15462         (build_pred_graph): Adjust.
15463         (solve_graph): Likewise.  Simplify and speedup.
15464         (get_constraint_for_ssa_var, get_constraint_for_ptr_offset,
15465         get_constraint_for_component_ref, get_constraint_for_1,
15466         first_vi_for_offset, first_or_preceding_vi_for_offset,
15467         create_function_info_for, create_variable_info_for_1,
15468         create_variable_info_for, intra_create_variable_infos): Adjust.
15469         (init_base_vars): Push NULL for ID zero.
15470         (compute_points_to_sets): Adjust.
15472 2013-03-20  Richard Biener  <rguenther@suse.de>
15474         * cfgloop.c (verify_loop_structure): Streamline and avoid
15475         ICEing on corrupt loop tree.
15476         * graph.c (draw_cfg_nodes_for_loop): Avoid ICEing on corrupt
15477         loop tree.
15479 2013-03-20  Richard Biener  <rguenther@suse.de>
15481         * tree-vect-loop-manip.c (slpeel_can_duplicate_loop_p): Do not
15482         check whether an SSA update is needed.
15484 2013-03-20  Richard Sandiford  <rdsandiford@googlemail.com>
15486         * config/mips/constraints.md (T): Rename to...
15487         (Yf): ...this.
15488         (U): Rename to...
15489         (Yd): ...this.
15490         * config/mips/mips.md (*movdi_64bit, *movdi_64bit_mips16)
15491         (*mov<mode>_internal, *mov<mode>_mips16): Update accordingly.
15493 2013-03-19  Ian Bolton  <ian.bolton@arm.com>
15495         * config/aarch64/aarch64.md (*sub<mode>3_carryin): New pattern.
15496         (*subsi3_carryin_uxtw): Likewise.
15498 2013-03-19  Ian Bolton  <ian.bolton@arm.com>
15500         * config/aarch64/aarch64.md (*ror<mode>3_insn): New pattern.
15501         (*rorsi3_insn_uxtw): Likewise.
15503 2013-03-19  Ian Bolton  <ian.bolton@arm.com>
15505         * config/aarch64/aarch64.md (*extr<mode>5_insn): New pattern.
15506         (*extrsi5_insn_uxtw): Likewise.
15508 2013-03-19  Richard Biener  <rguenther@suse.de>
15510         PR tree-optimization/56273
15511         * passes.c (init_optimization_passes): Move second VRP after DOM.
15513 2013-03-19  Uros Bizjak  <ubizjak@gmail.com>
15515         * config/i386/i386.md (*movti_internal): Merge from
15516         *movti_internal_rex64 and *movti_internal_sse.  Use x64 isa attribute.
15517         (*movdi_internal): Merge with *movdi_internal_rex64.  Use x64 and
15518         nox64 isa attributes.
15520 2013-03-18  Richard Biener  <rguenther@suse.de>
15522         * tree-ssa-structalias.c (find): Use gcc_checking_assert.
15523         (unite): Likewise.
15524         (merge_node_constraints): Likewise.
15525         (build_succ_graph): Likewise.
15526         (valid_graph_edge): Inline into single caller.
15527         (unify_nodes): Likewise.  Use bitmap_set_bit return value
15528         and cache varinfo.
15529         (scc_visit): Fix formatting and variable use.
15530         (do_sd_constraint): Use gcc_checking_assert.
15531         (do_ds_constraint): Likewise.
15532         (do_complex_constraint): Likewise.
15533         (condense_visit): Likewise.  Cleanup.
15534         (dump_pred_graph): New function.
15535         (perform_var_substitution): Dump the pred-graph before
15536         variable substitution.
15537         (find_equivalent_node): Use gcc_checking_assert.
15538         (rewrite_constraints): Guard checking loop with ENABLE_CHECKING.
15540 2013-03-18  Richard Biener  <rguenther@suse.de>
15542         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
15543         Remove cond_expr_stmt_list argument and do not gimplify the
15544         built expression.
15545         (vect_loop_versioning): Adjust.
15546         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
15547         Cleanup to use less temporaries.
15548         (vect_create_data_ref_ptr): Cleanup.
15550 2013-03-18  Jakub Jelinek  <jakub@redhat.com>
15552         PR tree-optimization/56635
15553         * fold-const.c (operand_equal_p): For MEM_REF and TARGET_MEM_REF,
15554         require types_compatible_p types.
15556 2013-03-18  Nick Clifton  <nickc@redhat.com>
15558         * config/stormy16/stormy16.c (xstormy16_expand_prologue): Remove
15559         spurious backslash.
15561         * config/mn10300/mn10300.c (mn10300_get_live_callee_saved_regs):
15562         Add missing line to comment describing function.
15564 2013-03-18  Richard Biener  <rguenther@suse.de>
15566         PR tree-optimization/56210
15567         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
15568         Handle string / character search functions.
15569         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
15571 2013-03-18  Richard Biener  <rguenther@suse.de>
15573         PR middle-end/56483
15574         * cfgexpand.c (expand_gimple_cond): Inline gimple_cond_single_var_p
15575         and implement properly.
15576         * gimple.h (gimple_cond_single_var_p): Remove.
15578 2013-03-18  Richard Biener  <rguenther@suse.de>
15580         * tree-data-ref.h (find_data_references_in_loop): Declare.
15581         * tree-data-ref.c (get_references_in_stmt): Use a stack
15582         vector pre-allocated in the callers.
15583         (find_data_references_in_stmt): Adjust.
15584         (graphite_find_data_references_in_stmt): Likewise.
15585         (create_rdg_vertices): Likewise.
15586         (find_data_references_in_loop): Export.
15587         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15588         Compute dependences here...
15589         (vect_analyze_data_refs): ...not here.  When we encounter
15590         a non-vectorizable data reference in basic-block vectorization
15591         truncate the data reference vector.  Do not bother to
15592         fixup data-dependence information for gather loads.
15593         * tree-vect-slp.c (vect_slp_analyze_bb_1): Check the number
15594         of data references, as reported.
15596 2013-03-18  Richard Biener  <rguenther@suse.de>
15598         PR tree-optimization/3713
15599         * tree-ssa-sccvn.c (visit_copy): Simplify.  Always propagate
15600         has_constants and expr.
15601         (stmt_has_constants): Properly valueize SSA names when deciding
15602         whether the stmt has constants.
15604 2013-03-18  Richard Biener  <rguenther@suse.de>
15606         * tree-ssa-loop-manip.c (find_uses_to_rename): Do not scan the
15607         whole function when there is nothing to do.
15608         * tree-ssa-loop.c (pass_vectorize): Remove TODO_update_ssa.
15609         * tree-vectorizer.c (vectorize_loops): Update virtual and
15610         loop-closed SSA once.
15611         * tree-vect-loop.c (vect_transform_loop): Do not update SSA here.
15613 2013-03-18  Richard Biener  <rguenther@suse.de>
15615         PR middle-end/56113
15616         * domwalk.c (bb_postorder): New global static.
15617         (cmp_bb_postorder): New function.
15618         (walk_dominator_tree): Replace scheme imposing an order for
15619         visiting dominator sons by one sorting them at the time they
15620         are pushed on the stack.
15622 2013-03-18  Richard Biener  <rguenther@suse.de>
15624         PR tree-optimization/39326
15625         * tree-ssa-loop-im.c (refs_independent_p): Exploit symmetry.
15626         (struct mem_ref): Replace mem member with ao_ref typed member.
15627         (MEM_ANALYZABLE): Adjust.
15628         (memref_eq): Likewise.
15629         (mem_ref_alloc): Likewise.
15630         (gather_mem_refs_stmt): Likewise.
15631         (mem_refs_may_alias_p): Use the ao_ref to query the alias oracle.
15632         (execute_sm_if_changed_flag_set): Adjust.
15633         (execute_sm): Likewise.
15634         (ref_always_accessed_p): Likewise.
15635         (refs_independent_p): Likewise.
15636         (can_sm_ref_p): Likewise.
15638 2013-03-18  Jakub Jelinek  <jakub@redhat.com>
15640         PR c/56566
15641         * tree.c (tree_int_cst_min_precision): For integer_zerop (value)
15642         return 1 even for !unsignedp.
15644 2013-03-17  Uros Bizjak  <ubizjak@gmail.com>
15646         * config/i386/i386.md (isa): Add x64 and nox64.
15647         (enabled): Define x64 for TARGET_64BIT and nox64 for !TARGET_64BIT.
15648         (*pushtf): Enable *roF alternative for x64 isa only.
15649         (*pushxf): Merge with *pushxf_nointeger.  Use Yx*r constraint. Set
15650         mode attribute of integer alternatives to DImode for TARGET_64BIT.
15651         (*pushdf): Merge with *pushdf_rex64.  Use x64 and nox64 isa attributes.
15652         (*movtf_internal): Merge from *movtf_internal_rex64 and
15653         *movtf_internal_sse.  Use x64 and nox64 isa attributes.
15654         (*movxf_internal): Merge with *movxf_internal_rex64.  Use x64 and
15655         nox64 isa attributes.
15656         (*movdf_internal): Merge with *movdf_internal_rex64.  Use x64 and
15657         nox64 isa attributes.
15658         * config/i386/constraints.md (Yd): Do not set for TARGET_64BIT.
15660 2013-03-17  Uros Bizjak  <ubizjak@gmail.com>
15662         * config/alpha/alpha.c (TARGET_LRA_P): New define.
15664 2013-03-17  Jakub Jelinek  <jakub@redhat.com>
15666         PR target/56640
15667         * config/arm/arm.h (REG_CLASS_NAMES): Add "SFP_REG" and "AFP_REG"
15668         class names.  Remove trailing comma after "ALL_REGS".
15670 2013-03-16  Jan Hubicka  <jh@suse.cz>
15672         * cgraph.h (cgraph_get_create_real_symbol_node): Declare.
15673         * cgraph.c (cgraph_get_create_real_symbol_node): New function.
15674         * cgrpahbuild.c: Use cgraph_get_create_real_symbol_node instead
15675         of cgraph_get_create_node.
15676         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15678 2013-03-16  Jason Merrill  <jason@redhat.com>
15680         PR debug/49090
15681         * dwarf2out.c (gen_generic_params_dies): Indicate default arguments
15682         with DW_AT_default_value.
15684 2013-03-16  Jakub Jelinek  <jakub@redhat.com>
15686         * BASE-VER: Set to 4.9.0.
15688 2013-03-14  Andi Kleen  <ak@linux.intel.com>
15690         PR target/56619
15691         * doc/extend.texi: Document __ATOMIC_HLE_ACQUIRE,
15692         __ATOMIC_HLE_RELEASE. Document __builtin_ia32 TSX intrincs.
15693         Document _x* TSX intrinsics.
15695 2013-03-14  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
15696             David Holsgrove  <david.holsgrove@xilinx.com>
15698         * configure.ac: Add MicroBlaze TLS support detection.
15699         * configure: Regenerate.
15700         * config/microblaze/microblaze-protos.h
15701         (microblaze_cannot_force_const_mem, microblaze_tls_referenced_p,
15702         symbol_mentioned_p, label_mentioned_p): Add prototypes.
15703         * config/microblaze/microblaze.c (microblaze_address_type): Add
15704         ADDRESS_TLS and tls_reloc address types.
15705         (microblaze_address_info): Add tls_reloc.
15706         (TARGET_HAVE_TLS): Define.
15707         (get_tls_get_addr, microblaze_tls_symbol_p, microblaze_tls_operand_p_1,
15708          microblaze_tls_referenced_p, microblaze_cannot_force_const_mem,
15709          symbol_mentioned_p, label_mentioned_p, tls_mentioned_p,
15710         load_tls_operand,  microblaze_call_tls_get_addr,
15711         microblaze_legitimize_tls_address): New functions.
15712         (microblaze_classify_unspec): Handle UNSPEC_TLS.
15713         (get_base_reg): Use microblaze_tls_symbol_p.
15714         (microblaze_classify_address): Handle TLS.
15715         (microblaze_legitimate_pic_operand): Use symbol_mentioned_p,
15716         label_mentioned_p and microblaze_tls_referenced_p.
15717         (microblaze_legitimize_address): Handle TLS.
15718         (microblaze_address_insns): Handle ADDRESS_TLS.
15719         (pic_address_needs_scratch): Handle TLS.
15720         (print_operand_address): Handle TLS.
15721         (microblaze_expand_prologue): Check TLS_NEEDS_GOT.
15722         (microblaze_expand_move): Handle TLS.
15723         (microblaze_legitimate_constant_p): Check
15724         microblaze_cannot_force_const_mem and microblaze_tls_symbol_p.
15725         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
15726         * config/microblaze/microblaze.h (TLS_NEEDS_GOT): Define
15727         (PIC_OFFSET_TABLE_REGNUM): Set.
15728         * config/microblaze/linux.h (TLS_NEEDS_GOT): Define.
15729         * config/microblaze/microblaze.md (UNSPEC_TLS): Define.
15730         (addsi3, movsi_internal2, movdf_internal): Update constraints
15731         * config/microblaze/predicates.md (arith_plus_operand): Define
15732         (move_operand): Redefine as move_src_operand,
15733         check microblaze_tls_referenced_p.
15735 2013-03-14  Ian Bolton  <ian.bolton@arm.com>
15737         * config/aarch64/aarch64.md: (*and<mode>3nr_compare0): Use CC_NZ.
15738         (*and_<SHIFT:optab><mode>3nr_compare0): Likewise.
15740 2013-03-14  Ian Bolton  <ian.bolton@arm.com>
15742         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return correct
15743         CC mode for AND.
15745 2013-03-14  Jakub Jelinek  <jakub@redhat.com>
15747         PR tree-optimization/53265
15748         * common.opt (Waggressive-loop-optimizations): New option.
15749         * tree-ssa-loop-niter.c: Include tree-pass.h.
15750         (do_warn_aggressive_loop_optimizations): New function.
15751         (record_estimate): Call it.  Don't add !is_exit bounds to loop->bounds
15752         if number_of_latch_executions returned constant.
15753         (estimate_numbers_of_iterations_loop): Call number_of_latch_executions
15754         early.  If number_of_latch_executions returned constant, set
15755         nb_iterations_upper_bound back to it.
15756         * cfgloop.h (struct loop): Add warned_aggressive_loop_optimizations
15757         field.
15758         * Makefile.in (tree-ssa-loop-niter.o): Depend on $(TREE_PASS_H).
15759         * doc/invoke.texi (-Wno-aggressive-loop-optimizations): Document.
15761         * config/aarch64/t-aarch64-linux (MULTARCH_DIRNAME): Remove.
15762         (MULTILIB_OSDIRNAMES): Set.
15763         * genmultilib: If defaultosdirname doesn't start with :: , set
15764         defaultosdirname2 instead, clear it and emit two . multilib_raw
15765         entries instead of just one.
15767 2013-03-14  Kaz Kojima  <kkojima@gcc.gnu.org>
15769         * config/sh/linux.h (TARGET_DEFAULT): Remove MASK_USERMODE.
15770         (SUBTARGET_OVERRIDE_OPTIONS): Set TARGET_USERMODE as default.
15771         * config/sh/netbsd-elf.h (TARGET_DEFAULT): Remove MASK_USERMODE.
15772         (SUBTARGET_OVERRIDE_OPTIONS): New.
15774 2013-03-13  Oleg Endo  <olegendo@gcc.gnu.org>
15776         PR target/49880
15777         * config/sh/sh.opt (FPU_SINGLE_ONLY): New mask.
15778         (musermode): Convert to Var(TARGET_USERMODE).
15779         * config/sh/sh.h (SELECT_SH2A_SINGLE_ONLY, SELECT_SH4_SINGLE_ONLY,
15780         MASK_ARCH): Add MASK_FPU_SINGLE_ONLY.
15781         * config/sh/sh.c (sh_option_override): Use
15782         TARGET_FPU_DOUBLE || TARGET_FPU_SINGLE_ONLY for call-fp case.
15783         * config/sh/sh.md (udivsi3_i1, divsi3_i1): Remove ! TARGET_SH4
15784         condition.
15785         (udivsi3_i4, divsi3_i4): Use TARGET_FPU_DOUBLE condition instead of
15786         TARGET_SH4.
15787         (udivsi3_i4_single, divsi3_i4_single): Use
15788         TARGET_FPU_SINGLE_ONLY || TARGET_FPU_DOUBLE instead of TARGET_HARD_SH4.
15790 2013-03-13  Dave Korn  <dave.korn.cygwin@....>
15792         * config/i386/cygwin.h (SHARED_LIBGCC_SPEC): Make shared libgcc the
15793         default setting.
15795 2013-03-13  Richard Biener  <rguenther@suse.de>
15797         PR tree-optimization/56608
15798         * tree-vect-slp.c (vect_schedule_slp): Do not remove scalar
15799         calls when vectorizing basic-blocks.
15801 2013-03-13  Jakub Jelinek  <jakub@redhat.com>
15803         PR plugins/45078
15804         * config.gcc: On arm, mips, sh and sparc add vxworks-dummy.h to
15805         tm_file.
15807 2013-03-12  Jakub Jelinek  <jakub@redhat.com>
15809         * doc/invoke.texi (-Waddr-space-convert): Move into the table earlier.
15811 2013-03-11  Jan Hubicka  <jh@suse.cz>
15813         PR lto/56557
15814         * lto-streamer-out.c (output_symbol_p): Skip references from
15815         constructors of external variables.
15817 2013-03-11  Jan Hubicka  <jh@suse.cz>
15819         PR middle-end/56571
15820         * valtrack.c (cleanup_auto_inc_dec): Unshare clobbers originating
15821         from pseudos.
15822         * emit-rtl.c (verify_rtx_sharing): Likewise.
15823         (copy_insn_1): Likewise.
15824         * rtl.c (copy_rtx): Likewise.
15826 2013-03-11  Georg-Johann Lay  <avr@gjlay.de>
15828         PR target/56591
15829         * config/avr/avr.c (avr_print_operand): Add space after '%c' in
15830         output_operand_lossage message.
15832 2013-03-11  Richard Earnshaw  <rearnsha@arm.com>
15834         PR target/56470
15835         * arm.c (shift_op): Validate RTL pattern on the fly.
15836         (arm_print_operand, case 'S'): Don't use shift_operator to validate
15837         the RTL.
15839 2013-03-10  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
15841         PR target/56347
15842         * config/pa/pa.md (call_value): Check for calls to powf and direct to
15843         new call patterns that clobber %fr12.
15844         (call_val_powf, call_val_powf_pic, call_val_powf_64bit): New insn,
15845         split and postreload patterns.
15846         * config/pa/pa.c (pa_conditional_register_usage): Revert marking
15847         registers %fr12 and %fr12R as call used.
15849 2013-03-09  Steven Bosscher  <steven@gcc.gnu.org>
15851         * dse.c (delete_dead_store_insn): Respect TDF_DETAILS.
15852         (canon_address, record_store, replace_read, check_mem_read_rtx,
15853         scan_insn, dse_step1, dse_step2_init, dse_step2_spill,
15854         dse_step4, dse_step5_nospill, dse_step5_spill, dse_step6,
15855         rest_of_handle_dse): Likewise.
15857 2013-03-09  Richard Sandiford  <rdsandiford@googlemail.com>
15859         PR middle-end/56524
15860         * tree.h (tree_optimization_option): Rename target_optabs to optabs.
15861         Add base_optabs.
15862         (TREE_OPTIMIZATION_OPTABS): Update after previous field change.
15863         (TREE_OPTIMIZATION_BASE_OPTABS): New macro.
15864         (save_optabs_if_changed): Replace with...
15865         (init_tree_optimization_optabs): ...this.
15866         * optabs.c (save_optabs_if_changed): Rename to...
15867         (init_tree_optimization_optabs): ...this.  Take the optimization node
15868         as argument.  Do nothing if the base optabs are already correct.
15869         Reuse the existing TREE_OPTIMIZATION_OPTABS memory if we need
15870         to recompute optabs.
15871         * function.h (function): Remove optabs field.
15872         * function.c (invoke_set_current_function_hook): Call
15873         init_tree_optimization_optabs.  Use the result to initialize
15874         this_fn_optabs.
15876 2013-02-27  Aldy Hernandez  <aldyh@redhat.com>
15878         * trans-mem.c (expand_transaction): Do not set PR_INSTRUMENTEDCODE
15879         if GTMA_HAS_NO_INSTRUMENTATION.
15880         (generate_tm_state): Keep GTMA_HAS_NO_INSTRUMENTATION bit.
15881         (ipa_tm_transform_transaction): Set GTMA_HAS_NO_INSTRUMENTATION.
15882         * gimple.h (GTMA_HAS_NO_INSTRUMENTATION): Define.
15883         * gimple-pretty-print.c (dump_gimple_transaction): Handle
15884         GTMA_HAS_NO_INSTRUMENTATION.
15886 2013-03-08  Jakub Jelinek  <jakub@redhat.com>
15888         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Don't link against
15889         libasan_preinit.o.
15891 2013-03-08  Marek Polacek  <polacek@redhat.com>
15892             Jakub Jelinek  <jakub@redhat.com>
15894         PR tree-optimization/56478
15895         * predict.c (is_comparison_with_loop_invariant_p): Change the
15896         type of loop_step to tree.
15897         (predict_loops): Adjust.
15898         (predict_iv_comparison): Perform the computations on double_ints.
15900 2013-03-08  Richard Biener  <rguenther@suse.de>
15902         PR tree-optimization/56570
15903         * tree-cfg.c (verify_expr_location_1): Verify locations for
15904         DECL_DEBUG_EXPR.
15905         * tree-sra.c (create_access_replacement): Strip locations
15906         from DECL_DEBUG_EXPRs.
15908 2013-03-08  Richard Biener  <rguenther@suse.de>
15910         * tree-inline.c (expand_call_inline): Do not associate
15911         a BLOCK with the location in BLOCK_SOURCE_LOCATION.
15912         * tree-cfg.c (verify_location): Verify BLOCK_SOURCE_LOCATION.
15914 2013-03-08  Richard Biener  <rguenther@suse.de>
15916         * tree-ssa-ter.c (is_replaceable_p): Do not TER across location
15917         or block changes with -Og.  Fix for location / block encoding
15918         changes and PHI arguments with locations.
15920 2013-03-07  Steven Bosscher  <steven@gcc.gnu.org>
15922         * bitmap.c (struct bitmap_descriptor_d): Use unsigned HOST_WIDEST_INT
15923         for all counters.
15924         (struct output_info): Likewise.
15925         (register_overhead): Remove bad gcc_assert.
15926         (bitmap_find_bit): If there is only a single bitmap element, do not
15927         count a miss as a search.
15928         (print_statistics): Update for counter type changes.
15929         (dump_bitmap_statistics): Likewise.  Print headers such that they
15930         are properly lined up with the printed counters.
15932 2013-03-07  Jakub Jelinek  <jakub@redhat.com>
15934         PR tree-optimization/56559
15935         * tree-ssa-reassoc.c (zero_one_operation): When looking at rhs2,
15936         check that it has only a single use.
15938 2013-03-07  Richard Biener  <rguenther@suse.de>
15940         * doc/invoke.texi (fwhole-program): Discourage use in combination
15941         with -flto.
15943 2013-03-06  Jakub Jelinek  <jakub@redhat.com>
15945         * config/arm/t-arm (TM_H, OPTIONS_H_EXTRA): Add arm-cores.def.
15947         PR tree-optimization/56539
15948         * tree-tailcall.c (adjust_return_value_with_ops): Use GSI_SAME_STMT
15949         instead of GSI_CONTINUE_LINKING as last argument to
15950         force_gimple_operand_gsi.  Adjust function comment.
15952         * config/aarch64/t-aarch64 (TM_H, OPTIONS_H_EXTRA): Add
15953         aarch64-cores.def.
15955         PR middle-end/56548
15956         * expr.c (expand_cond_expr_using_cmove): When expanding cmove in
15957         promoted mode, convert the result back to the original mode.
15959 2013-03-06  Richard Biener  <rguenther@suse.de>
15961         PR middle-end/56294
15962         * tree-into-ssa.c (insert_phi_nodes_for): Add dumping.
15963         (insert_updated_phi_nodes_compare_uids): New function.
15964         (update_ssa): Sort symbols_to_rename after UID before
15965         traversing it to insert PHI nodes.
15967 2013-03-06  Richard Biener  <rguenther@suse.de>
15969         PR middle-end/50494
15970         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p):
15971         Do not adjust alignment of DECL_IN_CONSTANT_POOL decls.
15973         Revert
15974         2013-02-13  Richard Biener  <rguenther@suse.de>
15976         PR lto/50494
15977         * varasm.c (output_constant_def_1): Get the decl representing
15978         the constant as argument.
15979         (output_constant_def): Wrap output_constant_def_1.
15980         (make_decl_rtl): Use output_constant_def_1 with the decl
15981         representing the constant.
15982         (build_constant_desc): Optionally re-use a decl already
15983         representing the constant.
15984         (tree_output_constant_def): Adjust.
