2014-04-15 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blobbfba5a5fdf405757ae4c877fb7158bb3c909aa0b
1 2014-04-15  Richard Biener  <rguenther@suse.de>
3         * alias.c (ncr_compar): New function.
4         (nonoverlapping_component_refs_p): Re-implement in O (n log n).
6 2014-04-15  Richard Biener  <rguenther@suse.de>
8         * alias.c (record_component_aliases): Do not walk BINFOs.
10 2014-04-15  Richard Biener  <rguenther@suse.de>
12         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13         Add struct function argument and adjust.
14         (find_func_aliases_for_call): Likewise.
15         (find_func_aliases): Likewise.
16         (find_func_clobbers): Likewise.
17         (intra_create_variable_infos): Likewise.
18         (compute_points_to_sets): Likewise.
19         (ipa_pta_execute): Adjust.  Do not push/pop cfun.
21 2014-04-15  Richard Biener  <rguenther@suse.de>
23         * tree.c (iterative_hash_expr): Use enum tree_code_class
24         to store TREE_CODE_CLASS.
25         (tree_block): Likewise.
26         (tree_set_block): Likewise.
27         * tree.h (fold_build_pointer_plus_loc): Use
28         convert_to_ptrofftype_loc.
30 2014-04-15  Jakub Jelinek  <jakub@redhat.com>
32         PR plugins/59335
33         * Makefile.in (PLUGIN_HEADERS): Add various headers that have been
34         added in 4.9.
36 2014-04-15  Eric Botcazou  <ebotcazou@adacore.com>
38         * cfgloop.h (struct loop): Move force_vectorize down.
39         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Handle new kinds.
40         (gimplify_expr) <ANNOTATE_EXPR>: Minor tweak.
41         * lto-streamer-in.c (input_cfg): Read dont_vectorize field.
42         * lto-streamer-out.c (output_cfg): Write dont_vectorize field.
43         * tree-cfg.c (replace_loop_annotate): Revamp and handle new kinds.
44         * tree-core.h (enum annot_expr_kind): Add new kind values.
45         * tree-inline.c (copy_loops): Copy dont_vectorize field and reorder.
46         * tree-pretty-print.c (dump_generic_node) <ANNOTATE_EXPR>: Handle new
47         kinds.
48         * tree.def (ANNOTATE_EXPR): Tweak comment.
50 2014-04-14  Jan Hubicka  <hubicka@ucw.cz>
52         * ipa-devirt.c (maybe_record_node): Ignore all non-methods (including
53         cxa_pure_virtual).
55 2014-04-14  Paolo Carlini  <paolo.carlini@oracle.com>
57         * tree.h (TYPE_IDENTIFIER): Declare.
58         * tree.c (subrange_type_for_debug_p): Use it.
59         * godump.c (go_format_type): Likewise.
60         * dwarf2out.c (is_cxx_auto, modified_type_die,
61         gen_type_die_with_usage, gen_type_die_with_usage): Likewise.
62         * dbxout.c (dbxout_type, dbxout_symbol): Likewise.
64 2014-04-14  Jan Hubicka   <hubicka@ucw.cz>
66         PR lto/60820
67         * varpool.c (varpool_remove_node): Do not alter decls when streaming.
69 2014-04-14  Uros Bizjak  <ubizjak@gmail.com>
71         * config/i386/i386.c (examine_argument): Return bool.  Return true if
72         parameter should be passed in memory.
73         <case X86_64_COMPLEX_X87_CLASS>: Adjust.
74         (construct_container): Update calls to examine_argument.
75         (function_arg_advance_64): Ditto.
76         (return_in_memory_32): Merge with ix86_return_in_memory.
77         (return_in_memory_64): Ditto.
78         (return_in_memory_ms_64): Ditto.
80 2014-04-14  Jan Hubicka  <hubicka@ucw.cz>
82         * ipa-utils.c (ipa_merge_profiles): Merge profile_id.
83         * coverage.c (coverage_compute_profile_id): Handle externally visible
84         symbols.
86 2014-04-14  Martin Jambor  <mjambor@suse.cz>
88         * tree-sra.c (ipa_sra_preliminary_function_checks): Skip
89         DECL_DISREGARD_INLINE_LIMITS functions.
91 2014-04-14  H.J. Lu  <hongjiu.lu@intel.com>
93         PR target/60827
94         * config/i386/i386.md (*fixuns_trunc<mode>_1): Revert the last change.
96 2014-04-14  H.J. Lu  <hongjiu.lu@intel.com>
98         PR target/60827
99         * config/i386/i386.md (*fixuns_trunc<mode>_1): Check
100         optimize_insn_for_speed_p instead of
101         optimize_function_for_speed_p.
103 2014-04-14  Yufeng Zhang  <yufeng.zhang@arm.com>
105         * doc/invoke.texi (free): Document AArch64.
107 2014-04-14  Richard Biener  <rguenther@suse.de>
109         PR tree-optimization/60042
110         * tree-ssa-pre.c (inhibit_phi_insertion): Remove.
111         (insert_into_preds_of_block): Do not prevent PHI insertion
112         for REFERENCE exprs here ...
113         (eliminate_dom_walker::before_dom_children): ... but prevent
114         their use here under similar conditions when applied to the
115         IL after PRE optimizations.
117 2014-04-14  Richard Biener  <rguenther@suse.de>
119         * passes.def: Move early points-to after early SRA.
121 2014-04-14  Richard Biener  <rguenther@suse.de>
123         * tree-ssa-forwprop.c (simplify_gimple_switch): Enhance
124         check for which sign-changes we allow when forwarding
125         a converted value into a switch.
127 2014-04-14  Eric Botcazou  <ebotcazou@adacore.com>
129         * stor-layout.c (place_field): Finalize non-constant offset for the
130         field, if any.
132 2014-04-14  Richard Biener  <rguenther@suse.de>
134         * tree-switch-conversion.c (lshift_cheap_p): Get speed_p
135         as argument.
136         (expand_switch_using_bit_tests_p): Likewise.
137         (process_switch): Compute and pass on speed_p based on the
138         switch stmt.
139         * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Use
140         optimize_bb_for_speed_p.
142 2014-04-14  Eric Botcazou  <ebotcazou@adacore.com>
144         * cfgloop.h (struct loop): Rename force_vect into force_vectorize.
145         * function.h (struct function): Rename has_force_vect_loops into
146         has_force_vectorize_loops.
147         * lto-streamer-in.c (input_cfg): Adjust for renaming.
148         (input_struct_function_base): Likewise.
149         * lto-streamer-out.c (output_cfg): Likewise.
150         (output_struct_function_base): Likewise.
151         * omp-low.c (expand_omp_simd): Likewise.
152         * tree-cfg.c (move_sese_region_to_fn): Likewise.
153         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
154         (version_loop_for_if_conversion): Likewise.
155         (tree_if_conversion): Likewise.
156         (main_tree_if_conversion): Likewise.
157         (gate_tree_if_conversion): Likewise.
158         * tree-inline.c (copy_loops): Likewise.
159         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Likewise.
160         * tree-ssa-loop.c (tree_loop_vectorize): Likewise.
161         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Likewise.
162         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Likewise.
163         * tree-vectorizer.c (vectorize_loops): Likewise.
164         * tree-vectorizer.h (unlimited_cost_model): Likewise.
166 2014-04-14  Richard Biener  <rguenther@suse.de>
168         PR lto/60720
169         * lto-streamer-out.c (wrap_refs): New function.
170         (lto_output): Wrap symbol references in global initializes in
171         type-preserving MEM_REFs.
173 2014-04-14  Christian Bruel  <christian.bruel@st.com>
175         * config/sh/sh-mem.cc (sh_expand_strlen): Unroll last word.
177 2014-04-14  Christian Bruel  <christian.bruel@st.com>
179         * config/sh/sh.md (setmemqi): New expand pattern.
180         * config/sh/sh.h (CLEAR_RATIO): Define.
181         * config/sh/sh-mem.cc (sh_expand_setmem): Define.
182         * config/sh/sh-protos.h (sh_expand_setmem): Declare.
184 2014-04-14  Richard Biener  <rguenther@suse.de>
186         PR middle-end/55022
187         * fold-const.c (negate_expr_p): Don't negate directional rounding
188         division.
189         (fold_negate_expr): Likewise.
191 2014-04-14  Richard Biener  <rguenther@suse.de>
193         PR tree-optimization/59817
194         PR tree-optimization/60453
195         * graphite-scop-detection.c (graphite_can_represent_scev): Complete
196         recursion to catch all CHRECs in the scalar evolution and restrict
197         the predicate for the remains appropriately.
199 2014-04-12  Catherine Moore  <clm@codesourcery.com>
201         * config/mips/constraints.md: Add new register constraint "kb".
202         * config/mips/mips.md (*mov<mode>_internal): Use constraint "kb".
203         (*movhi_internal): Likewise.
204         (*movqi_internal): Likewise.
205         * config/mips/mips.h (M16_STORE_REGS): New register class.
206         (REG_CLASS_NAMES): Add M16_STORE_REGS.
207         (REG_CLASS_CONTENTS): Likewise.
208         * config/mips/mips.c (mips_regno_to_class): Add M16_STORE_REGS.
210 2014-04-11  Tobias Burnus  <burnus@net-b.de>
212         PR c/60194
213         * doc/invoke.texi (-Wformat-signedness): Document it.
214         (Wformat=2): Mention that this enables -Wformat-signedness.
216 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
218         * common/config/epiphany/epiphany-common.c
219         (epiphany_option_optimization_table): Enable section anchors by
220         default at -O1 or higher.
221         * config/epiphany/epiphany.c (TARGET_MAX_ANCHOR_OFFSET): Define.
222         (TARGET_MIN_ANCHOR_OFFSET): Likewise.
223         (epiphany_rtx_costs) <SET>: For binary operators, the set as such
224         carries no extra cost.
225         (epiphany_legitimate_address_p): For BLKmode, apply SImode check.
226         * config/epiphany/epiphany.h (ASM_OUTPUT_DEF): Define.
227         * config/epiphany/predicates.md (memclob_operand): New predicate.
228         * config/epiphany/epiphany.md (stack_adjust_add, stack_adjust_str):
229         Use memclob_operand predicate and X constraint for operand 3.
231 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
233         * config/epiphany/epiphany.c (epiphany_rtx_cost): Compare
234         with CC_N_NE / CC_C_LTU / CC_C_GTU carries no extra cost for
235         its operands.
237 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
239         PR rtl-optimization/60651
240         * mode-switching.c (optimize_mode_switching): Make sure to emit
241         sets of a lower numbered entity before sets of a higher numbered
242         entity to a mode of the same or lower priority.
243         When creating a seginfo for a basic block that starts with a code
244         label, move the insertion point past the code label.
245         (new_seginfo): Document and enforce requirement that
246         NOTE_INSN_BASIC_BLOCK only appears for empty blocks.
247         * doc/tm.texi.in: Document ordering constraint for emitted mode sets.
248         * doc/tm.texi: Regenerate.
250 2014-01-11  Joern Rennecke  <joern.rennecke@embecosm.com>
252         PR target/60811
253         * config/arc/arc.c (arc_save_restore): Fix assert typo.
255 2013-04-11  Jakub Jelinek  <jakub@redhat.com>
257         * BASE-VER: Set to 4.10.0.
259 2014-04-11  Tobias Burnus  <burnus@net-b.de>
261         PR other/59055
262         * doc/bugreport.texi (Bugs): Remove nodes pointing to the nirvana.
263         * doc/gcc.texi (Service): Update description in the @menu
264         * doc/invoke.texi (Option Summary): Remove misplaced and
265         duplicated @menu.
267 2014-04-11  Steve Ellcey  <sellcey@mips.com>
268             Jakub Jelinek  <jakub@redhat.com>
270         PR middle-end/60556
271         * expr.c (convert_move): Use emit_store_flag_force instead of
272         emit_store_flag.  Pass lowpart_mode instead of VOIDmode as 5th
273         argument to it.
275 2014-04-11  Richard Biener  <rguenther@suse.de>
277         PR middle-end/60797
278         * varasm.c (assemble_alias): Avoid endless error reporting
279         recursion by setting TREE_ASM_WRITTEN.
281 2014-04-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
283         * config/s390/s390.md: Add a splitter for NOT rtx.
285 2014-04-11  Jakub Jelinek  <jakub@redhat.com>
287         PR rtl-optimization/60663
288         * cse.c (cse_insn): Set src_volatile on ASM_OPERANDS in PARALLEL.
290 2014-04-10  Jan Hubicka  <hubicka@ucw.cz>
291             Jakub Jelinek  <jakub@redhat.com>
293         PR lto/60567
294         * ipa.c (function_and_variable_visibility): Copy forced_by_abi
295         flag from decl_node to node.
297 2014-04-10  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
299         PR debug/60655
300         * config/arm/arm.c (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Define
301         (arm_const_not_ok_for_debug_p): Reject MINUS with SYM_REF's
302         ameliorating the cases where it can be.
304 2014-04-09  David Edelsohn  <dje.gcc@gmail.com>
306         Revert
307         2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
309         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
310         (loadsync_<mode>): Change mode.
311         (load_quadpti, store_quadpti): New.
312         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
313         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
314         * config/rs6000/predicates.md (quad_memory_operand): !TARGET_SYNC_TI.
316 2014-04-09  Cong Hou  <congh@google.com>
318         PR testsuite/60773
319         * doc/sourcebuild.texi (vect_widen_mult_si_to_di_pattern): Add
320         documentation.
322 2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
324         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
325         instead of vnor to exploit possible fusion opportunity in the
326         future.
327         (altivec_expand_vec_perm_const_le): Likewise.
329 2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
331         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
332         (loadsync_<mode>): Change mode.
333         (load_quadpti, store_quadpti): New.
334         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
335         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
337 2014-04-08  Richard Sandiford  <rdsandiford@googlemail.com>
339         PR target/60763
340         * config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
341         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
342         Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.
344 2014-04-08  Richard Biener  <rguenther@suse.de>
346         PR middle-end/60706
347         * tree-pretty-print.c (pp_double_int): For HWI32 hosts with
348         a 64bit widest int print double-int similar to on HWI64 hosts.
350 2014-04-08  Richard Biener  <rguenther@suse.de>
352         PR tree-optimization/60785
353         * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
354         default defs properly.
356 2014-04-08  Nathan Sidwell  <nathan@codesourcery.com>
358         * doc/invoke (Wnon-virtual-dtor): Update to match implementation.
359         (Weffc++): Likewise.
361 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
363         * ipa-devirt.c (maybe_record_node): When node is not recorded,
364         set completep to false rather than true.
366 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
368         PR target/60504
369         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
370         ARM_TARGET2_DWARF_FORMAT.
372 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
374         PR target/60609
375         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
376         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
377         ADDR_DIFF_VEC.
379 2014-04-07  Richard Biener  <rguenther@suse.de>
381         PR tree-optimization/60766
382         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
383         (may_eliminate_iv): Convert cand_value_at result to desired type.
385 2014-04-07  Jason Merrill  <jason@redhat.com>
387         PR c++/60731
388         * common.opt (-fno-gnu-unique): Add.
389         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
391 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
393         * haifa-sched.c: Fix outdated function reference and minor
394         grammar errors in introductory comment.
396 2014-04-07  Richard Biener  <rguenther@suse.de>
398         PR middle-end/60750
399         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
400         for noreturn calls.
401         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
403 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
405         PR debug/55794
406         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
407         size accounting for thunks.
408         (pa_asm_output_mi_thunk): Use final_start_function() and
409         final_end_function() to output function start and end directives.
411 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
413         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have
414         device specific ISA/ feature information. Remove short_sp and
415         errata_skip ds.  Add avr_device_specific_features enum to have device
416         specific info.
417         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
418         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
419         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
420         updated device specific info.
421         * config/avr/avr-mcus.def: Merge device specific details to
422         dev_attribute field.
423         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
424         errata_skip.
425         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
426         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
427         assembler if RMW isa supported by current device.
428         * config/avr/genmultilib.awk: Update as device info structure changed.
429         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
431 2014-04-04  Cong Hou  <congh@google.com>
433         PR tree-optimization/60656
434         * tree-vect-stmts.c (supportable_widening_operation):
435         Fix a bug that elements in a vector with vect_used_by_reduction
436         property are incorrectly reordered when the operation on it is not
437         consistant with the one in reduction operation.
439 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
441         PR rtl-optimization/60155
442         * gcse.c (record_set_data): New function.
443         (single_set_gcse): New function.
444         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
445         (hoist_code): Likewise.
446         (get_pressure_class_and_nregs): Likewise.
448 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
450         * explow.c (probe_stack_range): Emit a final optimization blockage.
452 2014-04-04  Anthony Green  <green@moxielogic.com>
454         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
455         typos.
457 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
459         PR ipa/59626
460         * lto-cgraph.c (input_overwrite_node): Check that partitioning
461         flags are set only during streaming.
462         * ipa.c (process_references, walk_polymorphic_call_targets,
463         symtab_remove_unreachable_nodes): Drop bodies of always inline
464         after early inlining.
465         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
467 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
468         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
470         PR debug/60655
471         * dwarf2out.c (const_ok_for_output_1): Reject expressions
472         containing a NOT.
474 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
476         PR bootstrap/60743
477         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
478         duration.
479         (cortex_a53_fdivd): Likewise.
481 2014-04-04  Martin Jambor  <mjambor@suse.cz>
483         PR ipa/60640
484         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
485         Adjust all callers.
486         * cgraph.c (clone_of_p): Also return true if thunks match.
487         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
488         cgraph_function_or_thunk_node and an obsolete comment.
489         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
490         file.
491         (build_function_decl_skip_args): Likewise.
492         (set_new_clone_decl_and_node_flags): New function.
493         (duplicate_thunk_for_node): Likewise.
494         (redirect_edge_duplicating_thunks): Likewise.
495         (cgraph_clone_node): New parameter args_to_skip, pass it to
496         redirect_edge_duplicating_thunks which is called instead of
497         cgraph_redirect_edge_callee.
498         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
499         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
501 2014-04-04  Jeff Law  <law@redhat.com>
503         PR target/60657
504         * config/arm/predicates.md (const_int_I_operand): New predicate.
505         (const_int_M_operand): Similarly.
506         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
507         const_int_operand.
508         (insv_t2, extv_reg, extzv_t2): Likewise.
509         (load_multiple_with_writeback): Similarly for const_int_I_operand.
510         (pop_multiple_with_writeback_and_return): Likewise.
511         (vfp_pop_multiple_with_writeback): Likewise
513 2014-04-04  Richard Biener  <rguenther@suse.de>
515         PR ipa/60746
516         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
517         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
518         non-GIMPLE_LABELs.
519         * gimplify.h (gimple_add_tmp_var_fn): Declare.
520         * gimplify.c (gimple_add_tmp_var_fn): New function.
521         * gimple-expr.h (create_tmp_reg_fn): Declare.
522         * gimple-expr.c (create_tmp_reg_fn): New function.
523         * gimple-low.c (record_vars_into): Don't change cfun.
524         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
525         code generation without cfun.
527 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
529         PR bootstrap/60719
530         * Makefile.in (install-driver): Fix shell scripting.
532 2014-04-03  Cong Hou  <congh@google.com>
534         PR tree-optimization/60505
535         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
536         threshold of number of iterations below which no vectorization
537         will be done.
538         * tree-vect-loop.c (new_loop_vec_info):
539         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
540         * tree-vect-loop.c (vect_analyze_loop_operations):
541         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
542         * tree-vect-loop.c (vect_transform_loop):
543         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
544         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
545         of iterations of the loop and see if we should build the epilogue.
547 2014-04-03  Richard Biener  <rguenther@suse.de>
549         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx member.
550         (streamer_tree_cache_create): Adjust.
551         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
552         to allow optional nodes array.
553         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
554         (streamer_tree_cache_append): Likewise.
555         (streamer_tree_cache_create): Create nodes array optionally
556         as specified by parameter.
557         * lto-streamer-out.c (create_output_block): Avoid maintaining
558         the node array in the writer cache.
559         (DFS_write_tree): Remove assertion.
560         (produce_asm_for_decls): Free the out decl state hash table early.
561         * lto-streamer-in.c (lto_data_in_create): Adjust for
562         streamer_tree_cache_create prototype change.
564 2014-04-03  Richard Biener  <rguenther@suse.de>
566         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
567         set TREE_CHAIN to NULL_TREE.
569 2014-04-03  Richard Biener  <rguenther@suse.de>
571         PR tree-optimization/60740
572         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
573         over all GIMPLE_COND operands.
575 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
577         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
578         (Weffc++): Remove Scott's numbering, merge lists and reference
579         Wnon-virtual-dtor.
581         c-family/
583         cp/
584         * class.c (accessible_nvdtor_p): New.
585         (check_bases): Don't check base destructor here ...
586         (check_bases_and_members): ... check them here.  Trigger on
587         Wnon-virtual-dtor flag.
588         (finish_struct_1): Use accessible_nvdtor_p.
590         testsuite/
591         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
592         * g++.dg/warn/Wnvdtor-2.C: New.
593         * g++.dg/warn/Wnvdtor-3.C: New.
594         * g++.dg/warn/Wnvdtor-4.C: New.
595         * g++.dg/warn/Weff1.C: Delete.
596         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
597         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
599 2014-04-03  Nick Clifton  <nickc@redhat.com>
601         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
602         properly.
604 2014-04-03  Martin Jambor  <mjambor@suse.cz>
606         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
607         mention gcc_unreachable before failing.
608         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
609         removed symbols.
611 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
613         PR ipa/60659
614         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type
615         inconsistent code and instead mark the context inconsistent.
616         (possible_polymorphic_call_targets): For inconsistent contexts
617         return empty complete list.
619 2014-04-02  Anthony Green  <green@moxielogic.com>
621         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
622         (extendqisi2, extendhisi2): Define.
623         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
624         (WCHAR_TYPE): Change to unsigned int.
626 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
628         PR tree-optimization/60733
629         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
630         insertion point for PHI candidates to be the end of the feeding
631         block for the PHI argument.
633 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
635         PR rtl-optimization/60650
636         * lra-constraints.c (process_alt_operands): Decrease reject for
637         earlyclobber matching.
639 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
641         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
643 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
645         * config/spu/spu.c (pad_bb): Do not crash when the last
646         insn is CODE_FOR_blockage.
648 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
650         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
651         lies outside the target mode.
653 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
655         PR target/60735
656         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
657         software floating point or no floating point registers, do not
658         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
659         in GPRs that occurs after we tested for GPRs that would never be
660         true.
662         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
663         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
664         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
665         specifically allow DDmode, since that does not use the SPE SIMD
666         instructions.
668 2014-04-02  Richard Biener  <rguenther@suse.de>
670         PR middle-end/60729
671         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
672         MODE_INTs.  Properly use negv_optab.
673         (expand_abs): Likewise.