15986 2013-03-06  Joey Ye  <joey.ye@arm.com>
15988         PR lto/50293
15989         * gcc.c (convert_white_space): New function.
15990         (main): Handles white space in function name.
15992 2013-03-06  Oleg Endo  <olegendo@gcc.gnu.org>
15994         PR target/56529
15995         * config/sh/sh.c (sh_option_override): Check for TARGET_DYNSHIFT
15996         instead of TARGET_SH2 for call-table case.  Do not set sh_div_strategy
15997         to SH_DIV_CALL_TABLE for TARGET_SH2.
15998         * config.gcc (sh_multilibs): Add m2 and m2a to sh*-*-linux* multilib
15999         list.
16000         * doc/invoke.texi (SH options): Document mdiv= call-div1, call-fp,
16001         call-table options.
16003 2013-03-05  Sterling Augustine  <saugustine@google.com>
16004             Cary Coutant  <ccoutant@google.com>
16006         PR debug/55364
16007         * dwarf2out.c (resolve_addr): Don't call
16008         remove_loc_list_addr_table_entries a second time for the same
16009         expression.
16011 2013-03-05  Jakub Jelinek  <jakub@redhat.com>
16013         PR debug/56510
16014         * cfgexpand.c (expand_debug_parm_decl): Call copy_rtx on incoming.
16015         (avoid_complex_debug_insns): New function.
16016         (expand_debug_locations): Call it.
16018         PR rtl-optimization/56484
16019         * ifcvt.c (noce_process_if_block): If else_bb is NULL, avoid extending
16020         lifetimes of hard registers on small register class machines.
16022 2013-03-05  David Holsgrove  <david.holsgrove@xilinx.com>
16024         * config/microblaze/microblaze-protos.h: Rename
16025         microblaze_is_interrupt_handler to microblaze_is_interrupt_variant.
16026         * config/microblaze/microblaze.c (microblaze_attribute_table): Add
16027         fast_interrupt.
16028         (microblaze_fast_interrupt_function_p): New function.
16029         (microblaze_is_interrupt_handler): Rename to
16030         microblaze_is_interrupt_variant and add fast_interrupt check.
16031         (microblaze_must_save_register): Use microblaze_is_interrupt_variant.
16032         (save_restore_insns): Likewise.
16033         (compute_frame_size): Likewise.
16034         (microblaze_function_prologue): Add FAST_INTERRUPT_NAME.
16035         (microblaze_globalize_label): Likewise.
16036         * config/microblaze/microblaze.h: Define FAST_INTERRUPT_NAME.
16037         * config/microblaze/microblaze.md: Use wrapper
16038         microblaze_is_interrupt_variant.
16040 2013-03-05  Kai Tietz  <ktietz@redhat.com>
16042         * sdbout.c (sdbout_one_type): Switch to current function's section
16043         supporting cold/hot.
16045 2013-03-05  David Holsgrove  <david.holsgrove@xilinx.com>
16047         * doc/invoke.texi (MicroBlaze): Add -mbig-endian, -mlittle-endian,
16048         -mxl-reorder.
16050 2013-03-05  Jakub Jelinek  <jakub@redhat.com>
16052         PR middle-end/56461
16053         * ggc-common.c (gt_pch_save): For ENABLE_VALGRIND_CHECKING,
16054         if VALGRIND_GET_VBITS is defined, temporarily make object
16055         memory all defined, and restore previous valgrind addressability
16056         and definability afterwards.  Free this_object at the end.
16058         PR middle-end/56461
16059         * lra.c (lra): Call lra_clear_live_ranges if live_p,
16060         right before calling lra_create_live_ranges, also call it
16061         when clearing live_p.  Only call lra_clear_live_ranges
16062         at the end if live_p.
16064         PR middle-end/56461
16065         * sched-deps.c (delete_dep_node): Free DEP_REPLACE.
16067 2013-03-05  Richard Biener  <rguenther@suse.de>
16069         PR tree-optimization/56521
16070         * tree-ssa-sccvn.c (set_value_id_for_result): Always initialize
16071         value-id.
16073 2013-03-05  Steven Bosscher  <steven@gcc.gnu.org>
16075         PR c++/55135
16076         * except.h (remove_unreachable_eh_regions): New prototype.
16077         * except.c (remove_eh_handler_splicer): New function, split out
16078         of remove_eh_handler.
16079         (remove_eh_handler): Use remove_eh_handler_splicer.  Add comment
16080         warning about running it on many EH regions one at a time.
16081         (remove_unreachable_eh_regions_worker): New function, walk the
16082         EH tree in depth-first order and remove non-marked regions.
16083         (remove_unreachable_eh_regions): New function.
16084         * tree-eh.c (mark_reachable_handlers): New function, split out
16085         from remove_unreachable_handlers.
16086         (remove_unreachable_handlers): Use mark_reachable_handlers and
16087         remove_unreachable_eh_regions.
16088         (remove_unreachable_handlers_no_lp): Use mark_reachable_handlers
16089         and remove_unreachable_eh_regions.
16091 2013-03-05  Richard Biener  <rguenther@suse.de>
16093         PR middle-end/56525
16094         * loop-init.c (fix_loop_structure): Remove loops in two stages,
16095         not freeing them until the end.
16097 2013-03-05  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
16099         * config/s390/s390.h: Define DWARF2_ASM_LINE_DEBUG_INFO.
16101 2013-03-05  Richard Biener  <rguenther@suse.de>
16103         PR tree-optimization/56270
16104         * tree-vect-slp.c (vect_schedule_slp): Clear vectorized stmts
16105         of loads after scheduling an SLP instance.
16107 2013-03-05  Jakub Jelinek  <jakub@redhat.com>
16109         * Makefile.in (dg_target_exps): Add aarch64.exp, epiphany.exp and
16110         tic6x.exp.
16111         (check_gcc_parallelize): Run guality.exp as a separate job from
16112         vect.exp with unsorted.exp and $(dg_target_exps) separately from
16113         struct-layout-1.exp with stackalign.exp.
16115         * alias.c (init_alias_analysis): Clear reg_known_equiv_p bitmap.
16117         PR middle-end/56461
16118         * tree-vect-slp.c (vect_supported_load_permutation_p): Free
16119         load_index sbitmap even if some bit in it isn't set.
16121         PR middle-end/56461
16122         * tree-ssa-loop-niter.c (bb_queue): Remove typedef.
16123         (discover_iteration_bound_by_body_walk): Change queues to
16124         vec<vec<basic_block> > and queue to vec<basic_block>.  Fix up
16125         spelling in comment.  Call safe_push on queues[bound_index] directly.
16126         Release queues[queue_index] in every iteration unconditionally.
16127         Release bounds vector.
16129         PR middle-end/56461
16130         * tree-vect-stmts.c (free_stmt_vec_info_vec): Call
16131         free_stmt_vec_info on any left-over stmt_vec_info in the vector.
16132         * tree-vect-loop.c (vect_create_epilog_for_reduction): Release
16133         inner_phis vector.
16135 2013-03-05  Richard Biener  <rguenther@suse.de>
16137         PR lto/56515
16138         * tree-inline.c (remap_blocks_to_null): New function.
16139         (expand_call_inline): When expanding a call stmt without
16140         an associated block inline remap all callee blocks to NULL.
16142 2013-03-05  Jakub Jelinek  <jakub@redhat.com>
16144         PR rtl-optimization/56494
16145         * simplify-rtx.c (simplify_truncation): If C is narrower than A,
16146         optimize (truncate:A (subreg:B (truncate:C X) 0)) into
16147         (subreg:A (truncate:C X) 0) instead of (truncate:A X).
16149         PR middle-end/56461
16150         * sel-sched-ir.c (free_sched_pools): Release
16151         succs_info_pool.stack[succs_info_pool.max_top] vectors too
16152         if succs_info_pool.max_top isn't -1.
16154         PR bootstrap/56509
16155         * opts.c (opts_obstack, opts_concat): Moved to...
16156         * opts-common.c (opts_obstack, opts_concat): ... here.
16158 2013-03-04  Jakub Jelinek  <jakub@redhat.com>
16160         PR middle-end/56461
16161         * diagnostic.c (diagnostic_append_note): Save and restore old prefix.
16163 2013-03-04  Martin Jambor  <mjambor@suse.cz>
16165         * tree-dfa.c (get_or_create_ssa_default_def): Use parameter fn in
16166         all appropriate places.
16168 2013-01-04  Eric Botcazou  <ebotcazou@adacore.com>
16170         PR tree-optimization/56424
16171         * ipa-split.c (split_function): Do not set the RSO flag if result is
16172         not by reference and its type is a register type.
16174 2013-03-04  David Holsgrove  <david.holsgrove@xilinx.com>
16176         * config/microblaze/microblaze.c (microblaze_valid_pic_const): New
16177         (microblaze_legitimate_pic_operand): Likewise
16178         * config/microblaze/microblaze.h (LEGITIMATE_PIC_OPERAND_P): calls
16179         new function microblaze_legitimate_pic_operand
16180         * config/microblaze/microblaze-protos.h
16181         (microblaze_legitimate_pic_operand): Declare.
16183 2013-03-04  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
16185         * config/microblaze/predicates.md (call_insn_simple_operand):
16186         New predicate for supported rtx code types.
16187         * config/microblaze/microblaze.md (call_internal1): Use
16188         call_insn_simple_operand predicate.
16190 2013-03-04  Jakub Jelinek  <jakub@redhat.com>
16192         PR middle-end/56461
16193         * tree-loop-distribution.c (ldist_gen): Call partition_free after each
16194         partitions.ordered_remove.
16196         PR middle-end/56461
16197         * tree-vect-stmts.c (vectorizable_conversion): Don't call
16198         vec_oprnds0.create (1) for modifier == NONE.
16200         PR middle-end/56461
16201         * tree-vect-stmts.c (vectorizable_shift): Don't call create methods
16202         on vec_oprnds0 or vec_oprnds1 before loop, only call it on
16203         vec_oprnds1 right before pushing anything to it for
16204         scalar_shift_arg.
16206         PR middle-end/56461
16207         * tree-vect-loop.c (destroy_loop_vec_info): For !clean_stmts, just
16208         set nbbs to 0 instead of having separate code path.
16209         (vect_analyze_loop_form): Call destroy_loop_vec_info with true
16210         instead of false as last argument if returning NULL.
16212 2013-03-03  Sandra Loosemore  <sandra@codesourcery.com>
16214         * target.def (TARGET_OPTION_VALID_ATTRIBUTE_P): Update comments;
16215         the attribute is now called "target" instead of "option".
16216         (TARGET_OPTION_PRAGMA_PARSE): Likewise, for the pragma.
16217         * doc/tm.texi.in (Target Attributes):  Likewise document the correct
16218         attribute/pragma name for TARGET_OPTION_VALID_P and
16219         TARGET_OPTION_PRAGMA_PARSE.  Also copy-edit and correct markup.
16220         * doc/tm.texi: Regenerated.
16222 2013-03-02  David Holsgrove  <david.holsgrove@xilinx.com>
16224         * config/microblaze/microblaze.c:
16225         Check mcpu, pcmp requirement and set TARGET_REORDER to 0 if not met.
16226         * config/microblaze/microblaze.h: Add -mxl-reorder to
16227         DRIVER_SELF_SPECS.
16228         * config/microblaze/microblaze.md: New bswapsi2 and bswaphi2.
16229         instructions emitted if TARGET_REORDER.
16230         * config/microblaze/microblaze.opt: New option -mxl-reorder set to 1
16231         or 0 for -m/-mno case, but initialises as 2 to detect default use case
16232         separately.
16234 2013-03-01  Xinliang David Li  <davidxl@google.com>
16236         * tree-ssa-uninit.c (compute_control_dep_chain): Limit post-dom
16237         walk length.
16239 2013-03-01  Jakub Jelinek  <jakub@redhat.com>
16241         PR middle-end/56461
16242         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Release path
16243         vector even when returning true.  Fix up function comment formatting.
16245         PR middle-end/56461
16246         * ira-build.c (ira_loop_nodes_count): New variable.
16247         (create_loop_tree_nodes): Initialize it.
16248         (finish_loop_tree_nodes): Use it instead of looking at current_loops.
16250         PR middle-end/56461
16251         * tree-vect-data-refs.c (vect_permute_store_chain): Avoid using copy
16252         method on dr_chain and result_chain.
16253         * tree-vect-stmts.c (vectorizable_store): Only call
16254         result_chain.create if j == 0.
16256         PR middle-end/56461
16257         * tree-vect-stmts.c (vect_create_vectorized_promotion_stmts): Call
16258         vec_oprnds0->release (); rather than vec_oprnds0->truncate (0)
16259         before overwriting it.
16261 2013-03-01  Tobias Burnus  <burnus@net-b.de>
16263         * doc/extended.texi (C Extensions): Change order in @menu
16264         to match @node.
16265         (Other MIPS Built-in Functions): Move last MIPS entry before
16266         "picoChip Built-in Functions".
16267         (SH Built-in Functions): Move after RX Built-in Functions.
16268         * doc/gcc.texi (Introduction): Change order in @menu
16269         to match @node.
16270         * doc/md.texi (Constraints): Ditto.
16271         * gty.texi (Type Information): Ditto.
16272         (User-provided marking routines for template types): Make
16273         subsection.
16274         * doc/invoke.texi (AArch64 Options): Move before
16275         "Adapteva Epiphany Options".
16277 2013-02-28  Konstantin Serebryany  <konstantin.s.serebryany@gmail.com>
16278             Jakub Jelinek  <jakub@redhat.com>
16280         PR sanitizer/56454
16281         * asan.c (gate_asan): Lookup no_sanitize_address instead of
16282         no_address_safety_analysis attribute.
16283         * doc/extend.texi (no_address_safety_attribute): Rename to
16284         no_sanitize_address attribute, mention no_address_safety_analysis
16285         attribute as deprecated alias.
16287 2013-02-28  Jakub Jelinek  <jakub@redhat.com>
16289         PR middle-end/56461
16290         * tree-vectorizer.h (vect_get_slp_defs): Change 3rd argument
16291         type to vec<vec<tree> > *.
16292         * tree-vect-slp.c (vect_get_slp_defs): Likewise.  Change vec_defs
16293         to be vec<tree> instead of vec<tree> *, set vec_defs
16294         to vNULL and call vec_defs.create (number_of_vects), adjust other
16295         uses of vec_defs.
16296         * tree-vect-stmts.c (vect_get_vec_defs, vectorizable_call,
16297         vectorizable_condition): Adjust vect_get_slp_defs callers.
16299 2013-02-28  James Greenhalgh  <james.greenhalgh@arm.com>
16301         * config/aarch64/aarch64.c
16302         (aarch64_float_const_representable): Remove unused variable.
16304 2013-02-28  James Greenhalgh  <james.greenhalgh@arm.com>
16306         * config/aarch64/aarch64.c (aarch64_mangle_type): Make static.
16308 2013-02-28  James Greenhalgh  <james.greenhalgh@arm.com>
16310         * config/aarch64/aarch64-builtins.c
16311         (aarch64_init_simd_builtins): Make static.
16313 2013-02-28  James Greenhalgh  <james.greenhalgh@arm.com>
16315         * config/aarch64/aarch64.c
16316         (aarch64_simd_make_constant): Make static.
16318 2013-02-28  Martin Jambor  <mjambor@suse.cz>
16320         * tree-sra.c (load_assign_lhs_subreplacements): Do not put replacements
16321         with no initialization to the RHS of debug statements.
16323 2013-02-28  Martin Jambor  <mjambor@suse.cz>
16325         PR tree-optimization/56294
16326         * tree-sra.c (analyze_access_subtree): Create replacement declarations.
16327         Adjust dumping.
16328         (get_access_replacement): Do not call create_access_replacement.
16329         Assert a replacement exists.
16330         (get_repl_default_def_ssa_name): Create the replacement declaration
16331         itself.
16333 2013-02-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16335         * config/arm/arm.c (arm_output_mi_thunk): Call final_start_function and
16336         final_end_function.
16338 2013-02-28  Marek Polacek  <polacek@redhat.com>
16340         PR rtl-optimization/56466
16341         * loop-unroll.c (unroll_and_peel_loops): Call fix_loop_structure
16342         if we're changing a loop.
16343         (peel_loops_completely): Likewise.
16345 2013-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
16347         PR c++/55813
16348         * doc/invoke.texi ([-Wctor-dtor-privacy]): Complete.
16350 2013-02-28  Georg-Johann Lay  <avr@gjlay.de>
16352         PR target/56445
16353         * config/avr/avr.c (avr_init_builtins): Use 'n' instead of empty
16354         macro parameters with: FX_FTYPE_FX, FX_FTYPE_FX_INT, INT_FTYPE_FX,
16355         INTX_FTYPE_FX, FX_FTYPE_INTX.
16356         * config/avr/builtins.def: Adjust respective DEF_BUILTIN.
16358 2013-02-28  Georg-Johann Lay  <avr@gjlay.de>
16360         * avr/avr-mcus.def (ata5272, ata5505, attiny1634, ata6285)
16361         (ata6286, atmega8a, atmega48pa, ata5790, ata5790n, ata5795)
16362         (atmega164pa, atmega165pa, atmega168pa, atmega16hva, atmega16hvb)
16363         (atmega16hvbrevb, atmega16m1, atmega16u4, atmega26hvg, atmega32a)
16364         (atmega32a, atmega3250pa, atmega3290pa, atmega32c1, atmega32m1)
16365         (atmega32u4, atmega32u6, atmega64a, atmega6490a, atmega6490p)
16366         (atmega64c1, atmega64m1, atmega64rfa2, atmega64rfr2, atmega32hvb)
16367         (atmega32hvbrevb, atmega16hva2, atmega48hvf, at90pwm161)
16368         (atmega128a, atmega1284, atmxt112sl, atmxt224, atmxt224e)
16369         (atmxt336s, atxmega16a4u, atxmega16c4, atxmega32a4u, atxmega32c4)
16370         (atxmega32e5, atxmega64a3u, atxmega64a4u, atxmega64b1, atxmega64b3)
16371         (atxmega64c3, atxmega64d4, atxmega128a3u, atxmega128b1)
16372         (atxmega128b3, atxmega128c3, atxmega128d4, atmxt540s, atmxt540sreva)
16373         (atxmega192a3u, atxmega192c3, atxmega256a3u, atxmega256c3)
16374         (atxmega384c3, atxmega384d3, atxmega128a4u): New AVR_MCU.
16375         (avrxmega6): Increase max flash segments from 5 to 6.
16376         * config/avr/t-multilib: Regenerate.
16377         * config/avr/avr-tables.opt: Regenerate.
16378         * doc/avr-mmcu.texi: Regenerate.
16380 2013-02-28  Georg-Johann Lay  <avr@gjlay.de>
16382         * config/avr/avr.h (device_to_arch): Rename to device_to_ld.
16383         (avr_device_to_arch): Rename to avr_device_to_ld.
16384         (avr_device_to_as): New prototype.
16385         (EXTRA_SPEC_FUNCTIONS): Add device_to_as.
16386         (ASM_SPEC): Use device_to_as to get -mmcu= and -mno-skip-bug=.
16387         * config/avr/driver-avr.c (avr_device_to_as): New.
16388         (avr_device_to_arch): Rename to avr_device_to_ld.
16390 2013-02-27  Jakub Jelinek  <jakub@redhat.com>
16392         PR middle-end/56461
16393         * tree-vect-data-refs.c (vect_permute_load_chain): Avoid using copy
16394         method on dr_chain and result_chain.
16396         PR middle-end/56461
16397         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Call
16398         pointer_set_destroy on not_executed_last_iteration.
16400         PR middle-end/56461
16401         * tree-vect-loop.c (vectorizable_reduction): Release vect_defs vector.
16403         PR middle-end/56461
16404         * ipa-pure-const.c (propagate): Use FOR_EACH_FUNCTION instead of
16405         FOR_EACH_DEFINED_FUNCTION when freeing state.
16407         PR middle-end/56461
16408         * df-scan.c (df_insn_delete): Use df_scan_free_mws_vec before
16409         pool_free.
16410         (df_insn_rescan_debug_internal): Use df_scan_free_mws_vec before
16411         overwriting it.
16413         PR middle-end/56461
16414         * ipa-cp.c (decide_whether_version_node): Call vec_free on
16415         known_aggs[i].items and release known_aggs vector.
16417         PR middle-end/56461
16418         * ipa-reference.c (propagate): Free node_info even for alias nodes.
16420 2013-02-27  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
16422         * config/microblaze/microblaze.c (microblaze_emit_compare):
16423         Use xor for EQ/NE comparisions.
16424         * config/microblaze/microblaze.md (cstoresf4): Add constraints
16425         (cbranchsf4): Adjust operator to comparison_operator.
16427 2013-02-27  Jakub Jelinek  <jakub@redhat.com>
16429         PR middle-end/56461
16430         * tree-flow.h (edge_var_map_vector): Change into va_heap, vl_embed
16431         vector.
16432         * tree-ssa.c (redirect_edge_var_map_add): Use vec_safe_reserve and
16433         vec_safe_push, always update *slot.
16434         (redirect_edge_var_map_clear): Use vec_free.
16435         (redirect_edge_var_map_dup): Use vec_safe_copy and vec_safe_reserve.
16436         (free_var_map_entry): Use vec_free.
16437         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Use
16438         FOR_EACH_VEC_SAFE_ELT instead of FOR_EACH_VEC_ELT.
16440 2013-02-27  Andrey Belevantsev  <abel@ispras.ru>
16442         PR middle-end/45472
16443         * sel-sched-ir.c (merge_expr): Also change vinsn of merged expr
16444         when the may_trap_p bit of the exprs being merged differs.
16445         Reorder tests for speculativeness in the logical and operator.
16447 2013-02-27  Jakub Jelinek  <jakub@redhat.com>
16449         * incpath.c (add_standard_paths): Use reconcat instead of concat
16450         where appropriate and avoid leaking memory.
16452         * opts.h: Include obstack.h.
16453         (opts_concat): New prototype.
16454         (opts_obstack): New declaration.
16455         * opts.c (opts_concat): New function.
16456         (opts_obstack): New variable.
16457         (init_options_struct): Call gcc_init_obstack on opts_obstack.
16458         (finish_options): Use opts_concat instead of concat
16459         and XOBNEWVEC instead of XNEWVEC.
16460         * opts-common.c (generate_canonical_option, decode_cmdline_option,
16461         generate_option): Likewise.
16462         * Makefile.in (OPTS_H): Depend on $(OBSTACK_H).
16463         * lto-wrapper.c (main): Call gcc_init_obstack on opts_obstack.
16465         PR target/56455
16466         * stmt.c (expand_switch_as_decision_tree_p): If flag_pic
16467         and ASM_OUTPUT_ADDR_DIFF_ELT isn't defined, return true.
16469 2013-02-26  Jakub Jelinek  <jakub@redhat.com>
16471         PR middle-end/56461
16472         * lra-spills.c (lra_spill): Free spill_hard_reg at the end.
16474 2013-02-26  Joern Rennecke  <joern.rennecke@embecosm.com>
16476         * config/arm/arm.c (const_ok_for_dimode_op): Back out last change.
16477         (arm_block_move_unaligned_straight): Likewise.
16478         (arm_adjust_block_mem): Likewise.
16480 2013-02-26  Joern Rennecke  <joern.rennecke@embecosm.com>
16482         PR target/48901
16483         * config/lm32/lm32.c (gen_int_relational): Remove unused variables
16484         temp, cond and label.
16485         * config/lm32/lm32.md (ashlsi3): Remove unused variable one.
16487         PR target/52500
16488         * config/c6x/c6x.c (dbx_register_map): Change to unsigned.
16489         * config/c6x/c6x.h (dbx_register_map): Update declaration.
16491         PR target/52501
16492         * config/cr16/cr16-protos.h: Move end of RTX_CODE guard below end
16493         of prologue/epilogue functions.
16495         PR target/52550
16496         * config/tilegx/tilegx.c (tilegx_expand_prologue):
16497         Remove unused variable cfa_offset.
16498         * config/tilepro/tilepro.c (tilepro_expand_prologue): Likewise.
16500         PR target/54639
16501         * config/mn10300/mn10300.c (mn10300_expand_epilogue): Avoid offset
16502         type promotion to unsigned.
16504         PR target/54640
16505         * config/arm/arm.c (const_ok_for_dimode_op): Make code consistent
16506         for HOST_WIDE_INT of 32 bit / same size as int.
16507         (arm_block_move_unaligned_straight): Likewise.
16508         (arm_adjust_block_mem): Likewise.
16510         PR target/54662
16511         * config/mep/t-mep (mep-pragma.o): Use ALL_COMPILERFLAGS instead of
16512         ALL_CFLAGS.
16514 2013-02-26  Marek Polacek  <polacek@redhat.com>
16516         PR tree-optimization/56426
16517         * tree-ssa-loop.c (tree_ssa_loop_init): Always call scev_initialize.