675 2014-04-02  Richard Biener  <rguenther@suse.de>
677         PR bootstrap/60719
678         * Makefile.in (install-driver): Guard extra installs with special
679         names properly.
681 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
683         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
684         Document vec_vgbbd.
686 2014-04-01  Richard Henderson  <rth@redhat.com>
688         PR target/60704
689         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
690         alternative enabled before register allocation.
692 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
694         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
695         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
696         typo.
697         (nios2_large_got_address): Remove unneeded 'sym' parameter.
698         (nios2_got_address): Update nios2_large_got_address call site.
699         (nios2_delegitimize_address): New function.
700         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
701         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
702         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
704 2014-04-01  Martin Husemann  <martin@duskware.de>
706         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
707         for -mabi=32.
709 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
711         PR rtl-optimization/60604
712         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
713         check from register_operand.
714         (register_operand): Redefine in terms of general_operand.
715         (nonmemory_operand): Use register_operand for the non-constant cases.
717 2014-04-01  Richard Biener  <rguenther@suse.de>
719         * gimple.h (struct gimple_statement_base): Align subcode to 16 bits.
721 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
723         * doc/invoke.texi (mapp-regs): Clarify.
725 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
727         * config/i386/avx512fintrin.h (__v32hi): Define type.
728         (__v64qi): Likewise.
729         (_mm512_set1_epi8): Define.
730         (_mm512_set1_epi16): Define.
731         (_mm512_set4_epi32): Define.
732         (_mm512_set4_epi64): Define.
733         (_mm512_set4_pd): Define.
734         (_mm512_set4_ps): Define.
735         (_mm512_setr4_epi64): Define.
736         (_mm512_setr4_epi32): Define.
737         (_mm512_setr4_pd): Define.
738         (_mm512_setr4_ps): Define.
739         (_mm512_setzero_epi32): Define.
741 2014-03-31  Martin Jambor  <mjambor@suse.cz>
743         PR middle-end/60647
744         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
745         callsite_arguments_match_p.  Updated all callers.  Also check types of
746         corresponding formal parameters and actual arguments.
747         (not_all_callers_have_enough_arguments_p) Renamed to
748         some_callers_have_mismatched_arguments_p.
750 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
752         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
754 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
756         PR target/60034
757         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
758         section anchor.
760 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
762         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
763         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
764         Split out
765         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
766         Use FMAMODE_NOVF512 mode iterator.
767         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
768         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
769         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
770         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
771         Split out
772         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
773         Use VF_128_256 mode iterator.
774         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
775         Ditto.
777 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
779         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
780         static chain if needed.
782 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
784         PR target/60697
785         * lra-constraints.c (index_part_to_reg): New.
786         (process_address): Use it.
788 2014-03-27  Jeff Law  <law@redhat.com>
789             Jakub Jelinek  <jakub@redhat.com>
791         PR target/60648
792         * expr.c (do_tablejump): Use simplify_gen_binary rather than
793         gen_rtx_{PLUS,MULT} to build up the address expression.
795         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
796         creating non-canonical RTL.
798 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
800         PR ipa/60243
801         * ipa-inline.c (want_inline_small_function_p): Short circuit large
802         functions; reorganize to make cheap checks first.
803         (inline_small_functions): Do not estimate growth when dumping;
804         it is expensive.
805         * ipa-inline.h (inline_summary): Add min_size.
806         (growth_likely_positive): New function.
807         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
808         (set_cond_stmt_execution_predicate): Cleanup.
809         (estimate_edge_size_and_time): Compute min_size.
810         (estimate_calls_size_and_time): Likewise.
811         (estimate_node_size_and_time): Likewise.
812         (inline_update_overall_summary): Update min_size.
813         (do_estimate_edge_time): Likewise.
814         (do_estimate_edge_size): Update.
815         (do_estimate_edge_hints): Update.
816         (growth_likely_positive): New function.
818 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
820         PR target/60693
821         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
822         also if addr has VOIDmode.
824 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
826         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
827         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
828         Declare extern.
829         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
830         instructions as well as AdvancedSIMD loads.
832 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
834         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
835         Use crypto_aese type.
836         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
837         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
838         crypto_aese, crypto_aesmc.  Move to types.md.
839         * config/arm/types.md (crypto_aes): Split into crypto_aese,
840         crypto_aesmc.
841         * config/arm/iterators.md (crypto_type): Likewise.
843 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
845         * cgraph.c: Include expr.h and tree-dfa.h.
846         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
847         remove LHS.
849 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
851         PR target/60675
852         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
853         regs from checking multi-reg pseudos.
855 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
857         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
859 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
861         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
862         if it would clobber the stack pointer, even temporarily.
864 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
866         * mode-switching.c: Make small adjustments to the top comment.
868 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
870         * config/rs6000/constraints.md (wD constraint): New constraint to
871         match the constant integer to get the top DImode/DFmode out of a
872         vector in a VSX register.
874         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
875         match the constant integer to get the top DImode/DFmode out of a
876         vector in a VSX register.
878         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
879         for ISA 2.07.
881         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
882         vbpermq builtins.
884         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
885         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
887         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
888         Optimize vec_extract of 64-bit values, where the value being
889         extracted is in the top word, where we can use scalar
890         instructions.  Add direct move and store support.  Combine the big
891         endian/little endian vector select load support into a single insn.
892         (vsx_extract_<mode>_internal1): Likewise.
893         (vsx_extract_<mode>_internal2): Likewise.
894         (vsx_extract_<mode>_load): Likewise.
895         (vsx_extract_<mode>_store): Likewise.
896         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
897         combined into vsx_extract_<mode>_load.
898         (vsx_extract_<mode>_one_le): Likewise.
900         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
901         define the top 64-bit vector element.
903         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
904         constraint.
906         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
907         Document vec_vbpermq builtin.
909         PR target/60672
910         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
911         enable use of xxsldwi and xxpermdi builtin functions.
912         (vec_xxpermdi): Likewise.
914         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
915         Document use of vec_xxsldwi and vec_xxpermdi builtins.
917 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
919         PR rtl-optimization/60650
920         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
921         first_p.  Use it.
922         (find_spills_for): New.
923         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
924         Spill all pseudos on the second iteration.
926 2014-03-27  Marek Polacek  <polacek@redhat.com>
928         PR c/50347
929         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
930         types.
932 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
934         * config/s390/s390.c (s390_can_use_return_insn): Check for
935         call-saved FPRs on 31 bit.
937 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
939         PR middle-end/60682
940         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
941         if they need regimplification, just drop them instead of
942         calling gimple_regimplify_operands on them.
944 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
946         PR target/60580
947         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
948         (aarch64_frame_pointer_required): Adjust logic.
949         (aarch64_can_eliminate): Adjust logic.
950         (aarch64_override_options_after_change): Adjust logic.
952 2014-03-27  Dehao Chen  <dehao@google.com>
954         * ipa-inline.c (early_inliner): Update node's inline info.
956 2014-03-26  Dehao Chen  <dehao@google.com>
958         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
959         compiler inserted conditional jumps for NAN float check.
961 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
963         * ubsan.h (ubsan_create_data): Change second argument's type
964         to const location_t *.
965         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
966         _("<unknown>").
967         (ubsan_create_data): Change second argument to const location_t *PLOC.
968         Create Loc field whenever PLOC is non-NULL.
969         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
970         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
971         callers.
973         PR other/59545
974         * real.c (real_to_integer2): Change type of low to UHWI.
976 2014-03-26  Tobias Burnus  <burnus@net-b.de>
978         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
979         (CILK_SELF_SPECS): New define.
980         (driver_self_specs): Use it.
982 2014-03-26  Richard Biener  <rguenther@suse.de>
984         * tree-pretty-print.c (percent_K_format): Implement special
985         case for LTO and its stripped down BLOCK tree.
987 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
989         PR sanitizer/60636
990         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
992         * tree-vrp.c (simplify_internal_call_using_ranges): If only
993         one range is range_int_cst_p, but not both, at least optimize
994         addition/subtraction of 0 and multiplication by 0 or 1.
995         * gimple-fold.c (gimple_fold_call): Fold
996         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
997         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
998         INTEGER_CSTs, try to fold at least x * 0 and y - y.
1000 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
1002         PR rtl-optimization/60452
1003         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
1004         <case REG>: Return 1 for invalid offsets from the frame pointer.
1006 2014-03-26  Marek Polacek  <polacek@redhat.com>
1008         PR c/37428
1009         * doc/extend.texi (C Extensions): Mention variable-length arrays in
1010         a structure/union.
1012 2014-03-26  Marek Polacek  <polacek@redhat.com>
1014         PR c/39525
1015         * doc/extend.texi (Designated Inits): Describe what happens to omitted
1016         field members.
1018 2014-03-26  Marek Polacek  <polacek@redhat.com>
1020         PR other/59545
1021         * ira-color.c (update_conflict_hard_regno_costs): Perform the
1022         multiplication in unsigned type.
1024 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
1026         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
1028 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
1030         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
1032 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
1034         PR ipa/60315
1035         * cif-code.def (UNREACHABLE) New code.
1036         * ipa-inline.c (inline_small_functions): Skip edges to
1037         __builtlin_unreachable.
1038         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
1039         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
1040         predicate to __bulitin_unreachable.
1041         (set_cond_stmt_execution_predicate): Fix issue when
1042         invert_tree_comparison returns ERROR_MARK.
1043         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
1044         propagate to inline clones.
1045         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
1046         to unreachable.
1047         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
1048         * cgraphclones.c (cgraph_clone_node): If call destination is already
1049         ureachable, do not redirect it back.
1050         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
1051         unreachable.
1053 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
1055         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
1056         Do not modify inline clones.
1058 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
1060         * config/i386/i386.md (general_sext_operand): New mode attr.
1061         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
1062         don't generate (sign_extend (const_int)).
1063         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
1064         operands[2].  Use We constraint instead of <i> and
1065         <general_sext_operand> predicate instead of <general_operand>.
1066         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
1067         * config/i386/constraints.md (We): New constraint.
1068         * config/i386/predicates.md (x86_64_sext_operand,
1069         sext_operand): New predicates.
1071 2014-03-25  Martin Jambor  <mjambor@suse.cz>
1073         PR ipa/60600
1074         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
1075         inconsistent devirtualizations to __builtin_unreachable.
1077 2014-03-25  Marek Polacek  <polacek@redhat.com>
1079         PR c/35449
1080         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
1082 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
1084         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
1085         order of elements for big-endian.
1087 2014-03-25  Richard Biener  <rguenther@suse.de>
1089         PR middle-end/60635
1090         * gimplify-me.c (gimple_regimplify_operands): Update the
1091         re-gimplifed stmt.
1093 2014-03-25  Martin Jambor  <mjambor@suse.cz>
1095         PR ipa/59176
1096         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
1097         (lto_output_varpool_node): Likewise.
1098         (input_overwrite_node): Likewise.
1099         (input_varpool_node): Likewise.
1101 2014-03-25  Richard Biener  <rguenther@suse.de>
1103         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
1104         (run_gcc): Likewise.
1106 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
1108         * combine.c (simplify_compare_const): Add MODE argument.
1109         Handle mode_width 0 as very large mode_width.
1110         (try_combine, simplify_comparison): Adjust callers.
1112         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
1113         type to avoid signed integer overflow.
1114         * explow.c (plus_constant): Likewise.
1116 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1118         * doc/generic.texi: Correct typos.
1120 2014-03-24  Tobias Burnus  <burnus@net-b.de>
1122         * doc/invoke.texi (-flto): Expand section about
1123         using static libraries with LTO.
1125 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1127         PR rtl-optimization/60501
1128         * optabs.def (addptr3_optab): New optab.
1129         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
1130         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
1131         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
1133         * lra.c (emit_add3_insn): Use the addptr pattern if available.
1135         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
1137 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
1139         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
1140         _mm512_set1_pd.
1142         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
1143         (_mm256_undefined_ps): Define.
1144         (_mm256_undefined_pd): Define.
1145         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
1146         (_mm_undefined_pd): Define.
1147         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
1148         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
1149         (_mm512_undefined_ps): Define.
1150         (_mm512_undefined_pd): Define.
1151         Use _mm*_undefined_*.
1152         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
1154 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
1156         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
1157         (lshr_simd): DI mode added.
1158         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
1159         (aarch64_ushr_simddi): Likewise.
1160         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
1161         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
1162         (vshrd_n_u64): Likewise.
1164 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1166         * Makefile.in (s-macro_list): Depend on cc1.
1168 2014-03-23  Teresa Johnson  <tejohnson@google.com>
1170         * ipa-utils.c (ipa_print_order): Use specified dump file.
1172 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
1174         PR rtl-optimization/60601
1175         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
1177         * gcc.c (eval_spec_function): Initialize save_growing_value.
1179 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
1181         PR sanitizer/60613
1182         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
1183         code == MINUS_EXPR, never swap op0 with op1.
1185         * toplev.c (init_local_tick): Avoid signed integer multiplication
1186         overflow.
1187         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
1188         shift by first operand's bitsize.
1190 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
1192         PR target/60610
1193         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
1194         redefine to 1 or 0.
1195         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
1196         TARGET_ISA_64BIT_P(x).
1198 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1200         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
1201         pattern for vector nor instead of subtract from splat(-1).
1202         (altivec_expand_vec_perm_const_le): Likewise.
1204 2014-03-21  Richard Henderson  <rth@twiddle.net>
1206         PR target/60598
1207         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
1208         related insns after epilogue_completed.
1210 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1212         PR ipa/59176
1213         * cgraph.h (symtab_node): New flag body_removed.
1214         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
1215         when removing bodies.
1216         * symtab.c (dump_symtab_base): Dump body_removed flag.
1217         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
1218         had their bodies removed.
1220 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1222         PR ipa/60419
1223         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
1224         in the border.
1226 2014-03-21  Richard Biener  <rguenther@suse.de>
1228         PR tree-optimization/60577
1229         * tree-core.h (struct tree_base): Document nothrow_flag use
1230         in VAR_DECL_NONALIASED.
1231         * tree.h (VAR_DECL_NONALIASED): New.
1232         (may_be_aliased): Adjust.
1233         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
1235 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1237         * expr.c (expand_expr_real_1): Remove outdated comment.
1239 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
1241         PR middle-end/60597
1242         * ira.c (adjust_cleared_regs): Call copy_rtx on
1243         *reg_equiv[REGNO (loc)].src_p before passing it to
1244         simplify_replace_fn_rtx.
1246         PR target/60568
1247         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
1248         into CONST, put pic register as first operand of PLUS.  Use
1249         gen_const_mem for both 32-bit and 64-bit PIC got loads.
1251 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1253         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
1255 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1257         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
1258         around for store forwarding issue in the FPU on the UT699.
1259         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
1260         loads and operations if -mfix-ut699 is specified.
1261         (divtf3_hq): Tweak attribute.
1262         (sqrttf2_hq): Likewise.
1264 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1266         * calls.c (store_one_arg): Remove incorrect const qualification on the
1267         type of the temporary.
1268         * cfgexpand.c (expand_return): Likewise.
1269         * expr.c (expand_constructor): Likewise.
1270         (expand_expr_real_1): Likewise.
1272 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1274         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
1275         of parts.
1277 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
1279         PR target/60039
1280         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
1282 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
1284         * config/arm/aarch-common-protos.h
1285         (alu_cost_table): Fix spelling of "extend".
1286         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
1288 2014-03-19  Richard Biener  <rguenther@suse.de>
1290         PR middle-end/60553
1291         * tree-core.h (tree_type_common): Re-order pointer members
1292         to reduce recursion depth during GC walks.
1294 2014-03-19  Marek Polacek  <polacek@redhat.com>
1296         PR sanitizer/60569
1297         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
1298         before accessing it.
1300 2014-03-19  Richard Biener  <rguenther@suse.de>
1302         PR lto/59543
1303         * lto-streamer-in.c (input_function): In WPA stage do not drop
1304         debug stmts.
1306 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
1308         PR tree-optimization/60559
1309         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
1310         with build_zero_cst assignment.
1312 2014-03-18  Kai Tietz  <ktietz@redhat.com>
1314         PR rtl-optimization/56356
1315         * sdbout.c (sdbout_parms): Verify that parms'
1316         incoming argument is valid.
1317         (sdbout_reg_parms): Likewise.
1319 2014-03-18  Richard Henderson  <rth@redhat.com>
1321         PR target/60562
1322         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
1323         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
1324         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
1326 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
1328         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
1329         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
1330         Italicize plugin event names in description.  Explain that
1331         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
1332         Remind that no GCC functions should be called after PLUGIN_FINISH.
1333         Explain what pragmas with expansion are.
1335 2014-03-18  Martin Liska  <mliska@suse.cz>
1337         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
1338         gimple call statement is update.
1339         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
1340         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
1342 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1344         PR sanitizer/60557
1345         * ubsan.c (ubsan_instrument_unreachable): Call
1346         initialize_sanitizer_builtins.
1347         (ubsan_pass): Likewise.
1349         PR sanitizer/60535
1350         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1351         varpool_finalize_decl instead of rest_of_decl_compilation.
1353 2014-03-18  Richard Biener  <rguenther@suse.de>
1355         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1356         by using bitmap_and_compl instead of bitmap_and_compl_into.
1357         (df_rd_transfer_function): Likewise.
1359 2014-03-18  Richard Biener  <rguenther@suse.de>
1361         * doc/lto.texi (fresolution): Fix typo.
1363 2014-03-18  Richard Biener  <rguenther@suse.de>
1365         * doc/invoke.texi (flto): Update for changes in 4.9.
1367 2014-03-18  Richard Biener  <rguenther@suse.de>
1369         * doc/loop.texi: Remove section on the removed lambda framework.
1370         Update loop docs with recent changes in preserving loop structure.
1372 2014-03-18  Richard Biener  <rguenther@suse.de>
1374         * doc/lto.texi (-fresolution): Document.
1376 2014-03-18  Richard Biener  <rguenther@suse.de>
1378         * doc/contrib.texi: Adjust my name.
1380 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1382         PR ipa/58721
1383         * internal-fn.c: Include diagnostic-core.h.
1384         (expand_BUILTIN_EXPECT): New function.
1385         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1386         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1387         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1388         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1389         IFN_BUILTIN_EXPECT.
1390         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1391         Revert 3 argument __builtin_expect code.
1392         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1393         * gimple-fold.c (gimple_fold_call): Likewise.
1394         * tree.h (fold_builtin_expect): New prototype.
1395         * builtins.c (build_builtin_expect_predicate): Add predictor
1396         argument, if non-NULL, create 3 argument __builtin_expect.
1397         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1398         pass it through to build_builtin_expect_predicate.
1399         (fold_builtin_2): Adjust caller.
1400         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1401         * internal-fn.def (BUILTIN_EXPECT): New.
1403 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1405         PR ipa/58721
1406         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1407         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1408         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1410 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1412         PR ipa/58721
1413         * predict.c (combine_predictions_for_bb): Fix up formatting.
1414         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1415         fill what it points to if non-NULL.
1416         (tree_predict_by_opcode): Adjust caller, use the predictor.
1417         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1419 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1421         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1422         proper constant for the store mode.
1424 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1426         * symtab.c (change_decl_assembler_name): Fix transparent alias
1427         chain construction.
1429 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1431         * config/aarch64/aarch64.c: Correct the comments about the
1432         aarch64 stack layout.
1434 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1436         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1437         check for GF_OMP_FOR_KIND_FOR.
1439 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1441         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1442         ymm and zmm register names.
1444 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1446         PR target/60516
1447         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1448         note creation for the 2010-08-31 changes.
1450 2014-03-17  Marek Polacek  <polacek@redhat.com>
1452         PR middle-end/60534
1453         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1454         as -fno-tree-loop-vectorize.
1455         (expand_omp_simd): Likewise.
1457 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1459         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1460         (eligible_for_call_delay): New prototype.
1461         * config/sparc/sparc.c (tls_call_delay): Rename into...
1462         (eligible_for_call_delay): ...this.  Return false if the instruction
1463         cannot be put in the delay slot of a branch.
1464         (eligible_for_restore_insn): Simplify.
1465         (eligible_for_return_delay): Return false if the instruction cannot be
1466         put in the delay slot of a branch and simplify.
1467         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1468         put in the delay slot of a branch.
1469         * config/sparc/sparc.md (fix_ut699): New attribute.
1470         (tls_call_delay): Delete.
1471         (in_call_delay): Reimplement.
1472         (eligible_for_sibcall_delay): Rename into...
1473         (in_sibcall_delay): ...this.
1474         (eligible_for_return_delay): Rename into...
1475         (in_return_delay): ...this.
1476         (in_branch_delay): Reimplement.
1477         (in_uncond_branch_delay): Delete.
1478         (in_annul_branch_delay): Delete.
1480 2014-03-14  Richard Henderson  <rth@redhat.com>
1482         PR target/60525
1483         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1484         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1485         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1486         (floathi splitters): Remove.
1487         (float<SWI48x>xf2): New pattern.
1488         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1489         code that tried to handle DImode for 32-bit, but which was excluded
1490         by the pattern's condition.  Drop allocation of stack temporary.
1491         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1492         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1493         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1494         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1495         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1496         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1497         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1498         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1499         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1500         (*float<SWI48x><X87MODEF>2_i387): Remove.
1501         (all float _with_temp splitters): Remove.
1502         (*float<SWI48x><MODEF>2_i387): New pattern.
1503         (*float<SWI48><MODEF>2_sse): New pattern.
1504         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1505         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1507 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1508             Marek Polacek  <polacek@redhat.com>
1510         PR middle-end/60484
1511         * common.opt (dump_base_name_prefixed): New Variable.
1512         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1513         if x_dump_base_name_prefixed is already set, set it at the end.
1515 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1517         PR rtl-optimization/60508
1518         * lra-constraints.c (get_reload_reg): Add new parameter
1519         in_subreg_p.
1520         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1521         Pass the new parameter values.
1523 2014-03-14  Richard Biener  <rguenther@suse.de>
1525         * common.opt: Revert unintented changes from r205065.
1526         * opts.c: Likewise.
1528 2014-03-14  Richard Biener  <rguenther@suse.de>
1530         PR middle-end/60518
1531         * cfghooks.c (split_block): Properly adjust all loops the
1532         block was a latch of.
1534 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1536         PR lto/60461
1537         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1538         and simplify it.
1540 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1542         PR target/59396
1543         * config/avr/avr.c (avr_set_current_function): Pass function name
1544         through default_strip_name_encoding before sanity checking instead
1545         of skipping the first char of the assembler name.
1547 2014-03-13  Richard Henderson  <rth@redhat.com>
1549         PR debug/60438
1550         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1551         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1552         * config/i386/i386-protos.h: Likewise.
1553         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1554         in the expander instead of a splitter.
1555         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1556         any possibility of requiring a memory.
1557         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1558         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1559         (fp branch splitters): Update for ix86_split_fp_branch.
1560         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1561         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1562         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1563         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1564         (*fop_<MODEF>_3_i387): Likewise.
1565         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1566         (splitters for the fop_* register patterns): Remove.
1567         (fscalexf4_i387): Rename from *fscalexf4_i387.
1568         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1570 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1572         PR tree-optimization/59779
1573         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1574         type for bitsize and maxsize instead of HOST_WIDE_INT.
1576 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1578         PR rtl-optimization/57320
1579         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1580         the CFG after thread_prologue_and_epilogue_insns.
1582 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1584         PR rtl-optimization/57189
1585         * lra-constraints.c (process_alt_operands): Disfavor spilling
1586         vector pseudos.
1588 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1590         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1592 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1594         PR tree-optimization/59025
1595         PR middle-end/60418
1596         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1597         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1599 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1601         PR target/60486
1602         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1603         calls of avr_out_plus_1.
1605 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1607         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1608         BB's single pred and update the father loop's latch info later.
1610 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1612         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1613         (VEC_M): Likewise.
1614         (VEC_N): Likewise.
1615         (VEC_R): Likewise.
1616         (VEC_base): Likewise.
1617         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1618         registers, we need to swap double words in little endian mode.
1620         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1621         to be a container mode for 128-bit integer operations added in ISA
1622         2.07.  Unlike TImode and PTImode, the preferred register set is
1623         the Altivec/VMX registers for the 128-bit operations.
1625         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1626         declarations.
1627         (rs6000_split_128bit_ok_p): Likewise.
1629         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1630         macros for creating ISA 2.07 normal and overloaded builtin
1631         functions with 3 arguments.
1632         (BU_P8V_OVERLOAD_3): Likewise.
1633         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1634         for use as overloaded functions.
1635         (VPERM_1TI_UNS): Likewise.
1636         (VSEL_1TI): Likewise.
1637         (VSEL_1TI_UNS): Likewise.
1638         (ST_INTERNAL_1ti): Likewise.
1639         (LD_INTERNAL_1ti): Likewise.
1640         (XXSEL_1TI): Likewise.
1641         (XXSEL_1TI_UNS): Likewise.
1642         (VPERM_1TI): Likewise.
1643         (VPERM_1TI_UNS): Likewise.
1644         (XXPERMDI_1TI): Likewise.
1645         (SET_1TI): Likewise.
1646         (LXVD2X_V1TI): Likewise.
1647         (STXVD2X_V1TI): Likewise.
1648         (VEC_INIT_V1TI): Likewise.
1649         (VEC_SET_V1TI): Likewise.
1650         (VEC_EXT_V1TI): Likewise.
1651         (EQV_V1TI): Likewise.
1652         (NAND_V1TI): Likewise.
1653         (ORC_V1TI): Likewise.
1654         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1655         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1656         overloaded builtin.
1657         (VADDUQM): Likewise.
1658         (VSUBCUQ): Likewise.
1659         (VADDEUQM): Likewise.
1660         (VADDECUQ): Likewise.
1661         (VSUBEUQM): Likewise.
1662         (VSUBECUQ): Likewise.
1664         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1665         __int128_t and __uint128_t types.
1666         (__uint128_type): Likewise.
1667         (altivec_categorize_keyword): Add support for vector __int128_t,
1668         vector __uint128_t, vector __int128, and vector unsigned __int128
1669         as a container type for TImode operations that need to be done in
1670         VSX/Altivec registers.
1671         (rs6000_macro_to_expand): Likewise.
1672         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1673         to support 128-bit integer instructions vaddcuq, vadduqm,
1674         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1675         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1677         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1678         for V1TImode, and set up preferences to use VSX/Altivec registers.
1679         Setup VSX reload handlers.
1680         (rs6000_debug_reg_global): Likewise.
1681         (rs6000_init_hard_regno_mode_ok): Likewise.
1682         (rs6000_preferred_simd_mode): Likewise.
1683         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1684         (easy_altivec_constant): Likewise.
1685         (output_vec_const_move): Likewise.
1686         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1687         simple move.
1688         (rs6000_expand_vector_extract): Likewise.
1689         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1690         addressing.
1691         (rs6000_const_vec): Add support for V1TImode.
1692         (rs6000_emit_le_vsx_load): Swap double words when loading or
1693         storing TImode/V1TImode.
1694         (rs6000_emit_le_vsx_store): Likewise.
1695         (rs6000_emit_le_vsx_move): Likewise.
1696         (rs6000_emit_move): Add support for V1TImode.
1697         (altivec_expand_ld_builtin): Likewise.
1698         (altivec_expand_st_builtin): Likewise.
1699         (altivec_expand_vec_init_builtin): Likewise.
1700         (altivec_expand_builtin): Likewise.
1701         (rs6000_init_builtins): Add support for V1TImode type.  Add
1702         support for ISA 2.07 128-bit integer builtins.  Define type names
1703         for the VSX/Altivec vector types.
1704         (altivec_init_builtins): Add support for overloaded vector
1705         functions with V1TImode type.
1706         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1707         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1708         external function.
1709         (rs6000_split_128bit_ok_p): Likewise.
1710         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1711         __int128_t and vector __uint128_t.
1713         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1714         and mode attributes.
1715         (VSX_M): Likewise.
1716         (VSX_M2): Likewise.
1717         (VSm): Likewise.
1718         (VSs): Likewise.
1719         (VSr): Likewise.
1720         (VSv): Likewise.
1721         (VS_scalar): Likewise.
1722         (VS_double): Likewise.
1723         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1725         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1726         we support the ISA 2.07 128-bit integer arithmetic instructions.
1727         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1728         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1729         and TImode types for use with the builtin functions.
1730         (V1TI_type_node): Likewise.
1731         (unsigned_V1TI_type_node): Likewise.
1732         (intTI_type_internal_node): Likewise.
1733         (uintTI_type_internal_node): Likewise.
1735         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1736         128-bit builtin functions.
1737         (UNSPEC_VADDEUQM): Likewise.
1738         (UNSPEC_VADDECUQ): Likewise.
1739         (UNSPEC_VSUBCUQ): Likewise.
1740         (UNSPEC_VSUBEUQM): Likewise.
1741         (UNSPEC_VSUBECUQ): Likewise.
1742         (VM): Add V1TImode to vector mode iterators.
1743         (VM2): Likewise.
1744         (VI_unit): Likewise.
1745         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1746         (altivec_vaddcuq): Likewise.
1747         (altivec_vsubuqm): Likewise.
1748         (altivec_vsubcuq): Likewise.
1749         (altivec_vaddeuqm): Likewise.
1750         (altivec_vaddecuq): Likewise.
1751         (altivec_vsubeuqm): Likewise.
1752         (altivec_vsubecuq): Likewise.
1754         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1755         mode iterators.
1756         (BOOL_128): Likewise.
1757         (BOOL_REGS_OUTPUT): Likewise.
1758         (BOOL_REGS_OP1): Likewise.
1759         (BOOL_REGS_OP2): Likewise.
1760         (BOOL_REGS_UNARY): Likewise.
1761         (BOOL_REGS_AND_CR0): Likewise.
1763         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1764         128-bit integer builtin support.
1765         (vec_vadduqm): Likewise.
1766         (vec_vaddecuq): Likewise.
1767         (vec_vaddeuqm): Likewise.
1768         (vec_vsubecuq): Likewise.
1769         (vec_vsubeuqm): Likewise.
1770         (vec_vsubcuq): Likewise.
1771         (vec_vsubuqm): Likewise.
1773         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1774         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1775         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1776         128-bit integer add/subtract to ISA 2.07.
1778 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1780         * config/arc/arc.c (arc_predicate_delay_insns):
1781         Fix third argument passed to conditionalize_nonjump.
1783 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1785         * config/aarch64/aarch64-builtins.c
1786         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1787         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1788         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1789         instead of __builtin_lfloor.
1790         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1792 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1794         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1795         (tree_ssa_ifcombine_bb_1): New function.
1796         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1797         is an empty forwarder block to then_bb or vice versa and then_bb
1798         and else_bb are effectively swapped.
1800 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1802         PR target/60264
1803         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1804         REG_CFA_DEF_CFA note.
1805         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1806         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1808 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1810         PR tree-optimization/60454
1811         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1813 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1815         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1816         Do not define target_cpu_default2 to generic.
1817         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1818         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1819         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1821 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1822             Marc Glisse  <marc.glisse@inria.fr>
1824         PR tree-optimization/60502
1825         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1826         instead of build_low_bits_mask.
1828 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1830         PR middle-end/60482
1831         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1832         if there are multiple uses, but op doesn't live on E edge.
1833         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1834         clobber stmts before __builtin_unreachable.
1836 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1838         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1839         hard_frame_pointer_rtx.
1840         * cse.c (cse_insn): Remove volatile check.
1841         * cselib.c (cselib_process_insn): Likewise.
1842         * dse.c (scan_insn): Likewise.
1844 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1846         * config/arc/arc.c (conditionalize_nonjump): New function,
1847         broken out of ...
1848         (arc_ifcvt): ... this.
1849         (arc_predicate_delay_insns): Use it.
1851 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1853         * config/arc/predicates.md (extend_operand): During/after reload,
1854         allow const_int_operand.
1855         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1856         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1857         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1858         to "i".
1859         (umulsi3_highpart_i): Likewise.
1861 2014-03-11  Richard Biener  <rguenther@suse.de>
1863         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1864         Add asserts to guard possible wrong-code bugs.
1866 2014-03-11  Richard Biener  <rguenther@suse.de>
1868         PR tree-optimization/60429
1869         PR tree-optimization/60485
1870         * tree-ssa-structalias.c (set_union_with_increment): Properly
1871         take into account all fields that overlap the shifted vars.
1872         (do_sd_constraint): Likewise.
1873         (do_ds_constraint): Likewise.
1874         (get_constraint_for_ptr_offset): Likewise.
1876 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1878         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1879         (nios2_compute_frame_layout):
1880         Add calculation of cfun->machine->fp_save_offset.
1881         (nios2_expand_prologue): Correct setting of frame pointer register
1882         in prologue.
1883         (nios2_expand_epilogue): Update recovery of stack pointer from
1884         frame pointer accordingly.
1885         (nios2_initial_elimination_offset): Update calculation of offset
1886         for eliminating to HARD_FRAME_POINTER_REGNUM.
1888 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1890         PR ipa/60457
1891         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1892         cgraph_get_create_node on VAR_DECLs.
1894 2014-03-10  Richard Biener  <rguenther@suse.de>
1896         PR middle-end/60474
1897         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1899 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1901         * config/vms/vms.opt (vms_float_format): New variable.
1903 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1905         * doc/invoke.texi (-fcilkplus): Update implementation status.
1907 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1908             Richard Biener  <rguenther@suse.de>
1910         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1911         consistently accross all TUs.
1912         (run_gcc): Enable -fshort-double automatically at link at link-time
1913         and disallow override.
1915 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1917         PR target/58271
1918         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1919         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1920         if they can't be used.
1922 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1924         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1925         for Solaris 11/x86 ld.
1926         * configure: Regenerate.
1928 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1930         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1931         (LIB_TLS_SPEC): Save as ld_tls_libs.
1932         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1933         (HAVE_AS_IX86_TLSLDM): New test.
1934         * configure, config.in: Regenerate.
1935         * config/i386/i386.c (legitimize_tls_address): Fall back to
1936         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1937         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1938         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1939         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1941 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1943         * common.opt (fira-loop-pressure): Mark as optimization.
1945 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1947         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1948         an OpenMP mappable type.
1950 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1952         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1953         MULTILIB_OSDIRNAMES is not defined.
1955 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1956             Meador Inge  <meadori@codesourcery.com>
1958         PR target/58595
1959         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1960         (arm_legitimize_address): Call legitimize_tls_address for any
1961         arm_tls_referenced_p expression, handle constant addend.  Call it
1962         before testing for !TARGET_ARM.
1963         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1965 2014-03-06  Richard Biener  <rguenther@suse.de>
1967         PR middle-end/60445
1968         PR lto/60424
1969         PR lto/60427
1970         Revert
1971         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1973         * tree-streamer.c (record_common_node): Assert we don't record
1974         nodes with type double.
1975         (preload_common_node): Skip type double, complex double and double
1976         pointer since it is now frontend dependent due to fshort-double option.
1978 2014-03-06  Richard Biener  <rguenther@suse.de>
1980         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1981         or -fno-lto is specified and the linker has full plugin support.
1982         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1983         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1984         * lto-wrapper.c (merge_and_complain): Merge compile-time
1985         optimization levels.
1986         (run_gcc): And pass it through to the link options.
1988 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1990         PR debug/60381
1991         Revert:
1992         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1993         PR debug/59992
1994         * cselib.c (remove_useless_values): Skip to avoid quadratic
1995         behavior if the condition moved from...
1996         (cselib_process_insn): ... here holds.
1998 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
2000         PR plugins/59335
2001         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
2002         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
2004         PR plugins/59335
2005         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
2006         (TM_H): Add x86-tune.def.
2008 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2010         * config/aarch64/aarch64.c (generic_tunings):
2011         Use cortexa57_extra_costs.
2013 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
2015         PR lto/60404
2016         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
2017         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
2018         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
2019         cost for in_lto_p.
2021 2014-03-04  Heiher  <r@hev.cc>
2023         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
2024         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
2026 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
2028         * config/i386/predicates.md (const2356_operand): Change to ...
2029         (const2367_operand): ... this.
2030         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
2031         const2367_operand.
2032         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2033         (*avx512pf_scatterpf<mode>sf): Ditto.
2034         (avx512pf_scatterpf<mode>df): Ditto.
2035         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2036         (*avx512pf_scatterpf<mode>df): Ditto.
2037         * config/i386/i386.c (ix86_expand_builtin): Update
2038         incorrect hint operand error message.
2040 2014-03-04  Richard Biener  <rguenther@suse.de>
2042         * lto-section-in.c (lto_get_section_data): Fix const cast.
2044 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
2046         * tree-streamer.c (record_common_node): Assert we don't record
2047         nodes with type double.
2048         (preload_common_node): Skip type double, complex double and double
2049         pointer since it is now frontend dependent due to fshort-double option.
2051 2014-03-04  Richard Biener  <rguenther@suse.de>
2053         PR lto/60405
2054         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
2055         (lto_input_toplevel_asms): Likewise.
2056         * lto-section-in.c (lto_get_section_data): Instead do it here
2057         for every section.
2059 2014-03-04  Richard Biener  <rguenther@suse.de>
2061         PR tree-optimization/60382
2062         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
2063         dead PHIs a reduction.
2065 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
2067         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
2068         hint value.
2069         (_mm_prefetch): Move out of GCC target("sse") pragma.
2070         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
2071         GCC target("prfchw") pragma.
2072         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
2073         for locality <= 2.
2074         * config/i386/i386.c (ix86_option_override_internal): Enable
2075         -mprfchw with -mprefetchwt1.
2077 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
2079         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
2080         Mark as varying.
2082 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
2084         * opts.h (CL_PCH_IGNORE): Define.
2085         * targhooks.c (option_affects_pch_p):
2086         Return false for options that have CL_PCH_IGNORE set.
2087         * opt-functions.awk: Process PchIgnore.
2088         * doc/options.texi: Document PchIgnore.
2090         * config/arc/arc.opt (misize): Add PchIgnore property.
2092 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2094         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
2095         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
2096         constraint on constants to permit them being loaded into
2097         GENERAL_REGS or BASE_REGS.
2099 2014-03-03  Nick Clifton  <nickc@redhat.com>
2101         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
2102         anti-cacnonical alternatives.
2103         (negandhi3_real): New pattern.
2104         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
2106 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2108         * config/avr/avr-mcus.def: Remove atxmega16x1.
2109         * config/avr/avr-tables.opt: Regenerate.
2110         * config/avr/t-multilib: Regenerate.
2111         * doc/avr-mmcu.texi: Regenerate.
2113 2014-03-03  Tobias Grosser  <tobias@grosser.es>
2114             Mircea Namolaru  <mircea.namolaru@inria.fr>
2116         PR tree-optimization/58028
2117         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
2118         scalar dimensions.
2120 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
2122         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
2123         not handled by recognizers.
2125 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
2127         PR middle-end/60175
2128         * function.c (expand_function_end): Don't emit
2129         clobber_return_register sequence if clobber_after is a BARRIER.
2130         * cfgexpand.c (construct_exit_block): Append instructions before
2131         return_label to prev_bb.
2133 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2135         * config/rs6000/constraints.md: Document reserved use of "wc".
2137 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
2139         PR ipa/60150
2140         * ipa.c (function_and_variable_visibility): When dissolving comdat
2141         group, also set all symbols to local.
2143 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
2145         PR ipa/60306
2147         Revert:
2148         2013-12-14   Jan Hubicka  <jh@suse.cz>
2149         PR middle-end/58477
2150         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
2152 2014-03-02  Jon Beniston  <jon@beniston.com>
2154         PR bootstrap/48230
2155         PR bootstrap/50927
2156         PR bootstrap/52466
2157         PR target/46898
2158         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
2159           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
2160         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
2161         (simple_return, *simple_return): New patterns
2162         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
2163         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
2165 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
2167         * dwarf2out.c (gen_subprogram_die): Tidy.
2169 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
2171         PR target/60071
2172         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
2173         (*mov_t_msb_neg_negc): ... this new insn.
2175 2014-02-28  Jason Merrill  <jason@redhat.com>
2177         PR c++/58678
2178         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
2179         function.
2181 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
2183         PR c++/60314
2184         * dwarf2out.c (decltype_auto_die): New static.
2185         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
2186         (gen_type_die_with_usage): Handle 'decltype(auto)'.
2187         (is_cxx_auto): Likewise.
2189 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
2191         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
2192         we are not using general regs only.
2194 2014-02-28  Richard Biener  <rguenther@suse.de>
2196         PR target/60280
2197         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
2198         previous fix and only allow to remove trivial pre-headers
2199         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
2200         (remove_forwarder_block): Properly update the latch of a loop.
2202 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2204         PR debug/59992
2205         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
2206         (cselib_preserved_hash_table): New.
2207         (preserve_constants_and_equivs): Move preserved vals to it.
2208         (cselib_find_slot): Look it up first.
2209         (cselib_init): Initialize it.
2210         (cselib_finish): Release it.
2211         (dump_cselib_table): Dump it.
2213 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2215         PR debug/59992
2216         * cselib.c (remove_useless_values): Skip to avoid quadratic
2217         behavior if the condition moved from...
2218         (cselib_process_insn): ... here holds.
2220 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2222         PR debug/57232
2223         * var-tracking.c (vt_initialize): Apply the same condition to
2224         preserve the CFA base value.
2226 2014-02-28  Joey Ye  <joey.ye@arm.com>
2228         PR target/PR60169
2229         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
2230         if reload in progress or completed.
2232 2014-02-28  Tobias Burnus  <burnus@net-b.de>
2234         PR middle-end/60147
2235         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
2236         NAMELIST_DECL.
2238 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
2240         * doc/tm.texi.in (Condition Code Status): Update documention for
2241         relative locations of cc0-setter and cc0-user.
2243 2014-02-27  Jeff Law  <law@redhat.com>
2245         PR rtl-optimization/52714
2246         * combine.c (try_combine): When splitting an unrecognized PARALLEL
2247         into two independent simple sets, if I3 is a jump, ensure the
2248         pattern we place into I3 is a (set (pc) ...).
2250 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
2251             Jeff Law  <law@redhat.com>
2253         PR rtl-optimization/49847
2254         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
2255         are in different blocks.
2256         * doc/tm.texi (Condition Code Status): Update documention for
2257         relative locations of cc0-setter and cc0-user.
2259 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
2261         PR target/59222
2262         * lra.c (lra_emit_add): Check SUBREG too.
2264 2014-02-27  Andreas Schwab  <schwab@suse.de>
2266         * config/m68k/m68k.c (m68k_option_override): Disable
2267         -flive-range-shrinkage for classic m68k.
2268         (m68k_override_options_after_change): Likewise.
2270 2014-02-27  Marek Polacek  <polacek@redhat.com>
2272         PR middle-end/59223
2273         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
2274         -Wmaybe-uninitialized.
2276 2014-02-27  Alan Modra  <amodra@gmail.com>
2278         PR target/57936
2279         * reload1.c (emit_input_reload_insns): When reload_override_in,
2280         set old to rl->in_reg when rl->in_reg is a subreg.
2282 2014-02-26  Richard Biener  <rguenther@suse.de>
2284         PR bootstrap/60343
2285         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
2287 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2289         * common/config/i386/predicates.md (const1256_operand): Remove.
2290         (const2356_operand): New.
2291         (const_1_to_2_operand): Remove.
2292         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2293         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2294         (*avx512pf_gatherpf<mode>sf): Ditto.
2295         (avx512pf_gatherpf<mode>df): Ditto.
2296         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2297         (*avx512pf_gatherpf<mode>df): Ditto.
2298         (avx512pf_scatterpf<mode>sf): Ditto.
2299         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2300         (*avx512pf_scatterpf<mode>sf): Ditto.
2301         (avx512pf_scatterpf<mode>df): Ditto.
2302         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2303         (*avx512pf_scatterpf<mode>df): Ditto.
2304         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
2306 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
2308         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
2309         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
2310         (_mm512_mask_testn_epi64_mask): Move to ...
2311         * config/i386/avx512cdintrin.h: Here.
2312         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
2313         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
2314         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
2315         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
2316         TARGET_AVX512F from TARGET_AVX512CD.
2318 2014-02-26  Richard Biener  <rguenther@suse.de>
2320         PR ipa/60327
2321         * ipa.c (walk_polymorphic_call_targets): Properly guard
2322         call to inline_update_overall_summary.
2324 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
2326         PR target/60280
2327         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
2328         and latches only if requested.  Fix latch if it is removed.
2329         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
2330         LOOPS_HAVE_PREHEADERS.
2332 2014-02-25  Andrew Pinski  <apinski@cavium.com>
2334         * builtins.c (expand_builtin_thread_pointer): Create a new target
2335         when the target is NULL.
2337 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
2339         PR rtl-optimization/60317
2340         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2341         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2342         * lra-assigns.c: Include params.h.
2343         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
2344         other reload pseudos considerations.
2346 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2348         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
2349         to use canonical form for nor<mode>3.
2351 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2353         PR target/55426
2354         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2355         conversions.
2357 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2359         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2360         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2361         (ix86_handle_option): Handle OPT_mprefetchwt1.
2362         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2363         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2364         PREFETCHWT1 CPUID.
2365         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2366         OPTION_MASK_ISA_PREFETCHWT1.
2367         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2368         (PTA_PREFETCHWT1): New.
2369         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2370         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2371         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2372         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2373         (*prefetch_avx512pf_<mode>_: Change into ...
2374         (*prefetch_prefetchwt1_<mode>: This.