16519 2013-02-26  Richard Biener  <rguenther@suse.de>
16521         PR target/56444
16522         * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Remove
16523         unused variable loops.
16525 2013-02-26  Jakub Jelinek  <jakub@redhat.com>
16527         PR tree-optimization/56448
16528         * fold-const.c (operand_equal_p) <case tcc_reference>: Don't look at
16529         TREE_SIDE_EFFECTS if flags contain OEP_CONSTANT_ADDRESS_OF.
16530         Clear OEP_CONSTANT_ADDRESS_OF from flags before recursing on second or
16531         later operands of the references, or even first operand for
16532         INDIRECT_REF, TARGET_MEM_REF or MEM_REF.
16534         PR tree-optimization/56443
16535         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): For
16536         overaligned types, pass TYPE_UNSIGNED (scalar_type) as second argument
16537         to type_for_mode langhook.
16539 2013-02-25  Matt Turner  <mattst88@gmail.com>
16541         * doc/invoke.texi: Document r4700.
16543 2013-02-25  Richard Biener  <rguenther@suse.de>
16545         PR tree-optimization/56175
16546         * tree-ssa-forwprop.c (hoist_conversion_for_bitop_p): New predicate,
16547         split out from ...
16548         (simplify_bitwise_binary): ... here.  Also guard the conversion
16549         of (type) X op CST to (type) (X op ((type-x) CST)) with it.
16551 2013-02-25  Catherine Moore  <clm@codesourcery.com>
16553         Revert:
16554         2013-02-24  Catherine Moore  <clm@codesourcery.com>
16555             Maciej W. Rozycki  <macro@codesourcery.com>
16556             Tom de Vries  <tom@codesourcery.com>
16557             Nathan Sidwell  <nathan@codesourcery.com>
16558             Iain Sandoe  <iain@codesourcery.com>
16559             Nathan Froyd  <froydnj@codesourcery.com>
16560             Chao-ying Fu  <fu@mips.com>
16562         * doc/extend.texi: (micromips, nomicromips, nocompression):
16563         Document new function attributes.
16564         * doc/invoke.texi (minterlink-compressed, mmicromips,
16565         m14k, m14ke, m14kec): Document new options.
16566         (minterlink-mips16): Update documentation.
16567         * doc/md.texi (ZC, ZD): Document new constraints.
16568         * configure.ac (gcc_cv_as_micromips): Check if linker
16569         supports the .set micromips directive.
16570         * configure: Regenerate.
16571         * config.in: Regenerate.
16572         * config/mips/mips-tables.opt: Regenerate.
16573         * config/mips/micromips.md: New file.
16574         * constraints.md (ZC, AD): New constraints.
16575         * config/mips/predicates.md (movep_src_register): New predicate.
16576         (movep_src_operand): New predicate.
16577         (non_volatile_mem_operand): New predicate.
16578         * config/mips/mips.md (multimem): New type.
16579         (length): Differentiate between 17-bit and 18-bit branch offsets.
16580         (MOVEP1, MOVEP2): New mode iterator.
16581         (mov_<load>l): Use ZC constraint.
16582         (mov_<load>r): Likewise.
16583         (mov_<store>l): Likewise.
16584         (mov_<store>r): Likewise.
16585         (*branch_equality<mode>_inverted): Add microMIPS support.
16586         (*branch_equality<mode>): Likewise.
16587         (*jump_absolute): Likewise.
16588         (indirect_jump_<mode>): Likewise.
16589         (tablejump_<mode>): Likewise.
16590         (<optab>_internal): Likewise.
16591         (sibcall_internal): Likewise.
16592         (sibcall_value_internal): Likewise.
16593         (prefetch): Use constraint ZD.
16594         * config/mips/mips.opt (minterlink-compressed): New option.
16595         (minterlink-mips16): Now an alias for minterlink-compressed.
16596         (mmicromips): New option.
16597         * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
16598         (compare_and_swap_12): Likewise.
16599         (sync_add<mode>): Likewise.
16600         (sync_<optab>_12): Likewise.
16601         (sync_old_<optab>_12): Likewise.
16602         (sync_new_<optab>_12): Likewise.
16603         (sync_nand_12): Likewise.
16604         (sync_old_nand_12): Likewise.
16605         (sync_new_nand_12): Likewise.
16606         (sync_sub<mode>): Likewise.
16607         (sync_old_add<mode>): Likewise.
16608         (sync_old_sub<mode>): Likewise.
16609         (sync_new_add<mode>): Likewise.
16610         (sync_new_sub<mode>): Likewise.
16611         (sync_<optab><mode>): Likewise.
16612         (sync_old_<optab><mode>): Likewise.
16613         (sync_new_<optab><mode>): Likewise.
16614         (sync_nand<mode>): Likewise.
16615         (sync_old_nand<mode>): Likewise.
16616         (sync_new_nand<mode>): Likewise.
16617         (sync_lock_test_and_set<mode>): Likewise.
16618         (test_and_set_12): Likewise.
16619         (atomic_compare_and_swap<mode>): Likewise.
16620         (atomic_exchange<mode>_llsc): Likewise.
16621         (atomic_fetch_add<mode>_llsc): Likewise.
16622         * config/mips/mips-cpus.def (m14kc, m14k): New processors.
16623         * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
16624         (umips_save_restore_pattern_p): Likewise.
16625         (umips_load_store_pair_p): Likewise.
16626         (umips_output_load_store_pair): Likewise.
16627         (umips_movep_target_p): Likewise.
16628         (umips_12bit_offset_address_p): Likewise.
16629         * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
16630         (mips_base_mips16): Rename this...
16631         (mips_base_compression_flags): ...to this. Update all uses.
16632         (mips_attribute_table): Add micromips, nomicromips and nocompression.
16633         (mips_mips16_decl_p): Delete.
16634         (mips_nomips16_decl_p): Delete.
16635         (mips_get_compress_on_flags): New function.
16636         (mips_get_compress_off_flags): New function.
16637         (mips_get_compress_mode): New function.
16638         (mips_get_compress_on_name): New function.
16639         (mips_get_compress_off_name): New function.
16640         (mips_insert_attributes): Support multiple compression types.
16641         (mips_merge_decl_attributes): Likewise.
16642         (umips_12bit_offset_address_p): New function.
16643         (mips_start_function_definition): Emit .set micromips directive.
16644         (mips_call_may_need_jalx_p): New function.
16645         (mips_function_ok_for_sibcall): Add microMIPS support.
16646         (mips_print_operand_punctuation): Support short delay slots and
16647         compact jumps.
16648         (umips_swm_mask, umips_swm_encoding): New.
16649         (umips_build_save_restore): New function.
16650         (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
16651         (was_mips16_p): Remove.
16652         (old_compression_mode): New.
16653         (mips_set_compression_mode): New function.
16654         (mips_set_current_function): Add microMIPS support.
16655         (mips_option_override): Likewise.
16656         (umips_save_restore_pattern_p): New function.
16657         (umips_output_save_restore): New function.
16658         (umips_load_store_pair_p_1): New function.
16659         (umips_load_store_pair_p): New function.
16660         (umips_output_load_store_pair_1): New function.
16661         (umips_output_load_store_pair): New function.
16662         (umips_movep_target_p) New function.
16663         (mips_prepare_pch_save): Add microMIPS support.
16664         * config/mips/mips.h (TARGET_COMPRESSION): New.
16665         (TARGET_CPU_CPP_BUILTINS): Update macro
16666         to use new compression flags and to support microMIPS.
16667         (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
16668         (MIPS_ARCH_FLOAT_SPEC): Likewise.
16669         (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
16670         (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
16671         (ASM_SPEC): Support mmicromips and mno-micromips.
16672         (M16STORE_REG_P): New macro.
16673         (MIPS_CALL): Support TARGET_MICROMIPS.
16674         (MICROMIPS_J): New macro.
16675         (mips_base_mips16): Rename this...
16676         (mips_base_compression_flags): ...to this.
16677         (UMIPS_12BIT_OFFSET_P): New macro.
16678         * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
16679         (MULTILIB_DIRNAMES): Likewise.
16681 2013-02-25  Tom de Vries  <tom@codesourcery.com>
16683         PR rtl-optimization/56131
16684         * insn-notes.def (INSN_NOTE_BASIC_BLOCK): Update comment.
16685         * cfgrtl.c (delete_insn): Don't reorder NOTE_INSN_DELETED_LABEL and
16686         NOTE_INSN_BASIC_BLOCK if BLOCK_FOR_INSN == NULL.
16688 2013-02-25  Tobias Burnus  <burnus@net-b.de>
16690         * doc/invoke.texi (-fsanitize=): Move from optimization
16691         to debugging options.
16693 2013-02-25  Andrey Belevantsev  <abel@ispras.ru>
16695         * sched-deps.c (sched_analyze_insn): Fix typo in comment.
16697 2013-02-25  Andrey Belevantsev  <abel@ispras.ru>
16698             Alexander Monakov  <amonakov@ispras.ru>
16700         PR middle-end/56077
16701         * sched-deps.c (sched_analyze_insn): When reg_pending_barrier,
16702         flush pending lists also on non-jumps.  Adjust comment.
16704 2013-02-24  Catherine Moore  <clm@codesourcery.com>
16705             Maciej W. Rozycki  <macro@codesourcery.com>
16706             Tom de Vries  <tom@codesourcery.com>
16707             Nathan Sidwell  <nathan@codesourcery.com>
16708             Iain Sandoe  <iain@codesourcery.com>
16709             Nathan Froyd  <froydnj@codesourcery.com>
16710             Chao-ying Fu  <fu@mips.com>
16712         * doc/extend.texi: (micromips, nomicromips, nocompression):
16713         Document new function attributes.
16714         * doc/invoke.texi (minterlink-compressed, mmicromips,
16715         m14k, m14ke, m14kec): Document new options.
16716         (minterlink-mips16): Update documentation.
16717         * doc/md.texi (ZC, ZD): Document new constraints.
16718         * configure.ac (gcc_cv_as_micromips): Check if linker
16719         supports the .set micromips directive.
16720         * configure: Regenerate.
16721         * config.in: Regenerate.
16722         * config/mips/mips-tables.opt: Regenerate.
16723         * config/mips/micromips.md: New file.
16724         * constraints.md (ZC, AD): New constraints.
16725         * config/mips/predicates.md (movep_src_register): New predicate.
16726         (movep_src_operand): New predicate.
16727         (non_volatile_mem_operand): New predicate.
16728         * config/mips/mips.md (multimem): New type.
16729         (length): Differentiate between 17-bit and 18-bit branch offsets.
16730         (MOVEP1, MOVEP2): New mode iterator.
16731         (mov_<load>l): Use ZC constraint.
16732         (mov_<load>r): Likewise.
16733         (mov_<store>l): Likewise.
16734         (mov_<store>r): Likewise.
16735         (*branch_equality<mode>_inverted): Add microMIPS support.
16736         (*branch_equality<mode>): Likewise.
16737         (*jump_absolute): Likewise.
16738         (indirect_jump_<mode>): Likewise.
16739         (tablejump_<mode>): Likewise.
16740         (<optab>_internal): Likewise.
16741         (sibcall_internal): Likewise.
16742         (sibcall_value_internal): Likewise.
16743         (prefetch): Use constraint ZD.
16744         * config/mips/mips.opt (minterlink-compressed): New option.
16745         (minterlink-mips16): Now an alias for minterlink-compressed.
16746         (mmicromips): New option.
16747         * config/mips/sync.md (sync_compare_and_swap<mode>): Use ZR constraint.
16748         (compare_and_swap_12): Likewise.
16749         (sync_add<mode>): Likewise.
16750         (sync_<optab>_12): Likewise.
16751         (sync_old_<optab>_12): Likewise.
16752         (sync_new_<optab>_12): Likewise.
16753         (sync_nand_12): Likewise.
16754         (sync_old_nand_12): Likewise.
16755         (sync_new_nand_12): Likewise.
16756         (sync_sub<mode>): Likewise.
16757         (sync_old_add<mode>): Likewise.
16758         (sync_old_sub<mode>): Likewise.
16759         (sync_new_add<mode>): Likewise.
16760         (sync_new_sub<mode>): Likewise.
16761         (sync_<optab><mode>): Likewise.
16762         (sync_old_<optab><mode>): Likewise.
16763         (sync_new_<optab><mode>): Likewise.
16764         (sync_nand<mode>): Likewise.
16765         (sync_old_nand<mode>): Likewise.
16766         (sync_new_nand<mode>): Likewise.
16767         (sync_lock_test_and_set<mode>): Likewise.
16768         (test_and_set_12): Likewise.
16769         (atomic_compare_and_swap<mode>): Likewise.
16770         (atomic_exchange<mode>_llsc): Likewise.
16771         (atomic_fetch_add<mode>_llsc): Likewise.
16772         * config/mips/mips-cpus.def (m14kc, m14k): New processors.
16773         * config/mips/mips-protos.h (umips_output_save_restore): New prototype.
16774         (umips_save_restore_pattern_p): Likewise.
16775         (umips_load_store_pair_p): Likewise.
16776         (umips_output_load_store_pair): Likewise.
16777         (umips_movep_target_p): Likewise.
16778         (umips_12bit_offset_address_p): Likewise.
16779         * config/mips/mips.c (MIPS_MAX_FIRST_STEP): Update for microMIPS.
16780         (mips_base_mips16): Rename this...
16781         (mips_base_compression_flags): ...to this. Update all uses.
16782         (mips_attribute_table): Add micromips, nomicromips and nocompression.
16783         (mips_mips16_decl_p): Delete.
16784         (mips_nomips16_decl_p): Delete.
16785         (mips_get_compress_on_flags): New function.
16786         (mips_get_compress_off_flags): New function.
16787         (mips_get_compress_mode): New function.
16788         (mips_get_compress_on_name): New function.
16789         (mips_get_compress_off_name): New function.
16790         (mips_insert_attributes): Support multiple compression types.
16791         (mips_merge_decl_attributes): Likewise.
16792         (umips_12bit_offset_address_p): New function.
16793         (mips_start_function_definition): Emit .set micromips directive.
16794         (mips_call_may_need_jalx_p): New function.
16795         (mips_function_ok_for_sibcall): Add microMIPS support.
16796         (mips_print_operand_punctuation): Support short delay slots and
16797         compact jumps.
16798         (umips_swm_mask, umips_swm_encoding): New.
16799         (umips_build_save_restore): New function.
16800         (mips_for_each_saved_gpr_and_fpr): Add microMIPS support.
16801         (was_mips16_p): Remove.
16802         (old_compression_mode): New.
16803         (mips_set_compression_mode): New function.
16804         (mips_set_current_function): Add microMIPS support.
16805         (mips_option_override): Likewise.
16806         (umips_save_restore_pattern_p): New function.
16807         (umips_output_save_restore): New function.
16808         (umips_load_store_pair_p_1): New function.
16809         (umips_load_store_pair_p): New function.
16810         (umips_output_load_store_pair_1): New function.
16811         (umips_output_load_store_pair): New function.
16812         (umips_movep_target_p) New function.
16813         (mips_prepare_pch_save): Add microMIPS support.
16814         * config/mips/mips.h (TARGET_COMPRESSION): New.
16815         (TARGET_CPU_CPP_BUILTINS): Update macro
16816         to use new compression flags and to support microMIPS.
16817         (MIPS_ISA_LEVEL_SPEC): Add m14k processors.
16818         (MIPS_ARCH_FLOAT_SPEC): Likewise.
16819         (ISA_HAS_LWXS): Include TARGET_MICROMIPS.
16820         (ISA_HAS_LOAD_DELAY): Exclude TARGET_MICROMIPS.
16821         (ASM_SPEC): Support mmicromips and mno-micromips.
16822         (M16STORE_REG_P): New macro.
16823         (MIPS_CALL): Support TARGET_MICROMIPS.
16824         (MICROMIPS_J): New macro.
16825         (mips_base_mips16): Rename this...
16826         (mips_base_compression_flags): ...to this.
16827         (UMIPS_12BIT_OFFSET_P): New macro.
16828         * config/mips/t-sde: (MULTILIB_OPTIONS): Add microMIPS.
16829         (MULTILIB_DIRNAMES): Likewise.
16831 2013-02-24  Jakub Jelinek  <jakub@redhat.com>
16833         PR target/52555
16834         * target-globals.c (save_target_globals): For init_reg_sets and
16835         target_reinit remporarily set this_fn_optabs to this_target_optabs.
16837 2013-02-22  James Grennahlgh  <james.greenhalgh@arm.com>
16839         * config/aarch64/aarch64-simd-builtins.def: Add copyright header.
16840         * config/aarch64/t-aarch64
16841         (aarch64-builtins.o): Depend on aarch64-simd-builtins.def.
16843 2013-02-22  Vladimir Makarov  <vmakarov@redhat.com>
16845         PR inline-asm/56148
16846         * lra-constraints.c (process_alt_operands): Reload operand
16847         conflicting with earlier clobber only if no more other conflicting
16848         operands.
16850 2013-02-22  Jakub Jelinek  <jakub@redhat.com>
16852         PR sanitizer/56393
16853         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Link in libasan_preinit.o
16854         if not linking a shared library.
16856 2013-02-22  Seth LaForge  <sethml@google.com>
16858         * config.gcc (arm*-*-eabi*): Treat arm*eb as big-endian.
16860 2013-02-22  Greta Yorsh  <Greta.Yorsh@arm.com>
16862         * config/arm/arm.md (split for extendsidi): Update condition.
16863         (zero_extend<mode>di2,extend<mode>di2): Add an alternative.
16864         * config/arm/iterators.md (qhs_extenddi_cstr): Likewise.
16865         (qhs_zextenddi_cstr): Likewise.
16867 2013-02-21  Jakub Jelinek  <jakub@redhat.com>
16869         PR middle-end/56420
16870         * expmed.c (EXACT_POWER_OF_2_OR_ZERO_P): Do subtraction in uhwi, to
16871         avoid signed wrapping.
16872         (expand_mult): Handle properly multiplication by
16873         ((dword_type) -1) << (BITS_PER_WORD - 1).  Improve multiplication by
16874         ((dword_type) 1) << (BITS_PER_WORD - 1).  Avoid undefined behavior
16875         in the compiler if coeff is HOST_WIDE_INT_MIN.
16876         (expand_divmod): Don't make ext_op1 static, change it's type to
16877         uhwi.  Avoid undefined behavior in -INTVAL (op1).
16879         PR rtl-optimization/50339
16880         * lower-subreg.h (struct lower_subreg_choices): Add splitting_ashiftrt
16881         field.
16882         * lower-subreg.c (compute_splitting_shift): Handle ASHIFTRT.
16883         (compute_costs): Call compute_splitting_shift also for ASHIFTRT
16884         into splitting_ashiftrt field.
16885         (find_decomposable_shift_zext, resolve_shift_zext): Handle also
16886         ASHIFTRT.
16887         (dump_choices): Fix up printing LSHIFTRT choices, print ASHIFTRT
16888         choices.
16890 2013-02-20  Aldy Hernandez  <aldyh@redhat.com>
16892         PR middle-end/56108
16893         * trans-mem.c (execute_tm_mark): Do not expand transactions that
16894         are sure to go irrevocable.
16896 2013-02-21  Hans-Peter Nilsson  <hp@axis.com>
16898         * doc/rtl.texi (vec_concat, vec_duplicate): Mention that
16899         scalars are valid operands.
16901 2013-02-21  Martin Jambor  <mjambor@suse.cz>
16903         PR tree-optimization/56310
16904         * ipa-cp.c (agg_replacements_to_vector): New parameter index, copy
16905         only matching indices and non-negative final offsets.
16906         (intersect_aggregates_with_edge): Pass src_idx to
16907         agg_replacements_to_vector.  Pass src_idx insstead of index to
16908         intersect_with_agg_replacements.
16910 2013-02-21  Martin Jambor  <mjambor@suse.cz>
16912         * ipa-cp.c (good_cloning_opportunity_p): Dump the real threshold
16913         instead of hard-wired defaults.
16915 2013-02-21  Maciej W. Rozycki  <macro@codesourcery.com>
16917         * doc/invoke.texi (MIPS Options): Update documentation of the
16918         floating-point multiply-accumulate instruction restrictions.
16920 2013-02-21  Kostya Serebryany  <kcc@google.com>
16922         * config/i386/i386.c (ix86_asan_shadow_offset): Use 0x7fff8000 as
16923         asan_shadow_offset on x86_64 linux.
16925 2013-02-21  Richard Biener  <rguenther@suse.de>
16927         PR tree-optimization/56415
16928         Revert
16929         2013-02-11  Richard Biener  <rguenther@suse.de>
16931         PR tree-optimization/56273
16932         * tree-vrp.c (simplify_cond_using_ranges): Disable for the
16933         first VRP run.
16935 2013-02-21  Jakub Jelinek  <jakub@redhat.com>
16937         PR bootstrap/56258
16938         * doc/invoke.texi (-fdump-rtl-pro_and_epilogue): Use @item
16939         instead of @itemx.
16941         PR inline-asm/56405
16942         * expr.c (expand_expr_real_1) <case TARGET_MEM_REF, MEM_REF>: Don't
16943         use movmisalign or extract_bit_field for EXPAND_MEMORY modifier.
16945 2013-02-20  Jan Hubicka  <jh@suse.cz>
16947         PR tree-optimization/56265
16948         * ipa-prop.c (ipa_make_edge_direct_to_target): Fixup callgraph
16949         when target is referenced for first time.
16951 2013-02-20  Richard Biener  <rguenther@suse.de>
16953         * tree-call-cdce.c (tree_call_cdce): Do not remove unused locals.
16954         * tree-ssa-forwprop.c (ssa_forward_propagate_and_combine): Likewise.
16955         * tree-ssa-dce.c (perform_tree_ssa_dce): Likewise.
16956         * tree-ssa-copyrename.c (copy_rename_partition_coalesce): Do
16957         not return anything.
16958         (rename_ssa_copies): Do not remove unused locals.
16959         * tree-ssa-ccp.c (do_ssa_ccp): Likewise.
16960         * tree-ssanames.c (pass_release_ssa_names): Remove unused locals first.
16961         * passes.c (execute_function_todo): Do not schedule unused locals
16962         removal if cleanup_tree_cfg did something.
16963         * tree-ssa-live.c (remove_unused_locals): Dump statistics
16964         about the number of removed locals.
16966 2013-02-20  Richard Biener  <rguenther@suse.de>
16968         PR tree-optimization/56398
16969         * tree-vect-loop-manip.c (adjust_debug_stmts): Skip SSA default defs.
16971 2013-02-20  Martin Jambor  <mjambor@suse.cz>
16973         PR tree-optimization/55334
16974         * ipa-cp.c (initialize_node_lattices): Disable IPA-CP through and to
16975         restricted pointers to arrays.
16977 2013-02-20  Richard Biener  <rguenther@suse.de>
16978             Jakub Jelinek  <jakub@redhat.com>
16980         PR tree-optimization/56396
16981         * tree-ssa-ccp.c (n_const_val): New static variable.
16982         (get_value): Return NULL for SSA names we don't have a lattice
16983         entry for.
16984         (ccp_initialize): Initialize n_const_val.
16985         * tree-ssa-copy.c (n_copy_of): New static variable.
16986         (init_copy_prop): Initialize n_copy_of.
16987         (get_value): Return NULL_TREE for SSA names we don't have a
16988         lattice entry for.
16990 2013-02-20  Martin Jambor  <mjambor@suse.cz>
16992         * ipa-cp.c (initialize_node_lattices): Fix dumping condition.
16994 2013-02-20  Richard Biener  <rguenther@suse.de>
16996         * genpreds.c (write_lookup_constraint): Do not compare first
16997         letter of the constraint again.
16999 2013-02-20  Richard Biener  <rguenther@suse.de>
17001         * tree-ssa-loop-ivopts.c (alloc_use_cost_map): Use bitmap_count_bits
17002         and ceil_log2.
17003         (get_use_iv_cost): Terminate hashtable walk when coming across
17004         an empty entry.
17006 2013-02-20  Igor Zamyatin  <igor.zamyatin@intel.com>
17008         * config/i386/i386.c (initial_ix86_tune_features): Turn on fp
17009         reassociation for avx2 targets.
17011 2012-02-19  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17013         * config/microblaze/microblaze.c: microblaze_has_clz = 0
17014         Add version check for v8.10.a to enable microblaze_has_clz
17015         * config/microblaze/microblaze.h: Add TARGET_HAS_CLZ as combined
17016         version and TARGET_PATTERN_COMPARE check
17017         * config/microblaze/microblaze.md: New clzsi2 instruction
17019 2012-02-19  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17021         * config/microblaze/microblaze.md (call_value_intern): Check symbol is
17022         function before branching.
17024 2012-02-19  Andrey Belevantsev  <abel@ispras.ru>
17026         * sel-sched-dump.c (dump_insn_rtx_flags): Explicitly set
17027         DUMP_INSN_RTX_UID.
17028         (dump_insn_rtx_1): Pass PATTERN (insn) to str_pattern_slim.