2375         * config/i386/i386.opt (mprefetchwt1): New.
2376         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2377         (_mm_prefetch): Handle intent to write.
2378         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2380 2014-02-25  Richard Biener  <rguenther@suse.de>
2382         PR middle-end/60291
2383         * emit-rtl.c (mem_attrs_htab): Remove.
2384         (mem_attrs_htab_hash): Likewise.
2385         (mem_attrs_htab_eq): Likewise.
2386         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2387         (init_emit_once): Do not allocate mem_attrs_htab.
2389 2014-02-25  Richard Biener  <rguenther@suse.de>
2391         PR lto/60319
2392         * lto-opts.c (lto_write_options): Output non-explicit conservative
2393         -fwrapv, -fno-trapv and -fno-strict-overflow.
2394         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2395         (run_gcc): And pass them through.
2397 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2399         * sel-sched.c (calculate_new_fences): New parameter ptime.
2400         Calculate it as a maximum over all fence cycles.
2401         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2402         Print the final schedule timing when sched_verbose.
2404 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2406         PR rtl-optimization/60292
2407         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2408         bit fot the fence instruction.
2410 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2412         * calls.h: Fix typo in comment.
2414 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2416         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2417         adjusting offsetable addresses.
2419 2014-02-24  Guozhi Wei  <carrot@google.com>
2421         * sparseset.h (sparseset_pop): Fix the wrong index.
2423 2014-02-24  Walter Lee  <walt@tilera.com>
2425         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2426         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2427         triplet.
2428         * common/config/tilegx/tilegx-common.c
2429         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2430         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2431         (LINK_SPEC): Ditto.
2432         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2433         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2434         (tilegx_gimplify_va_arg_expr): Handle big endian.
2435         (tilegx_expand_unaligned_load): Ditto.
2436         (tilegx_expand_unaligned_store): Ditto.
2437         (TARGET_RETURN_IN_MSB): New.
2438         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2439         (TARGET_ENDIAN_DEFAULT): New.
2440         (TARGET_BIG_ENDIAN): Handle big endian.
2441         (BYTES_BIG_ENDIAN): Ditto.
2442         (WORDS_BIG_ENDIAN): Ditto.
2443         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2444         (ENDIAN_SPEC): New.
2445         (EXTRA_SPECS): New.
2446         * config/tilegx/tilegx.md (extv): Handle big endian.
2447         (extzv): Ditto.
2448         (insn_st<n>): Ditto.
2449         (insn_st<n>_add<bitsuffix>): Ditto.
2450         (insn_stnt<n>): Ditto.
2451         (insn_stnt<n>_add<bitsuffix>):Ditto.
2452         (vec_interleave_highv8qi): Handle big endian.
2453         (vec_interleave_highv8qi_be): New.
2454         (vec_interleave_highv8qi_le): New.
2455         (insn_v1int_h): Handle big endian.
2456         (vec_interleave_lowv8qi): Handle big endian.
2457         (vec_interleave_lowv8qi_be): New.
2458         (vec_interleave_lowv8qi_le): New.
2459         (insn_v1int_l): Handle big endian.
2460         (vec_interleave_highv4hi): Handle big endian.
2461         (vec_interleave_highv4hi_be): New.
2462         (vec_interleave_highv4hi_le): New.
2463         (insn_v2int_h): Handle big endian.
2464         (vec_interleave_lowv4hi): Handle big endian.
2465         (vec_interleave_lowv4hi_be): New.
2466         (vec_interleave_lowv4hi_le): New.
2467         (insn_v2int_l): Handle big endian.
2468         (vec_interleave_highv2si): Handle big endian.
2469         (vec_interleave_highv2si_be): New.
2470         (vec_interleave_highv2si_le): New.
2471         (insn_v4int_h): Handle big endian.
2472         (vec_interleave_lowv2si): Handle big endian.
2473         (vec_interleave_lowv2si_be): New.
2474         (vec_interleave_lowv2si_le): New.
2475         (insn_v4int_l): Handle big endian.
2476         * config/tilegx/tilegx.opt (mbig-endian): New option.
2477         (mlittle-endian): New option.
2478         * doc/install.texi: Document tilegxbe-linux.
2479         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2481 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2483         PR ipa/60266
2484         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2485         there are no parameter descriptors.
2487 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2489         PR rtl-optimization/60268
2490         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2491         initialization to ...
2492         (sched_rgn_init): ... here.
2493         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2495 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2497         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2498         names.
2500 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2502         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2503         definition.
2505 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2507         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2508         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2510 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2512         * config/microblaze/predicates.md: Add cmp_op predicate.
2513         * config/microblaze/microblaze.md: Add branch_compare instruction
2514         which uses cmp_op predicate and emits cmp insn before branch.
2515         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2516         to microblaze_expand_conditional_branch and consolidate logic.
2517         (microblaze_expand_conditional_branch): emit branch_compare
2518         insn instead of handling cmp op separate from branch insn.
2520 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2522         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2523         to permit subregs.
2525 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2527         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2528         define_insn with define_expand and new define_insn
2529         *altivec_lve<VI_char>x_internal.
2530         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2531         and new define_insn *altivec_stve<VI_char>x_internal.
2532         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2533         prototype.
2534         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2535         lve*x built-ins.
2536         (altivec_expand_stvex_be): New function.
2538 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2540         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2541         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2542         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2543         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2545 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2547         PR target/60298
2548         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2549         instead of emit_move_insn.
2551 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2553         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2554         vspltw with vsldoi.
2555         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2556         gen_altivec_vsumsws.
2558 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2560         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2561         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2562         (altivec_lvxl_<mode>): New define_expand incorporating
2563         -maltivec=be semantics where needed.
2564         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2565         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2566         semantics where needed.
2567         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2568         (altivec_stvx_<mode>): New define_expand incorporating
2569         -maltivec=be semantics where needed.
2570         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2571         VM2 iterator instead of V4SI.
2572         (altivec_stvxl_<mode>): New define_expand incorporating
2573         -maltivec=be semantics where needed.
2574         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2575         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2576         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2577         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2578         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2579         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2580         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2581         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2582         ALTIVEC_BUILTIN_STVXL.
2583         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2584         (altivec_expand_stvx_be): Likewise.
2585         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2586         (altivec_expand_lvx_be): Likewise.
2587         (altivec_expand_stvx_be): Likewise.
2588         (altivec_expand_builtin): Add cases for
2589         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2590         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2591         (altivec_init_builtins): Add definitions for
2592         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2593         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2595 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2597         * doc/invoke.texi (mvirt, mno-virt): Document.
2598         * config/mips/mips.opt (mvirt): New option.
2599         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2601 2014-02-21  Richard Biener  <rguenther@suse.de>
2603         PR tree-optimization/60276
2604         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2605         (STMT_VINFO_MIN_NEG_DIST): New macro.
2606         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2607         STMT_VINFO_MIN_NEG_DIST.
2608         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2609         made for negative dependence distances still hold.
2611 2014-02-21  Richard Biener  <rguenther@suse.de>
2613         PR middle-end/60291
2614         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2615         DECL_INITIAL for globals not in the current function context.
2617 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2619         PR tree-optimization/56490
2620         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2621         * tree-ssa-uninit.c: Include params.h.
2622         (compute_control_dep_chain): Add num_calls argument, return false
2623         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2624         num_calls to recursive call.
2625         (find_predicates): Change dep_chain into normal array,
2626         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2627         variable and adjust compute_control_dep_chain caller.
2628         (find_def_preds): Likewise.
2630 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2632         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2633         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2635 2014-02-21  Nick Clifton  <nickc@redhat.com>
2637         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2638         (pushhi1): Likewise.
2639         (popqi1): Add mode to pre_dec.
2640         (pophi1): Likewise.
2642 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2644         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2645         mode for mask of V8SFmode permutation.
2647 2014-02-20  Richard Henderson  <rth@redhat.com>
2649         PR c++/60272
2650         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2651         a new pseudo for OLDVAL.
2653 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2655         PR target/57896
2656         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2657         gen_reg_rtx if d->testing_p.
2658         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2659         if d->testing_p and we will certainly return true.
2660         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2661         if d->testing_p.
2663 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2665         * emit-rtl.c (gen_reg_rtx): Assert that
2666         crtl->emit.regno_pointer_align_length is non-zero.
2668 2014-02-20  Richard Henderson  <rth@redhat.com>
2670         PR c++/60272
2671         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2672         on failure the store back into EXPECT.
2674 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2675             Sandra Loosemore  <sandra@codesourcery.com>
2677         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2678         * config/nios2/nios2.c (nios2_function_profiler): Add
2679         -fPIC (flag_pic == 2) support.
2680         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2681         (nios2_large_offset_p): New function.
2682         (nios2_unspec_reloc_p): Move up position, update to use
2683         nios2_large_offset_p.
2684         (nios2_unspec_address): Remove function.
2685         (nios2_unspec_offset): New function.
2686         (nios2_large_got_address): New function.
2687         (nios2_got_address): Add large offset support.
2688         (nios2_legitimize_tls_address): Update usage of removed and new
2689         functions.
2690         (nios2_symbol_binds_local_p): New function.
2691         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2692         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2693         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2694         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2695         processing for (const (unspec ...)).
2696         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2698 2014-02-20  Richard Biener  <rguenther@suse.de>
2700         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2701         doing the substitution.
2702         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2704 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2706         PR ipa/55260
2707         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2708         info when checking whether lattices are bottom.
2710 2014-02-20  Richard Biener  <rguenther@suse.de>
2712         PR middle-end/60221
2713         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2714         regions at -O0.
2716 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2718         PR ipa/58555
2719         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2720         parameter specifying the scaling.
2721         (inline_call): Update.
2722         (want_inline_recursively): Guard division by zero.
2723         (recursive_inlining): Update.
2724         * ipa-inline.h (clone_inlined_nodes): Update.
2726 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2728         PR target/60204
2729         * config/i386/i386.c (classify_argument): Pass structures of size
2730         64 bytes or less in register.
2732 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2733             Kirill Yukhin  <kirill.yukhin@intel.com>
2735         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2736         (_mm_rcp28_round_ss): Ditto.
2737         (_mm_rsqrt28_round_sd): Ditto.
2738         (_mm_rsqrt28_round_ss): Ditto.
2739         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2740         (_mm_rcp14_round_ss): Ditto.
2741         (_mm_rsqrt14_round_sd): Ditto.
2742         (_mm_rsqrt14_round_ss): Ditto.
2743         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2744         the first input operand, get rid of match_dup.
2745         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2746         attribute to sse.
2747         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2748         Ditto.
2749         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2750         operand as the first input operand, set type attribute.
2751         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2752         Set type attribute.
2753         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2754         operand as the first input operand, set type attribute.
2756 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2758         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2759         bit of zero.
2761 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2763         PR target/60207
2764         * config/i386/i386.c (construct_container): Remove TFmode check
2765         for X86_64_INTEGER_CLASS.
2767 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2769         PR target/59794
2770         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2771         only when -Wpsabi is enabled.
2773 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2775          PR target/59799
2776         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2777         passing arrays in registers are the same as for structs, so remove the
2778         special case for them.
2780 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2782         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2783         destination type, extract only the valid bits if the source type is not
2784         integral and has a different mode.
2786 2014-02-19  Richard Biener  <rguenther@suse.de>
2788         PR ipa/60243
2789         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2790         for all calls.
2792 2014-02-19  Richard Biener  <rguenther@suse.de>
2794         PR ipa/60243
2795         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2796         (ipa_modify_call_arguments): Emit an argument load explicitely and
2797         preserve virtual SSA form there and for the replacement call.
2798         Do not update SSA form nor free dominance info.
2800 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2802         * ipa.c (function_and_variable_visibility): Also clear WEAK
2803         flag when disolving COMDAT_GROUP.
2805 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2807         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2808         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2809         not devirtualizing.
2810         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2811         do more sanity checks.
2812         (detect_type_change): Return true when giving up early.
2813         (compute_complex_assign_jump_func): Fix type parameter of
2814         ipa_set_ancestor_jf.
2815         (compute_complex_ancestor_jump_func): Likewise.
2816         (update_jump_functions_after_inlining): Fix updating of
2817         ancestor function.
2818         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2820 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2822         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2823         inline clones when edge disappears.
2825 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2827         PR target/60203
2828         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2829         Split 64-bit moves into 2 patterns.  Do not allow the use of
2830         direct move for TDmode in little endian, since the decimal value
2831         has little endian bytes within a word, but the 64-bit pieces are
2832         ordered in a big endian fashion, and normal subreg's of TDmode are
2833         not allowed.
2834         (mov<mode>_64bit_dm): Likewise.
2835         (movtd_64bit_nodm): Likewise.
2837 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2839         PR tree-optimization/60174
2840         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2841         statement of an SSA_NAME that occurs in an abnormal PHI node.
2843 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2845         PR sanitizer/60142
2846         * final.c (SEEN_BB): Remove.
2847         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2848         (final_scan_insn): Don't force_source_line on second
2849         NOTE_INSN_BASIC_BLOCK.
2851 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2853         PR target/60205
2854         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2855         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2856         (type_natural_mode): Warn ABI change when %zmm register is not
2857         available for AVX512F vector value passing.
2859 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2861         PR target/60193
2862         * config/i386/i386.c (ix86_expand_prologue): Use value in
2863         rax register as displacement when restoring %r10 or %rax.
2864         Fix wrong offset when restoring both registers.
2866 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2868         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2869         assertion with conditional return.
2871 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2872             Uros Bizjak  <ubizjak@gmail.com>
2874         PR driver/60233
2875         * config/i386/driver-i386.c (host_detect_local_cpu): If
2876         YMM state is not saved by the OS, also clear has_f16c.  Move
2877         CPUID 0x80000001 handling before YMM state saving checking.
2879 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2881         PR rtl-optimization/58960
2882         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2883         factored out from ...
2884         (sched_init): ... here.
2885         (free_global_sched_pressure_data): New, factored out from ...
2886         (sched_finish): ... here.
2887         * sched-int.h (free_global_sched_pressure_data): Declare.
2888         * sched-rgn.c (nr_regions_initial): New static global.
2889         (haifa_find_rgns): Initialize it.
2890         (schedule_region): Disable sched-pressure for the newly
2891         generated regions.
2893 2014-02-17  Richard Biener  <rguenther@suse.de>
2895         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2896         release SSA defs of pattern stmts.
2898 2014-02-17  Richard Biener  <rguenther@suse.de>
2900         * tree-inline.c (expand_call_inline): Release the virtual
2901         operand defined by the call we are about to inline.
2903 2014-02-17  Richard Biener  <rguenther@suse.de>
2905         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2907 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2908             Ilya Tocar  <ilya.tocar@intel.com>
2910         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2911         arguments order in builtin.
2912         (_mm512_permutexvar_epi64): Ditto.
2913         (_mm512_mask_permutexvar_epi64): Ditto
2914         (_mm512_maskz_permutexvar_epi32): Ditto
2915         (_mm512_permutexvar_epi32): Ditto
2916         (_mm512_mask_permutexvar_epi32): Ditto
2918 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2920         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2921         (p8_vmrgow): Likewise.
2923 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2925         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2926         endian targets.
2928 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2930         PR target/60203
2931         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2932         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2933         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2934         using direct move instructions on ISA 2.07.  Also adjust
2935         instruction length for 64-bit.
2936         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2937         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2939 2014-02-15  Alan Modra  <amodra@gmail.com>
2941         PR target/58675
2942         PR target/57935
2943         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2944         find_replacement on parts of insn rtl that might be reloaded.
2946 2014-02-15  Richard Biener  <rguenther@suse.de>
2948         PR tree-optimization/60183
2949         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2950         (tree_ssa_phiprop): Calculate and free post-dominators.
2952 2014-02-14  Jeff Law  <law@redhat.com>
2954         PR rtl-optimization/60131
2955         * ree.c (get_extended_src_reg): New function.
2956         (combine_reaching_defs): Use it rather than assuming location of REG.
2957         (find_and_remove_re): Verify first operand of extension is
2958         a REG before adding the insns to the copy list.
2960 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2962         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2963         * configure: Regenerated.
2964         * config.in: Regenerated.
2965         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2966         instead of ASM_SHORT.
2968 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2969             Richard Earnshaw  <rearnsha@arm.com>
2971         PR rtl-optimization/59535
2972         * lra-constraints.c (process_alt_operands): Encourage alternative
2973         when unassigned pseudo class is superset of the alternative class.
2974         (inherit_reload_reg): Don't inherit when optimizing for code size.
2975         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2976         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2977         modes not less than 4 for Thumb1.
2979 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2981         PR pch/60010
2982         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2984 2014-02-14  Richard Biener  <rguenther@suse.de>
2986         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2987         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2988         Do not strip INDIRECT_REFs.
2990 2014-02-14  Richard Biener  <rguenther@suse.de>
2992         PR lto/60179
2993         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2994         DECL_FUNCTION_SPECIFIC_TARGET.
2995         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2996         * tree-streamer-out.c (pack_ts_target_option): Remove.
2997         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2998         (write_ts_function_decl_tree_pointers): Do not stream
2999         DECL_FUNCTION_SPECIFIC_TARGET.
3000         * tree-streamer-in.c (unpack_ts_target_option): Remove.
3001         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
3002         (lto_input_ts_function_decl_tree_pointers): Do not stream
3003         DECL_FUNCTION_SPECIFIC_TARGET.
3005 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
3007         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
3008         (get_initial_def_for_induction, vectorizable_induction): Ignore
3009         debug stmts when looking for exit_phi.
3010         (vectorizable_live_operation): Fix up condition.
3012 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3014         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
3015         nreverse() because it changes the content of original tree list.
3017 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3019         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
3020         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
3022 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3024         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
3025         GNU coding standards.
3027 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
3029         PR debug/60152
3030         * dwarf2out.c (gen_subprogram_die): Don't call
3031         add_calling_convention_attribute if subr_die is old_die.
3033 2014-02-13  Sharad Singhai  <singhai@google.com>
3035         * doc/optinfo.texi: Fix order of nodes.
3037 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
3039         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
3040         operands[2], not operands[3].
3042 2014-02-13  Richard Biener  <rguenther@suse.de>
3044         PR bootstrap/59878
3045         * doc/install.texi (ISL): Update recommended version to 0.12.2,
3046         mention the possibility of an in-tree build.
3047         (CLooG): Update recommended version to 0.18.1, mention the
3048         possibility of an in-tree build and clarify that the ISL
3049         bundled with CLooG does not work.
3051 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
3053         PR target/43546
3054         * expr.c (compress_float_constant): If x is a hard register,
3055         extend into a pseudo and then move to x.
3057 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
3059         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
3060         caused by bad second argument to warning_at() with -mhotpatch and
3061         nested functions (e.g. with gfortran).
3063 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
3065         * opts.c (option_name): Remove "enabled by default" rider.
3067 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
3069         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
3071 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
3072             Uros Bizjak  <ubizjak@gmail.com>
3074         PR target/60151
3075         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
3076         * configure: Regenerated.
3078 2014-02-12  Richard Biener  <rguenther@suse.de>
3080         * vec.c (vec_prefix::calculate_allocation): Move as
3081         inline variant to vec.h.
3082         (vec_prefix::calculate_allocation_1): New out-of-line version.
3083         * vec.h (vec_prefix::calculate_allocation_1): Declare.
3084         (vec_prefix::m_has_auto_buf): Rename to ...
3085         (vec_prefix::m_using_auto_storage): ... this.
3086         (vec_prefix::calculate_allocation): Inline the easy cases
3087         and dispatch to calculate_allocation_1 which doesn't need the
3088         prefix address.
3089         (va_heap::reserve): Use gcc_checking_assert.
3090         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
3091         m_using_auto_storage.
3092         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
3093         member and adjust.
3094         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
3095         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
3096         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
3098 2014-02-12  Richard Biener  <rguenther@suse.de>
3100         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
3101         when we found a dependence.
3103 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
3105         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
3106         common code...
3107         (maybe_fold_stmt): ... into this new function.
3108         * omp-low.c (lower_omp): Update comment.
3110         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
3111         last use.
3113         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
3114         dereference.
3116 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
3118         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
3119         identifiers in comments.
3120         (cortexa53_extra_costs): Likewise.
3121         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
3122         (cortexa7_extra_costs): Likewise.
3123         (cortexa12_extra_costs): Likewise.
3124         (cortexa15_extra_costs): Likewise.
3125         (v7m_extra_costs): Likewise.
3127 2014-02-12  Richard Biener  <rguenther@suse.de>
3129         PR middle-end/60092
3130         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
3131         of posix_memalign being successful.
3132         (lower_stmt): Restrict lowering of posix_memalign to when
3133         -ftree-bit-ccp is enabled.
3135 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
3137         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
3138         arg_loc.
3139         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
3141 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
3143         PR rtl-optimization/60116
3144         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
3145         other_insn once the combination has been validated.
3147 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
3149         PR lto/59468
3150         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
3151         and wrapper.
3152         * ipa-devirt.c: Include demangle.h
3153         (odr_violation_reported): New static variable.
3154         (add_type_duplicate): Update odr_violations.
3155         (maybe_record_node): Add completep parameter; update it.
3156         (record_target_from_binfo): Add COMPLETEP parameter;
3157         update it as needed.
3158         (possible_polymorphic_call_targets_1): Likewise.
3159         (struct polymorphic_call_target_d): Add nonconstruction_targets;
3160         rename FINAL to COMPLETE.
3161         (record_targets_from_bases): Sanity check we found the binfo;
3162         fix COMPLETEP updating.
3163         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
3164         parameter, fix computing of COMPLETEP.
3165         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
3166         at LTO time do demangling.
3167         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
3168         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
3169         parameter.
3170         (gimple_get_virt_method_for_binfo): Likewise.
3171         * gimple-fold.h (gimple_get_virt_method_for_binfo,
3172         gimple_get_virt_method_for_vtable): Update prototypes.
3174 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
3176         PR target/49008
3177         * genautomata.c (add_presence_absence): Fix typo with
3178         {final_}presence_list.
3180 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
3182         PR target/60137
3183         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
3184         for VSX/Altivec vectors that land in GPR registers.
3186 2014-02-11  Richard Henderson  <rth@redhat.com>
3187             Jakub Jelinek  <jakub@redhat.com>
3189         PR debug/59776
3190         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
3191         around drhs if type conversion to lacc->type is not useless.
3193 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3195         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
3196         tuning struct.
3197         (cortex-a57.cortex-a53): Likewise.
3198         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
3200 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3202         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
3203         arm_restrict_it.
3205 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
3207         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
3208         add_options_for_arm_vfp3.
3210 2014-02-11  Jeff Law  <law@redhat.com>
3212         PR middle-end/54041
3213         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
3214         object with an undesirable mode.
3216 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3218         PR libgomp/60107
3219         * config/i386/sol2-9.h: New file.
3220         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
3221         *-*-solaris2.9*): Use it.
3223 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3225         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
3226         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
3228 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3230         * config/microblaze/microblaze.c: Extend mcpu version format
3232 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
3234         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
3236 2014-02-10  Richard Henderson  <rth@redhat.com>
3238         PR target/59927
3239         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
3240         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
3241         ms-abi vs -mno-accumulate-outgoing-args.
3242         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
3243         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
3244         respect to ms-abi.
3246 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3248         PR middle-end/60080
3249         * cfgexpand.c (expand_asm_operands): Attach source location to
3250         ASM_INPUT rtx objects.
3251         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
3253 2014-02-10  Nick Clifton  <nickc@redhat.com>
3255         * config/mn10300/mn10300.c (popcount): New function.
3256         (mn10300_expand_prologue): Include saved registers in stack usage
3257         count.
3259 2014-02-10  Jeff Law  <law@redhat.com>
3261         PR middle-end/52306
3262         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
3263         when changing the SET_DEST of a prior insn to avoid an input reload.
3265 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
3267         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
3268         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
3269         -mcall-openbsd, or -mcall-linux.
3270         (CC1_ENDIAN_BIG_SPEC): Remove.
3271         (CC1_ENDIAN_LITTLE_SPEC): Remove.
3272         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3273         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
3274         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
3275         and %cc1_endian_default.
3276         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3278 2014-02-10  Richard Biener  <rguenther@suse.de>
3280         PR tree-optimization/60115
3281         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3282         MEM_REF handling.  Properly verify that the accesses are not
3283         out of the objects bound.
3285 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3287         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
3288         coretex to cortex.
3290 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
3292         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
3293         proper constants and fix formatting.
3294         (possible_polymorphic_call_targets): Fix formatting.
3296 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
3297             Ilya Tocar  <ilya.tocar@intel.com>
3299         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
3300         (_mm512_loadu_epi32): Renamed into...
3301         (_mm512_loadu_si512): This.
3302         (_mm512_storeu_epi32): Renamed into...
3303         (_mm512_storeu_si512): This.
3304         (_mm512_maskz_ceil_ps): Removed.
3305         (_mm512_maskz_ceil_pd): Ditto.
3306         (_mm512_maskz_floor_ps): Ditto.
3307         (_mm512_maskz_floor_pd): Ditto.
3308         (_mm512_floor_round_ps): Ditto.
3309         (_mm512_floor_round_pd): Ditto.
3310         (_mm512_ceil_round_ps): Ditto.
3311         (_mm512_ceil_round_pd): Ditto.
3312         (_mm512_mask_floor_round_ps): Ditto.
3313         (_mm512_mask_floor_round_pd): Ditto.
3314         (_mm512_mask_ceil_round_ps): Ditto.
3315         (_mm512_mask_ceil_round_pd): Ditto.
3316         (_mm512_maskz_floor_round_ps): Ditto.
3317         (_mm512_maskz_floor_round_pd): Ditto.
3318         (_mm512_maskz_ceil_round_ps): Ditto.
3319         (_mm512_maskz_ceil_round_pd): Ditto.
3320         (_mm512_expand_pd): Ditto.
3321         (_mm512_expand_ps): Ditto.
3322         * config/i386/i386.c (ix86_builtins): Remove
3323         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
3324         (bdesc_args): Ditto.
3325         * config/i386/predicates.md (const1256_operand): New.
3326         (const_1_to_2_operand): Ditto.
3327         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
3328         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3329         (*avx512pf_gatherpf<mode>sf): Ditto.
3330         (avx512pf_gatherpf<mode>df): Ditto.
3331         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3332         (*avx512pf_gatherpf<mode>df): Ditto.
3333         (avx512pf_scatterpf<mode>sf): Ditto.
3334         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3335         (*avx512pf_scatterpf<mode>sf): Ditto.
3336         (avx512pf_scatterpf<mode>df): Ditto.
3337         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3338         (*avx512pf_scatterpf<mode>df): Ditto.
3339         (avx512f_expand<mode>): Removed.
3340         (<shift_insn><mode>3<mask_name>): Change predicate type.
3342 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3344         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
3345         not at the end of datarefs vector use ordered_remove to avoid
3346         reordering datarefs vector.
3348         PR c/59984
3349         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
3350         mark local addressable non-static vars as GOVD_PRIVATE
3351         instead of GOVD_LOCAL.
3352         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3353         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3354         than copying them.
3356         PR middle-end/60092
3357         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3358         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3359         assume_aligned or alloc_align attributes.
3360         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3361         arguments.  Handle also assume_aligned and alloc_align attributes.
3362         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3363         calls to functions with assume_aligned or alloc_align attributes.
3364         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3366 2014-02-08  Terry Guo  <terry.guo@arm.com>
3368         * doc/invoke.texi: Document ARM -march=armv7e-m.
3370 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3372         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3373         flag on __cilkrts_rethrow builtin.
3375         PR ipa/60026
3376         * ipa-cp.c (determine_versionability): Fail at -O0
3377         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3378         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3380         Revert:
3381         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3383         PR ipa/60026
3384         * tree-inline.c (copy_forbidden): Fail for
3385         __attribute__((optimize (0))) functions.
3387 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3389         * varpool.c: Include pointer-set.h.
3390         (varpool_remove_unreferenced_decls): Variables in other partitions
3391         will not be output; be however careful to not lose information
3392         about partitioning.
3394 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3396         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3397         lookup in the vtable constructor.
3399 2014-02-07  Jeff Law  <law@redhat.com>
3401         PR target/40977
3402         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3403         define_insn_and_split.
3405         * ipa-inline.c (inline_small_functions): Fix typos.
3407 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3409         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3410         (s390_can_use_return_insn): Declare.
3411         * config/s390/s390.h (EPILOGUE_USES): Define.
3412         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3413         instructions.
3414         (s390_chunkify_start): Handle return JUMP_LABELs.
3415         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3416         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3417         (s390_can_use_return_insn): New functions.
3418         (s390_fix_long_loop_prediction): Handle conditional returns.
3419         (TARGET_SET_UP_BY_PROLOGUE): Define.
3420         * config/s390/s390.md (ANY_RETURN): New code iterator.
3421         (*creturn, *csimple_return, return, simple_return): New patterns.
3423 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3425         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3426         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3427         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3428         REG_CFA_RESTORE list when deciding not to restore a register.
3430 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3432         * config/s390/s390.c: Include tree-pass.h and context.h.
3433         (s390_early_mach): New function, split out from...
3434         (s390_emit_prologue): ...here.
3435         (pass_data_s390_early_mach): New pass structure.
3436         (pass_s390_early_mach): New class.
3437         (s390_option_override): Create and register early_mach pass.
3438         Move to end of file.
3440 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3442         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3443         to match for the exit block.
3445 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3447         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3448         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3449         Reject misaligned operands.
3451 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3453         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3455 2014-02-07  Richard Biener  <rguenther@suse.de>
3457         PR middle-end/60092
3458         * gimple-low.c (lower_builtin_posix_memalign): New function.
3459         (lower_stmt): Call it to lower posix_memalign in a way
3460         to make alignment info accessible.
3462 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3464         PR c++/60082
3465         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3466         __builtin_setjmp_receiver.
3468 2014-02-07  Richard Biener  <rguenther@suse.de>
3470         PR middle-end/60092
3471         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3472         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3473         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3474         Handle BUILT_IN_POSIX_MEMALIGN.
3475         (find_func_clobbers): Likewise.
3476         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3477         (call_may_clobber_ref_p_1): Likewise.
3479 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3481         PR ipa/59918
3482         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3483         sanity check.
3485 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3487         PR ipa/59469
3488         * lto-cgraph.c (lto_output_node): Use
3489         symtab_get_symbol_partitioning_class.
3490         (lto_output_varpool_node): likewise.
3491         (symtab_get_symbol_partitioning_class): Move here from
3492         lto/lto-partition.c
3493         * cgraph.h (symbol_partitioning_class): Likewise.
3494         (symtab_get_symbol_partitioning_class): Declare.
3496 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3498         * ggc.h (ggc_internal_cleared_alloc): New macro.
3499         * vec.h (vec_safe_copy): Handle memory stats.
3500         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3501         * target-globals.c (save_target_globals): Likewise.
3503 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3505         PR target/60077
3506         * expr.c (emit_move_resolve_push): Export; be bit more selective
3507         on when to clear alias set.
3508         * expr.h (emit_move_resolve_push): Declare.
3509         * function.h (struct function): Add tail_call_marked.
3510         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3511         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3512         * config/i386/i386.md (TImode move expander): De not call
3513         ix86_expand_push.
3514         (FP push expanders): Preserve memory attributes.
3515         * config/i386/sse.md (push<mode>1): Remove.
3516         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3517         (ix86_expand_push): Remove.
3518         * config/i386/mmx.md (push<mode>1): Remove.
3520 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3522         PR rtl-optimization/60030
3523         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3524         lopart with paradoxical subreg before shifting it up by hprec.
3526 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3528         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3529         Remove extra newline at end of file.
3530         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3531         (arm_issue_rate): Handle cortexa57.
3532         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3533         (cortex-a57.cortex-a53): Likewise.
3535 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3537         PR target/59575
3538         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3539         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3540         bitmask.
3541         (arm_expand_prologue): Adjust all callers.
3542         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3543         info, registers also at the lowest numbered registers side.  Use
3544         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3545         XEXP.
3547         PR debug/59992
3548         * var-tracking.c (adjust_mems): Before adding a SET to
3549         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3551 2014-02-06  Alan Modra  <amodra@gmail.com>
3553         PR target/60032
3554         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3555         change SDmode to DDmode when lra_in_progress.
3557 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3559         PR middle-end/59150
3560         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3561         free_data_ref on the dr first, and before goto again also set dr
3562         to the next dr.  For simd_lane_access, free old datarefs[i] before
3563         overwriting it.  For get_vectype_for_scalar_type failure, don't
3564         free_data_ref if simd_lane_access.
3566         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3568         PR target/60062
3569         * tree.h (opts_for_fn): New inline function.
3570         (opt_for_fn): Define.
3571         * config/i386/i386.c (ix86_function_regparm): Use
3572         opt_for_fn (decl, optimize) instead of optimize.
3574 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3576         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3577         for SYMBOL_REF in large memory model.
3579 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3581         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3582         and crypto support.
3583         (cortex-a57): Likewise.
3584         (cortex-a57.cortex-a53): Likewise.
3586 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3587             Kugan Vivekanandarajah  <kuganv@linaro.org>
3589         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3590         unaligned_access.
3591         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3593 2014-02-06  Richard Biener  <rguenther@suse.de>
3595         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3596         set_loop_copy and initialize_original_copy_tables.
3598 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3600         * config/aarch64/aarch64-simd.md
3601         (aarch64_ashr_simddi): Change QI to SI.
3603 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3604             Jakub Jelinek  <jakub@redhat.com>
3606         PR middle-end/60013
3607         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3608         of the dataflow.
3610 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3612         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3613         CODE_FOR_altivec_vpku[hw]um to
3614         CODE_FOR_altivec_vpku[hw]um_direct.
3615         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3616         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3617         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3618         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3620 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3622         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3623         generation for -maltivec=be.
3624         (altivec_vsumsws): Simplify redundant test.
3626 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3628         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3629         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3630         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3631         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3632         gen_altivec_vpkuwum.
3633         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3634         BYTES_BIG_ENDIAN.
3635         (altivec_vpks<VI_char>ss): Likewise.
3636         (altivec_vpks<VI_char>us): Likewise.
3637         (altivec_vpku<VI_char>us): Likewise.
3638         (altivec_vpku<VI_char>um): Likewise.
3639         (altivec_vpku<VI_char>um_direct): New (copy of
3640         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3641         internal use).
3642         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3643         target is little endian and -maltivec=be is not specified.
3644         (*altivec_vupkhs<VU_char>_direct): New (copy of
3645         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3646         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3647         target is little endian and -maltivec=be is not specified.
3648         (*altivec_vupkls<VU_char>_direct): New (copy of
3649         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3650         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3651         little endian and -maltivec=be is not specified.
3652         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3653         little endian and -maltivec=be is not specified.
3655 2014-02-05  Richard Henderson  <rth@redhat.com>
3657         PR debug/52727
3658         * combine-stack-adj.c: Revert r206943.
3659         * sched-int.h (struct deps_desc): Add last_args_size.
3660         * sched-deps.c (init_deps): Initialize it.
3661         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3662         contain REG_ARGS_SIZE notes.
3664 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3666         * lto-cgraph.c (asm_nodes_output): Make global.
3667         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3668         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3669         (driver_handle_option): Handle OPT_fwpa.
3671 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3673         PR ipa/59947
3674         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3675         a comment typo and formatting issue.  If odr_hash hasn't been
3676         created, return vNULL and set *completep to false.
3678         PR middle-end/57499
3679         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3680         bb with no successors.
3682 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3684         PR target/59718
3685         * doc/invoke.texi (-march): Clarify documentation for ARM.
3686         (-mtune): Likewise.
3687         (-mcpu): Likewise.
3689 2014-02-05  Richard Biener  <rguenther@suse.de>
3691         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3692         when not vectorizing because of too many alias checks.
3693         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3694         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3696 2014-02-05  Nick Clifton  <nickc@redhat.com>
3698         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3699         accept extended registers in any mode when compiling for the MN10300.
3701 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3703         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3704         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3705         sanitization attributes.
3706         (can_inline_edge_p): Likewise.
3707         (sanitize_attrs_match_for_inline_p): New function.
3709 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3711         * ipa-prop.c (detect_type_change): Shor circuit testing of
3712         type changes on THIS pointer.
3714 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3716         PR target/59777
3717         * config/pa/pa.c (legitimize_tls_address): Return original address
3718         if not passed a SYMBOL_REF rtx.
3719         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3720         addresses.
3721         (pa_emit_move_sequence): Simplify TLS source operands.
3722         (pa_legitimate_constant_p): Reject all TLS constants.
3723         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3724         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3726 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3728         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3729         groups when we know they are controlled by LTO.
3730         * varasm.c (default_binds_local_p_1): If object is in other partition,
3731         it will be resolved locally.
3733 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3735         * config/host-linux.c (linux_gt_pch_use_address): Don't
3736         use SSIZE_MAX because it is not always defined.
3738 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3740         PR bootstrap/59913
3741         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3742         threshold for pseudo splitting.
3743         (update_ebb_live_info): Process call argument hard registers and
3744         hard registers from insn definition too.
3745         (max_small_class_regs_num): New constant.
3746         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3747         reloads_num only for small register classes.  Don't split for
3748         outputs of jumps.
3750 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3752         PR ipa/60058
3753         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3754         is non-null.
3756 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3758         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3759         visibility is safe.
3761 2014-02-04  Marek Polacek  <polacek@redhat.com>
3763         * gdbinit.in (pel): Define.
3765 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3767         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3768         behavior.
3770 2014-02-04  Richard Biener  <rguenther@suse.de>
3772         PR lto/59723
3773         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3774         in function context local.
3775         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3776         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3777         similar to LTO_imported_decl_ref.
3779 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3781         PR tree-optimization/60002
3782         * cgraphclones.c (build_function_decl_skip_args): Clear
3783         DECL_LANG_SPECIFIC.
3785         PR tree-optimization/60023
3786         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3787         false to gsi_replace.
3788         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3789         has been in some EH region and vec_stmt could throw, add
3790         vec_stmt into the same EH region.
3791         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3792         has no lhs, ignore it.
3793         * internal-fn.c (expand_MASK_LOAD): Likewise.
3795         PR ipa/60026
3796         * tree-inline.c (copy_forbidden): Fail for
3797         __attribute__((optimize (0))) functions.
3799         PR other/58712
3800         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3801         is set, copy one less argument.
3802         (expand_simd_clones): Don't subtract clone_info->inbranch
3803         from simd_clone_struct_alloc argument.
3805         PR rtl-optimization/57915
3806         * recog.c (simplify_while_replacing): If all unary/binary/relational
3807         operation arguments are constant, attempt to simplify those.
3809         PR middle-end/59261
3810         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3811         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3813 2014-02-04  Richard Biener  <rguenther@suse.de>
3815         PR tree-optimization/60012
3816         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3817         TBAA disambiguation to all DDRs.
3819 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3821         PR target/59788
3822         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3823         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3825 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3827         PR ipa/59882
3828         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3830 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3832         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3833         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3835 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3837         PR ipa/59831
3838         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3839         to figure out targets of polymorphic calls with known decl.
3840         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3841         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3842         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3843         (get_polymorphic_call_info): ... here.
3844         (get_polymorphic_call_info_from_invariant): New function.
3846 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3848         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3849         lookup via vtable pointer; check for type consistency
3850         and turn inconsitent facts into UNREACHABLE.
3851         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3852         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3853         type inconsistent querries; return UNREACHABLE instead.
3855 2014-02-03  Richard Henderson  <rth@twiddle.net>
3857         PR tree-opt/59924
3858         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3859         already processed this node.
3860         (normalize_one_pred_1): Pass along mark_set.
3861         (normalize_one_pred): Create and destroy a pointer_set_t.
3862         (normalize_one_pred_chain): Likewise.
3864 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3866         PR gcov-profile/58602
3867         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3869 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3871         PR ipa/59831
3872         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3873         -fno-devirtualize; try to devirtualize by the knowledge of
3874         virtual table pointer given by aggregate propagation.
3875         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3876         (ipa_print_node_jump_functions): Dump also offset that
3877         is relevant for polymorphic calls.
3878         (determine_known_aggregate_parts): Add arg_type parameter; use it
3879         instead of determining the type from pointer type.
3880         (ipa_compute_jump_functions_for_edge): Update call of
3881         determine_known_aggregate_parts.
3882         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3883         (gimple_get_virt_method_for_binfo): ... here; simplify using
3884         vtable_pointer_value_to_vtable.
3885         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3886         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3887         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3888         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3889         POINTER_PLUS_EXPR.
3890         (vtable_pointer_value_to_binfo): ... here.
3891         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3893 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3895         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3896         redef of outer loop index variable.
3898 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3900         PR c++/53017
3901         PR c++/59211
3902         * doc/extend.texi (Function Attributes): Typo.
3904 2014-02-03  Cong Hou  <congh@google.com>
3906         PR tree-optimization/60000
3907         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3908         if the vectorized statement is a store.  A store statement can only
3909         appear at the end of pattern statements.
3911 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3913         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3914         (ix86_option_override_internal): Default long double to 64-bit for
3915         32-bit Bionic and to 128-bit for 64-bit Bionic.
3917         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3918         TARGET_LONG_DOUBLE_128 is true.
3919         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3921         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3922         (mlong-double-64): Negate -mlong-double-128.
3923         (mlong-double-128): New option.
3925         * config/i386/i386-c.c (ix86_target_macros): Define
3926         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3928         * doc/invoke.texi: Document -mlong-double-128.
3930 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3932         PR rtl-optimization/60024
3933         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3935 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3937         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3939 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3941         PR rtl-optimization/57662
3942         * sel-sched.c (code_motion_path_driver): Do not mark already not
3943         existing blocks in the visiting bitmap.
3945 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3947         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3948         on the insn being emitted.
3950 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3951             Will Deacon  <will.deacon@arm.com>
3953         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3955 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3957         * config/arm/arm-tables.opt: Regenerate.
3959 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3961         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3962         for vector types other than V16QImode.
3963         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3964         define_expand, and call altivec_expand_vec_perm_le when producing
3965         code with little endian element order.
3966         (*altivec_vperm_<mode>_internal): New insn having previous
3967         behavior of altivec_vperm_<mode>.
3968         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3969         altivec_expand_vec_perm_le when producing code with little endian
3970         element order.
3971         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3972         behavior of altivec_vperm_<mode>_uns.
3974 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3976         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3977         (altivec_vsumsws): Add handling for -maltivec=be with a little
3978         endian target.
3979         (altivec_vsumsws_direct): New.
3980         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3981         gen_altivec_vsumsws.
3983 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3985         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3986         vtable_pointer_value_to_binfo): New functions.
3987         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3988         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3990 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3992         * config/nios2/nios2.md (load_got_register): Initialize GOT
3993         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3994         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3996 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3998         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3999         preserverd by passthrough, do not propagate the type.
4001 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
4003         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
4004         (mips_atomic_assign_expand_fenv): New function.
4005         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
4007 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
4009         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
4010         (__builtin_mips_set_fcsr): Likewise.
4011         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
4012         MIPS_USI_FTYPE_VOID.
4013         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
4014         (mips16_expand_set_fcsr): Likewise.
4015         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
4016         (mips16_set_fcsr_stub): Likewise.
4017         (mips16_get_fcsr_one_only_stub): New class.
4018         (mips16_set_fcsr_one_only_stub): Likewise.
4019         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
4020         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
4021         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
4022         (hard_float): New availability predicate.
4023         (mips_builtins): Add get_fcsr and set_fcsr.
4024         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
4025         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
4026         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
4027         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
4028         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
4029         patterns.
4031 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
4033         * config/mips/mips.c (mips_one_only_stub): New class.
4034         (mips_need_mips16_rdhwr_p): Replace with...
4035         (mips16_rdhwr_stub): ...this new variable.
4036         (mips16_stub_call_address): New function.
4037         (mips16_rdhwr_one_only_stub): New class.
4038         (mips_expand_thread_pointer): Use mips16_stub_call_address.
4039         (mips_output_mips16_rdhwr): Delete.
4040         (mips_finish_stub): New function.
4041         (mips_code_end): Use it to handle rdhwr stubs.
4043 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
4045         PR target/60017
4046         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
4047         when calculating size of integer atomic types.
4049 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
4051         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
4053 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
4055         PR tree-optimization/60003
4056         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
4057         * profile.c (branch_prob): Use gimple_call_builtin_p
4058         to check for BUILT_IN_SETJMP_RECEIVER.
4059         * tree-inline.c (copy_bb): Call notice_special_calls.
4061 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
4063         PR bootstrap/59985
4064         * lra-constraints.c (process_alt_operands): Update reload_sum only
4065         on the first pass.
4067 2014-01-31  Richard Henderson  <rth@redhat.com>
4069         PR middle-end/60004
4070         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
4071         until after else_eh is processed.
4073 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
4075         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
4076         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
4077         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
4078         in smmintrin.h, remove them.
4079         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
4080         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
4081         * config/i386/i386.md (ROUND_SAE): Fix value.
4082         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
4083         (const48_operand): New.
4084         * config/i386/subst.md (round), (round_expand): Use
4085         const_4_or_8_to_11_operand.
4086         (round_saeonly), (round_saeonly_expand): Use const48_operand.
4088 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
4090         * config/i386/constraints.md (Yk): Swap meaning with k.
4091         * config/i386/i386.md (movhi_internal): Change Yk to k.
4092         (movqi_internal): Ditto.
4093         (*k<logic><mode>): Ditto.