17030 2012-02-19  Andrey Belevantsev  <abel@ispras.ru>
17032         PR middle-end/55889
17033         * sel-sched.c: Include ira.h.
17034         (implicit_clobber_conflict_p): New function.
17035         (moveup_expr): Use it.
17036         * Makefile.in (sel-sched.o): Depend on ira.h.
17038 2013-02-19  Richard Biener  <rguenther@suse.de>
17040         PR tree-optimization/56384
17041         * tree-ssa-sccvn.h (struct vn_phi_s): Add type member.
17042         (vn_hash_type): Split out from ...
17043         (vn_hash_constant_with_type): ... here.
17044         * tree-ssa-sccvn.c (vn_phi_compute_hash): Use vn_hash_type.
17045         (vn_phi_eq): Compare types from vn_phi_s structure.
17046         (vn_phi_lookup): Populate vn_phi_s type.
17047         (vn_phi_insert): Likewise.
17049 2013-02-19  Jakub Jelinek  <jakub@redhat.com>
17051         PR tree-optimization/56350
17052         * tree-vect-loop.c (vectorizable_reduction): If orig_stmt, return false
17053         if haven't found reduction or nested cycle operand, rather than
17054         asserting we must find it.
17056         PR tree-optimization/56381
17057         * tree-ssa-pre.c (create_expression_by_pieces): Fix up last argument
17058         to fold_build3.
17060 2013-02-18  Aldy Hernandez  <aldyh@redhat.com>
17061             Jakub Jelinek  <jakub@redhat.com>
17063         PR target/52555
17064         * genopinit.c (raw_optab_handler): Use this_fn_optabs.
17065         (swap_optab_enable): Same.
17066         (init_all_optabs): Use argument instead of global.
17067         * tree.h (struct tree_optimization_option): New field target_optabs.
17068         * expr.h (init_all_optabs): Add argument to prototype.
17069         (TREE_OPTIMIZATION_OPTABS): New.
17070         (save_optabs_if_changed): Protoize.
17071         * optabs.h: Declare this_fn_optabs.
17072         * optabs.c (save_optabs_if_changed): New.
17073         Declare this_fn_optabs.
17074         (init_optabs): Add argument to init_all_optabs() call.
17075         * function.c (invoke_set_current_function_hook): Handle per
17076         function optabs.
17077         * function.h (struct function): New field optabs.
17078         * config/mips/mips.c (mips_set_mips16_mode): Handle when
17079         optimization_current_node has changed.
17080         * target-globals.h (save_target_globals_default_opts): Protoize.
17081         * target-globals.c (save_target_globals_default_opts): New.
17083 2013-02-18  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
17085         PR target/56347
17086         * config/pa/pa.c (pa_conditional_register_usage): On HP-UX, mark
17087         registers %fr12 and %fr12R as call used.
17089         PR target/56214
17090         * config/pa/predicates.md (base14_operand): Except for BLKmode, QImode
17091         and HImode, require all displacements to be an integer multiple of
17092         their mode size.
17093         * config/pa/pa.c (pa_legitimate_address_p): For REG+BASE addresses,
17094         only allow QImode and HImode when reload is in progress and strict is
17095         true.  Likewise for symbolic addresses.  Use base14_operand to check
17096         displacements in REG+BASE addresses.
17098 2013-02-18  Richard Biener  <rguenther@suse.de>
17100         PR tree-optimization/56366
17101         * tree-vect-loop.c (get_initial_def_for_induction): Properly
17102         handle sign-conversion of outer-loop initial induction value.
17104 2013-02-18  Richard Biener  <rguenther@suse.de>
17106         PR middle-end/56349
17107         * cfghooks.c (merge_blocks): If we merge a latch into another
17108         block adjust references to it.
17109         * cfgloop.c (flow_loops_find): Reset latch before recomputing it.
17110         (verify_loop_structure): Verify that a recorded latch is in fact
17111         a latch.
17113 2013-02-18  Richard Biener  <rguenther@suse.de>
17115         PR tree-optimization/56321
17116         * tree-ssa-reassoc.c (propagate_op_to_single_use): Properly
17117         order SSA name release and virtual operand unlinking.
17119 2013-02-17  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17121         * config/microblaze/microblaze.md (save_stack_block): Define.
17122         (restore_stack_block): Likewise.
17124 2013-02-16  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17126         * config/microblaze/linux.h (TARGET_SUPPORTS_PIC): Define as 1.
17127         * config/microblaze/microblaze.h (TARGET_SUPPORTS_PIC): Define as 1.
17128         * config/microblaze/microblaze.c (microblaze_option_override):
17129         Bail out early for PIC modes when target does not support PIC.
17131 2013-02-16  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17133         * config/microblaze/microblaze.c (microblaze_asm_trampoline_template):
17134         Replace with a microblaze version.
17135         (microblaze_trampoline_init): Adapt for microblaze.
17136         * config/microblaze/microblaze.h (TRAMPOLINE_SIZE): Adapt for
17137         microblaze.
17139 2013-02-16  Jakub Jelinek  <jakub@redhat.com>
17140             Dodji Seketeli  <dodji@redhat.com>
17142         PR asan/56330
17143         * asan.c (get_mem_refs_of_builtin_call): White space and style cleanup.
17144         (instrument_mem_region_access): Do not forget to always put
17145         instrumentation of the of 'base' and 'base + len' in a "if (len !=
17146         0) statement, even for cases where either 'base' or 'base + len'
17147         are not instrumented -- because they have been previously
17148         instrumented.  Simplify the logic by putting all the statements
17149         instrument 'base + len' inside a sequence, and then insert that
17150         sequence right before the current insertion point.  Then, to
17151         instrument 'base + len', just get an iterator on that statement.
17152         And do not forget to update the pointer to iterator the function
17153         received as argument.
17155 2013-02-15  Vladimir Makarov  <vmakarov@redhat.com>
17157         PR rtl-optimization/56348
17158         * lra-assigns.c (reload_pseudo_compare_func): Prefer bigger pseudos.
17160 2013-02-15  Steven Bosscher  <steven@gcc.gnu.org>
17162         * graph.c (start_graph_dump): Print dumpfile base as digraph label.
17163         (clean_graph_dump_file): Pass base to start_graph_dump.
17165 2013-02-14  Richard Henderson  <rth@redhat.com>
17167         PR target/55941
17168         * lower-subreg.c (simple_move): Check dest mode instead of src mode.
17170 2013-02-14  Steven Bosscher  <steven@gcc.gnu.org>
17172         * collect2-aix.h: Define F_LOADONLY.
17174 2013-02-14  Richard Biener  <rguenther@suse.de>
17176         PR lto/50494
17177         * varasm.c (output_constant_def_1): Get the decl representing
17178         the constant as argument.
17179         (output_constant_def): Wrap output_constant_def_1.
17180         (make_decl_rtl): Use output_constant_def_1 with the decl
17181         representing the constant.
17182         (build_constant_desc): Optionally re-use a decl already
17183         representing the constant.
17184         (tree_output_constant_def): Adjust.
17186 2013-02-14  Dodji Seketeli  <dodji@redhat.com>
17188         Fix an asan crash
17189         * asan.c (instrument_builtin_call):  Really put the length of the
17190         second source argument into src1_len.
17192 2013-02-13  Jakub Jelinek  <jakub@redhat.com>
17194         * asan.c (create_cond_insert_point): Add create_then_fallthru_edge
17195         argument.  If it is false, don't create edge from then_bb to
17196         fallthru_bb.
17197         (insert_if_then_before_iter): Pass true to it.
17198         (build_check_stmt): Pass false to it.
17199         (transform_statements): Flush hash table only on extended basic
17200         block boundaries, rather than at the beginning of every bb.
17201         Don't flush hash table on nonfreeing_call_p calls.
17202         * tree-flow.h (nonfreeing_call_p): New prototype.
17203         * tree-ssa-phiopt.c (nonfreeing_call_p): No longer static.
17205 2013-02-13  David S. Miller  <davem@davemloft.net>
17207         * expmed.c (expand_shift_1): Only strip scalar integer subregs.
17209 2013-02-13  Vladimir Makarov  <vmakarov@redhat.com>
17211         PR target/56184
17212         * ira.c (max_regno_before_ira): Move from ...
17213         (ira): ... here.
17214         (fix_reg_equiv_init): Use max_regno_before_ira instead of
17215         vec_safe_length.
17217 2013-02-13  Jakub Jelinek  <jakub@redhat.com>
17219         * config/i386/i386.c (ix86_asan_shadow_offset): Revert last change.
17221 2013-02-13  Richard Biener  <rguenther@suse.de>
17223         PR lto/56295
17224         * gimple-streamer-out.c (output_gimple_stmt): Undo wrapping
17225         globals in MEM_REFs.
17227 2013-02-13  Richard Biener  <rguenther@suse.de>
17229         * loop-init.c (loop_optimizer_init): Clear loop state when
17230         re-initializing preserved loops.
17231         * loop-unswitch.c (unswitch_single_loop): Return whether
17232         we unswitched the loop.  Do not verify loop state here.
17233         (unswitch_loops): When we unswitched a loop discover new loops.
17235 2013-02-13  Kostya Serebryany  <kcc@google.com>
17237         * config/i386/i386.c: Use 0x7fff8000 as asan_shadow_offset
17238         on x86_64 linux.
17239         * sanitizer.def: Rename __asan_init to __asan_init_v1.
17241 2013-02-12  Dodji Seketeli  <dodji@redhat.com>
17243         Avoid instrumenting duplicated memory access in the same basic block
17244         * Makefile.in (asan.o): Add new dependency on hash-table.h
17245         * asan.c (struct asan_mem_ref, struct mem_ref_hasher): New types.
17246         (asan_mem_ref_init, asan_mem_ref_get_end, get_mem_ref_hash_table)
17247         (has_stmt_been_instrumented_p, empty_mem_ref_hash_table)
17248         (free_mem_ref_resources, has_mem_ref_been_instrumented)
17249         (has_stmt_been_instrumented_p, update_mem_ref_hash_table)
17250         (get_mem_ref_of_assignment): New functions.
17251         (get_mem_refs_of_builtin_call): Extract from
17252         instrument_builtin_call and tweak a little bit to make it fit with
17253         the new signature.
17254         (instrument_builtin_call): Use the new
17255         get_mem_refs_of_builtin_call.  Use gimple_call_builtin_p instead
17256         of is_gimple_builtin_call.
17257         (instrument_derefs, instrument_mem_region_access): Insert the
17258         instrumented memory reference into the hash table.
17259         (maybe_instrument_assignment): Renamed instrument_assignment into
17260         this, and change it to advance the iterator when instrumentation
17261         actually happened and return true in that case.  This makes it
17262         homogeneous with maybe_instrument_assignment, and thus give a
17263         chance to callers to be more 'regular'.
17264         (transform_statements): Clear the memory reference hash table
17265         whenever we enter a new BB, when we cross a function call, or when
17266         we are done transforming statements.  Use
17267         maybe_instrument_assignment instead of instrumentation.  No more
17268         need to special case maybe_instrument_assignment and advance the
17269         iterator after calling it; it's now handled just like
17270         maybe_instrument_call.  Update comment.
17272 2013-02-13  Richard Biener  <rguenther@suse.de>
17274         * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc):
17275         Fix loop discovery code.
17277 2013-02-12  Vladimir Makarov  <vmakarov@redhat.com>
17279         PR inline-asm/56148
17280         * lra-constraints.c (process_alt_operands): Match early clobber
17281         operand with itself.  Check conflicts with earlyclobber only if
17282         the operand is not reloaded.  Prefer to reload conflicting operand
17283         if earlyclobber and matching operands are the same.
17285 2013-02-12  Richard Biener  <rguenther@suse.de>
17287         PR lto/56297
17288         * lto-streamer-out.c (write_symbol): Do not output symbols
17289         for hard register variables.
17291 2013-02-12  Georg-Johann Lay  <avr@gjlay.de>
17293         PR target/54222
17294         * config/avr/avr-dimode.md (umulsidi3, mulsidi3): New expanders.
17295         (umulsidi3_insn, mulsidi3_insn): New insns.
17297 2013-02-12  Christophe Lyon  <christophe.lyon@linaro.org>
17299         * config/arm/arm-protos.h (struct cpu_vec_costs): New struct type.
17300         (struct tune_params): Add vec_costs field.
17301         * config/arm/arm.c (arm_builtin_vectorization_cost)
17302         (arm_add_stmt_cost): New functions.
17303         (TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST)
17304         (TARGET_VECTORIZE_ADD_STMT_COST): Define.
17305         (arm_default_vec_cost): New struct of type cpu_vec_costs.
17306         (arm_slowmul_tune, arm_fastmul_tune, arm_strongarm_tune)
17307         (arm_xscale_tune, arm_9e_tune, arm_v6t2_tune, arm_cortex_tune)
17308         (arm_cortex_a15_tune, arm_cortex_a5_tune, arm_cortex_a9_tune)
17309         (arm_v6m_tune, arm_fa726te_tune): Define new vec_costs field.
17311 2013-02-12  Richard Biener  <rguenther@suse.de>
17313         PR lto/56295
17314         * gimple-streamer-in.c (input_gimple_stmt): Strip MEM_REFs off
17315         decls again if possible.
17317 2013-02-12  Richard Biener  <rguenther@suse.de>
17319         PR middle-end/56288
17320         * tree-ssa.c (verify_ssa_name): Fix check, move
17321         SSA_NAME_IN_FREE_LIST check up.
17323 2013-02-12  Jakub Jelinek  <jakub@redhat.com>
17324             Steven Bosscher   <steven@gcc.gnu.org>
17326         PR rtl-optimization/56151
17327         * optabs.c (add_equal_note): Don't return 0 if target is a MEM,
17328         equal to op0 or op1, and last_insn pattern is CODE operation
17329         with MEM dest and one of the operands matches that MEM.
17331 2013-02-11  Sriraman Tallam  <tmsriramgoogle.com>
17333         * doc/extend.texi: Document Function Multiversioning and "default"
17334         parameter string to target attribute.
17335         * config/i386/i386.c (get_builtin_code_for_version): Return 0 if
17336         target attribute parameter is "default".
17337         (ix86_compare_version_priority): Remove checks for target attribute.
17338         (ix86_mangle_function_version_assembler_name): Change error to sorry.
17339         Remove check for target attribute equal to NULL. Add assert.
17340         (ix86_generate_version_dispatcher_body): Change error to sorry.
17342 2013-02-11  Iain Sandoe  <iain@codesourcery.com>
17343             Jack Howarth  <howarth@bromo.med.uc.edu>
17344             Patrick Marlier  <patrick.marlier@gmail.com>
17346         PR libitm/55693
17347         * config/darwin.h: Replace ENDFILE_SPEC with TM_DESTRUCTOR and
17348         define ENDFILE_SPEC as TM_DESTRUCTOR.
17349         * config/i386/darwin.h (ENDFILE_SPEC): Use TM_DESTRUCTOR.
17351 2013-02-11  Alexander Potapenko  <glider@google.com>
17352             Jack Howarth  <howarth@bromo.med.uc.edu>
17353             Jakub Jelinek  <jakub@redhat.com>
17355         PR sanitizer/55617
17356         * config/darwin.c (cdtor_record): Rename ctor_record.
17357         (sort_cdtor_records): Rename sort_ctor_records.
17358         (finalize_dtors): New routine to sort destructors by
17359         priority before use in assemble_integer.
17360         (machopic_asm_out_destructor): Use finalize_dtors if needed.
17362 2013-02-11  Uros Bizjak  <ubizjak@gmail.com>
17364         PR rtl-optimization/56275
17365         * simplify-rtx.c (avoid_constant_pool_reference): Check that
17366         offset is non-negative and less than cmode size before
17367         calling simplify_subreg.
17369 2013-02-11  Richard Biener  <rguenther@suse.de>
17371         PR tree-optimization/56264
17372         * cfgloop.h (fix_loop_structure): Adjust prototype.
17373         * loop-init.c (fix_loop_structure): Return the number of
17374         newly discovered loops.
17375         * tree-cfgcleanup.c (repair_loop_structures): When new loops
17376         are discovered, do a full loop-closed SSA rewrite.
17378 2013-02-11  Richard Biener  <rguenther@suse.de>
17380         PR tree-optimization/56273
17381         * tree-vrp.c (simplify_cond_using_ranges): Disable for the
17382         first VRP run.
17383         (check_array_ref): Fix missing newline in dumps.
17384         (search_for_addr_array): Likewise.
17386 2013-02-09  David Edelsohn  <dje.gcc@gmail.com>
17388         * config/rs6000/aix61.h (OS_MISSING_ALTIVEC): Undefine.
17390 2013-02-09  Jakub Jelinek  <jakub@redhat.com>
17392         PR target/56256
17393         * config/rs6000/rs6000.h (ASSEMBLER_DIALECT): Define.
17395 2013-02-08  Vladimir Makarov  <vmakarov@redhat.com>
17397         PR rtl-optimization/56246
17398         * lra-constraints.c (simplify_operand_subreg): Try to reuse
17399         reload pseudo.
17400         * lra.c (lra): Clear lra_optional_reload_pseudos only when all
17401         constraints are satisfied.
17403 2013-02-08  Jeff Law  <law@redhat.com>
17405         PR debug/53948
17406         * emit-rtl.c (reg_is_parm_p): New function.
17407         * regs.h (reg_is_parm_p): New prototype.
17408         * ira-conflicts.c (ira_build_conflicts): Allow parameters in
17409         callee-clobbered registers.
17411 2013-02-08  Michael Meissner  <meissner@linux.vnet.ibm.com>
17413         PR target/56043
17414         * config/rs6000/rs6000.c (rs6000_builtin_vectorized_libmass):
17415         If there is no implicit builtin declaration, just return NULL.
17417 2013-02-08  Uros Bizjak  <ubizjak@gmail.com>
17419         * config/i386/sse.md (FMAMODEM): New mode iterator.
17420         (fma<mode>4, fms<mode>4, fnma<mode>4, fnms<mode>4): Use FMAMODEM
17421         mode iterator. Do not use TARGET_SSE_MATH in insn constraint.
17423 2013-02-08  Uros Bizjak  <ubizjak@gmail.com>
17425         * config/i386/gnu-user.h (TARGET_CAN_SPLIT_STACK): Define only
17426         when HAVE_GAS_CFI_PERSONALITY_DIRECTIVE is set.
17427         * config/i386/gnu-user64.h (TARGET_CAN_SPLIT_STACK): Ditto.
17429 2013-02-08  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
17431         * config.gcc (microblaze*-linux*): Add TARGET_BIG_ENDIAN_DEFAULT.
17432         (microblaze*-*-elf): Likewise.
17433         * config/microblaze/linux.h: Add -mbig-endian / -mlittle-endian to
17434         LINK_SPEC.
17435         * config/microblaze/microblaze-c.c: Add builtin defines for
17436         _LITTLE_ENDIAN and _BIG_ENDIAN.
17437         * config/microblaze/microblaze.h: Add TARGET_ENDIAN_DEFAULT and
17438         add to TARGET_DEFAULT flags.
17439         Expand ASM_SPEC and LINK_SPEC.
17440         Update BYTES_BIG_ENDIAN and WORDS_BIG_ENDIAN.
17441         * config/microblaze/microblaze.md: Update extendsidi2 and
17442         movdi_internal instructions to use low-order / high-order reg
17443         print_operands.
17444         * config/microblaze/microblaze.opt: Add mbig-endian and mlittle-endian
17445         options and inversemask / mask of LITTLE_ENDIAN.
17446         * config/microblaze/t-microblaze: Expand multilib options to
17447         include mlittle-endian (le) and update exceptions patterns.
17449 2013-02-08  Jakub Jelinek  <jakub@redhat.com>
17451         PR rtl-optimization/56195
17452         * lra-constraints.c (get_reload_reg): Don't reuse regs
17453         if they have smaller mode than requested, if they have
17454         wider mode than requested, try to return a SUBREG.
17456         PR tree-optimization/56250
17457         * fold-const.c (extract_muldiv_1) <case NEGATE_EXPR>: Don't optimize
17458         if type is unsigned and code isn't MULT_EXPR.
17460 2013-02-08  Georg-Johann Lay  <avr@gjlay.de>
17462         PR tree-optimization/56064
17463         * fixed-value.c (fixed_from_double_int): Sign/zero extend payload
17464         bits according to mode.
17465         * fixed-value.h (fixed_from_double_int)
17466         (const_fixed_from_double_int): Adjust comments.
17468 2013-02-08  Richard Biener  <rguenther@suse.de>
17470         PR lto/56231
17471         * lto-streamer.h (struct data_in): Remove current_file, current_line
17472         and current_col members.
17473         * lto-streamer-out.c (lto_output_location): Stream changed bits
17474         en-block for efficiency.
17475         * lto-streamer-in.c (clear_line_info): Remove.
17476         (lto_input_location): Cache current file, line and column
17477         globally via local statics.  Read changed bits en-block.
17478         (input_function): Do not call clear_line_info.
17479         (lto_read_body): Likewise.
17480         (lto_input_toplevel_asms): Likewise.
17482 2013-02-08  Michael Matz  <matz@suse.de>
17484         PR tree-optimization/52448
17485         * tree-ssa-phiopt.c (struct name_to_bb): Add phase member.
17486         (nt_call_phase): New static.
17487         (add_or_mark_expr): Only mark accesses with newer phase than any
17488         call seen.
17489         (nonfreeing_call_p): New.
17490         (nt_init_block): Update nt_call_phase, mark blocks as visited.
17491         (nt_fini_block): Keep blocks marked as visited.
17492         (get_non_trapping): Initialize nt_call_phase, and reset aux pointer.
17494 2013-02-08  Richard Biener  <rguenther@suse.de>
17496         * ira.c (ira): Free broken dominator information.
17498 2013-02-08  Uros Bizjak  <ubizjak@gmail.com>
17500         * config/i386/i386.c (ix86_spill_class): Use INTEGER_CLASS_P macro.
17502 2013-02-08  Marek Polacek  <polacek@redhat.com>
17504         * cfgloop.c (verify_loop_structure): Add more checking of headers.
17506 2013-02-08  Richard Biener  <rguenther@suse.de>
17508         PR middle-end/56181
17509         * cfgloop.h (flow_loops_find): Adjust.
17510         (bb_loop_header_p): Declare.
17511         * cfgloop.c (bb_loop_header_p): New function split out from ...
17512         (flow_loops_find): ... here.  Adjust function signature,
17513         support incremental loop structure update.
17514         (verify_loop_structure): Cleanup.  Verify a loop is a loop.
17515         * cfgloopmanip.c (fix_loop_structure): Move ...
17516         * loop-init.c (fix_loop_structure): ... here.
17517         (apply_loop_flags): Split out from ...
17518         (loop_optimizer_init): ... here.
17519         (fix_loop_structure): Use apply_loop_flags.  Use flow_loops_find
17520         in incremental mode, only remove dead loops here.
17522 2013-02-08  Georg-Johann Lay  <avr@gjlay.de>
17524         PR target/54222
17525         * config/avr/avr.md (unspec) <UNSPEC_ROUND>: Add.
17526         * config/avr/avr-fixed.md (ALL4QA, ALL124QA): New mode iterators.
17527         (round<mode>3, round<mode>3_const): New expanders for fixed-mode.
17528         (*round<mode>3.libgcc): New insns for fixed-modes.
17529         * config/avr/builtins.def (ABSxx): Use a non-NULL LIBNAME.
17530         (ROUNDxx, COUNTLSxx, BITSxx, xxBITS): New DEF_BUILTINs.
17531         (ROUNDFX, COUNTLSFX, ABSFX): New DEF_BUILTINs.
17532         * config/avr/stdfix.h (absFX, bitsFX, FXbits): Remove inline
17533         implementations.  Define to __builtin_avr_absFX,
17534         __builtin_avr_bitsFX, __builtin_avr_FXbits, respectively.
17535         (roundFX, countlsFX): Define to __builtin_avr_roundFX,
17536         __builtin_avr_countlsFX, respectively.
17537         * config/avr/avr-c.c (target.h): Include it.
17538         (enum avr_builtin_id): New enum.
17539         (avr_resolve_overloaded_builtin): New static function.
17540         (avr_register_target_pragmas): Use it to set
17541         targetm.resolve_overloaded_builtin.
17542         * config/avr/avr.c (avr_init_builtins): Supply myriads of local
17543         tree nodes used by DEF_BUILTIN.
17544         (avr_expand_builtin) <AVR_BUILTIN_ROUNDxx>: Sanity-check them.
17545         (avr_fold_builtin) <AVR_BUILTIN_BITSxx>: Fold to VIEW_COVERT_EXPR.
17546         <AVR_BUILTIN_xxBITS>: Same.
17548 2013-02-08  Richard Biener  <rguenther@suse.de>
17550         * cfgloop.c (verify_loop_structure): Properly handle
17551         a loop exiting to another loop header.
17552         * ira-int.h (ira_loops): Remove.