4094         (*andhi_1): Ditto.
4095         (*andqi_1): Ditto.
4096         (kandn<mode>): Ditto.
4097         (*<code>hi_1): Ditto.
4098         (*<code>qi_1): Ditto.
4099         (kxnor<mode>): Ditto.
4100         (kortestzhi): Ditto.
4101         (kortestchi): Ditto.
4102         (kunpckhi): Ditto.
4103         (*one_cmplhi2_1): Ditto.
4104         (*one_cmplqi2_1): Ditto.
4105         * config/i386/sse.md (): Change k to Yk.
4106         (avx512f_load<mode>_mask): Ditto.
4107         (avx512f_blendm<mode>): Ditto.
4108         (avx512f_store<mode>_mask): Ditto.
4109         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
4110         (avx512f_storedqu<mode>_mask): Ditto.
4111         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4112         Ditto.
4113         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
4114         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
4115         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
4116         (avx512f_maskcmp<mode>3): Ditto.
4117         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
4118         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
4119         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
4120         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
4121         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
4122         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
4123         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
4124         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
4125         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
4126         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
4127         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
4128         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
4129         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
4130         (vec_extract_lo_<mode>_maskm): Ditto.
4131         (vec_extract_hi_<mode>_maskm): Ditto.
4132         (avx512f_vternlog<mode>_mask): Ditto.
4133         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
4134         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
4135         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
4136         (avx512f_<code>v8div16qi2_mask): Ditto.
4137         (avx512f_<code>v8div16qi2_mask_store): Ditto.
4138         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
4139         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
4140         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
4141         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
4142         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
4143         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4144         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
4145         (*avx512pf_scatterpf<mode>df_mask): Ditto.
4146         (avx512cd_maskb_vec_dupv8di): Ditto.
4147         (avx512cd_maskw_vec_dupv16si): Ditto.
4148         (avx512f_vpermi2var<mode>3_maskz): Ditto.
4149         (avx512f_vpermi2var<mode>3_mask): Ditto.
4150         (avx512f_vpermi2var<mode>3_mask): Ditto.
4151         (avx512f_vpermt2var<mode>3_maskz): Ditto.
4152         (*avx512f_gathersi<mode>): Ditto.
4153         (*avx512f_gathersi<mode>_2): Ditto.
4154         (*avx512f_gatherdi<mode>): Ditto.
4155         (*avx512f_gatherdi<mode>_2): Ditto.
4156         (*avx512f_scattersi<mode>): Ditto.
4157         (*avx512f_scatterdi<mode>): Ditto.
4158         (avx512f_compress<mode>_mask): Ditto.
4159         (avx512f_compressstore<mode>_mask): Ditto.
4160         (avx512f_expand<mode>_mask): Ditto.
4161         * config/i386/subst.md (mask): Change k to Yk.
4162         (mask_scalar_merge): Ditto.
4163         (sd): Ditto.
4165 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
4167         * doc/extend.texi (Vector Extensions): Document ?: in C++.
4169 2014-01-31  Richard Biener  <rguenther@suse.de>
4171         PR middle-end/59990
4172         * builtins.c (fold_builtin_memory_op): Make sure to not
4173         use a floating-point mode or a boolean or enumeral type for
4174         the copy operation.
4176 2014-01-30  DJ Delorie  <dj@redhat.com>
4178         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
4179         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
4180         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
4181         whenever main() has an epilogue.
4183 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4185         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
4186         unused variable "field".
4187         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
4188         (vsx_mergeh_<mode>): Likewise.
4189         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
4190         (altivec_vmrghh): Likewise.
4191         (altivec_vmrghw): Likewise.
4192         (altivec_vmrglb): Likewise.
4193         (altivec_vmrglh): Likewise.
4194         (altivec_vmrglw): Likewise.
4195         (altivec_vspltb): Add missing uses.
4196         (altivec_vsplth): Likewise.
4197         (altivec_vspltw): Likewise.
4198         (altivec_vspltsf): Likewise.
4200 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4202         PR target/59923
4203         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
4204         frame related instructions.
4206 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
4208         PR rtl-optimization/59959
4209         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
4210         any reload of register whose subreg is invalid.
4212 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4214         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
4215         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
4216         Add missing return type - void.
4218 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4220         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
4221         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
4222         remove element index adjustment for endian (now handled in vsx.md
4223         and altivec.md).
4224         (altivec_expand_vec_perm_const): Use
4225         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
4226         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
4227         (vsx_xxspltw_<mode>): Adjust element index for little endian.
4228         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
4229         define_expand and a new define_insn *altivec_vspltb_internal;
4230         adjust for -maltivec=be on a little endian target.
4231         (altivec_vspltb_direct): New.
4232         (altivec_vsplth): Divide into a define_expand and a new
4233         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
4234         little endian target.
4235         (altivec_vsplth_direct): New.
4236         (altivec_vspltw): Divide into a define_expand and a new
4237         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
4238         little endian target.
4239         (altivec_vspltw_direct): New.
4240         (altivec_vspltsf): Divide into a define_expand and a new
4241         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
4242         a little endian target.
4244 2014-01-30  Richard Biener  <rguenther@suse.de>
4246         PR tree-optimization/59993
4247         * tree-ssa-forwprop.c (associate_pointerplus): Check we
4248         can propagate form the earlier stmt and avoid the transform
4249         when the intermediate result is needed.
4251 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
4253         * README.Portability: Fix typo.
4255 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
4257         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
4258         comparison_operator with ordered_comparison_operator.
4260 2014-01-30  Nick Clifton  <nickc@redhat.com>
4262         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
4263         Rename to mn10300_store_multiple_regs.
4264         * config/mn10300/mn10300.c: Likewise.
4265         * config/mn10300/mn10300.md (store_movm): Fix typo: call
4266         store_multiple_regs.
4267         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
4268         Call mn10300_store_multiple_regs.
4270 2014-01-30  Nick Clifton  <nickc@redhat.com>
4271             DJ Delorie  <dj@redhat.com>
4273         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
4274         %fp 2 to keep registers after it properly word-aligned.
4275         (rl78_alloc_physical_registers_umul): Handle the case where both
4276         input operands are the same.
4278 2014-01-30  Richard Biener  <rguenther@suse.de>
4280         PR tree-optimization/59903
4281         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4282         check properly.
4284 2014-01-30  Jason Merrill  <jason@redhat.com>
4286         PR c++/59633
4287         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
4289         PR c++/59645
4290         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
4292 2014-01-30  Richard Biener  <rguenther@suse.de>
4294         PR tree-optimization/59951
4295         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
4297 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
4299         PR target/59784
4300         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
4301         SFmode to DFmode case.
4303 2014-01-29  DJ Delorie  <dj@redhat.com>
4305         * config/msp430/msp430.opt (-minrt): New.
4306         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
4307         if -minrt given.
4308         (ENDFILE_SPEC): Likewise.
4310 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
4312         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
4313         (estimate_function_body_sizes): Use it.
4315 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
4317         PR c++/58561
4318         * dwarf2out.c (is_cxx_auto): New.
4319         (is_base_type): Use it.
4320         (gen_type_die_with_usage): Likewise.
4322 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4324         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
4325         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
4326         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
4327         -maltivec=be with LE targets.
4328         (vsx_mergeh_<mode>): Likewise.
4329         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
4330         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
4331         (altivec_vmrghb): Replace with define_expand and new
4332         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
4333         (altivec_vmrghb_direct): New define_insn.
4334         (altivec_vmrghh): Replace with define_expand and new
4335         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
4336         (altivec_vmrghh_direct): New define_insn.
4337         (altivec_vmrghw): Replace with define_expand and new
4338         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
4339         (altivec_vmrghw_direct): New define_insn.
4340         (*altivec_vmrghsf): Adjust for endianness.
4341         (altivec_vmrglb): Replace with define_expand and new
4342         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
4343         (altivec_vmrglb_direct): New define_insn.
4344         (altivec_vmrglh): Replace with define_expand and new
4345         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
4346         (altivec_vmrglh_direct): New define_insn.
4347         (altivec_vmrglw): Replace with define_expand and new
4348         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
4349         (altivec_vmrglw_direct): New define_insn.
4350         (*altivec_vmrglsf): Adjust for endianness.
4351         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4352         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4353         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4354         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4355         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4356         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4357         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4358         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4360 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4362         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4363         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4364         whitespace.
4366 2014-01-29  Richard Biener  <rguenther@suse.de>
4368         PR tree-optimization/58742
4369         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4370         associate_pointerplus_align.
4371         (associate_pointerplus_diff): New function.
4372         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4373         and associate_pointerplus_diff.
4375 2014-01-29  Richard Biener  <rguenther@suse.de>
4377         * lto-streamer.h (LTO_major_version): Bump to 3.
4378         (LTO_minor_version): Reset to 0.
4380 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4382         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4383         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4384         (arm_file_start): Generate correct asm header for armv7ve.
4385         * config/arm/bpabi.h: Add multilib support for armv7ve.
4386         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4387         and cortex-a15 to armv7ve.
4388         * config/arm/t-aprofile: Add multilib support for armv7ve.
4389         * doc/invoke.texi: Document -march=armv7ve.
4391 2014-01-29  Richard Biener  <rguenther@suse.de>
4393         PR tree-optimization/58742
4394         * tree-ssa-forwprop.c (associate_plusminus): Return true
4395         if we changed sth, defer EH cleanup to ...
4396         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4397         (simplify_mult): New function.
4399 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4401         PR middle-end/59917
4402         PR tree-optimization/59920
4403         * tree.c (build_common_builtin_nodes): Remove
4404         __builtin_setjmp_dispatcher initialization.
4405         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4406         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4407         instead of gsi_after_labels + manually skipping debug stmts.
4408         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4409         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4410         * tree-inline.c (copy_edges_for_bb): Remove
4411         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4412         argument.  Ignore computed_goto_p stmts.  Don't call
4413         make_abnormal_goto_edges.  If a call might need abnormal edges
4414         for non-local gotos, see if it already has an edge to
4415         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4416         with true argument, don't do anything then, otherwise add
4417         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4418         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4419         caller.
4420         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4421         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4422         (lower_stmt): Don't set data->calls_builtin_setjmp.
4423         (lower_builtin_setjmp): Adjust comment.
4424         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4425         * tree-cfg.c (found_computed_goto): Remove.
4426         (factor_computed_gotos): Remove.
4427         (make_goto_expr_edges): Return bool, true for computed gotos.
4428         Don't call make_abnormal_goto_edges.
4429         (build_gimple_cfg): Don't set found_computed_goto, don't call
4430         factor_computed_gotos.
4431         (computed_goto_p): No longer static.
4432         (make_blocks): Don't set found_computed_goto.
4433         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4434         (make_edges): If make_goto_expr_edges returns true, push bb
4435         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4436         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4437         vector.  Record mapping between bbs and OpenMP regions if there
4438         are any, adjust make_gimple_omp_edges caller.  Call
4439         handle_abnormal_edges.
4440         (make_abnormal_goto_edges): Remove.
4441         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4442         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4443         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4444         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4445         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4446         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4447         filling *region also set *region_idx to (*region)->entry->index.
4449         PR other/58712
4450         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4451         For REGs set ORIGINAL_REGNO.
4453 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4455         * doc/md.texi: Mention that a target shouldn't implement
4456         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4457         than hi/lo pair.
4459 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4461         PR tree-optimization/59594
4462         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4463         a copy of the datarefs vector rather than the vector itself.
4465 2014-01-28  Jason Merrill  <jason@redhat.com>
4467         PR c++/53756
4468         * dwarf2out.c (auto_die): New static.
4469         (gen_type_die_with_usage): Handle C++1y 'auto'.
4470         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4471         on definition.
4473 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4475         PR target/59672
4476         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4477         (SPEC_X32): Likewise.
4478         (SPEC_64): Likewise.
4479         * config/i386/i386.c (ix86_option_override_internal): Turn off
4480         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4481         for TARGET_16BIT.
4482         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4483         * config/i386/i386.h (TARGET_16BIT): New macro.
4484         (TARGET_16BIT_P): Likewise.
4485         * config/i386/i386.opt: Add m16.
4486         * doc/invoke.texi: Document -m16.
4488 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4490         PR preprocessor/59935
4491         * input.c (location_get_source_line): Bail out on when line number
4492         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4494 2014-01-28  Richard Biener  <rguenther@suse.de>
4496         PR tree-optimization/58742
4497         * tree-ssa-forwprop.c (associate_plusminus): Handle
4498         pointer subtraction of the form (T)(P + A) - (T)P.
4500 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4502         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4503         at const_int_cost.
4505 2014-01-28  Richard Biener  <rguenther@suse.de>
4507         Revert
4508         2014-01-28  Richard Biener  <rguenther@suse.de>
4510         PR rtl-optimization/45364
4511         PR rtl-optimization/59890
4512         * var-tracking.c (local_get_addr_clear_given_value): Handle
4513         already cleared slot.
4514         (val_reset): Handle not allocated local_get_addr_cache.
4515         (vt_find_locations): Use post-order on the inverted CFG.
4517 2014-01-28  Richard Biener  <rguenther@suse.de>
4519         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4521 2014-01-28  Richard Biener  <rguenther@suse.de>
4523         PR rtl-optimization/45364
4524         PR rtl-optimization/59890
4525         * var-tracking.c (local_get_addr_clear_given_value): Handle
4526         already cleared slot.
4527         (val_reset): Handle not allocated local_get_addr_cache.
4528         (vt_find_locations): Use post-order on the inverted CFG.
4530 2014-01-28  Alan Modra  <amodra@gmail.com>
4532         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4533         * configure.ac <recursive call for build != host>: Define
4534         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4535         and LD_FOR_BUILD too.
4536         * configure: Regenerate.
4538 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4540         * config/i386/i386.c (get_builtin_code_for_version): Separate
4541         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4542         Broadwell from Haswell.
4544 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4546         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4547         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4548         * config/mips/mips.c (mips_option_override): Change setting
4549         of TARGET_DSP.
4550         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4551         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4552         Change from Mask to Var.
4554 2014-01-27  Jeff Law  <law@redhat.com>
4556         * ipa-inline.c (inline_small_functions): Fix typo.
4558 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4560         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4561         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4562         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4563         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4564         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4565         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4566         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4567         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4568         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4569         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4570         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4571         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4572         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4573         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4574         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4575         (_mm512_storeu_epi64): Ditto.
4576         (_mm512_cmpge_epi32_mask): Ditto.
4577         (_mm512_cmpge_epu32_mask): Ditto.
4578         (_mm512_cmpge_epi64_mask): Ditto.
4579         (_mm512_cmpge_epu64_mask): Ditto.
4580         (_mm512_cmple_epi32_mask): Ditto.
4581         (_mm512_cmple_epu32_mask): Ditto.
4582         (_mm512_cmple_epi64_mask): Ditto.
4583         (_mm512_cmple_epu64_mask): Ditto.
4584         (_mm512_cmplt_epi32_mask): Ditto.
4585         (_mm512_cmplt_epu32_mask): Ditto.
4586         (_mm512_cmplt_epi64_mask): Ditto.
4587         (_mm512_cmplt_epu64_mask): Ditto.
4588         (_mm512_cmpneq_epi32_mask): Ditto.
4589         (_mm512_cmpneq_epu32_mask): Ditto.
4590         (_mm512_cmpneq_epi64_mask): Ditto.
4591         (_mm512_cmpneq_epu64_mask): Ditto.
4592         (_mm512_expand_pd): Ditto.
4593         (_mm512_expand_ps): Ditto.
4594         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4595         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4596         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4597         * config/i386/i386.c (ix86_builtins): Add
4598         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4599         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4600         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4601         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4602         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4603         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4604         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4605         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4606         IX86_BUILTIN_PMOVUSQW512_MEM.
4607         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4608         __builtin_ia32_pmovsqd512mem_mask,
4609         __builtin_ia32_pmovqd512mem_mask,
4610         __builtin_ia32_pmovusqw512mem_mask,
4611         __builtin_ia32_pmovsqw512mem_mask,
4612         __builtin_ia32_pmovqw512mem_mask,
4613         __builtin_ia32_pmovusdw512mem_mask,
4614         __builtin_ia32_pmovsdw512mem_mask,
4615         __builtin_ia32_pmovdw512mem_mask,
4616         __builtin_ia32_pmovqb512mem_mask,
4617         __builtin_ia32_pmovusqb512mem_mask,
4618         __builtin_ia32_pmovsqb512mem_mask,
4619         __builtin_ia32_pmovusdb512mem_mask,
4620         __builtin_ia32_pmovsdb512mem_mask,
4621         __builtin_ia32_pmovdb512mem_mask.
4622         (bdesc_args): Add __builtin_ia32_expanddf512,
4623         __builtin_ia32_expandsf512.
4624         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4625         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4626         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4627         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4628         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4629         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4630         (avx512f_<code>v8div16qi2_mask_store): This.
4631         (avx512f_expand<mode>): New.
4633 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4635         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4636         New.
4637         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4638         (_mm512_prefetch_i32scatter_pd): Ditto.
4639         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4640         (_mm512_prefetch_i64scatter_pd): Ditto.
4641         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4642         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4643         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4644         (_mm512_prefetch_i32scatter_ps): Ditto.
4645         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4646         (_mm512_prefetch_i64scatter_ps): Ditto.
4647         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4648         * config/i386/i386-builtin-types.def: Define
4649         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4650         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4651         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4652         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4653         IX86_BUILTIN_SCATTERPFQPD.
4654         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4655         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4656         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4657         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4658         __builtin_ia32_scatterpfqps.
4659         (ix86_expand_builtin): Expand new built-ins.
4660         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4661         fix memory access data type.
4662         (*avx512pf_gatherpf<mode>_mask): Ditto.
4663         (*avx512pf_gatherpf<mode>): Ditto.
4664         (avx512pf_scatterpf<mode>): Ditto.
4665         (*avx512pf_scatterpf<mode>_mask): Ditto.
4666         (*avx512pf_scatterpf<mode>): Ditto.
4667         (GATHER_SCATTER_SF_MEM_MODE): New.
4668         (avx512pf_gatherpf<mode>df): Ditto.
4669         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4670         (*avx512pf_scatterpf<mode>df): Ditto.
4672 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4674         PR bootstrap/59934
4675         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4676         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4677         reached.
4679 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4681         * common/config/arm/arm-common.c
4682         (arm_rewrite_mcpu): Handle multiple names.
4683         * config/arm/arm.h
4684         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4686 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4688         * gimple-builder.h (create_gimple_tmp): Delete.
4690 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4692         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4693         words comparisons.
4695 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4697         * config/pa/pa.md (call): Generate indirect long calls to non-local
4698         functions when outputing 32-bit code.
4699         (call_value): Likewise except for special call to buggy powf function.
4701         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4702         portable runtime and PIC indirect calls.
4703         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4704         and PIC call sequences.  Use ldo instead of blr to set return register
4705         in PIC call sequence.
4707 2014-01-25  Walter Lee  <walt@tilera.com>
4709         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4710         avoid clobbering a live register.
4712 2014-01-25  Walter Lee  <walt@tilera.com>
4714         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4715         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4716         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4717         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4719 2014-01-25  Walter Lee  <walt@tilera.com>
4721         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4722         arguments on even registers.
4723         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4724         STACK_BOUNDARY.
4725         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4726         (BIGGEST_ALIGNMENT): Ditto.
4727         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4729 2014-01-25  Walter Lee  <walt@tilera.com>
4731         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4732         insns before bundling.
4733         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4735 2014-01-25  Walter Lee  <walt@tilera.com>
4737         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4738         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4739         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4741 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4743         * config/mips/constraints.md (kl): Delete.
4744         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4745         define expands, using...
4746         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4747         instructions for MIPS16.
4748         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4749         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4751 2014-01-25  Walter Lee  <walt@tilera.com>
4753         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4754         (clzdi2): Ditto.
4755         (ffsdi2): Ditto.
4757 2014-01-25  Walter Lee  <walt@tilera.com>
4759         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4760         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4762 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4764         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4765         Handle XOR.
4767 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4769         * print-rtl.c (in_call_function_usage): New var.
4770         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4771         EXPR_LIST mode as mode and not as reg note name.
4773         PR middle-end/59561
4774         * cfgloopmanip.c (copy_loop_info): If
4775         loop->warned_aggressive_loop_optimizations, make sure
4776         the flag is set in target loop too.
4778 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4780         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4781         flag_cilkplus.
4782         * builtins.def: Likewise.
4783         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4784         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4785         * ira.c (ira_setup_eliminable_regset): Likewise.
4786         * omp-low.c (gate_expand_omp): Likewise.
4787         (execute_lower_omp): Likewise.
4788         (diagnose_sb_0): Likewise.
4789         (gate_diagnose_omp_blocks): Likewise.
4790         (simd_clone_clauses_extract): Likewise.
4791         (gate): Likewise.
4793 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4795         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4796         correction for little endian...
4797         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4798         here.
4800 2014-01-24  Jeff Law  <law@redhat.com>
4802         PR tree-optimization/59919
4803         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4804         for non-returning calls.
4806 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4808         * common/config/aarch64/aarch64-common.c
4809         (aarch64_rewrite_mcpu): Handle multiple names.
4810         * config/aarch64/aarch64.h
4811         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4813 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4815         * input.c (add_file_to_cache_tab): Handle the case where fopen
4816         returns NULL.
4818 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4820         PR target/59929
4821         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4822         from push operand if code of push isn't PRE_DEC.
4824 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4826         PR target/59909
4827         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4828         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4829         it is only used for non-atomic loads/stores.
4831         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4832         -mquad-memory or -mquad-memory-atomic switches.
4834         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4835         -mquad-memory-atomic to ISA 2.07 support.
4837         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4838         to separate support of normal quad word memory operations (ldq, stq)
4839         from the atomic quad word memory operations.
4841         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4842         support to separate non-atomic quad word operations from atomic
4843         quad word operations.  Disable non-atomic quad word operations in
4844         little endian mode so that we don't have to swap words after the
4845         load and before the store.
4846         (quad_load_store_p): Add comment about atomic quad word support.
4847         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4848         options printed with -mdebug=reg.
4850         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4851         -mquad-memory-atomic as the test for whether we have quad word
4852         atomic instructions.
4853         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4854         or -mp8-vector are used, allow byte/half-word atomic operations.
4856         * config/rs6000/sync.md (load_lockedti): Insure that the address
4857         is a proper indexed or indirect address for the lqarx instruction.
4858         On little endian systems, swap the hi/lo registers after the lqarx
4859         instruction.
4860         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4861         insure the address is valid for the lqarx instruction.
4862         (store_conditionalti): Insure that the address is a proper indexed
4863         or indirect address for the stqcrx. instruction.  On little endian
4864         systems, swap the hi/lo registers before doing the stqcrx.
4865         instruction.
4866         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4867         insure the address is valid for the stqcrx. instruction.