17553         * ira.c (ira_loops): Remove.
17554         (ira): Use loop_optimizer_init and loop_optimizer_finalize.
17555         (do_reload): Use loop_optimizer_finalize.
17556         * ira-build.c (create_loop_tree_nodes): Use get_loops and
17557         number_of_loops to access the loop tree.
17558         (more_one_region_p): Likewise.
17559         (finish_loop_tree_nodes): Likewise.
17560         (rebuild_regno_allocno_maps): Likewise.
17561         (mark_loops_for_removal): Likewise.
17562         (mark_all_loops_for_removal): Likewise.
17563         (remove_unnecessary_regions): Likewise.
17564         (ira_build): Likewise.
17565         * ira-emit.c (setup_entered_from_non_parent_p): Likewise.
17567 2013-02-08  Richard Biener  <rguenther@suse.de>
17569         * Makefile.in (tree-tailcall.o): Add $(CFGLOOP_H) dependency.
17570         * ipa-pure-const.c (analyze_function): Avoid calling
17571         mark_irreducible_loops twice.
17572         * tree-tailcall.c (tree_optimize_tail_calls_1): Mark loops for fixup.
17574 2013-02-07  David S. Miller  <davem@davemloft.net>
17576         * dwarf2out.c (based_loc_descr): Perform leaf register remapping
17577         on 'reg'.
17578         * var-tracking.c (vt_add_function_parameter): Test the presence of
17579         HAVE_window_save properly and do not remap argument registers when
17580         we have a leaf function.
17582 2013-02-07  Uros Bizjak  <ubizjak@gmail.com>
17584         PR bootstrap/56227
17585         * ggc-page.c (ggc_print_statistics): Use HOST_LONG_LONG_FORMAT
17586         instead of "ll".
17587         * config/i386/i386.c (ix86_print_operand): Ditto.
17589 2013-02-07  Vladimir Makarov  <vmakarov@redhat.com>
17591         * lra-constraints.c (process_alt_operands): Fix recently added comment.
17593 2013-02-07  Vladimir Makarov  <vmakarov@redhat.com>
17595         PR rtl-optimization/56225
17596         * lra-constraints.c (process_alt_operands): Check that reload hard
17597         reg can hold value for strict_low_part.
17599 2013-02-07  Jakub Jelinek  <jakub@redhat.com>
17601         PR debug/56154
17602         * dwarf2out.c (dwarf2_debug_hooks): Set end_function hook to
17603         dwarf2out_end_function.
17604         (in_first_function_p, maybe_at_text_label_p,
17605         first_loclabel_num_not_at_text_label): New variables.
17606         (dwarf2out_var_location): In the first function find out
17607         lowest loclabel_num N where .LVLN is known not to be equal to .Ltext0.
17608         (find_empty_loc_ranges_at_text_label, dwarf2out_end_function): New
17609         functions.
17611 2013-02-07  Eric Botcazou  <ebotcazou@adacore.com>
17613         PR rtl-optimization/56178
17614         * cse.c (cse_insn): Do not create a REG_EQUAL note if the source is a
17615         SUBREG of a register.  Tidy up related block of code.
17616         * fwprop.c (forward_propagate_and_simplify): Do not create a REG_EQUAL
17617         note if the source is a register or a SUBREG of a register.
17619 2013-02-07  Jakub Jelinek  <jakub@redhat.com>
17621         PR target/56228
17622         * config/rs6000/rs6000.md (ptrm): New mode attr.
17623         (call_indirect_aix<ptrsize>, call_indirect_aix<ptrsize>_nor11,
17624         call_value_indirect_aix<pttrsize>,
17625         call_value_indirect_aix<pttrsize>_nor11): Use <ptrm> instead of
17626         m in constraints.
17628 2013-02-07  Michael Haubenwallner  <michael.haubenwallner@salomon.at>
17630         * collect2.c (main): Set aix64_flag for -G and -bsvr4 too, disable
17631         if -bnortl. Convert to strcmp and strncmp.
17633 2013-02-07  Alan Modra  <amodra@gmail.com>
17635         PR target/54009
17636         * config/rs6000/rs6000.c (mem_operand_gpr): Check that LO_SUM
17637         addresses won't wrap when offsetting.
17638         (rs6000_secondary_reload): Provide secondary reloads needed for
17639         wrapping LO_SUM addresses.
17641 2013-02-06  Thomas Schwinge  <thomas@codesourcery.com>
17643         * config/gnu.h (GNU_USER_TARGET_OS_CPP_BUILTINS): Never define
17644         MACH, just __MACH__.
17646 2013-02-06  Richard Biener  <rguenther@suse.de>
17648         * tracer.c (tracer): Mark loops with LOOPS_NEED_FIXUP
17649         instead of calling fix_loop_structure.
17651 2013-02-06  Jakub Jelinek  <jakub@redhat.com>
17653         PR middle-end/56217
17654         * omp-low.c (use_pointer_for_field): Return false if
17655         lower_send_shared_vars doesn't generate any copy-out code.
17657 2013-02-06  Tom de Vries  <tom@codesourcery.com>
17659         PR rtl-optimization/56131
17660         * cfgrtl.c (delete_insn): Use NOTE_BASIC_BLOCK instead of BLOCK_FOR_INSN
17661         to get the bb of a NOTE_INSN_BASIC_BLOCK.  Handle the case that the bb
17662         of the label is NULL.  Add comment.
17664 2013-02-05  Jakub Jelinek  <jakub@redhat.com>
17666         * tree.h (struct tree_decl_with_vis): Remove thread_local field.
17668         PR sanitizer/55374
17669         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Define.
17670         (STATIC_LIBTSAN_LIBS): Likewise.
17671         * gcc.c (ADD_STATIC_LIBTSAN_LIBS, LIBTSAN_EARLY_SPEC): Define.
17672         (LIBTSAN_SPEC): Add ADD_STATIC_LIBTSAN_LIBS, if LIBTSAN_EARLY_SPEC
17673         is defined, don't add anything else beyond that.
17674         (SANITIZER_EARLY_SPEC, SANITIZER_SPEC): Define.
17675         (LINK_COMMAND_SPEC): Use them.
17677         PR tree-optimization/56205
17678         * tree-stdarg.c (check_all_va_list_escapes): Return true if
17679         there are any PHI nodes that set non-va_list_escape_vars SSA_NAME
17680         and some va_list_escape_vars SSA_NAME appears in some PHI argument.
17682 2013-02-05  Richard Biener  <rguenther@suse.de>
17684         PR tree-optimization/53342
17685         PR tree-optimization/53185
17686         * tree-vectorizer.h (vect_check_strided_load): Remove.
17687         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Do
17688         not disallow peeling for vectorized strided loads.
17689         (vect_check_strided_load): Make static and simplify.
17690         (vect_analyze_data_refs): Adjust.
17691         * tree-vect-stmts.c (vectorizable_load): Handle peeled loops
17692         correctly when vectorizing strided loads.
17694 2013-02-05  Richard Biener  <rguenther@suse.de>
17696         * doc/install.texi: Refer to ISL, not PPL.
17698 2013-02-05  Jan Hubicka  <jh@suse.cz>
17700         PR tree-optimization/55789
17701         * params.def (PARAM_EARLY_INLINER_MAX_ITERATIONS): Drop to 1.
17703 2013-02-05  Jan Hubicka  <jh@suse.cz>
17705         PR tree-optimization/55789
17706         * cgraphclones.c (cgraph_remove_node_and_inline_clones): Remove
17707         the dead call anyway.
17709 2013-02-05  Eric Botcazou  <ebotcazou@adacore.com>
17711         PR sanitizer/55374
17712         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Add missing guard.
17714 2013-02-04  Alexander Potapenko  <glider@google.com>
17715             Jack Howarth  <howarth@bromo.med.uc.edu>
17716             Jakub Jelinek  <jakub@redhat.com>
17718         PR sanitizer/55617
17719         * config/darwin.c (sort_ctor_records): Stabilized qsort
17720         on constructor priority by using original position.
17721         (finalize_ctors): New routine to sort constructors by
17722         priority before use in assemble_integer.
17723         (machopic_asm_out_constructor): Use finalize_ctors if needed.
17725 2013-02-04  Jakub Jelinek  <jakub@redhat.com>
17727         PR libstdc++/54314
17728         * config/i386/winnt.c (i386_pe_assemble_visibility): Don't warn
17729         about visibility on artificial decls.
17730         * config/sol2.c (solaris_assemble_visibility): Likewise.
17732 2013-02-04  Kai Tietz  <ktietz@redhat.com>
17734         PR target/56186
17735         * config/i386/i386.c (function_value_ms_64): Add additional valtype
17736         argument and improve checking of return-argument types for 16-byte
17737         modes.
17738         (ix86_function_value_1): Add additional valtype argument on call
17739         of function_value_64.
17740         (return_in_memory_ms_64): Sync 16-byte sized mode handling with
17741         handling infunction_value_64 function.
17743 2013-02-04  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
17745         * reload.c (subst_reloads): Fix DEBUG_RELOAD build issue.
17747 2013-02-04  Richard Biener  <rguenther@suse.de>
17749         PR tree-optimization/56188
17750         * tree-ssa-structalias.c (label_visit): Consider case with
17751         initially non-empty points-to set.
17752         (perform_var_substitution): Dump node mapping and clean up.
17754 2013-02-04  Richard Guenther  <rguenther@suse.de>
17756         PR lto/56168
17757         * lto-symtab.c (lto_symtab_merge_decls_1): Make non-builtin
17758         node prevail as last resort.
17759         (lto_symtab_merge_decls): Remove guard on LTRANS here.
17760         (lto_symtab_prevailing_decl): Builtins are their own prevailing decl.
17762 2013-02-04  Richard Biener  <rguenther@suse.de>
17764         PR tree-optimization/56113
17765         * tree-ssa-structalias.c (equiv_class_lookup, equiv_class_add):
17766         Merge into ...
17767         (equiv_class_lookup_or_add): ... this.
17768         (label_visit): Adjust and fix error in previous patch.
17769         (perform_var_substitution): Adjust.
17771 2013-02-03  Oleg Endo  <olegendo@gcc.gnu.org>
17773         * config/sh/divtab.c: Fix formatting and comments throughout the file.
17774         * config/sh/sh4-300.md: Likewise.
17775         * config/sh/sh4a.md: Likewise.
17776         * config/sh/constraints.md: Likewise.
17777         * config/sh/sh.md: Likewise.
17778         * config/sh/netbsd-elf.h: Likewise.
17779         * config/sh/predicates.md: Likewise.
17780         * config/sh/sh-protos.h: Likewise.
17781         * config/sh/ushmedia.h: Likewise.
17782         * config/sh/linux.h: Likewise.
17783         * config/sh/sh.c: Likewise.
17784         * config/sh/superh.h: Likewise.
17785         * config/sh/elf.h: Likewise.
17786         * config/sh/sh4.md: Likewise.
17787         * config/sh/sh.h: Likewise.
17789 2013-02-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
17791         * config/pa/constraints.md: Adjust unused letters.  Change "T"
17792         constraint to match_test floating_point_store_memory_operand().
17793         * config/pa/predicates.md (reg_plus_base_memory_operand): New.
17794         (base14_operand): New.
17795         (floating_point_store_memory_operand): New.
17796         (integer_store_memory_operand): Revise to use base14_operand and
17797         reg_plus_base_memory_operand.
17798         (move_dest_operand): Allow symbolic_memory_operands.
17799         (symbolic_memory_operand): Check for LO_SOM.
17800         (symbolic_operand): Change default case to break.
17801         * config/pa/pa.md: Remove unamed DFmode and SFmode patterns to force
17802         CONST_DOUBLE values to be reloaded by putting them into memory when
17803         the destination is a floating point register.
17804         (movdf): Remove code to handle CONST_DOUBLE.
17805         (movsf): Likewise.
17806         (reload_indf_r1): New.
17807         (reload_insf_r1): New.
17808         Consistently use "Q" and "T" constraints with integer and floating
17809         point move instructions, respectively.
17810         (movdi): Remove FAIL.
17811         Change predicate for source operand unamed DImode move from
17812         general_operand to move_src_operand.
17813         (umulsidi3): Change predicate for destination operand to
17814         register_operand.
17815         Likewise for similar unamed patterns.
17816         * config/pa/pa-protos.h (pa_legitimize_reload_address): Declare.
17817         * config/pa/pa.c (pa_symbolic_expression_p): Remove extra parenthesis.
17818         (hppa_legitimize_address): Simplify mask calculation.
17819         (pa_emit_move_sequence): Revised handling of secondary reloads from
17820         REG+D addresses for floating point loads and stores.  Directly handle
17821         loading CONST0_RTX (mode) to a floating point register.
17822         (pa_secondary_reload): Handle reloading DF and SFmode constant values
17823         to floating point registers.  Don't restrict secondary reloads to
17824         floating point registers to integer modes.  Revise some comments and
17825         cleanup some code.
17826         (TARGET_LEGITIMATE_ADDRESS_P): Define.
17827         (pa_legitimate_address_p): New.
17828         (pa_legitimize_reload_address): New.
17829         * config/pa/pa.h (STRICT_REG_OK_FOR_INDEX_P): New.
17830         (STRICT_REG_OK_FOR_BASE_P): New.
17831         (GO_IF_LEGITIMATE_ADDRESS): Delete.  Update some related comments.
17832         (LEGITIMIZE_RELOAD_ADDRESS): Revise to use pa_legitimize_reload_address.
17834 2013-02-03  David Edelsohn  <dje.gcc@gmail.com>
17835             Andrew Dixie  <andrewd@gentrack.com>
17837         * collect2.c (GCC_CHECK_HDR): Do not scan objects with F_LOADONLY
17838         flag set.
17840 2013-02-03  Richard Sandiford  <rdsandiford@googlemail.com>
17842         * expmed.c (extract_bit_field_1): Pass the full width of the
17843         structure to get_best_reg_extraction_insn.
17845 2013-02-01  David Edelsohn  <dje.gcc@gmail.com>
17847         PR target/54601
17848         * configure.ac (use_cxa_atexit): Add AIX.
17849         * configure: Regenerate.
17851         * config/rs6000/aix61.h (STARTFILE_SPEC): Add crtcxa.o.
17853 2013-02-01  Jakub Jelinek  <jakub@redhat.com>
17855         PR debug/54793
17856         * final.c (need_profile_function): New variable.
17857         (final_start_function): Drop ATTRIBUTE_UNUSED from first argument.
17858         If first of NOTE_INSN_BASIC_BLOCK or NOTE_INSN_FUNCTION_BEG
17859         is only preceeded by NOTE_INSN_VAR_LOCATION or NOTE_INSN_DELETED
17860         notes, targetm.asm_out.function_prologue doesn't emit anything,
17861         HAVE_prologue and profiler should be emitted before prologue,
17862         set need_profile_function instead of emitting it.
17863         (final_scan_insn): If need_profile_function, emit
17864         profile_function on the first NOTE_INSN_BASIC_BLOCK or
17865         NOTE_INSN_FUNCTION_BEG note.
17867 2013-02-01  Richard Henderson  <rth@redhat.com>
17869         * config/rs6000/rs6000.md (smulditi3): New.
17870         (umulditi3): New.
17872         * config/alpha/alpha.md (umulditi3): New.
17874 2013-02-01  David Edelsohn  <dje.gcc@gmail.com>
17876         * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_COMMON): Use floor_log2.
17877         (ASM_OUTPUT_ALIGNED_LOCAL): New.
17879 2013-02-01  Richard Biener  <rguenther@suse.de>
17881         PR tree-optimization/56113
17882         * tree-ssa-structalias.c (label_visit): Reduce work for
17883         single-predecessor nodes.
17885 2013-02-01  Eric Botcazou  <ebotcazou@adacore.com>
17887         * fold-const.c (make_range_step) <TRUTH_NOT_EXPR>: Bail out if the
17888         range isn't testing for zero.
17890 2013-01-31  Steven Bosscher  <steven@gcc.gnu.org>
17892         PR middle-end/56113
17893         * fwprop.c (fwprop_init): Set up loops without CFG modifications.
17895 2013-01-31  Hiroyuki Ono  <hiroyuki.ono.jc@renesas.com>
17896             Nick Clifton  <nickc@redhat.com>
17898         * config/v850/constraints.md (Q): Define as a memory constraint.
17899         * config/v850/predicates.md (label_ref_operand): New predicate.
17900         (e3v5_shift_operand): New predicate.
17901         (ior_operator): New predicate.
17902         * config/v850/t-v850: Add e3v5 multilib.
17903         * config/v850/v850-protos.h (v850_adjust_insn_length): Prototype.
17904         (v850_gen_movdi): Prototype.
17905         * config/v850/v850.c: Add support for e3v5 architecture.
17906         Rename all uses of TARGET_V850E || TARGET_V850E2_ALL to
17907         TARGET_V850E_UP.
17908         (construct_save_jarl): Add e3v5 long JARL support.
17909         (v850_adjust_insn_length): New function.  Adjust length of call
17910         insns when using e3v5 instructions.
17911         (v850_gen_movdi): New function: Generate instructions to move a
17912         DImode value.
17913         * config/v850/v850.h (TARGET_CPU_v850e3v5): Define.
17914         (CPP_SPEC): Define __v850e3v5__ as appropriate.
17915         (TARGET_USE_FPU): Enable for e3v5.
17916         (CONST_OK_FOR_W): New macro.
17917         (ADJUST_INSN_LENGTH): Define.
17918         * config/v850/v850.md (UNSPEC_LOOP): Define.
17919         (attr cpu): Add v850e3v5.
17920         Rename all uses of TARGET_V850E2 to TARGET_V850E2V3_UP.
17921         (movdi): New pattern.
17922         (movdi_internal): New pattern.
17923         (cbranchsf4): Conditionalize on TARGET_USE_FPU.
17924         (cbranchdf4): Conditionalize on TARGET_USE_FPU.
17925         (cstoresf4): Likewise.
17926         (cstoredf4): Likewise.
17927         (insv): New pattern.
17928         (rotlso3_a): New pattern.
17929         (rotlsi3_b): New pattern
17930         (rotlsi3_v850e3v5): New pattern.
17931         (doloop_begin): New pattern.
17932         (fix_loop_counter): New pattern.
17933         (doloop_end): New pattern.
17934         (branch_normal): Add e3v5 long branch support.
17935         (branch_invert): Likewise.
17936         (branch_z_normal): Likewise.
17937         (branch_z_invert): Likewise.
17938         (branch_nz_normal): Likewise.
17939         (branch_nz_invert): Likewise.
17940         (call_internal_short): Add e3v5 register-indirect JARL support.
17941         (call_internal_long): Likewise.
17942         (call_value_internal_short): Likewise.
17943         (call_value_internal_long): Likewise.
17944         * config/v850/v850.opt (mv850e3v5, mv850e2v4): New options.
17945         (mloop): New option.
17946         * config.gcc: Add support for configuring v840e3v5 target.
17947         * doc/invoke.texi: Document new v850 specific command line options.
17949 2013-01-31  Paul Koning  <ni1d@arrl.net>
17951         PR debug/55059
17952         PR debug/54508
17953         * dwarf2out.c (prune_unused_types_mark): Mark all of parent's
17954         children if parent is a class.
17955         (prune_unused_types_prune): Don't add DW_AT_declaration.
17957 2013-01-31  Richard Biener  <rguenther@suse.de>
17959         PR tree-optimization/56157
17960         * tree-vect-slp.c (vect_get_slp_defs): More thoroughly try to
17961         match up operand with SLP child.
17963 2013-01-31  Jason Merrill  <jason@redhat.com>
17965         PR debug/54410
17966         * dwarf2out.c (gen_struct_or_union_type_die): Always schedule template
17967         parameters the first time.
17968         (gen_scheduled_generic_parms_dies): Check completeness here.
17970 2013-01-31  Richard Biener  <rguenther@suse.de>
17972         PR middle-end/53073
17973         * common.opt (faggressive-loop-optimizations): New flag,
17974         enabled by default.
17975         * doc/invoke.texi (faggressive-loop-optimizations): Document.
17976         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Guard
17977         infer_loop_bounds_from_undefined by it.
17979 2013-01-31  Richard Biener  <rguenther@suse.de>
17981         PR tree-optimization/56150
17982         * tree-ssa-loop-manip.c (find_uses_to_rename_stmt): Do not
17983         visit virtual operands.
17984         (find_uses_to_rename_bb): Likewise.
17986 2013-01-31  Richard Biener  <rguenther@suse.de>
17988         PR tree-optimization/56150
17989         * tree-ssa-tail-merge.c (gimple_equal_p): Properly handle
17990         mixed store non-store stmts.
17992 2013-01-30  Jakub Jelinek  <jakub@redhat.com>
17994         PR sanitizer/55374
17995         * gcc.c (LIBASAN_SPEC): Define just to ADD_STATIC_LIBASAN_LIBS if
17996         LIBASAN_EARLY_SPEC is defined.
17997         (LIBASAN_EARLY_SPEC): Define to empty string if not already defined.
17998         (LINK_COMMAND_SPEC): Add LIBASAN_EARLY_SPEC for -fsanitize=address,
17999         before %o.
18000         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Define.
18002         PR c++/55742
18003         * config/i386/i386.c (ix86_valid_target_attribute_inner_p): Diagnose
18004         invalid args instead of ICEing on it.
18005         (ix86_valid_target_attribute_tree): Return error_mark_node if
18006         ix86_valid_target_attribute_inner_p failed.
18007         (ix86_valid_target_attribute_p): Return false only if
18008         ix86_valid_target_attribute_tree returned error_mark_node.  Allow
18009         target("default") attribute.
18010         (sorted_attr_string): Change argument from const char * to tree,
18011         merge in all target attribute arguments rather than just one.
18012         Formatting fix.  Use XNEWVEC instead of xmalloc and XDELETEVEC
18013         instead of free.  Avoid using strcat.
18014         (ix86_mangle_function_version_assembler_name): Mangle
18015         target("default") as if no target attribute is present.  Adjust
18016         sorted_attr_string caller.  Avoid leaking memory.  Use XNEWVEC
18017         instead of xmalloc and XDELETEVEC instead of free.
18018         (ix86_function_versions): Don't return true if one of the decls
18019         doesn't have target attribute.  If they don't and one of the decls
18020         is DECL_FUNCTION_VERSIONED, report an error.  Adjust
18021         sorted_attr_string caller.  Use XDELETEVEC instead of free.
18022         (ix86_supports_function_versions): Remove.
18023         (make_name): Fix up formatting.
18024         (make_dispatcher_decl): Remove resolver_name and its initialization.
18025         Avoid leaking memory.
18026         (is_function_default_version): Return true if there is
18027         target("default") attribute rather than no target attribute at all.
18028         (make_resolver_func): Avoid leaking memory.
18029         (ix86_generate_version_dispatcher_body): Likewise.
18030         (TARGET_OPTION_SUPPORTS_FUNCTION_VERSIONS): Remove.
18031         * target.def (supports_function_versions): Remove.
18032         * doc/tm.texi.in (SUPPORTS_FUNCTION_VERSIONS): Remove.
18033         * doc/tm.texi: Regenerated.
18035 2013-01-30  Vladimir Makarov  <vmakarov@redhat.com>
18037         PR rtl-optimization/56144
18038         * lra-constraints.c (get_reload_reg): Don't reuse reload pseudo
18039         for values with side effects.
18041 2013-01-30  Richard Biener  <rguenther@suse.de>
18043         * sparseset.h (sparseset_bit_p): Use gcc_checking_assert.
18044         (sparseset_pop): Likewise.
18045         * cfganal.c (compute_idf): Likewise.  Increase work-stack size
18046         to be able to use quick_push in the worker loop.
18048 2013-01-30  Marek Polacek  <polacek@redhat.com>
18050         * cfgcleanup.c (cleanup_cfg): Don't mark affected BBs.
18052 2013-01-30  Richard Biener  <rguenther@suse.de>
18054         PR lto/56147
18055         * lto-symtab.c (lto_symtab_merge_decls_1): Guard DECL_BUILT_IN check.
18057 2013-01-30  Georg-Johann Lay  <avr@gjlay.de>
18059         PR tree-optimization/56064
18060         * fixed-value.c (fixed_from_double_int): New function.
18061         * fixed-value.h (fixed_from_double_int): New prototype.
18062         (const_fixed_from_double_int): New static inline function.
18063         * fold-const.c (native_interpret_fixed): New static function.
18064         (native_interpret_expr) <FIXED_POINT_TYPE>: Use it.
18065         (can_native_interpret_type_p) <FIXED_POINT_TYPE>: Return true.
18066         (native_encode_fixed): New static function.
18067         (native_encode_expr) <FIXED_CST>: Use it.
18068         (native_interpret_int): Move double_int worker code to...
18069         * double-int.c (double_int::from_buffer): ...this new static method.
18070         * double-int.h (double_int::from_buffer): Prototype it.
18072 2013-01-30  Richard Biener  <rguenther@suse.de>
18074         * tree-ssa-structalias.c (final_solutions, final_solutions_obstack):
18075         New pointer-map and obstack.
18076         (init_alias_vars): Allocate pointer-map and obstack.
18077         (delete_points_to_sets): Free them.
18078         (find_what_var_points_to): Cache result.
18079         (find_what_p_points_to): Adjust for changed interface of
18080         find_what_var_points_to.
18081         (compute_points_to_sets): Likewise.
18082         (ipa_pta_execute): Likewise.
18084 2013-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18086         * configure.ac (HAVE_AS_SPARC_NOBITS): New test.
18087         * configure: Regenerate.
18088         * config.in: Regenerate.
18089         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section): Emit
18090         #nobits/#progbits if supported.
18092 2013-01-29  Oleg Endo  <olegendo@gcc.gnu.org>
18094         PR target/56121
18095         * config/sh/sh.md (bclr_m2a, bset_m2a, bst_m2a, bld_m2a, bldsign_m2a,
18096         bld_reg, *bld_regqi, band_m2a, bandreg_m2a, bor_m2a, borreg_m2a,
18097         bxor_m2a, bxorreg_m2a): Add satisfies_constraint_K03 condition.
18099 2013-01-29  Greta Yorsh  <Greta.Yorsh@arm.com>
18101         * config/arm/cortex-a7.md (cortex_a7_neon, cortex_a7_all): Remove.
18102         (cortex_a7_idiv): Use cortex_a7_both instead of cortex_a7_all.
18104 2013-01-29  Greta Yorsh  <Greta.Yorsh@arm.com>
18106         * config/arm/arm.c (cortexa7_younger): Return true for TYPE_CALL.
18107         * config/arm/cortex-a7.md (cortex_a7_call): Update required units.
18109 2013-01-29  Greta Yorsh  <Greta.Yorsh@arm.com>
18111         * config/arm/arm-protos.h (arm_mac_accumulator_is_result): New
18112         declaration.
18113         * config/arm/arm.c (arm_mac_accumulator_is_result): New function.
18114         * config/arm/cortex-a7.md: New bypasses using
18115         arm_mac_accumulator_is_result.
18117 2013-01-29  Greta Yorsh  <Greta.Yorsh@arm.com>
18119         * config/arm/cortex-a7.md (cortex_a7_neon_mul):  New reservation.
18120         (cortex_a7_neon_mla): Likewise.
18121         (cortex_a7_fpfmad): New reservation.
18122         (cortex_a7_fpmacs): Use ffmas and update required units.
18123         (cortex_a7_fpmuld): Update required units and latency.
18124         (cortex_a7_fpmacd): Likewise.
18125         (cortex_a7_fdivs, cortex_a7_fdivd): Likewise.
18126         (cortex_a7_neon). Likewise.
18127         (bypass) Update participating units.
18129 2013-01-29  Greta Yorsh  <Greta.Yorsh@arm.com>
18131         * config/arm/arm.md (type): Add ffmas and ffmad to "type" attribute.
18132         * config/arm/vfp.md (fma,fmsub,fnmsub,fnmadd): Change type
18133         from fmac to ffma.
18134         * config/arm/vfp11.md (vfp_farith): Use ffmas.
18135         (vfp_fmul): Use ffmad.
18136         * config/arm/cortex-r4f.md (cortex_r4_fmacs): Use ffmas.
18137         (cortex_r4_fmacd): Use ffmad.
18138         * config/arm/cortex-m4-fpu.md (cortex_m4_fmacs): Use ffmas.
18139         * config/arm/cortex-a9.md (cortex_a9_fmacs):  Use ffmas.
18140         (cortex_a9_fmacd): Use ffmad.
18141         * config/arm/cortex-a8-neon.md (cortex_a8_vfp_macs): Use ffmas.
18142         (cortex_a8_vfp_macd): Use ffmad.
18143         * config/arm/cortex-a5.md (cortex_a5_fpmacs): Use ffmas.
18144         (cortex_a5_fpmacd): Use ffmad.
18145         * config/arm/cortex-a15-neon.md (cortex_a15_vfp_macs) Use ffmas.
18146         (cortex_a15_vfp_macd): Use ffmad.
18147         * config/arm/arm1020e.md (v10_fmul): Use ffmas and ffmad.
18149 2013-01-29  Jason Merrill  <jason@redhat.com>
18151         PR libstdc++/54314
18152         * varasm.c (default_assemble_visibility): Don't warn about
18153         visibility on artificial decls.
18155 2013-01-29  Richard Biener  <rguenther@suse.de>
18157         PR tree-optimization/56113
18158         * tree-ssa-structalias.c (equiv_class_lookup): Also return
18159         the bitmap leader.
18160         (label_visit): Free duplicate bitmaps and record the leader instead.
18161         (perform_var_substitution): Adjust.
18163 2013-01-29  Richard Biener  <rguenther@suse.de>
18165         PR tree-optimization/55270
18166         * tree-ssa-dom.c (eliminate_degenerate_phis): If we changed
18167         the CFG, schedule loops for fixup.
18169 2013-01-29  Nick Clifton  <nickc@redhat.com>
18171         * config/rl78/rl78.c (rl78_regno_mode_code_ok_for_base_p): Allow
18172         SP_REG.
18174 2013-01-28  Leif Ekblad  <leif@rdos.net>
18176         * config.gcc (i[34567]86-*-rdos*, x86_64-*-rdos*): New targets.
18177         * config/i386/i386.h (TARGET_RDOS): New macro.
18178         (DEFAULT_LARGE_SECTION_THRESHOLD): New macro.
18179         * config/i386/i386.c (ix86_option_override_internal): For 64bit
18180         TARGET_RDOS, set ix86_cmodel to CM_MEDIUM_PIC and flag_pic to 1.
18181         * config/i386/i386.opt (mlarge-data-threshold): Initialize to
18182         DEFAULT_LARGE_SECTION_THRESHOLD.
18183         * config/i386/i386.md (R14_REG, R15_REG): New constants.
18184         * config/i386/rdos.h: New file.
18185         * config/i386/rdos64.h: New file.
18187 2013-01-28  Bernd Schmidt  <bernds@codesourcery.com>
18189         PR other/54814
18190         * reload.c (find_valid_class_1): Use in_hard_reg_set_p instead of
18191         TEST_HARD_REG_BIT.
18193 2013-01-28  Jakub Jelinek  <jakub@redhat.com>
18195         PR rtl-optimization/56117
18196         * sched-deps.c (sched_analyze_2) <case PREFETCH>: For use_cselib
18197         call cselib_lookup_from_insn on the MEM before calling
18198         add_insn_mem_dependence.
18200 2013-01-28  Richard Biener  <rguenther@suse.de>
18202         * tree-inline.c (remap_gimple_stmt): Do not assing a BLOCK
18203         to a stmt that didn't have one.
18204         (copy_phis_for_bb): Likewise for PHI arguments.
18205         (copy_debug_stmt): Likewise for debug stmts.
18207 2013-01-28  Richard Biener  <rguenther@suse.de>
18209         PR tree-optimization/56034
18210         * tree-loop-distribution.c (enum partition_kind): Add PKIND_REDUCTION.
18211         (partition_builtin_p): Adjust.
18212         (generate_code_for_partition): Handle PKIND_REDUCTION.  Assert
18213         it is the last partition.
18214         (rdg_flag_uses): Check SSA_NAME_IS_DEFAULT_DEF before looking
18215         up the vertex for the definition.
18216         (classify_partition): Classify whether a partition is a
18217         PKIND_REDUCTION, thus has uses outside of the loop.
18218         (ldist_gen): Inherit PKIND_REDUCTION when merging partitions.
18219         Merge all PKIND_REDUCTION partitions into the last partition.
18220         (tree_loop_distribution): Seed partitions from reductions as well.
18222 2013-01-28  Jakub Jelinek  <jakub@redhat.com>
18224         PR tree-optimization/56125
18225         * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Don't optimize
18226         pow(x,c) into sqrt(x) * powi(x, n/2) or
18227         1.0 / (sqrt(x) * powi(x, abs(n/2))) if c is an integer or when
18228         optimizing for size.
18229         Don't optimize pow(x,c) into powi(x, n/3) * powi(cbrt(x), n%3) or
18230         1.0 / (powi(x, abs(n)/3) * powi(cbrt(x), abs(n)%3)) if 2c is an
18231         integer.
18233         PR tree-optimization/56094
18234         * gimplify.c (force_gimple_operand_1): Temporarily set input_location
18235         to UNKNOWN_LOCATION while gimplifying expr.
18237 2013-01-27  Uros Bizjak  <ubizjak@gmail.com>
18239         PR target/56114
18240         * config/i386/i386.md (*movabs<mode>_1): Add square brackets around
18241         operand 0 in movabs insn template for -masm=intel asm alternative.
18242         (*movabs<mode>_2): Ditto for operand 1.
18244 2013-01-26  David Holsgrove  <david.holsgrove@xilinx.com>
18246         PR target/54663
18247         * config.gcc (microblaze*-linux*): Add tmake_file to allow building
18248         of microblaze-c.o
18250 2013-01-26  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
18252         * config.gcc (microblaze*-*-*): Rename microblaze*-*-elf, update
18253         tm_file.
18255 2013-01-25  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
18257         * config/aarch64/aarch64.c (TARGET_FIXED_CONDITION_CODE_REGS):
18258         Undef to avoid warning.
18260 2013-01-25  Michael Haubenwallner  <michael.haubenwallner@salomon.at>
18262         * configure.ac (gcc_cv_ld_static_dynamic): Define for AIX native ld.
18263         * configure: Regenerate.
18265 2013-01-25  Jakub Jelinek  <jakub@redhat.com>
18267         PR tree-optimization/56098
18268         * tree-ssa-phiopt.c (nt_init_block): Don't call add_or_mark_expr
18269         for stmts with volatile ops.
18270         (cond_store_replacement): Don't optimize if assign has volatile ops.
18271         (cond_if_else_store_replacement_1): Don't optimize if either
18272         then_assign or else_assign have volatile ops.
18273         (hoist_adjacent_loads): Don't optimize if either def1 or def2 have
18274         volatile ops.
18276 2013-01-25  Georg-Johann Lay  <avr@gjlay.de>
18278         * doc/invoke.texi (AVR Built-in Macros): Document __XMEGA__.
18280 2013-01-25  Georg-Johann Lay  <avr@gjlay.de>
18282         * doc/extend.texi (Example of asm with clobbered asm reg): Fix
18283         missing ':' in asm example.
18285 2013-01-25  Tejas Belagod  <tejas.belagod@arm.com>
18287         * config/aarch64/aarch64-simd-builtins.def: Separate sq<r>dmulh_lane
18288         entries into lane and laneq entries.
18289         * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_lane<mode>):
18290         Remove AdvSIMD scalar modes.
18291         (aarch64_sq<r>dmulh_laneq<mode>): New.
18292         (aarch64_sq<r>dmulh_lane<mode>): New RTL pattern for Scalar AdvSIMD
18293         modes.
18294         * config/aarch64/arm_neon.h: Fix all the vq<r>dmulh_lane* intrinsics'
18295         builtin implementations to relfect changes in RTL in aarch64-simd.md.
18296         * config/aarch64/iterators.md (VCOND): New.
18297         (VCONQ): New.
18299 2013-01-25  Georg-Johann Lay  <avr@gjlay.de>
18301         PR target/54222
18302         * config/avr/builtins.def (DEF_BUILTIN): Add LIBNAME argument.
18303         Add NULL LIBNAME argument to existing definitions.
18304         (ABSHR, ABSR, ABSLR, ABSLLR, ABSHK, ABSK, ABSLK, ABSLLK): New.
18305         * config/avr/avr-c.c (DEF_BUILTIN): Add LIBNAME argument.
18306         * config/avr/avr.c (DEF_BUILTIN): Same.
18307         (avr_init_builtins): Pass down LIBNAME to add_builtin_function.
18308         (avr_expand_builtin): Expand to a vanilla call if a libgcc
18309         implementation is available (DECL_ASSEMBLER_NAME is set).
18310         (avr_fold_absfx): New static function.
18311         (avr_fold_builtin): Use it to handle: AVR_BUILTIN_ABSHR,
18312         AVR_BUILTIN_ABSR, AVR_BUILTIN_ABSLR, AVR_BUILTIN_ABSLLR,
18313         AVR_BUILTIN_ABSHK, AVR_BUILTIN_ABSK, AVR_BUILTIN_ABSLK,
18314         AVR_BUILTIN_ABSLLK.
18315         * config/avr/stdfix.h (abshr, absr, abslr, absllr)
18316         (abshk, absk, abslk, absllk): Provide as static inline functions.
18318 2013-01-25  Marek Polacek  <polacek@redhat.com>
18320         PR tree-optimization/56035
18321         * cfgloopmanip.c (fix_loop_structure): Remove redundant condition.
18323 2012-01-24  Uros Bizjak  <ubizjak@gmail.com>
18325         * config/i386/i386.md (*movti_internal_rex64): Add (o,e) alternative.
18326         (*movtf_internal_rex64): Add (!o,C) alternative
18327         (*movxf_internal_rex64): Ditto.
18328         (*movdf_internal_rex64): Add (?r,C) and (?m,C) alternatives.
18330 2013-01-24  Shenghou Ma  <minux.ma@gmail.com>
18332         * doc/invoke.texi: fix typo.
18333         * doc/objc.texi: fix typo.
18335 2013-01-24  Richard Sandiford  <rdsandiford@googlemail.com>
18337         * config/mips/mips.md (*and<mode>3_mips16): Use the "W" constraint
18338         for the first two alternatives.
18340 2013-01-24  Diego Novillo  <dnovillo@google.com>
18342         * Makefile.in (GGC): Remove.  Replace all instances with ggc-page.o.
18343         (ggc-zone.o): Remove.
18344         * configure.ac: Remove option --with-gc.
18345         * configure: Re-generate.
18346         * doc/install.texi: Remove documentation for --with-gc.
18347         * gengtype.c (write_enum_defn): Remove.  Update all users.
18348         (write_Types_process_field): Remove generation of gt_e_* argument.
18349         (output_type_enum): Remove.  Update all users.
18350         (write_enum_defn): Remove.  Update all users.
18351         (enum alloc_zone): Remove.  Update all users.
18352         (write_splay_tree_allocator_def): Remove generation of gt_e_* argument.
18353         * ggc-common.c (ggc_splay_alloc): Remove first argument.
18354         Update all callers.
18355         (struct ptr_data): Remove field TYPE.  Update all users.
18356         (gt_pch_note_object): Remove argument TYPE.  Update all users.
18357         * ggc-internal.h (ggc_pch_alloc_object): Remove last argument.
18358         Update all users.
18359         * ggc-none.c (ggc_alloc_typed_stat): Remove.
18360         (struct alloc_zone): Remove.
18361         (ggc_internal_alloc_zone_stat): Remove.
18362         (ggc_internal_cleared_alloc_zone_stat): Remove.
18363         * ggc-page.c (ggc_alloc_typed_stat): Remove.
18364         (ggc_pch_count_object): Remove last argument.  Update all users.
18365         (ggc_pch_alloc_object): Remove last argument.  Update all users.
18366         (struct alloc_zone): Remove.
18367         * ggc-zone.c: Remove.
18368         * ggc.h (gt_pch_note_object): Remove last argument.  Update all users.
18369         (struct alloc_zone): Remove.
18370         (ggc_alloc_typed_stat): Remove.
18371         (ggc_alloc_typed): Remove.
18372         (ggc_splay_alloc): Remove first argument.
18373         (rtl_zone): Remove.  Update all users.
18374         (tree_zone): Remove.  Update all users.
18375         (tree_id_zone): Remove.  Update all users.
18376         (ggc_internal_zone_alloc_stat): Remove.  Update all users.
18377         (ggc_internal_zone_cleared_alloc_stat): Remove.  Update all users.
18378         (ggc_internal_zone_vec_alloc_stat): Remove.  Update all users.
18379         * tree-ssanames.c: Remove references to zone allocator in comments.
18381 2013-01-24  Georg-Johann Lay  <avr@gjlay.de>
18383         * config/avr/avr.c (avr_out_fract): Make register numbers that
18384         might be outside of source operand signed.
18386 2013-01-24  Uros Bizjak  <ubizjak@gmail.com>
18388         * config/i386/constraints.md (Yf): New constraint.
18389         * config/i386/i386.md (*movdf_internal_rex64): Use Yf*f instead
18390         of f constraint to conditionaly disable x87 register preferences.
18391         (*movdf_internal): Ditto.
18392         (*movsf_internal): Ditto.
18394 2013-01-24  Steven Bosscher  <steven@gcc.gnu.org>
18396         PR inline-asm/55934
18397         * lra-assigns.c (assign_by_spills): Throw away the pattern of asms
18398         that have operands with impossible constraints.
18399         Add a FIXME for a speed-up opportunity.
18400         * lra-constraints.c (process_alt_operands): Verify that a class
18401         selected from constraints on asms is valid for the operand mode.
18402         (curr_insn_transform): Remove incorrect comment.
18404 2013-01-23  David Edelsohn  <dje.gcc@gmail.com>
18406         * config/rs6000/rs6000.c (rs6000_delegitimize_address): Check that
18407         TOC operand is a valid symbol ref in the constant pool.
18409 2013-01-23  Edgar E. Iglesias  <edgar.iglesias@gmail.com>
18411         * config/microblaze/linux.h: Add TARGET_OS_CPP_BUILTINS
18413 2013-01-23  Georg-Johann Lay  <avr@gjlay.de>
18415         PR target/54222
18416         * config/avr/stdfix.h: New file.
18417         * t-avr (stdfix-gcc.h): New rule to build it.
18418         (EXTRA_HEADERS): Set it to install stdfix.h, stdfix-gcc.h.
18420 2013-01-23  Kostya Serebryany  <kcc@google.com>
18422         * config/darwin.h: remove dependency on
18423         CoreFoundation (asan on Mac OS).
18425 2013-01-23  Jakub Jelinek  <jakub@redhat.com>
18427         PR target/49069
18428         * config/arm/arm.md (cbranchdi4, cstoredi4): Use s_register_operand
18429         instead of cmpdi_operand for first comparison operand.
18430         Don't assert that comparison operands aren't both constants.
18432 2013-01-22  Jonathan Wakely  <jwakely.gcc@gmail.com>
18434         * doc/install.texi (Downloading the Source): Update references to
18435         downloading separate components.
18437 2013-01-22  Jonathan Wakely  <jwakely.gcc@gmail.com>
18439         * doc/extend.texi (__int128): Improve grammar.
18441 2013-01-22  Uros Bizjak  <ubizjak@gmail.com>
18443         PR target/56028
18444         * config/i386/i386.md (*movti_internal_rex64): Change (o,riF)
18445         alternative to (o,r).
18446         (*movdi_internal_rex64): Remove (!o,n) alternative.
18447         (DImode immediate->memory splitter): Remove.
18448         (DImode immediate->memory peephole2): Remove.
18449         (movtf): Enable for TARGET_64BIT || TARGET_SSE.
18450         (*movtf_internal_rex64): Rename from *movtf_internal. Change (!o,F*r)
18451         alternative to (!o,*r).
18452         (*movtf_internal_sse): New pattern.
18453         (*movxf_internal_rex64): New pattern.
18454         (*movxf_internal): Disable for TARGET_64BIT.
18455         (*movdf_internal_rex64): Remove (!o,F) alternative.
18457 2013-01-22  Jakub Jelinek  <jakub@redhat.com>
18459         PR middle-end/56074
18460         * dumpfile.c (dump_loc): Only print loc if LOCATION_LOCUS (loc)
18461         isn't UNKNOWN_LOCATION nor BUILTINS_LOCATION.
18462         * tree-vect-loop-manip.c (find_loop_location): Also ignore
18463         stmt locations where LOCATION_LOCUS of the stmt location is
18464         UNKNOWN_LOCATION or BUILTINS_LOCATION.
18466         PR target/55686
18467         * config/i386/i386.md (UNSPEC_STOS): New.
18468         (strset_singleop, *strsetdi_rex_1, *strsetsi_1, *strsethi_1,
18469         *strsetqi_1): Add UNSPEC_STOS.
18471 2013-01-22  Paolo Carlini  <paolo.carlini@oracle.com>
18473         PR c++/56067
18474         * doc/invoke.texi: Remove left over -Wsynth example.
18476 2013-01-21  Jakub Jelinek  <jakub@redhat.com>
18478         PR tree-optimization/56051
18479         * fold-const.c (fold_binary_loc): Don't fold
18480         X < (cast) (1 << Y) into (X >> Y) != 0 if cast is either
18481         a narrowing conversion, or widening conversion from signed
18482         to unsigned.
18484 2013-01-21  Uros Bizjak  <ubizjak@gmail.com>
18486         PR rtl-optimization/56023
18487         * haifa-sched.c (fix_inter_tick): Do not update ticks of instructions,
18488         dependent on debug instruction.
18490 2013-01-21  Martin Jambor  <mjambor@suse.cz>
18492         PR middle-end/56022
18493         * function.c (allocate_struct_function): Call
18494         invoke_set_current_function_hook earlier.
18496 2013-01-21  Jakub Jelinek  <jakub@redhat.com>
18498         * reload1.c (init_reload): Only initialize reload_obstack
18499         during the first call.
18501 2013-01-21  Marek Polacek  <polacek@redhat.com>
18503         * cfgloop.c (verify_loop_structure): Fix up grammar.
18505 2013-01-21  Yi-Hsiu Hsu  <ahsu@marvell.com>
18507         * config/arm/marvell-pj4.md (pj4_shift_conds, pj4_alu_shift,
18508         pj4_alu_shift_conds, pj4_shift): Handle simple_alu_shift.
18510 2013-01-21  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18512         PR target/56058
18513         * config/arm/marvell-pj4.md: Update copyright year.
18514         Fix up use of alu to alu_reg and simple_alu_imm.
18516 2013-01-21  Uros Bizjak  <ubizjak@gmail.com>
18518         * config/i386/i386.md (enabled): Do not disable fma4 for TARGET_FMA.
18520 2013-01-20  Vladimir Makarov  <vmakarov@redhat.com>
18522         PR target/55433
18523         * lra-constraints.c (curr_insn_transform): Don't reuse original
18524         insn for secondary memory move when memory mode should be different.
18526 2013-01-20  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
18528         * config/pa/pa.md (atomic_loaddi, atomic_loaddi_1, atomic_storedi,
18529         atomic_storedi_1): New patterns.
18531 2013-01-20  Venkataramanan Kumar  <venkataramanan.kumar@amd.com>
18533         btver2 pipeline descriptions.
18534         * config/i386/i386.c: Enable CPU_BTVER2 to use btver2 pipeline
18535         descriptions.
18536         * config/i386/i386.md (btver2_decode): New type attributes.
18537         * config/i386/sse.md (btver2_decode, btver2_sse_attr): New
18538         type attributes.
18539         * config/i386/btver2.md: New file describing btver2 pipelines.
18541 2013-01-19  Andrew Pinski  <apinski@cavium.com>
18543         PR tree-optimization/52631
18544         * tree-ssa-sccvn (visit_use): Before looking up the original
18545         statement, try looking up the simplified expression.
18547 2013-01-19  Anthony Green  <green@moxielogic.com>
18549         * config/moxie/moxie.c (moxie_expand_prologue): Set
18550         current_function_static_stack_size.
18552 2013-01-18  Jakub Jelinek  <jakub@redhat.com>
18554         PR tree-optimization/56029
18555         * tree-phinodes.c (reserve_phi_args_for_new_edge): Set
18556         gimple_phi_arg_location for the new arg to UNKNOWN_LOCATION.
18558 2013-01-18  Sharad Singhai  <singhai@google.com>
18560         PR tree-optimization/55995
18561         * dumpfile.c (dump_loc): Print location only if available.
18562         * tree-vectorizer.c (increase_alignment): Intialize vect_location.
18564 2013-01-18  Vladimir Makarov  <vmakarov@redhat.com>
18566         PR target/55433
18567         * lra-constraints.c (curr_insn_transform): Reuse original insn for
18568         secondary memory move.
18569         (inherit_reload_reg): Use rclass instead of cl for
18570         check_secondary_memory_needed_p.
18572 2013-01-18  Jakub Jelinek  <jakub@redhat.com>
18574         PR middle-end/56015
18575         * expr.c (expand_expr_real_2) <case COMPLEX_EXPR>: Handle
18576         the case where writing real complex part of target modifies op1.
18578 2013-01-18  James Greenhalgh  <james.greenhalgh@arm.com>
18580         * config/aarch64/aarch64-simd.md
18581         (aarch64_vcond_internal<mode>): Handle unordered cases.
18582         * config/aarch64/iterators.md (v_cmp_result): New.
18584 2013-01-18  Yi-Hsiu Hsu  <ahsu@marvell.com>
18585             Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18587         * config/arm/marvell-pj4.md: New file.
18588         * config/arm/arm.c (arm_issue_rate): Add marvell_pj4.