4869         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4870         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4871         type of quad memory support is available.
4873 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4875         PR regression/59915
4876         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4877         there is a danger of looping.
4879 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4881         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4882         force flag_ira_loop_pressure if set via command line.
4884 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4886         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4887         (ashr_simd): New builtin handling DI mode.
4888         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4889         (aarch64_sshr_simddi): New match pattern.
4890         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4891         (vshrd_n_s64): Likewise.
4892         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4894 2014-01-23  Nick Clifton  <nickc@redhat.com>
4896         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4897         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4898         favour of mcu specific scripts.
4899         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4900         430x multilibs.
4902 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4903             Alex Velenko  <Alex.Velenko@arm.com>
4905         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4906         (vaddv_s16): Likewise.
4907         (vaddv_s32): Likewise.
4908         (vaddv_u8): Likewise.
4909         (vaddv_u16): Likewise.
4910         (vaddv_u32): Likewise.
4911         (vaddvq_s8): Likewise.
4912         (vaddvq_s16): Likewise.
4913         (vaddvq_s32): Likewise.
4914         (vaddvq_s64): Likewise.
4915         (vaddvq_u8): Likewise.
4916         (vaddvq_u16): Likewise.
4917         (vaddvq_u32): Likewise.
4918         (vaddvq_u64): Likewise.
4919         (vaddv_f32): Likewise.
4920         (vaddvq_f32): Likewise.
4921         (vaddvq_f64): Likewise.
4922         (vmaxv_f32): Likewise.
4923         (vmaxv_s8): Likewise.
4924         (vmaxv_s16): Likewise.
4925         (vmaxv_s32): Likewise.
4926         (vmaxv_u8): Likewise.
4927         (vmaxv_u16): Likewise.
4928         (vmaxv_u32): Likewise.
4929         (vmaxvq_f32): Likewise.
4930         (vmaxvq_f64): Likewise.
4931         (vmaxvq_s8): Likewise.
4932         (vmaxvq_s16): Likewise.
4933         (vmaxvq_s32): Likewise.
4934         (vmaxvq_u8): Likewise.
4935         (vmaxvq_u16): Likewise.
4936         (vmaxvq_u32): Likewise.
4937         (vmaxnmv_f32): Likewise.
4938         (vmaxnmvq_f32): Likewise.
4939         (vmaxnmvq_f64): Likewise.
4940         (vminv_f32): Likewise.
4941         (vminv_s8): Likewise.
4942         (vminv_s16): Likewise.
4943         (vminv_s32): Likewise.
4944         (vminv_u8): Likewise.
4945         (vminv_u16): Likewise.
4946         (vminv_u32): Likewise.
4947         (vminvq_f32): Likewise.
4948         (vminvq_f64): Likewise.
4949         (vminvq_s8): Likewise.
4950         (vminvq_s16): Likewise.
4951         (vminvq_s32): Likewise.
4952         (vminvq_u8): Likewise.
4953         (vminvq_u16): Likewise.
4954         (vminvq_u32): Likewise.
4955         (vminnmv_f32): Likewise.
4956         (vminnmvq_f32): Likewise.
4957         (vminnmvq_f64): Likewise.
4959 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4961         * config/aarch64/aarch64-simd.md
4962         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4963         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4964         (*aarch64_mul3_elt<mode>): Likewise.
4965         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4966         (*aarch64_mul3_elt_to_64v2df): Likewise.
4967         (*aarch64_mla_elt<mode>): Likewise.
4968         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4969         (*aarch64_mls_elt<mode>): Likewise.
4970         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4971         (*aarch64_fma4_elt<mode>): Likewise.
4972         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4973         (*aarch64_fma4_elt_to_64v2df): Likewise.
4974         (*aarch64_fnma4_elt<mode>): Likewise.
4975         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4976         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4977         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4978         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4979         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4980         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4981         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4982         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4983         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4985 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4987         * config/aarch64/aarch64-simd.md
4988         (aarch64_be_checked_get_lane<mode>): New define_expand.
4989         * config/aarch64/aarch64-simd-builtins.def
4990         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4991         New builtin definition.
4992         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4993         Use new safe be builtin.
4995 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4997         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4998         New define_insn.
4999         (aarch64_be_st1<mode>): Likewise.
5000         (aarch_ld1<VALL:mode>): Define_expand modified.
5001         (aarch_st1<VALL:mode>): Likewise.
5002         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
5003         (UNSPEC_ST1): Likewise.
5005 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
5007         * config/microblaze/microblaze.md: Add trap insn and attribute
5009 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
5011         PR preprocessor/58580
5012         * input.h (location_get_source_line): Take an additional line_size
5013         parameter.
5014         (void diagnostics_file_cache_fini): Declare new function.
5015         * input.c (struct fcache): New type.
5016         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
5017         New static constants.
5018         (diagnostic_file_cache_init, total_lines_num)
5019         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
5020         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
5021         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
5022         (get_next_line, read_next_line, goto_next_line, read_line_num):
5023         New static function definitions.
5024         (diagnostic_file_cache_fini): New function.
5025         (location_get_source_line): Take an additional output line_len
5026         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
5027         read_line_num.
5028         * diagnostic.c (diagnostic_finish): Call
5029         diagnostic_file_cache_fini.
5030         (adjust_line): Take an additional input parameter for the length
5031         of the line, rather than calculating it with strlen.
5032         (diagnostic_show_locus): Adjust the use of
5033         location_get_source_line and adjust_line with respect to their new
5034         signature.  While displaying a line now, do not stop at the first
5035         null byte.  Rather, display the zero byte as a space and keep
5036         going until we reach the size of the line.
5037         * Makefile.in: Add vec.o to OBJS-libcommon
5039 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
5040             Ilya Tocar     <ilya.tocar@intel.com>
5042         * config/i386/avx512fintrin.h (_mm512_kmov): New.
5043         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
5044         (__builtin_ia32_kmov16): Ditto.
5045         * config/i386/i386.md (UNSPEC_KMOV): New.
5046         (kmovw): Ditto.
5048 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
5050         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
5051         (_mm512_storeu_si512): Ditto.
5053 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
5055         PR target/52125
5056         * rtl.h (get_referenced_operands): Declare.
5057         * recog.c (get_referenced_operands): New function.
5058         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
5059         operands have been referenced when recording LO_SUM references.
5061 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
5063         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
5065 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
5067         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
5068         Enable for generic and recent AMD targets.
5070 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
5072         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
5073         ARG_SIZE note when adjustment was eliminated.
5075 2014-01-22  Jeff Law  <law@redhat.com>
5077         PR tree-optimization/59597
5078         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
5079         in file.  Accept new argument REGISTERING and use it to modify
5080         dump output appropriately.
5081         (register_jump_thread): Corresponding changes.
5082         (mark_threaded_blocks): Reinstate code to cancel unprofitable
5083         thread paths involving joiner blocks.  Add code to dump cancelled
5084         jump threading paths.
5086 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
5088         PR rtl-optimization/59477
5089         * lra-constraints.c (inherit_in_ebb): Process call for living hard
5090         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
5092 2014-01-22  Tom Tromey  <tromey@redhat.com>
5094         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
5095         PARAMS.
5096         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
5098 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
5100         PR rtl-optimization/59896
5101         * lra-constraints.c (process_alt_operands): Check unused note for
5102         matched operands of insn with no output reloads.
5104 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
5106         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
5107         (mips_move_from_gpr_cost): Likewise.
5109 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
5111         PR rtl-optimization/59858
5112         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
5113         ira_class_hard_regs_num.
5114         (process_alt_operands): Increase reject for dying matched operand.
5116 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
5118         PR target/59003
5119         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
5120         smaller than size, perform several stores or loads and stores
5121         at dst + count - size to store or copy all of size bytes, rather
5122         than just last modesize bytes.
5124 2014-01-20  DJ Delorie  <dj@redhat.com>
5126         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
5127         that CLOBBERs are REGs before propogating their values.
5129 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
5131         PR middle-end/59789
5132         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
5133         (cgraph_inline_failed_type): New function.
5134         * cgraph.h (DEFCIFCODE): Add type.
5135         (cgraph_inline_failed_type_t): New enum.
5136         (cgraph_inline_failed_type): New prototype.
5137         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
5138         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
5139         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
5140         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
5141         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
5142         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
5143         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
5144         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
5145         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
5146         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
5147         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
5148         OPTIMIZATION_MISMATCH.
5149         * tree-inline.c (expand_call_inline): Emit errors during
5150         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
5152 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
5154         PR target/59685
5155         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
5156         mode attribute in insn output.
5158 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
5160         * output.h (output_constant): Delete.
5161         * varasm.c (output_constant): Make private.
5163 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
5165         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
5167 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5169         PR middle-end/59860
5170         * tree.h (fold_builtin_strcat): New prototype.
5171         * builtins.c (fold_builtin_strcat): No longer static.  Add len
5172         argument, if non-NULL, don't call c_strlen.  Optimize
5173         directly into __builtin_memcpy instead of __builtin_strcpy.
5174         (fold_builtin_2): Adjust fold_builtin_strcat caller.
5175         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
5177 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
5179         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5180         for SImode_address_operand operands, having only a REG argument.
5182 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5184         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
5185         loader name using mbig-endian.
5186         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
5188 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5190         * doc/invoke.texi (-march): Clarify documentation for AArch64.
5191         (-mtune): Likewise.
5192         (-mcpu): Likewise.
5194 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
5196         * config/aarch64/aarch64-protos.h
5197         (aarch64_cannot_change_mode_class_ptr): Declare.
5198         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
5199         aarch64_cannot_change_mode_class_ptr): New.
5200         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
5201         backend hook aarch64_cannot_change_mode_class.
5203 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5205         * common/config/aarch64/aarch64-common.c
5206         (aarch64_handle_option): Don't handle any option order logic here.
5207         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
5208         selected_cpu, warn on architecture version mismatch.
5209         (aarch64_override_options): Fix parsing order for option strings.
5211 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5212             Iain Sandoe  <iain@codesourcery.com>
5214         PR bootstrap/59496
5215         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
5216         warning.  Amend comment to reflect current functionality.
5218 2014-01-20  Richard Biener  <rguenther@suse.de>
5220         PR middle-end/59860
5221         * builtins.c (fold_builtin_strcat): Remove case better handled
5222         by tree-ssa-strlen.c.
5224 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
5226         * config/aarch64/aarch64.opt
5227         (mcpu, march, mtune): Make case-insensitive.
5229 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5231         PR target/59880
5232         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5233         if operands[1] is a REG or ZERO_EXTEND of a REG.
5235 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
5237         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
5239 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
5241         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
5242         long non-pic millicode calls.
5244 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5246         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
5248 2014-01-19  Kito Cheng  <kito@0xlab.org>
5250         * builtins.c (expand_movstr): Check movstr expand done or fail.
5252 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5253             H.J. Lu  <hongjiu.lu@intel.com>
5255         PR target/59379
5256         * config/i386/i386.md (*lea<mode>): Zero-extend return register
5257         to DImode for zero-extended addresses.
5259 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
5261         PR rtl-optimization/57763
5262         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
5263         on the new indirect jump_insn and increment LABEL_NUSES (label).
5265 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
5267         PR bootstrap/59580
5268         PR bootstrap/59583
5269         * config.gcc (x86_archs): New variable.
5270         (x86_64_archs): Likewise.
5271         (x86_cpus): Likewise.
5272         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
5273         --with-arch/--with-cpu= options.
5274         Support --with-arch=/--with-cpu={nehalem,westmere,
5275         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
5277 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5279         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
5280         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
5282 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5284         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
5286 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
5288         PR target/58944
5289         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
5290         clear cpp_get_options (parse_in)->warn_unused_macros for
5291         ix86_target_macros_internal with cpp_define.
5293 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
5295         * jump.c (delete_related_insns): Keep (use (insn))s.
5296         * reorg.c (redundant_insn): Check for barriers too.
5298 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5300         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
5302 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
5304         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
5305         call to $$dyncall when TARGET_LONG_CALLS is true.
5307 2014-01-17  Jeff Law  <law@redhat.com>
5309         * ree.c (combine_set_extension): Temporarily disable test for
5310         changing number of hard registers.
5312 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5314         PR middle-end/58125
5315         * ipa-inline-analysis.c (inline_free_summary):
5316         Do not free summary of aliases.
5318 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5320         PR middle-end/59706
5321         * gimplify.c (gimplify_expr): Use create_tmp_var
5322         instead of create_tmp_var_raw.  If cond doesn't have
5323         integral type, don't add the IFN_ANNOTATE builtin at all.
5325 2014-01-17  Martin Jambor  <mjambor@suse.cz>
5327         PR ipa/59736
5328         * ipa-cp.c (prev_edge_clone): New variable.
5329         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
5330         Also resize prev_edge_clone vector.
5331         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
5332         (ipcp_edge_removal_hook): New function.
5333         (ipcp_driver): Register ipcp_edge_removal_hook.
5335 2014-01-17  Andrew Pinski  <apinski@cavium.com>
5336             Steve Ellcey  <sellcey@mips.com>
5338         PR target/59462
5339         * config/mips/mips.c (mips_print_operand): Check operand mode instead
5340         of operator mode.
5342 2014-01-17  Jeff Law  <law@redhat.com>
5344         PR middle-end/57904
5345         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
5346         so that pass_ccp runs first.
5348 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5350         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5351         (ix86_adjust_cost): Use !TARGET_XXX.
5352         (do_reorder_for_imul): Likewise.
5353         (swap_top_of_ready_list): Likewise.
5354         (ix86_sched_reorder): Likewise.
5356 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5358         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5359         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5360         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5361         (intel_memset): New.  Duplicate slm_memset.
5362         (intel_cost): New.  Duplicate slm_cost.
5363         (m_INTEL): New macro.
5364         (processor_target_table): Add "intel".
5365         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5366         with PROCESSOR_INTEL for "intel".
5367         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5368         PROCESSOR_SILVERMONT.
5369         (ix86_issue_rate): Likewise.
5370         (ix86_adjust_cost): Likewise.
5371         (ia32_multipass_dfa_lookahead): Likewise.
5372         (swap_top_of_ready_list): Likewise.
5373         (ix86_sched_reorder): Likewise.
5374         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5375         instead of TARGET_OPT_AGU.
5376         * config/i386/i386.h (TARGET_INTEL): New.
5377         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5378         (processor_type): Add PROCESSOR_INTEL.
5379         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5380         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5382 2014-01-17  Marek Polacek  <polacek@redhat.com>
5384         PR c/58346
5385         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5386         size is zero.
5388 2014-01-17  Richard Biener  <rguenther@suse.de>
5390         PR tree-optimization/46590
5391         * opts.c (default_options_table): Add entries for
5392         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5393         all enabled at -O1 but not for -Og.
5394         * common.opt (fbranch-count-reg): Remove Init(1).
5395         (fmove-loop-invariants): Likewise.
5396         (ftree-pta): Likewise.
5398 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5400         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5401         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5402         decls to at least the GCC 4.8 used alignments.
5404         PR fortran/59440
5405         * tree-nested.c (convert_nonlocal_reference_stmt,
5406         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5407         of GIMPLE_BIND stmts, adjust associated decls.
5409 2014-01-17  Richard Biener  <rguenther@suse.de>
5411         PR tree-optimization/46590
5412         * vec.h (vec<>::bseach): New member function implementing
5413         binary search according to C89 bsearch.
5414         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5415         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5416         bitmap pointer again.  Make accesses_in_loop a flat array.
5417         (mem_ref_obstack): New global.
5418         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5419         (mark_ref_stored): Likewise.
5420         (ref_indep_loop_p_2): Likewise.
5421         (set_ref_stored_in_loop): New helper function.
5422         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5423         (memref_free): Adjust.
5424         (record_mem_ref_loc): Simplify.
5425         (gather_mem_refs_stmt): Adjust.
5426         (sort_locs_in_loop_postorder_cmp): New function.
5427         (analyze_memory_references): Sort accesses_in_loop after
5428         loop postorder number.
5429         (find_ref_loc_in_loop_cmp): New function.
5430         (for_all_locs_in_loop): Find relevant cluster of locs in
5431         accesses_in_loop and iterate without recursion.
5432         (execute_sm): Avoid uninit warning.
5433         (struct ref_always_accessed): Simplify.
5434         (ref_always_accessed::operator ()): Likewise.
5435         (ref_always_accessed_p): Likewise.
5436         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5437         loop postorder numbers here.
5438         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5439         numbers.
5441 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5443         PR c++/57945
5444         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5445         on decls for which assemble_alias has been called.
5447 2014-01-17  Nick Clifton  <nickc@redhat.com>
5449         * config/msp430/msp430.opt: (mcpu): New option.
5450         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5451         (msp430_option_override): Parse target_cpu.  If the MCU name
5452         matches a generic string, clear target_mcu.
5453         (msp430_attr): Allow numeric interrupt values up to 63.
5454         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5455         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5456         option.
5457         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5458         Add mcpu matches.
5459         * config/msp430/msp430.md (popm): Use %J rather than %I.
5460         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5461         (addhi_cy_i): Use immediate_operand for operand 2.
5462         * doc/invoke.texi: Document -mcpu option.
5464 2014-01-17  Richard Biener  <rguenther@suse.de>
5466         PR rtl-optimization/38518
5467         * df.h (df_analyze_loop): Declare.
5468         * df-core.c: Include cfgloop.h.
5469         (df_analyze_1): Split out main part of df_analyze.
5470         (df_analyze): Adjust.
5471         (loop_inverted_post_order_compute): New function.
5472         (loop_post_order_compute): Likewise.
5473         (df_analyze_loop): New function avoiding whole-function
5474         postorder computes.
5475         * loop-invariant.c (find_defs): Use df_analyze_loop.
5476         (find_invariants): Adjust.
5477         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5479 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5481         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5482         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5484 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5486         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5487         traversal when removing references.
5489 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5491         PR ipa/59775
5492         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5494 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5496         PR middle-end/56791
5497         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5498         pushing a reload for an autoinc when we had previously reloaded an
5499         inner part of the address.
5501 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5503         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5504         field.
5505         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5506         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5507         when not giving up or versioning for alias only because of
5508         loop->safelen.
5509         (vect_analyze_data_ref_dependences): Set to true.
5510         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5511         is a GIMPLE_PHI.
5512         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5513         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5514         to the condition.
5516         PR middle-end/58344
5517         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5519         PR target/59839
5520         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5521         operand 0 predicate for gathers, use a new pseudo as subtarget.
5523 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5525         PR middle-end/59609
5526         * lra-constraints.c (process_alt_operands): Add printing debug info.
5527         Check absence of input/output reloads for matched operands too.
5529 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5531         PR rtl-optimization/59835
5532         * ira.c (ira_init_register_move_cost): Increase cost for
5533         impossible modes.
5535 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5537         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5539 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5541         PR target/59780
5542         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5543         non-register objects.  Use gen_(high/low)part more consistently.
5544         Fix assertions.
5546 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5548         PR target/59844
5549         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5550         endian support, remove tests for WORDS_BIG_ENDIAN.
5551         (p8_mfvsrd_3_<mode>): Likewise.
5552         (reload_gpr_from_vsx<mode>): Likewise.
5553         (reload_gpr_from_vsxsf): Likewise.
5554         (p8_mfvsrd_4_disf): Likewise.
5556 2014-01-16  Richard Biener  <rguenther@suse.de>
5558         PR rtl-optimization/46590
5559         * lcm.c (compute_antinout_edge): Use postorder iteration.
5560         (compute_laterin): Use inverted postorder iteration.
5562 2014-01-16  Nick Clifton  <nickc@redhat.com>
5564         PR middle-end/28865
5565         * varasm.c (output_constant): Return the number of bytes actually
5566         emitted.
5567         (output_constructor_array_range): Update the field size with the
5568         number of bytes emitted by output_constant.
5569         (output_constructor_regular_field): Likewise.  Also do not
5570         complain if the total number of bytes emitted is now greater
5571         than the expected fieldpos.
5572         * output.h (output_constant): Update prototype and descriptive comment.
5574 2014-01-16  Marek Polacek  <polacek@redhat.com>
5576         PR middle-end/59827
5577         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5578         it is error_mark_node.
5580 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5582         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5583         VALID_AVX256_REG_OR_OI_MODE.
5585 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5587         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5588         current procedure should be profiled.
5590 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5592         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5593         of moving from/to the STACK_REG register class.
5595 2014-01-15  Richard Henderson  <rth@redhat.com>
5597         PR debug/54694
5598         * reginfo.c (global_regs_decl): Globalize.
5599         * rtl.h (global_regs_decl): Declare.
5600         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5601         reserved via global_regs.
5603 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5605         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5607 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5609         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5610         and vmulosh rather than call gen_vec_widen_smult_*.
5611         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5612         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5613         (vec_widen_smult_even_v16qi): Likewise.
5614         (vec_widen_umult_even_v8hi): Likewise.
5615         (vec_widen_smult_even_v8hi): Likewise.
5616         (vec_widen_umult_odd_v16qi): Likewise.
5617         (vec_widen_smult_odd_v16qi): Likewise.
5618         (vec_widen_umult_odd_v8hi): Likewise.
5619         (vec_widen_smult_odd_v8hi): Likewise.
5620         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5621         vmuloub rather than call gen_vec_widen_umult_*.
5622         (vec_widen_umult_lo_v16qi): Likewise.
5623         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5624         vmulosb rather than call gen_vec_widen_smult_*.
5625         (vec_widen_smult_lo_v16qi): Likewise.
5626         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5627         rather than call gen_vec_widen_umult_*.
5628         (vec_widen_umult_lo_v8hi): Likewise.
5629         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5630         rather than call gen_vec_widen_smult_*.
5631         (vec_widen_smult_lo_v8hi): Likewise.
5633 2014-01-15  Jeff Law  <law@redhat.com>
5635         PR tree-optimization/59747
5636         * ree.c (find_and_remove_re): Properly handle case where a second
5637         eliminated extension requires widening a copy created for elimination
5638         of a prior extension.
5639         (combine_set_extension): Ensure that the number of hard regs needed
5640         for a destination register does not change when we widen it.
5642 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5644         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5645         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5646         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5647         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5648         (avr-*-rtems*): Likewise.
5649         (bfin*-rtems*): Likewise.
5650         (moxie-*-rtems*): Likewise.
5651         (h8300-*-rtems*): Likewise.
5652         (i[34567]86-*-rtems*): Likewise.
5653         (lm32-*-rtems*): Likewise.
5654         (m32r-*-rtems*): Likewise.
5655         (m68k-*-rtems*): Likewise.
5656         (microblaze*-*-rtems*): Likewise.
5657         (mips*-*-rtems*): Likewise.
5658         (powerpc-*-rtems*): Likewise.
5659         (sh-*-rtems*): Likewise.
5660         (sparc-*-rtems*): Likewise.
5661         (sparc64-*-rtems*): Likewise.