18589         * config/arm/arm.md (generic_sched): Add marvell_pj4.
18590         (generic_vfp): Likewise.
18591         * config/arm/arm-cores.def: Add marvell-pj4.
18592         * config/arm/arm-tune.md: Regenerate.
18593         * config/arm/arm-tables.opt: Regenerate.
18594         * config/arm/bpabi.h (BE8_LINK_SPEC): Add marvell_pj4.
18595         * doc/invoke.texi: Document marvell-pj4.
18597 2013-01-18  Tejas Belagod  <tejas.belagod@arm.com>
18599         * config/aarch64/arm_neon.h: Map scalar types to standard types.
18601 2013-01-18  Alexandre Oliva  <aoliva@redhat.com>
18603         PR debug/54114
18604         PR debug/54402
18605         PR debug/49888
18606         * var-tracking.c (negative_power_of_two_p): New.
18607         (global_get_addr_cache, local_get_addr_cache): New.
18608         (get_addr_from_global_cache, get_addr_from_local_cache): New.
18609         (vt_canonicalize_addr): Rewrite using the above.  Adjust the
18610         heading comment.
18611         (vt_stack_offset_p): Remove.
18612         (vt_canon_true_dep): Always canonicalize loc's address.
18613         (clobber_overlapping_mems): Make sure we have a MEM.
18614         (local_get_addr_clear_given_value): New.
18615         (val_reset): Clear local cached entries.
18616         (compute_bb_dataflow): Create and release the local cache.
18617         Disable duplicate MEMs clobbering.
18618         (emit_notes_in_bb): Clobber MEMs likewise.
18619         (vt_emit_notes): Create and release the local cache.
18620         (vt_initialize, vt_finalize): Create and release the global
18621         cache, respectively.
18622         * alias.c (rtx_equal_for_memref_p): Compare operands of ENTRY_VALUEs.
18624 2013-01-18  Alexandre Oliva  <aoliva@redhat.com>
18626         PR libmudflap/53359
18627         * tree-mudflap.c (mudflap_finish_file): Skip deferred decls
18628         not found in the symtab.
18630 2013-01-18  Alexandre Oliva  <aoliva@redhat.com>
18632         PR debug/56006
18633         PR rtl-optimization/55547
18634         PR rtl-optimization/53827
18635         PR debug/53671
18636         PR debug/49888
18637         * alias.c (offset_overlap_p): New, factored out of...
18638         (memrefs_conflict_p): ... this.  Use absolute sizes.  Retain
18639         the conservative special case for symbolic constants.  Don't
18640         adjust zero sizes on alignment.
18642 2013-01-18  Bernd Schmidt  <bernds@codesourcery.com>
18644         PR rtl-optimization/52573
18645         * regrename.c (build_def_use): Ignore REG_DEAD notes if there is a
18646         REG_UNUSED for the same register.
18648 2013-01-17  Richard Biener  <rguenther@suse.de>
18649             Marek Polacek  <polacek@redhat.com>
18651         PR rtl-optimization/55833
18652         * loop-unswitch.c (unswitch_loops): Move loop verification...
18653         (unswitch_single_loop): ...here.  Call mark_irreducible_loops.
18654         * cfgloopmanip.c (fix_loop_placement): Add IRRED_INVALIDATED parameter.
18655         Set it to true when we're removing a loop from hierarchy tree in
18656         an irreducible region.
18657         (fix_bb_placements): Adjust caller.
18658         (fix_loop_placements): Likewise.
18660 2013-01-17  Georg-Johann Lay  <avr@gjlay.de>
18662         * config/avr/builtins.def (DEF_BUILTIN): Factor out
18663         "__builtin_avr_" from NAME, turn NAME to an uppercase identifier.
18664         Factor out 'CODE_FOR_' from ICODE, use 'nothing' instead of '-1'.
18665         Remove ID.  Adjust comments.
18666         * config/avr/avr-c.c (avr_builtin_name): Remove.
18667         (avr_cpu_cpp_builtins): Use DEF_BUILTIN instead of for-loop.
18668         * config/avr/avr.c (avr_tolower): New static function.
18669         (DEF_BUILTIN): Remove parameter ID.  Prefix ICODE by 'CODE_FOR_'.
18670         Stringify NAME, prefix it with "__builtin_avr_" and lowercase it.
18671         (avr_expand_builtin): Assert insn_code != CODE_FOR_nothing for
18672         default expansion.
18674 2013-01-17  Jan Hubicka  <jh@suse.cz>
18676         PR tree-optimization/55273
18677         * loop-iv.c (iv_number_of_iterations): Consider zero iteration case.
18679 2013-01-17  Uros Bizjak  <ubizjak@gmail.com>
18681         PR target/55981
18682         * config/i386/sync.md (atomic_store<mode>): Always generate SWImode
18683         store through atomic_store<mode>_1.
18684         (atomic_store<mode>_1): Macroize insn using SWI mode iterator.
18686 2013-01-17  Martin Jambor  <mjambor@suse.cz>
18688         PR tree-optimizations/55264
18689         * ipa-inline-transform.c (can_remove_node_now_p_1): Never return true
18690         for virtual methods.
18691         * ipa.c (symtab_remove_unreachable_nodes): Never return true for
18692         virtual methods before inlining is over.
18693         * cgraph.h (cgraph_only_called_directly_or_aliased_p): Return false for
18694         virtual functions.
18695         * cgraphclones.c (cgraph_create_virtual_clone): Mark clones as
18696         non-virtual.
18698 2013-01-16  Vladimir Makarov  <vmakarov@redhat.com>
18700         PR rtl-optimization/56005
18701         * sched-deps.c (sched_analyze_2): Check deps->readonly for adding
18702         pending reads for prefetch.
18704 2013-01-16  Ian Bolton  <ian.bolton@arm.com>
18706         * config/aarch64/aarch64.md
18707         (*cstoresi_neg_uxtw): New pattern.
18708         (*cmovsi_insn_uxtw): New pattern.
18709         (*<optab>si3_uxtw): New pattern.
18710         (*<LOGICAL:optab>_<SHIFT:optab>si3_uxtw): New pattern.
18711         (*<optab>si3_insn_uxtw): New pattern.
18712         (*bswapsi2_uxtw): New pattern.
18714 2013-01-16  Richard Biener  <rguenther@suse.de>
18716         * tree-inline.c (tree_function_versioning): Remove set but
18717         never used variable.
18719 2013-01-16  Richard Biener  <rguenther@suse.de>
18721         PR tree-optimization/55964
18722         * tree-flow.h (rename_variables_in_loop): Remove.
18723         (rename_variables_in_bb): Likewise.
18724         * tree-loop-distribution.c (update_phis_for_loop_copy): Remove.
18725         (copy_loop_before): Adjust and delete update-ssa status.
18726         * tree-vect-loop-manip.c (rename_variables_in_bb): Make static.
18727         (rename_variables_in_bb): Likewise.  Properly walk over predecessors.
18728         (rename_variables_in_loop): Remove.
18729         (slpeel_update_phis_for_duplicate_loop): Likewise.
18730         (slpeel_tree_duplicate_loop_to_edge_cfg): Handle nested loops,
18731         use available cfg machinery instead of duplicating it.
18732         Update PHI nodes and perform poor-mans SSA update here.
18733         (slpeel_tree_peel_loop_to_edge): Adjust.
18735 2013-01-16  Richard Biener  <rguenther@suse.de>
18737         PR tree-optimization/54767
18738         PR tree-optimization/53465
18739         * tree-vrp.c (vrp_meet_1): Revert original fix for PR53465.
18740         (vrp_visit_phi_node): For PHI arguments coming via backedges
18741         drop all symbolical range information.
18742         (execute_vrp): Compute backedges.
18744 2013-01-16  Richard Biener  <rguenther@suse.de>
18746         * doc/install.texi: Update CLooG and ISL requirements to
18747         0.18.0 and 0.11.1.
18749 2013-01-16  Christian Bruel  <christian.bruel@st.com>
18751         PR target/55301
18752         * config/sh/sh.c (sh_expand_prologue): Postpone new_stack mem symbol.
18753         (broken_move): Handle UNSPECV_SP_SWITCH_B.
18754         * config/sh/sh.md (sp_switch_1): Use set (reg:SI SP_REG).
18756 2013-01-16  DJ Delorie  <dj@redhat.com>
18758         * config/sh/sh.md (UNSPECV_SP_SWITCH_B): New.
18759         (UNSPECV_SP_SWITCH_E): New.
18760         (sp_switch_1): Change to an unspec.
18761         (sp_switch_2): Change to an unspec.  Don't use post-inc when we
18762         replace $r15.
18764 2013-01-16  Uros Bizjak  <ubizjak@gmail.com>
18766         * emit-rtl.c (need_atomic_barrier_p): Mask memory model argument
18767         with MEMMODEL_MASK before comparing with MEMMODEL_* memory types.
18768         * optabs.c (maybe_emit_sync_lock_test_and_set): Ditto.
18769         (expand_mem_thread_fence): Ditto.
18770         (expand_mem_signal_fence): Ditto.
18771         (expand_atomic_load): Ditto.
18772         (expand_atomic_store): Ditto.
18774 2013-01-16  Alexandre Oliva  <aoliva@redhat.com>
18776         PR rtl-optimization/55547
18777         PR rtl-optimization/53827
18778         PR debug/53671
18779         PR debug/49888
18780         * alias.c (memrefs_conflict_p): Set sizes to negative after
18781         AND adjustments.
18783 2013-01-15  Jakub Jelinek  <jakub@redhat.com>
18785         PR target/55940
18786         * function.c (thread_prologue_and_epilogue_insns): Always
18787         add crtl->drap_reg to set_up_by_prologue.set, even if
18788         stack_realign_drap is false.
18790 2013-01-15  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
18792         * config/vax/vax.md (add<mode>3, sub<mode>3, mul<mode>3, div<mode>3,
18793         and<mode>3, *and<mode>_const_int, ior<mode>3, xor<mode>3, ashrsi3,
18794         *call): Fix indention.
18796 2013-01-15  Tom de Vries  <tom@codesourcery.com>
18798         PR target/55876
18799         * optabs.c (widen_operand): Use gen_lowpart instead of gen_rtx_SUBREG.
18800         Update comment.
18802 2013-01-15  Vladimir Makarov  <vmakarov@redhat.com>
18804         PR rtl-optimization/55153
18805         * sched-deps.c (sched_analyze_2): Add pending reads for prefetch.
18807 2013-01-15  Martin Jambor  <mjambor@suse.cz>
18809         PR tree-optimization/55920
18810         * tree-sra.c (analyze_access_subtree): Do not mark non-removable
18811         accesses as grp_to_be_debug_replaced.
18813 2013-01-15  Jakub Jelinek  <jakub@redhat.com>
18815         PR tree-optimization/55920
18816         * tree-sra.c (sra_modify_assign): If for lacc->grp_to_be_debug_replaced
18817         there is non-useless type conversion needed from debug rhs to lhs,
18818         use build_debug_ref_for_model and/or VIEW_CONVERT_EXPR.
18820 2013-01-15  Joseph Myers  <joseph@codesourcery.com>
18821             Mikael Pettersson  <mikpe@it.uu.se>
18823         PR target/43961
18824         * config/arm/arm.h (ADDR_VEC_ALIGN): Align SImode jump tables for
18825         Thumb.
18826         (ASM_OUTPUT_CASE_LABEL): Remove.
18827         (ASM_OUTPUT_BEFORE_CASE_LABEL): Define to empty.
18828         * final.c (shorten_branches): Update alignment of labels before
18829         jump tables if CASE_VECTOR_SHORTEN_MODE.
18831 2013-01-15  Richard Biener  <rguenther@suse.de>
18833         PR bootstrap/55961
18834         * system.h: Do not include gmp.h for building host tools.
18836 2013-01-15  Richard Biener  <rguenther@suse.de>
18838         PR middle-end/55882
18839         * emit-rtl.c (set_mem_attributes_minus_bitpos): Correctly
18840         account for bitpos when computing alignment.
18842 2013-01-15  Vladimir Yakovlev  <vladimir.b.yakovlev@intel.com>
18844         * config/i386/i386-c.c (ix86_target_macros_internal): New case.
18845         (ix86_target_macros_internal): Likewise.
18847         * config/i386/i386.c (m_CORE2I7): Removed.
18848         (m_CORE_HASWELL): New macro.
18849         (m_CORE_ALL): Likewise.
18850         (initial_ix86_tune_features): m_CORE2I7 is replaced by m_CORE_ALL.
18851         (initial_ix86_arch_features): Likewise.
18852         (processor_target_table): Initializations for Core avx2.
18853         (cpu_names): New names "core-avx2".
18854         (ix86_option_override_internal): Changed PROCESSOR_COREI7 by
18855         PROCESSOR_CORE_HASWELL.
18856         (ix86_issue_rate): New case.
18857         (ia32_multipass_dfa_lookahead): Likewise.
18858         (ix86_sched_init_global): Likewise.
18860         * config/i386/i386.h (TARGET_HASWELL): New macro.
18861         (target_cpu_default): New TARGET_CPU_DEFAULT_haswell.
18862         (processor_type): New PROCESSOR_HASWELL.
18864 2013-01-15  Jakub Jelinek  <jakub@redhat.com>
18866         PR tree-optimization/55955
18867         * tree-vect-loop.c (vectorizable_reduction): Give up early on
18868         *SHIFT_EXPR and *ROTATE_EXPR codes.
18870         PR tree-optimization/48766
18871         * opts.c (common_handle_option): For -fwrapv disable -ftrapv, for
18872         -ftrapv disable -fwrapv.
18874 2013-01-14  Georg-Johann Lay  <avr@gjlay.de>
18876         PR target/55974
18877         * config/avr/avr-c.c (avr_cpu_cpp_builtins): Define __FLASH
18878         etc. to 1 and not to __flash.
18879         Use LL suffix for __INT24_MAX__ with -mint8.
18880         Use ULL suffix for __UINT24_MAX__ with -mint8.
18882 2013-01-14  Georg-Johann Lay  <avr@gjlay.de>
18884         * config/avr/avr-arch.h
18885         (struct base_arch_s): Use typedef avr_arch_t instead.
18886         (struct arch_info_s): Use typedef avr_arch_info_t instead.
18887         (struct mcu_type_s): Use typedef avr_mcu_t instead.
18888         * config/avr/avr.c: Same.
18889         * config/avr/avr-devices.c: Same.
18890         * config/avr/driver-avr.c: Same.
18891         * config/avr/gen-avr-mmcu-texi.c: Same.
18892         * config/avr/avr-mcus.def: Adjust comment.
18894 2013-01-14  Tejas Belagod  <tejas.belagod@arm.com>
18896         * config/aarch64/aarch64-simd.md (*aarch64_simd_ld1r<mode>): New.
18897         * config/aarch64/iterators.md (VALLDI): New.
18899 2013-01-14  Uros Bizjak  <ubizjak@gmail.com>
18900             Andi Kleen  <ak@linux.intel.com>
18902         PR target/55948
18903         * config/i386/sync.md (atomic_store<mode>_1): New pattern.
18904         (atomic_store<mode>): Call atomic_store<mode>_1 for IX86_HLE_RELEASE
18905         memmodel flag.
18907 2013-01-14  Georg-Johann Lay  <avr@gjlay.de>
18909         * config/avr/avr-stdint.h: Remove trailing blanks.
18910         * config/avr/avr-log.h: Same.
18911         * config/avr/avr-arch.h: Same.
18912         * config/avr/avr-devices.c: Same.
18913         * config/avr/avr-dimode.md: Same.
18914         * config/avr/predicates.md: Same.
18915         * config/avr/avr-c.c: Same.  And fix typo.
18917         * config/avr/avr-protos.h: Same.  And:
18918         (function_arg_regno_p): Rename to avr_function_arg_regno_p.
18919         (init_cumulative_args): Rename to avr_init_cumulative_args.
18920         (expand_prologue): Rename to avr_expand_prologue.
18921         (expand_epilogue): Rename to avr_expand_epilogue.
18922         (adjust_insn_length): Rename to avr_adjust_insn_length.
18923         (notice_update_cc): Rename to avr_notice_update_cc.
18924         (final_prescan_insn): Rename to avr_final_prescan_insn.
18925         * config/avr/avr.c: Same.
18926         * config/avr/avr.h: Same.
18927         * config/avr/avr.md: Remove trailing blanks.
18928         (prologue): Use avr_expand_prologue.
18929         (epilogue, sibcall_epilogue): Use avr_expand_epilogue.
18931 2013-01-14  Richard Biener  <rguenther@suse.de>
18933         * tree-cfg.c (verify_expr_location, verify_expr_location_1,
18934         verify_location, collect_subblocks): New functions.
18935         (verify_gimple_in_cfg): Verify that locations only reference
18936         BLOCKs in the functions BLOCK tree.
18938 2013-01-14  Richard Biener  <rguenther@suse.de>
18940         * tree-cfgcleanup.c (remove_forwarder_block): Unshare propagated
18941         PHI argument.
18942         * graphite-sese-to-poly.c (insert_out_of_ssa_copy): Properly
18943         unshare reference.
18944         (insert_out_of_ssa_copy_on_edge): Likewise.
18945         (rewrite_close_phi_out_of_ssa): Likewise.
18946         * tree-ssa.c (insert_debug_temp_for_var_def): Properly unshare
18947         debug expressions.
18948         * tree-ssa-pre.c (insert_into_preds_of_block): Properly unshare
18949         propagated constants.
18950         * tree-cfg.c (tree_node_can_be_shared): Handled component-refs
18951         can not be shared.
18953 2013-01-14  Georg-Johann Lay  <avr@gjlay.de>
18955         * config/avr/avr-modes.def: Add GPL copyright notice.
18957 2013-01-13  Uros Bizjak  <ubizjak@gmail.com>
18959         * config/i386/sync.md (mem_thread_fence): Mask operands[0] with
18960         MEMMODEL_MASK to determine memory model.
18961         (atomic_store<mode>): Ditto from operands[2].
18962         * config/i386/i386.c (ix86_memmodel_check): Declare "strong" as bool.
18964 2013-01-13  Jakub Jelinek  <jakub@redhat.com>
18966         PR fortran/55935
18967         * gimple-fold.c (get_symbol_constant_value): Call unshare_expr.
18968         (fold_gimple_assign): Don't call unshare_expr here.
18969         (fold_ctor_reference): Call unshare_expr.
18971 2013-01-13  Terry Guo  <terry.guo@arm.com>
18973         * Makefile.in (s-mlib): New argument MULTILIB_REUSE.
18974         * doc/fragments.texi: Document MULTILIB_REUSE.
18975         * gcc.c (multilib_reuse): New internal spec.
18976         (set_multilib_dir): Also search multilib from multilib_reuse.
18977         * genmultilib (tmpmultilib3): Refactor code.
18978         (tmpmultilib4): Ditto.
18979         (multilib_reuse): New multilib argument.
18981 2013-01-13  Richard Sandiford  <rdsandiford@googlemail.com>
18983         * Makefile.in: Update copyright.
18985 2013-01-12  Tom de Vries  <tom@codesourcery.com>
18987         PR middle-end/55890
18988         * calls.c (expand_call): Check if arg_nr is valid.
18990 2013-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
18992         * doc/extend.texi (X86 Built-in Functions): Add whitespace in
18993         __builtin_ia32_paddb256 and __builtin_ia32_pavgb256
18994         documentation.  Add missing '__' in front of
18995         __builtin_ia32_packssdw256.
18997 2013-01-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
18999         PR target/55719
19000         * config/s390/s390.c (s390_preferred_reload_class): Do not return
19001         NO_REGS for larl operands.
19002         (s390_reload_larl_operand): Use s390_load_address instead of
19003         emit_move_insn.
19005 2013-01-11  Richard Biener  <rguenther@suse.de>
19007         * tree-cfg.c (verify_node_sharing_1): Split out from ...
19008         (verify_node_sharing): ... here.
19009         (verify_gimple_in_cfg): Use verify_node_sharing_1 for walk_tree.
19011 2013-01-11  Eric Botcazou  <ebotcazou@adacore.com>
19013         * configure.ac (Tree checking): Set TREECHECKING to yes if enabled.
19014         Substitute TREECHECKING.
19015         * configure: Regenerate.
19016         * Makefile.in (TREECHECKING): New.
19018 2013-01-11  Richard Guenther  <rguenther@suse.de>
19020         PR tree-optimization/44061
19021         * tree-vrp.c (extract_range_basic): Compute zero as
19022         value-range for __builtin_constant_p of function parameters.
19024 2013-01-10  Richard Sandiford  <rdsandiford@googlemail.com>
19026         Update copyright years.
19028 2013-01-10  Vladimir Makarov  <vmakarov@redhat.com>
19030         PR rtl-optimization/55672
19031         * lra-eliminations.c (mark_not_eliminable): Permit addition with
19032         const to be eliminable.
19034 2013-01-10  David Edelsohn  <dje.gcc@gmail.com>
19036         * configure.ac (HAVE_AS_TLS): Add check for powerpc-ibm-aix.
19037         * configure: Regenerate.
19039 2013-01-10  Richard Biener  <rguenther@suse.de>
19041         * builtins.c (expand_builtin_init_trampoline): Use set_mem_attributes.
19043 2013-01-10  Richard Biener  <rguenther@suse.de>
19045         PR bootstrap/55792
19046         * tree-into-ssa.c (rewrite_add_phi_arguments): Do not set
19047         locations for virtual PHI arguments.
19048         (rewrite_update_phi_arguments): Likewise.
19050 2013-01-10  Joel Sherrill  <joel.sherrill@OARcorp.com>
19052         * config/v850/rtems.h (ASM_SPEC): Pass -m8byte-align and -mgcc-abi
19053         on to assembler.
19055 2013-01-10  Jakub Jelinek  <jakub@redhat.com>
19057         PR tree-optimization/55921
19058         * tree-complex.c (expand_complex_asm): New function.
19059         (expand_complex_operations_1): Call it for GIMPLE_ASM.
19061 2013-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
19063         PR target/55718
19064         * config/s390/s390.c (s390_symref_operand_p)
19065         (s390_loadrelative_operand_p): Merge the two functions.
19066         (s390_check_qrst_address, print_operand_address): Add parameters
19067         to s390_loadrelative_operand_p invokation.
19068         (s390_check_symref_alignment): Use s390_loadrelative_operand_p.
19069         (s390_reload_larl_operand, s390_secondary_reload): Use
19070         s390_loadrelative_operand_p instead of s390_symref_operand_p.
19071         (legitimize_pic_address): Handle @GOTENT and @PLT + addend.
19073 2013-01-09  Mike Stump  <mikestump@comcast.net>
19075         * dse.c (record_store): Remove unnecessary assert.
19077 2013-01-09  Jan Hubicka  <jh@suse.cz>
19079         PR tree-optimization/55569
19080         * cfgloopmanip.c (scale_loop_profile): Make ITERATION_BOUND gcov_type.
19081         * cfgloop.h (scale_loop_profile): Likewise.
19083 2013-01-09  Jan Hubicka  <jh@suse.cz>
19085         PR lto/45375
19086         * ipa-inline.c (ipa_inline): Remove extern inlines and virtual
19087         functions.
19088         * cgraphclones.c (cgraph_clone_node): Cpoy also LTO file data.
19090 2013-01-09  Richard Sandiford  <rdsandiford@googlemail.com>
19092         PR middle-end/55114
19093         * expr.h (maybe_emit_group_store): Declare.
19094         * expr.c (maybe_emit_group_store): New function.
19095         * builtins.c (expand_builtin_int_roundingfn): Call it.
19096         (expand_builtin_int_roundingfn_2): Likewise.
19098 2013-01-09  Vladimir Makarov  <vmakarov@redhat.com>
19100         PR rtl-optimization/55829
19101         * lra-constraints.c (match_reload): Add code for absent output.
19102         (curr_insn_transform): Add code for reloads of matched inputs
19103         without output.
19105 2013-01-09  Uros Bizjak  <ubizjak@gmail.com>
19107         * config/i386/sse.md (*vec_interleave_highv2df): Change mode
19108         attribute of movddup insn to DF.
19109         (*vec_interleave_lowv2df): Ditto.
19110         (vec_dupv2df): Ditto.
19112 2013-01-09  Jan Hubicka  <jh@suse.cz>
19114         PR tree-optimiation/55875
19115         * tree-ssa-loop-niter.c (number_of_iterations_cond): Add
19116         EVERY_ITERATION parameter.
19117         (number_of_iterations_exit): Check if exit is executed every iteration.
19118         (idx_infer_loop_bounds): Similarly here.
19119         (n_of_executions_at_most): Simplify
19120         to only test for cases where statement is dominated by the
19121         particular bound; handle correctly the "postdominance" test.
19122         (scev_probably_wraps_p): Use max loop iterations info
19123         as a global bound first.
19125 2013-01-09  Nguyen Duy Dat  <dat.nguyen.yn@rvc.renesas.com>
19126             Nick Clifton  <nickc@redhat.com>
19128         * config/v850/v850.md (cbranchsf4): New pattern.
19129         (cstoresf4): New pattern.
19130         (cbranchdf4): New pattern.
19131         (cstoredf4): New pattern.
19132         (movsicc): Disallow floating point comparisons.
19133         (cmpsf_le_insn): Fix order of operators.
19134         (cmpsf_lt_insn): Likewise.
19135         (cmpsf_eq_insn): Likewise.
19136         (cmpdf_le_insn): Likewise.
19137         (cmpdf_lt_insn): Likewise.
19138         (cmpdf_eq_insn): Likewise.
19139         (cmpsf_ge_insn): Use LE comparison.
19140         (cmpdf_ge_insn): Likewise.
19141         (cmpsf_gt_insn): Use LT comparison.
19142         (cmpdf_gt_insn): Likewise.
19143         (cmpsf_ne_insn): Delete pattern.
19144         (cmpdf_ne_insn): Delete pattern.
19145         * config/v850/v850.c (v850_gen_float_compare): Use
19146         gen_cmpdf_eq_insn for NE comparison.
19147         (v850_float_z_comparison_operator)
19148         (v850_float_nz_comparison_operator): Move from here ...
19149         * config/v850/predicates.md: ... to here.  Move GT and GE
19150         comparisons into v850_float_z_comparison_operator.
19151         * config/v850/v850-protos.h (v850_float_z_comparison_operator):
19152         Delete prototype.
19153         (v850_float_nz_comparison_operator): Likewise.
19155 2013-01-09  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
19157         * config/pa/pa.c (pa_emit_move_sequence): Replace calls to gen_insv
19158         with calls to gen_insvsi/gen_insvdi.
19160 2013-01-09  Venkataramanan Kumar  <venkataramanan.kumar@amd.com>
19162         * config/i386/i386.c (initial_ix86_tune_features): Set up
19163         X86_TUNE_AVX128_OPTIMAL for m_BTVER2.
19165 2013-01-09  Steven Bosscher  <steven@gcc.gnu.org>
19166             Jakub Jelinek  <jakub@redhat.com>
19168         PR tree-optimization/48189
19169         * predict.c (predict_loops): If max is 0, don't call compare_tree_int.
19170         If nitercst is 0, don't predict the exit edge.
19172 2013-01-08  Naveen H.S  <Naveen.Hurugalawadi@caviumnetworks.com>
19174         * config/aarch64/aarch64.c (aarch64_print_operand): Replace %r
19175         in asm_fprintf with reg_names.
19176         (aarch64_print_operand_address): Likewise.
19177         (aarch64_return_addr): Likewise.
19178         * config/aarch64/aarch64.h (ASM_FPRINTF_EXTENSIONS): Remove.
19180 2013-01-08  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
19182         * config/pa/pa.h (VAL_U6_BITS_P): Define.
19183         (INT_U6_BITS): Likewise.
19184         * config/pa/predicates.md (uint6_operand): New predicate.
19185         (shift5_operand, shift6_operand): Likewise.
19186         * config/pa/pa.md (lshrsi3, rotrsi3): Use shift5_operand instead of
19187         arith32_operand.
19188         (lshrdi3): Use shift6_operand.
19189         (shrpsi4, shrpdi4): New insn patterns.
19190         (extzv): Delete expander.
19191         (extzvsi, extzvdi): New expanders.  Use uint5_operand and uint6_operand
19192         predicates in unamed zero extract patterns.  Tighten common constraint.
19193         (extv): Delete expander.
19194         (extvsi, extvdi): New expanders.  Use uint5_operand and uint6_operand
19195         predicates in unamed sign extract patterns.  Tighten common constraint.
19196         (insv): Delete expander.
19197         (insvsi, insvdi): New expanders.  Use uint5_operand and uint6_operand
19198         predicates in unamed insert patterns.  Tighten common constraint.
19199         Change uint32_operand predicate to uint6_operand predicate in unamed
19200         DImode pattern to insert constant values of type 1...1xxxx.
19202 2013-01-04  Jan Hubicka  <jh@suse.cz>
19204         PR tree-optimization/55823
19205         * ipa-prop.c (update_indirect_edges_after_inlining): Fix ordering
19206         issue.
19208 2013-01-08  Jakub Jelinek  <jakub@redhat.com>
19209             Uros Bizjak  <ubizjak@gmail.com>
19211         PR rtl-optimization/55845
19212         * df-problems.c (can_move_insns_across): Stop scanning at
19213         volatile_insn_p source instruction or give up if
19214         across_from .. across_to range contains any volatile_insn_p
19215         instructions.
19217 2013-01-08  Tejas Belagod  <tejas.belagod@arm.com>
19219         * config/aarch64/aarch64-simd.md (vec_init<mode>): New.
19220         * config/aarch64/aarch64-protos.h (aarch64_expand_vector_init):
19221         Declare.
19222         * config/aarch64/aarch64.c (aarch64_simd_dup_constant,
19223         aarch64_simd_make_constant, aarch64_expand_vector_init): New.
19225 2013-01-08  Jakub Jelinek  <jakub@redhat.com>
19227         PR fortran/55341
19228         * asan.c (asan_clear_shadow): New function.
19229         (asan_emit_stack_protection): Use it.
19231 2013-01-08  Tejas Belagod  <tejas.belagod@arm.com>
19233         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_<su>mult_lo_<mode>,
19234         aarch64_simd_vec_<su>mult_hi_<mode>): Separate instruction and operand
19235         with tab instead of space.
19237 2013-01-08  Nick Clifton  <nickc@redhat.com>
19239         * config/rl78/rl78.c (rl78_expand_prologue): Always select
19240         register bank 0 at the start of an interrupt handler.
19241         * config/rl78/rl78.md (mulsi3_g13): Correct values for MDBL and
19242         MDBH registers.
19244 2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
19246         * config/aarch64/aarch64-simd.md
19247         (aarch64_simd_bsl<mode>_internal): Add floating-point modes.
19248         (aarch64_simd_bsl): Likewise.
19249         (aarch64_vcond_internal<mode>): Likewise.
19250         (vcond<mode><mode>): Likewise.
19251         (aarch64_cm<cmp><mode>): Fix constraints, add new modes.
19252         * config/aarch64/iterators.md (V_cmp_result): Add V2DF.
19254 2013-01-08  James Greenhalgh  <james.greenhalgh@arm.com>
19256         * config/aarch64/aarch64-builtins.c
19257         (aarch64_builtin_vectorized_function): Handle sqrt, sqrtf.
19259 2013-01-08  Martin Jambor  <mjambor@suse.cz>
19261         PR debug/55579
19262         * tree-sra.c (analyze_access_subtree): Return true also after
19263         potentially creating a debug-only replacement.
19265 2013-01-08  Jakub Jelinek  <jakub@redhat.com>
19267         PR middle-end/55890
19268         * tree-ssa-ccp.c (evaluate_stmt): Use gimple_call_builtin_p.
19270         PR tree-optimization/54120
19271         * tree-vrp.c (range_fits_type_p): Don't allow
19272         src_precision < precision from signed vr to unsigned_p
19273         if vr->min or vr->max is negative.
19274         (simplify_float_conversion_using_ranges): Test can_float_p
19275         against CODE_FOR_nothing.
19277 2013-01-08  Jakub Jelinek  <jakub@redhat.com>
19278             Richard Biener  <rguenther@suse.de>
19280         PR middle-end/55851
19281         * fold-const.c (int_binop_types_match_p): Allow all INTEGRAL_TYPE_P
19282         types instead of just INTEGER_TYPE types.
19284 2013-01-07  Mark Kettenis  <kettenis@openbsd.org>
19286         * config/i386/openbsdelf.h (LIBGCC2_HAS_TF_MODE, LIBGCC2_TF_CEXT,
19287         TF_SIZE): Define.
19289 2013-01-07  Steve Ellcey  <sellcey@mips.com>
19291         PR target/42661
19292         * config/mips/mips.opt: Change mad to mmad to match documentation.
19294 2013-01-07  Georg-Johann Lay  <avr@gjlay.de>
19296         PR target/55897
19297         * doc/extend.texi (AVR Named Address Spaces): __memx goes into
19298         .progmemx.data now.
19300 2013-01-07  Georg-Johann Lay  <avr@gjlay.de>
19302         PR target/55897
19303         * config/avr/avr.h (ADDR_SPACE_COUNT): New enum.
19304         (avr_addrspace_t): Add .section_name field.
19305         * config/avr/avr.c (progmem_section): Use ADDR_SPACE_COUNT as
19306         array size.
19307         (avr_addrspace): Same.  Initialize .section_name.  Remove last
19308         NULL entry.  Put __memx into .progmemx.data.
19309         (progmem_section_prefix): Remove.
19310         (avr_asm_init_sections): No need to initialize progmem_section.
19311         (avr_asm_named_section): Use avr_addrspace[].section_name to get
19312         section name prefix.
19313         (avr_asm_select_section): Ditto.  And use get_unnamed_section to
19314         retrieve the progmem section.
19315         * avr-c.c (avr_cpu_cpp_builtins): Use ADDR_SPACE_COUNT as loop
19316         boundary to run over avr_addrspace[].
19317         (avr_register_target_pragmas): Ditto.
19319 2013-01-06  Jakub Jelinek  <jakub@redhat.com>
19321         * varasm.c (output_constant_def_contents): For asan_protect_global
19322         protected strings, adjust DECL_ALIGN if needed, before testing for
19323         anchored symbols.
19324         (place_block_symbol): Adjust size for asan protected STRING_CSTs if
19325         TREE_CONSTANT_POOL_ADDRESS_P.  Increase alignment for asan protected
19326         normal decls.
19327         (output_object_block): For asan protected decls, emit asan padding
19328         after their contents.
19329         * asan.c (asan_protect_global): Don't check TREE_ASM_WRITTEN here.
19330         (asan_finish_file): Test it here instead.
19332 2013-01-07  Nick Clifton  <nickc@redhat.com>
19333             Matthias Klose  <doko@debian.org>
19334             Doug Kwan  <dougkwan@google.com>
19335             H.J. Lu  <hongjiu.lu@intel.com>
19337         PR driver/55470
19338         * collect2.c (main): Support -fuse-ld=bfd and -fuse-ld=gold.
19340         * common.opt: Add fuse-ld=bfd and fuse-ld=gold.
19342         * gcc.c (LINK_COMMAND_SPEC): Pass -fuse-ld=* to collect2.
19344         * opts.c (comman_handle_option): Ignore -fuse-ld=bfd and -fuse-ld=gold.
19346         * doc/invoke.texi: Document -fuse-ld=bfd and -fuse-ld=gold.
19348 2013-01-07  Georg-Johann Lay  <avr@gjlay.de>
19350         PR target/54461
19351         * doc/install.texi (Cross-Compiler-Specific Options): Document
19352         --with-avrlibc.
19354 2013-01-07  Tejas Belagod  <tejas.belagod@arm.com>
19356         * config/aarch64/arm_neon.h (vmovn_high_is16, vmovn_high_s32,
19357         vmovn_high_s64, vmovn_high_u16, vmovn_high_u32, vmovn_high_u64,
19358         vqmovn_high_s16, vqmovn_high_s32, vqmovn_high_s64, vqmovn_high_u16,
19359         vqmovn_high_u32, vqmovn_high_u64, vqmovun_high_s16, vqmovun_high_s32,
19360         vqmovun_high_s64): Fix source operand number and update copyright.
19362 2013-01-07  Richard Biener  <rguenther@suse.de>
19364         PR middle-end/55890
19365         * gimple.h (gimple_call_builtin_p): New overload.
19366         * gimple.c (validate_call): New function.
19367         (gimple_call_builtin_p): Likewise.
19368         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
19369         Use gimple_call_builtin_p.
19370         (find_func_clobbers): Likewise.
19371         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
19372         (strlen_optimize_stmt): Likewise.
19374 2013-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
19376         * config/aarch64/arm_neon.h (vld1_dup_*): Make argument const.
19377         (vld1q_dup_*): Likewise.
19378         (vld1_*): Likewise.
19379         (vld1q_*): Likewise.
19380         (vld1_lane_*): Likewise.
19381         (vld1q_lane_*): Likewise.
19383 2013-01-07  Richard Biener  <rguenther@suse.de>
19385         * lto-streamer.h (LTO_minor_version): Bump to 2.
19387 2013-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
19389         * config/aarch64/aarch64-protos.h
19390         (aarch64_const_double_zero_rtx_p): Rename to...
19391         (aarch64_float_const_zero_rtx_p): ...this.
19392         (aarch64_float_const_representable_p): New.
19393         (aarch64_output_simd_mov_immediate): Likewise.
19394         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>): Refactor
19395         move immediate case.
19396         * config/aarch64/aarch64.c
19397         (aarch64_const_double_zero_rtx_p): Rename to...
19398         (aarch64_float_const_zero_rtx_p): ...this.
19399         (aarch64_print_operand): Allow printing of new constants.
19400         (aarch64_valid_floating_const): New.
19401         (aarch64_legitimate_constant_p): Check for valid floating-point
19402         constants.
19403         (aarch64_simd_valid_immediate): Likewise.
19404         (aarch64_vect_float_const_representable_p): New.
19405         (aarch64_float_const_representable_p): Likewise.
19406         (aarch64_simd_imm_zero_p): Also allow for floating-point 0.0.
19407         (aarch64_output_simd_mov_immediate): New.
19408         * config/aarch64/aarch64.md (*movsf_aarch64): Add new alternative.
19409         (*movdf_aarch64): Likewise.
19410         * config/aarch64/constraints.md (Ufc): New.
19411         (Y): call aarch64_float_const_zero_rtx.
19412         * config/aarch64/predicates.md (aarch64_fp_compare_operand): New.
19414 2013-01-07  Richard Biener  <rguenther@suse.de>
19416         PR tree-optimization/55888
19417         PR tree-optimization/55862
19418         * tree-ssa-pre.c (phi_translate_1): Revert previous change.
19419         (valid_in_sets): Check if a NAME has a leader in AVAIL_OUT,
19420         not if it is contained therein.
19422 2013-01-07  Georg-Johann Lay  <avr@gjlay.de>
19424         * config/avr/t-avr: Typo.
19426 2013-01-07  Georg-Johann Lay  <avr@gjlay.de>
19428         PR55243
19429         * config/avr/t-avr: Don't automatically rebuild
19430         $(srcdir)/config/avr/t-multilib
19431         $(srcdir)/config/avr/avr-tables.opt
19432         $(srcdir)/doc/avr-mmcu.texi
19433         (avr-mcus): New phony target to build them on request.
19434         (s-avr-mlib, s-avr-mmcu-texi): Remove.
19435         * avr/avr-mcus.def: Adjust comments.
19437 2013-01-07  Uros Bizjak  <ubizjak@gmail.com>
19439         * config/i386/i386.c (DEFAULT_PCC_STRUCT_RETURN): Remove.
19441 2013-01-06  Richard Sandiford  <rdsandiford@googlemail.com>
19443         * file-find.c, file-find.h, realmpfr.c: Add FSF as copyright holder.
19445 2013-01-06  Richard Sandiford  <rdsandiford@googlemail.com>
19447         * config/tilepro/gen-mul-tables.cc: Put copyright on one line.
19449 2013-01-05  David Edelsohn  <dje.gcc@gmail.com>
19451         * config/rs6000/aix53.h (LIB_SPEC): Add -lpthreads when compiling
19452         to generate profiling.
19453         * config/rs6000/aix64.h (LIB_SPEC): Same.
19455 2013-01-04  Andrew Pinski  <apinski@cavium.com>
19457         * config/aarch64/aarch64.c (aarch64_fixed_condition_code_regs):
19458         New function.
19459         (TARGET_FIXED_CONDITION_CODE_REGS): Define.
19461 2013-01-04  Uros Bizjak  <ubizjak@gmail.com>
19463         * config/i386/i386.c (ix86_legitimize_address): Call convert_to_mode
19464         unconditionally.
19465         (ix86_expand_move): Ditto.
19466         (ix86_zero_extend_to_Pmode): Ditto.
19467         (ix86_expand_call): Ditto.
19468         (ix86_expand_special_args_builtin): Ditto.
19469         (ix86_expand_builtin): Ditto.
19471 2013-01-04  Richard Biener  <rguenther@suse.de>
19473         PR tree-optimization/55862
19474         * tree-ssa-pre.c (phi_translate_1): Valueize SSA names after
19475         translating them through PHI nodes.
19477 2013-01-04  Martin Jambor  <mjambor@suse.cz>
19479         PR tree-optimization/55755
19480         * tree-sra.c (sra_modify_assign): Do not check that an access has no
19481         children when trying to avoid producing a VIEW_CONVERT_EXPR.
19483 2013-01-04  Marek Polacek  <polacek@redhat.com>
19485         PR middle-end/55859
19486         * opts.c (default_options_optimization): Clarify error message.
19488 2013-01-04  Richard Biener  <rguenther@suse.de>
19490         PR middle-end/55863
19491         * fold-const.c (split_tree): Undo -X - 1 to ~X folding for
19492         reassociation.
19494 2013-01-03  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
19496         PR target/53789
19497         * config/pa/pa.md (movsi): Revert previous change.
19498         * config/pa/pa.c (pa_legitimate_constant_p): Reject all TLS symbol
19499         references.
19501 2013-01-03  Richard Henderson  <rth@redhat.com>
19503         * config/i386/i386.c (ix86_expand_move): Always assign to op1
19504         after eliminating TLS symbols.
19506 2013-01-03  Marc Glisse  <marc.glisse@inria.fr>
19508         PR bootstrap/50167
19509         * graphite-interchange.c (pdr_stride_in_loop): Use gmp_fprintf.
19510         * graphite-poly.c (debug_gmp_value): Likewise.
19512 2013-01-03  Uros Bizjak  <ubizjak@gmail.com>
19514         PR target/55712
19515         * config/i386/i386-c.c (ix86_target_macros_internal): Depending on
19516         selected code model, define __code_mode_small__, __code_model_medium__,
19517         __code_model_large__, __code_model_32__ or __code_model_kernel__.
19518         * config/i386/cpuid.h (__cpuid, __cpuid_count) [__i386__]: Prefix
19519         xchg temporary register with %k.  Declare temporary register as
19520         early clobbered.
19521         [__x86_64__]: For medium and large code models, preserve %rbx register.
19523 2013-01-03  Richard Biener  <rguenther@suse.de>
19525         * tree-data-ref.c (dump_conflict_function): Use less vertical spacing.
19526         (dump_subscript): Adjust.
19527         (finalize_ddr_dependent): Do not dump redundant info.
19528         (analyze_siv_subscript): Adjust.
19529         (subscript_dependence_tester): Likewise.
19530         (compute_affine_dependence): Likewise.
19532 2013-01-03  Richard Biener  <rguenther@suse.de>
19534         Revert
19535         2013-01-03  Richard Biener  <rguenther@suse.de>
19537         PR tree-optimization/55857
19538         * tree-vect-stmts.c (vectorizable_load): Do not setup
19539         re-alignment for invariant loads.
19541         2013-01-02  Richard Biener  <rguenther@suse.de>
19543         * tree-vect-stmts.c (vectorizable_load): When vectorizing an
19544         invariant load do not generate a vector load from the scalar location.
19546 2013-01-03  Richard Biener  <rguenther@suse.de>
19548         * tree-vect-loop.c (vect_analyze_loop_form): Clarify reason
19549         for not vectorizing.
19550         * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Do
19551         not build INDIRECT_REFs, call get_name once only.
19552         (vect_create_data_ref_ptr): Likewise.  Dump base object kind
19553         based on DR_BASE_OBJECT, not DR_BASE_ADDRESS.
19555 2013-01-03  Richard Biener  <rguenther@suse.de>
19557         PR tree-optimization/55857
19558         * tree-vect-stmts.c (vectorizable_load): Do not setup
19559         re-alignment for invariant loads.
19561 2013-01-03  Richard Biener  <rguenther@suse.de>
19563         PR lto/55848
19564         * lto-symtab.c (lto_symtab_merge_decls_1): As last resort, always
19565         prefer a built-in decl.
19567 2013-01-03  Jakub Jelinek  <jakub@redhat.com>
19569         * gcc.c (process_command): Update copyright notice dates.
19570         * gcov.c (print_version): Likewise.
19571         * gcov-dump.c (print_version): Likewise.
19573         PR rtl-optimization/55838
19574         * loop-iv.c (iv_number_of_iterations): Call lowpart_subreg on
19575         iv0.step, iv1.step and step.
19577 2013-01-03  Jakub Jelinek  <jakub@redhat.com>
19578             Marc Glisse  <marc.glisse@inria.fr>
19580         PR tree-optimization/55832
19581         * fold-const.c (fold_binary_loc): For ABS_EXPR<x> >= 0 and
19582         ABS_EXPR<x> < 0 folding use constant_boolean_node instead of
19583         integer_{one,zero}_node.
19585 2013-01-03  Jakub Jelinek  <jakub@redhat.com>
19587         PR debug/54402
19588         * params.def (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE): New param.
19589         * var-tracking.c (reverse_op): Don't add reverse ops to
19590         VALUEs that have already
19591         PARAM_VALUE (PARAM_MAX_VARTRACK_REVERSE_OP_SIZE) or longer locs list.
19593 2013-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
19595         * doc/contrib.texi: Note years as release manager for Mark Mitchell.
19597 2013-01-02  Teresa Johnson  <tejohnson@google.com>
19599         * dumpfile.c (dump_loc): Print filename with location.
19600         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Use
19601         new location_t parameter to emit complete unroll message with
19602         new dump framework.
19603         (canonicalize_loop_induction_variables): Compute loops location
19604         and pass to try_unroll_loop_completely.
19605         * loop-unroll.c (report_unroll_peel): New function.
19606         (peel_loops_completely): Use new dump format with location
19607         for main dumpfile message, and invoke report_unroll_peel on success.
19608         (decide_unrolling_and_peeling): Ditto.
19609         (decide_peel_once_rolling): Remove old dumpfile message subsumed
19610         by report_unroll_peel.
19611         (decide_peel_completely): Ditto.
19612         (decide_unroll_constant_iterations): Ditto.
19613         (decide_unroll_runtime_iterations): Ditto.
19614         (decide_peel_simple): Ditto.
19615         (decide_unroll_stupid): Ditto.
19616         * cfgloop.c (get_loop_location): New function.
19617         * cfgloop.h (get_loop_location): Declare.
19619 2013-01-02  Sriraman Tallam  <tmsriram@google.com>
19621         * config/i386/i386.c (fold_builtin_cpu): Remove unnecessary checks for
19622         NULL.
19624 2013-01-02  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
19626         PR middle-end/55198
19627         * expr.c (expand_expr_real_1): Don't use bitfield extraction for non
19628         BLKmode objects when EXPAND_MEMORY is specified.
19630 2013-01-02  Sriraman Tallam  <tmsriram@google.com>
19632         * config/i386/i386.c (ix86_get_function_versions_dispatcher): Fix bug
19633         in loop predicate.
19634         (fold_builtin_cpu): Do not share cpu model decls across statements.
19636 2013-01-02  Jason Merrill  <jason@redhat.com>
19638         PR c++/55804
19639         * tree.c (build_array_type_1): Revert earlier change.
19641 2013-01-02  Yufeng Zhang  <yufeng.zhang@arm.com>
19643         * config/aarch64/aarch64-cores.def: Add entries for "cortex-a53" and
19644         "cortex-a57".
19645         * config/aarch64/aarch64-tune.md: Re-generate.
19647 2013-01-02  Richard Biener  <rguenther@suse.de>
19649         * tree-vect-stmts.c (vectorizable_load): When vectorizing an
19650         invariant load do not generate a vector load from the scalar location.
19652 2013-01-02  Richard Biener  <rguenther@suse.de>
19654         PR bootstrap/55784
19655         * configure.ac: Add $GMPINC to CFLAGS/CXXFLAGS.
19656         * configure: Regenerate.
19658 2013-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
19660         * builtins.c (expand_builtin_mathfn, expand_builtin_mathfn_2)
19661         (expand_builtin_mathfn_ternary, expand_builtin_mathfn_3)
19662         (expand_builtin_int_roundingfn_2): Keep the original target around
19663         for the fallback case.
19665 2013-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
19667         * tree-vrp.c (range_fits_type_p): Require the MSB of the double_int
19668         to be clear for sign changes.
19670 2013-01-01  Jan Hubicka  <jh@suse.cz>
19672         * ipa-inline-analysis.c: Fix formatting.
19674 2013-01-01  Jakub Jelinek  <jakub@redhat.com>
19676         PR tree-optimization/55831
19677         * tree-vect-loop.c (get_initial_def_for_induction): Use
19678         gsi_after_labels instead of gsi_start_bb.
19680 Copyright (C) 2013 Free Software Foundation, Inc.
19682 Copying and distribution of this file, with or without modification,
19683 are permitted in any medium without royalty provided the copyright
19684 notice and this notice are preserved.