5662         (v850-*-rtems*): Likewise.
5663         (m32c-*-rtems*): Likewise.
5665 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5667         PR rtl-optimization/59511
5668         * ira.c (ira_init_register_move_cost): Use memory costs for some
5669         cases of register move cost calculations.
5670         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5671         instead of BB frequency.
5672         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5673         * lra-assigns.c (find_hard_regno_for): Ditto.
5675 2014-01-15  Richard Biener  <rguenther@suse.de>
5677         PR tree-optimization/59822
5678         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5679         (vectorizable_load): Use it to hoist defs of uses of invariant
5680         loads out of the loop.
5682 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5683             Kugan Vivekanandarajah  <kuganv@linaro.org>
5685         PR target/59695
5686         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5687         truncation.
5689 2014-01-15  Richard Biener  <rguenther@suse.de>
5691         PR rtl-optimization/59802
5692         * lcm.c (compute_available): Use inverted postorder to seed
5693         the initial worklist.
5695 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5697         PR target/59803
5698         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5699         ADDR_REGS for invalid symrefs in non-PIC code.
5701 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5703         PR other/58712
5704         * builtins.c (determine_block_size): Initialize *probable_max_size
5705         even if len_rtx is CONST_INT.
5707 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5709         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5710         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5711         (cortexa53_tunings): Likewise.
5712         (aarch64_sched_issue_rate): New function.
5713         (TARGET_SCHED_ISSUE_RATE): Define.
5715 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5717         * ira-costs.c (find_costs_and_classes): Add missed
5718         ira_init_register_move_cost_if_necessary.
5720 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5722         PR target/59787
5723         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5725 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5727         PR target/59794
5728         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5729         to indicate if type is used for function return value.  Warn ABI
5730         change if the vector mode isn't available for function return value.
5731         (ix86_function_arg_advance): Pass false to type_natural_mode.
5732         (ix86_function_arg): Likewise.
5733         (ix86_gimplify_va_arg): Likewise.
5734         (function_arg_32): Don't warn ABI change.
5735         (ix86_function_value): Pass true to type_natural_mode.
5736         (ix86_return_in_memory): Likewise.
5737         (ix86_struct_value_rtx): Removed.
5738         (TARGET_STRUCT_VALUE_RTX): Likewise.
5740 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5742         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5743         converting a conditional jump into a conditional return.
5745 2014-01-14  Richard Biener  <rguenther@suse.de>
5747         PR tree-optimization/58921
5748         PR tree-optimization/59006
5749         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5750         hoisting invariant stmts.
5751         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5752         invariant loads on the preheader edge if possible.
5754 2014-01-14  Joey Ye  <joey.ye@arm.com>
5756         * doc/plugin.texi (Building GCC plugins): Update to C++.
5758 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5760         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5761         (_mm_rcp28_round_ss): Ditto.
5762         (_mm_rsqrt28_round_sd): Ditto.
5763         (_mm_rsqrt28_round_ss): Ditto.
5764         (_mm_rcp28_sd): Ditto.
5765         (_mm_rcp28_ss): Ditto.
5766         (_mm_rsqrt28_sd): Ditto.
5767         (_mm_rsqrt28_ss): Ditto.
5768         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5769         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5770         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5771         (IX86_BUILTIN_RCP28SD): Ditto.
5772         (IX86_BUILTIN_RCP28SS): Ditto.
5773         (IX86_BUILTIN_RSQRT28SD): Ditto.
5774         (IX86_BUILTIN_RSQRT28SS): Ditto.
5775         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5776         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5777         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5778         (ix86_expand_special_args_builtin): Expand new FTYPE.
5779         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5780         (srcp14<mode>): Make insn unary.
5781         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5782         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5783         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5784         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5785         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5786         Fix rounding: make it SAE only.
5787         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5788         Ditto.
5789         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5790         Ditto.
5791         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5792         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5793         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5794         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5795         (round_saeonly_mask_scalar_operand4): Ditto.
5796         (round_saeonly_mask_scalar_op3): Ditto.
5797         (round_saeonly_mask_scalar_op4): Ditto.
5799 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5801         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5802         Implement -maltivec=be for vec_insert and vec_extract.
5804 2014-01-10  DJ Delorie  <dj@redhat.com>
5806         * config/msp430/msp430.md (call_internal): Don't allow memory
5807         references with SP as the base register.
5808         (call_value_internal): Likewise.
5809         * config/msp430/constraints.md (Yc): New.  For memory references
5810         that don't use SP as a base register.
5812         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5813         "an integer without a # prefix"
5814         * config/msp430/msp430.md (epilogue_helper): Use it.
5816 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5818         PR target/59617
5819         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5820         AVX512F gather builtins.
5821         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5822         on gather decls with INTEGER_TYPE masktype.
5823         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5824         directly into the builtin rather than hoisting it before loop.
5826         PR tree-optimization/59387
5827         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5828         (scev_const_prop): If folded_casts and type has undefined overflow,
5829         use force_gimple_operand instead of force_gimple_operand_gsi and
5830         for each added stmt if it is assign with
5831         arith_code_with_undefined_signed_overflow, call
5832         rewrite_to_defined_overflow.
5833         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5834         gimple-fold.h instead.
5835         (arith_code_with_undefined_signed_overflow,
5836         rewrite_to_defined_overflow): Moved to ...
5837         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5838         rewrite_to_defined_overflow): ... here.  No longer static.
5839         Include gimplify-me.h.
5840         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5841         rewrite_to_defined_overflow): New prototypes.
5843 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5845         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5847 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5849         * builtins.c (get_object_alignment_2): Minor tweak.
5850         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5852 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5854         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5855         optimized non constant lengths.
5857 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5859         PR libgomp/59194
5860         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5861         load as __atomic_load_N if possible.
5863 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5865         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5866         target parameter.
5867         (rs6000_expand_builtin): Adjust call.
5869 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5871         PR target/58115
5872         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5873         * config/rs6000/rs6000.c: Include target-globals.h.
5874         (rs6000_set_current_function): Instead of doing target_reinit
5875         unconditionally, use save_target_globals_default_opts and
5876         restore_target_globals.
5878         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5879         FPSCR.
5880         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5881         (rs6000_expand_builtin): Handle mffs and mtfsf.
5882         (rs6000_init_builtins): Define mffs and mtfsf.
5883         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5884         (rs6000_mffs): New pattern.
5885         (rs6000_mtfsf): New pattern.
5887 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5889         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5890         Start narrowing with START.  Apply candidate-use pair
5891         and check overall cost in narrowing.
5892         (iv_ca_prune): Pass new argument.
5894 2014-01-10  Jeff Law  <law@redhat.com>
5896         PR middle-end/59743
5897         * ree.c (combine_reaching_defs): Ensure the defining statement
5898         occurs before the extension when optimizing extensions with
5899         different source and destination hard registers.
5901 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5903         PR ipa/58585
5904         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5905         vtables into the type inheritance graph.
5907 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5909         PR rtl-optimization/59754
5910         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5911         modes in the REGNO != REGNO case.
5913 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5915         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5917 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5919         PR tree-optimization/59745
5920         * tree-predcom.c (tree_predictive_commoning_loop): Call
5921         free_affine_expand_cache if giving up because components is NULL.
5923         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5924         GC in payload of target_globals struct instead of allocating them on
5925         the heap and the larger structs separately using GC.
5926         * target-globals.h (struct target_globals): Make regs, hard_regs,
5927         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5928         of GTY((skip)) and change type to void *.
5929         (reset_target_globals): Cast loads from those fields to corresponding
5930         types.
5932 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5934         PR plugins/59335
5935         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5936         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5937         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5939 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5941         PR target/59744
5942         * aarch64-modes.def (CC_Zmode): New flags mode.
5943         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5944         represents an equality.
5945         (aarch64_get_condition_code): Handle CC_Zmode.
5946         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5948 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5950         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5951         extraction in good case.
5953 2014-01-10  Richard Biener  <rguenther@suse.de>
5955         PR tree-optimization/59374
5956         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5957         checking after SLP discovery.  Mark stmts not participating
5958         in any SLP instance properly.
5960 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5962         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5963         when handling a SET rtx.
5965 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5967         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5968         (cortex-a57): Likewise.
5969         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5971 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5973         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5974         non-iwmmxt builtins.
5976 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5978         PR ipa/58252
5979         PR ipa/59226
5980         * ipa-devirt.c record_target_from_binfo): Take as argument
5981         stack of binfos and lookup matching one for virtual inheritance.
5982         (possible_polymorphic_call_targets_1): Update.
5984 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5986         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5987         kernel strings for Loongson-2E/2F/3A.
5989 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5991         PR middle-end/59670
5992         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5993         is_gimple_call before calling gimple_call_internal_p.
5995 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5997         * Makefile.in (TREE_FLOW_H): Remove.
5998         (TREE_SSA_H): Add file names from tree-flow.h.
5999         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
6000         * tree.h: Remove tree-flow.h reference.
6001         * hash-table.h: Remove tree-flow.h reference.
6002         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
6003         reference with tree-ssa-loop.h.
6005 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6007         * doc/invoke.texi: Add -maltivec={be,le} options, and document
6008         default element-order behavior for -maltivec.
6009         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
6010         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
6011         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
6012         when targeting big endian, at least for now.
6013         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
6015 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6017         PR middle-end/47735
6018         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
6019         var satisfies use_register_for_decl, just take into account type
6020         alignment, rather than decl alignment.
6022         PR tree-optimization/59622
6023         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
6024         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
6025         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
6026         Don't devirtualize for inplace at all.  For targets.length () == 1,
6027         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
6029 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
6031         * config/i386/i386.md (cpu): Remove the unused btver1.
6033 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
6035         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
6037 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6039         PR target/58115
6040         * tree-core.h (struct target_globals): New forward declaration.
6041         (struct tree_target_option): Add globals field.
6042         * tree.h (TREE_TARGET_GLOBALS): Define.
6043         (prepare_target_option_nodes_for_pch): New prototype.
6044         * target-globals.h (struct target_globals): Define even if
6045         !SWITCHABLE_TARGET.
6046         * tree.c (prepare_target_option_node_for_pch,
6047         prepare_target_option_nodes_for_pch): New functions.
6048         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
6049         * config/i386/i386.c: Include target-globals.h.
6050         (ix86_set_current_function): Instead of doing target_reinit
6051         unconditionally, use save_target_globals_default_opts and
6052         restore_target_globals.
6054 2014-01-09  Richard Biener  <rguenther@suse.de>
6056         PR tree-optimization/59715
6057         * tree-cfg.h (split_critical_edges): Declare.
6058         * tree-cfg.c (split_critical_edges): Export.
6059         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
6061 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
6063         * cfgexpand.c (expand_stack_vars): Optionally disable
6064         asan stack protection.
6065         (expand_used_vars): Likewise.
6066         (partition_stack_vars): Likewise.
6067         * asan.c (asan_emit_stack_protection): Optionally disable
6068         after return stack usage.
6069         (instrument_derefs): Optionally disable memory access instrumentation.
6070         (instrument_builtin_call): Likewise.
6071         (instrument_strlen_call): Likewise.
6072         (asan_protect_global): Optionally disable global variables protection.
6073         * doc/invoke.texi: Added doc for new options.
6074         * params.def: Added new options.
6075         * params.h: Likewise.
6077 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6079         PR rtl-optimization/59724
6080         * ifcvt.c (cond_exec_process_if_block): Don't call
6081         flow_find_head_matching_sequence with 0 longest_match.
6082         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
6083         non-active insns if !stop_after.
6084         (try_head_merge_bb): Revert 2014-01-07 changes.
6086 2014-01-08  Jeff Law  <law@redhat.com>
6088         * ree.c (get_sub_rtx): New function, extracted from...
6089         (merge_def_and_ext): Here.
6090         (combine_reaching_defs): Use get_sub_rtx.
6092 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
6094         * cgraph.h (varpool_variable_node): Do not choke on null node.
6096 2014-01-08  Catherine Moore  <clm@codesourcery.com>
6098         * config/mips/mips.md (simple_return): Attempt to use JRC
6099         for microMIPS.
6100         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
6102 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
6104         PR rtl-optimization/59137
6105         * reorg.c (steal_delay_list_from_target): Call update_block for
6106         elided insns.
6107         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
6109 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6111         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6112         two duplicate entries.
6114 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
6116         Revert:
6117         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
6119         * config/mips/mips.c (mips_truncated_op_cost): New function.
6120         (mips_rtx_costs): Adjust test for BADDU.
6121         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
6123         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
6125         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
6126         (*baddu_si): ...this new pattern.
6128 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
6130         PR ipa/59722
6131         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
6133 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6135         PR middle-end/57748
6136         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
6137         inner_reference_p.
6138         (expand_expr, expand_normal): Adjust.
6139         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
6140         inner_reference_p. Use inner_reference_p to expand inner references.
6141         (store_expr): Adjust.
6142         * cfgexpand.c (expand_call_stmt): Adjust.
6144 2014-01-08  Rong Xu  <xur@google.com>
6146         * gcov-io.c (gcov_var): Move from gcov-io.h.
6147         (gcov_position): Ditto.
6148         (gcov_is_error): Ditto.
6149         (gcov_rewrite): Ditto.
6150         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
6151         only part to libgcc/libgcov.h.
6153 2014-01-08  Marek Polacek  <polacek@redhat.com>
6155         PR middle-end/59669
6156         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
6158 2014-01-08  Marek Polacek  <polacek@redhat.com>
6160         PR sanitizer/59667
6161         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
6163 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
6165         PR rtl-optimization/59649
6166         * stor-layout.c (get_mode_bounds): For BImode return
6167         0 and STORE_FLAG_VALUE.
6169 2014-01-08  Richard Biener  <rguenther@suse.de>
6171         PR middle-end/59630
6172         * gimple.h (is_gimple_builtin_call): Remove.
6173         (gimple_builtin_call_types_compatible_p): New.
6174         (gimple_call_builtin_p): New overload.
6175         * gimple.c (is_gimple_builtin_call): Remove.
6176         (validate_call): Rename to ...
6177         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
6178         check return types.
6179         (validate_type): New static function.
6180         (gimple_call_builtin_p): New overload and adjust.
6181         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
6182         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
6183         (gimple_fold_stmt_to_constant_1): Likewise.
6184         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
6186 2014-01-08  Richard Biener  <rguenther@suse.de>
6188         PR middle-end/59471
6189         * gimplify.c (gimplify_expr): Gimplify register-register type
6190         VIEW_CONVERT_EXPRs to separate stmts.
6192 2014-01-07  Jeff Law  <law@redhat.com>
6194         PR middle-end/53623
6195         * ree.c (combine_set_extension): Handle case where source
6196         and destination registers in an extension insn are different.
6197         (combine_reaching_defs): Allow source and destination registers
6198         in extension to be different under limited circumstances.
6199         (add_removable_extension): Remove restriction that the
6200         source and destination registers in the extension are the same.
6201         (find_and_remove_re): Emit a copy from the extension's
6202         destination to its source after the defining insn if
6203         the source and destination registers are different.
6205         PR middle-end/59285
6206         * ifcvt.c (merge_if_block): If we are merging a block with more than
6207         one successor with a block with no successors, remove any BARRIER
6208         after the second block.
6210 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
6212         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
6214 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
6216         PR target/59652
6217         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
6218         for 14-bit register offsets when INT14_OK_STRICT is false.
6220 2014-01-07  Roland Stigge  <stigge@antcom.de>
6221             Michael Meissner  <meissner@linux.vnet.ibm.com>
6223         PR 57386/target
6224         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
6225         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
6227 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
6229         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
6230         -mcpu.
6232 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
6234         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
6235         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
6236         rtx is const0_rtx or not.
6238 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
6240         PR target/58115
6241         * target-globals.c (save_target_globals): Remove this_fn_optab
6242         handling.
6243         * toplev.c: Include optabs.h.
6244         (target_reinit): Temporarily restore the global options if another
6245         set of options are in force.
6247 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
6249         PR rtl-optimization/58668
6250         * cfgcleanup.c (flow_find_cross_jump): Don't count
6251         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
6252         to determine what is counted.
6253         (flow_find_head_matching_sequence): Use active_insn_p to determine
6254         what is counted.
6255         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
6256         counting change.
6257         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
6258         determine what is counted.
6260         PR tree-optimization/59643
6261         * tree-predcom.c (split_data_refs_to_components): If one dr is
6262         read and one write, determine_offset fails and the write isn't
6263         in the bad component, just put the read into the bad component.
6265 2014-01-07  Mike Stump  <mikestump@comcast.net>
6266             Jakub Jelinek  <jakub@redhat.com>
6268         PR pch/59436
6269         * tree-core.h (struct tree_optimization_option): Change optabs
6270         type from unsigned char * to void *.
6271         * optabs.c (init_tree_optimization_optabs): Adjust
6272         TREE_OPTIMIZATION_OPTABS initialization.
6274 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
6276         PR target/59644
6277         * config/i386/i386.h (struct machine_function): Add
6278         no_drap_save_restore field.
6279         * config/i386/i386.c (ix86_save_reg): Use
6280         !cfun->machine->no_drap_save_restore instead of
6281         crtl->stack_realign_needed.
6282         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
6283         this function clears frame_pointer_needed.  Set
6284         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
6285         and DRAP reg is needed.
6287 2014-01-06  Marek Polacek  <polacek@redhat.com>
6289         PR c/57773
6290         * doc/implement-c.texi: Mention that other integer types are
6291         permitted as bit-field types in strictly conforming mode.
6293 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
6295         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
6296         is newly allocated.
6298 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
6300         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
6302 2014-01-06  Martin Jambor  <mjambor@suse.cz>
6304         PR ipa/59008
6305         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
6306         to int.
6307         * ipa-prop.c (ipa_print_node_params): Fix indentation.
6309 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
6311         PR debug/59350
6312         PR debug/59510
6313         * var-tracking.c (add_stores): Preserve the value of the source even if
6314         we don't record the store.
6316 2014-01-06  Terry Guo  <terry.guo@arm.com>
6318         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
6320 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
6322         PR bootstrap/59541
6323         * config/darwin.c (darwin_function_section): Adjust return values to
6324         correspond to optimisation changes made in r206070.
6326 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
6328         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
6329         from prefetch_block tune setting.
6330         (nocona_cost): Correct size of prefetch block to 64.
6332 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
6334         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
6335         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
6336         used to save the static chain register in the computation of the offset
6337         from which the FP registers need to be restored.
6339 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
6341         PR tree-optimization/59519
6342         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
6343         ICE if get_current_def (current_new_name) is already non-NULL, as long
6344         as it is a phi result of some other phi in *new_exit_bb that has
6345         the same argument.
6347         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
6348         or vmovdqu* for misaligned_operand.
6349         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
6350         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6351         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6352         aligned_mem for AVX512F masked aligned load and store builtins and for
6353         non-temporal moves.
6355 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6357         PR tree-optimization/59651
6358         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6359         Address range for negative step should be added by TYPE_SIZE_UNIT.
6361 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6363         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6364         overlapping registers also for registers other than the stack pointer.
6366 2014-01-03  Marek Polacek  <polacek@redhat.com>
6368         PR other/59661
6369         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6370         __builtin_FILE.
6372 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6374         PR target/59625
6375         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6376         asm goto as jump.
6378         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6379         (push splitter): Use <P:MODE_SIZE> instead of
6380         GET_MODE_SIZE (<P:MODE>mode).
6381         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6382         (mov -1, reg peephole2): Likewise.
6383         * config/i386/sse.md (*mov<mode>_internal,
6384         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6385         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6386         *<code><mode>3, *andnot<mode>3<mask_name>,
6387         <mask_codefor><code><mode>3<mask_name>): Likewise.
6388         * config/i386/subst.md (mask_mode512bit_condition,
6389         sd_mask_mode512bit_condition): Likewise.
6391 2014-01-02  Xinliang David Li  <davidxl@google.com>
6393         PR tree-optimization/59303
6394         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6395         (dump_predicates): Better output format.
6396         (pred_equal_p): New function.
6397         (is_neq_relop_p): Ditto.
6398         (is_neq_zero_form_p): Ditto.
6399         (pred_expr_equal_p): Ditto.
6400         (pred_neg_p): Ditto.
6401         (simplify_pred): Ditto.
6402         (simplify_preds_2): Ditto.
6403         (simplify_preds_3): Ditto.
6404         (simplify_preds_4): Ditto.
6405         (simplify_preds): Ditto.
6406         (push_pred): Ditto.
6407         (push_to_worklist): Ditto.
6408         (get_pred_info_from_cmp): Ditto.
6409         (is_degenerated_phi): Ditto.
6410         (normalize_one_pred_1): Ditto.
6411         (normalize_one_pred): Ditto.
6412         (normalize_one_pred_chain): Ditto.
6413         (normalize_preds): Ditto.
6414         (normalize_cond_1): Remove function.
6415         (normalize_cond): Ditto.
6416         (is_gcond_subset_of): Ditto.
6417         (is_subset_of_any): Ditto.
6418         (is_or_set_subset_of): Ditto.
6419         (is_and_set_subset_of): Ditto.
6420         (is_norm_cond_subset_of): Ditto.
6421         (pred_chain_length_cmp): Ditto.
6422         (convert_control_dep_chain_into_preds): Type change.
6423         (find_predicates): Ditto.
6424         (find_def_preds): Ditto.
6425         (destroy_predicates_vecs): Ditto.
6426         (find_matching_predicates_in_rest_chains): Ditto.
6427         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6428         (is_pred_expr_subset): Ditto.
6429         (is_pred_chain_subset_of): Ditto.
6430         (is_included_in): Ditto.
6431         (is_superset_of): Ditto.
6433 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6435         Update copyright years.
6437 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6439         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6440         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6441         config/arc/arc.md, config/arc/arc.opt,
6442         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6443         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6444         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6445         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6446         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6447         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6448         vtable-verify.c, vtable-verify.h: Use the standard form for the
6449         copyright notice.
6451 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6453         * gcc.c (process_command): Update copyright notice dates.
6454         * gcov-dump.c: Ditto.
6455         * gcov.c: Ditto.
6456         * doc/cpp.texi: Bump @copying's copyright year.
6457         * doc/cppinternals.texi: Ditto.
6458         * doc/gcc.texi: Ditto.
6459         * doc/gccint.texi: Ditto.
6460         * doc/gcov.texi: Ditto.
6461         * doc/install.texi: Ditto.
6462         * doc/invoke.texi: Ditto.
6464 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6466         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6468 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6470         * config/i386/sse.md (*mov<mode>_internal): Guard
6471         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6473         PR rtl-optimization/59647
6474         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6475         new_rtx into UNSIGNED_FLOAT rtxes.
6477 Copyright (C) 2014 Free Software Foundation, Inc.
6479 Copying and distribution of this file, with or without modification,
6480 are permitted in any medium without royalty provided the copyright
6481 notice and this notice are preserved.