2014-04-15 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob1a2f425e02adfe48c4f84fcc30963b0f358c1be4
1 2014-04-15  Richard Biener  <rguenther@suse.de>
3         * alias.c (record_component_aliases): Do not walk BINFOs.
5 2014-04-15  Richard Biener  <rguenther@suse.de>
7         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8         Add struct function argument and adjust.
9         (find_func_aliases_for_call): Likewise.
10         (find_func_aliases): Likewise.
11         (find_func_clobbers): Likewise.
12         (intra_create_variable_infos): Likewise.
13         (compute_points_to_sets): Likewise.
14         (ipa_pta_execute): Adjust.  Do not push/pop cfun.
16 2014-04-15  Richard Biener  <rguenther@suse.de>
18         * tree.c (iterative_hash_expr): Use enum tree_code_class
19         to store TREE_CODE_CLASS.
20         (tree_block): Likewise.
21         (tree_set_block): Likewise.
22         * tree.h (fold_build_pointer_plus_loc): Use
23         convert_to_ptrofftype_loc.
25 2014-04-15  Jakub Jelinek  <jakub@redhat.com>
27         PR plugins/59335
28         * Makefile.in (PLUGIN_HEADERS): Add various headers that have been
29         added in 4.9.
31 2014-04-15  Eric Botcazou  <ebotcazou@adacore.com>
33         * cfgloop.h (struct loop): Move force_vectorize down.
34         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Handle new kinds.
35         (gimplify_expr) <ANNOTATE_EXPR>: Minor tweak.
36         * lto-streamer-in.c (input_cfg): Read dont_vectorize field.
37         * lto-streamer-out.c (output_cfg): Write dont_vectorize field.
38         * tree-cfg.c (replace_loop_annotate): Revamp and handle new kinds.
39         * tree-core.h (enum annot_expr_kind): Add new kind values.
40         * tree-inline.c (copy_loops): Copy dont_vectorize field and reorder.
41         * tree-pretty-print.c (dump_generic_node) <ANNOTATE_EXPR>: Handle new
42         kinds.
43         * tree.def (ANNOTATE_EXPR): Tweak comment.
45 2014-04-14  Jan Hubicka  <hubicka@ucw.cz>
47         * ipa-devirt.c (maybe_record_node): Ignore all non-methods (including
48         cxa_pure_virtual).
50 2014-04-14  Paolo Carlini  <paolo.carlini@oracle.com>
52         * tree.h (TYPE_IDENTIFIER): Declare.
53         * tree.c (subrange_type_for_debug_p): Use it.
54         * godump.c (go_format_type): Likewise.
55         * dwarf2out.c (is_cxx_auto, modified_type_die,
56         gen_type_die_with_usage, gen_type_die_with_usage): Likewise.
57         * dbxout.c (dbxout_type, dbxout_symbol): Likewise.
59 2014-04-14  Jan Hubicka   <hubicka@ucw.cz>
61         PR lto/60820
62         * varpool.c (varpool_remove_node): Do not alter decls when streaming.
64 2014-04-14  Uros Bizjak  <ubizjak@gmail.com>
66         * config/i386/i386.c (examine_argument): Return bool.  Return true if
67         parameter should be passed in memory.
68         <case X86_64_COMPLEX_X87_CLASS>: Adjust.
69         (construct_container): Update calls to examine_argument.
70         (function_arg_advance_64): Ditto.
71         (return_in_memory_32): Merge with ix86_return_in_memory.
72         (return_in_memory_64): Ditto.
73         (return_in_memory_ms_64): Ditto.
75 2014-04-14  Jan Hubicka  <hubicka@ucw.cz>
77         * ipa-utils.c (ipa_merge_profiles): Merge profile_id.
78         * coverage.c (coverage_compute_profile_id): Handle externally visible
79         symbols.
81 2014-04-14  Martin Jambor  <mjambor@suse.cz>
83         * tree-sra.c (ipa_sra_preliminary_function_checks): Skip
84         DECL_DISREGARD_INLINE_LIMITS functions.
86 2014-04-14  H.J. Lu  <hongjiu.lu@intel.com>
88         PR target/60827
89         * config/i386/i386.md (*fixuns_trunc<mode>_1): Revert the last change.
91 2014-04-14  H.J. Lu  <hongjiu.lu@intel.com>
93         PR target/60827
94         * config/i386/i386.md (*fixuns_trunc<mode>_1): Check
95         optimize_insn_for_speed_p instead of
96         optimize_function_for_speed_p.
98 2014-04-14  Yufeng Zhang  <yufeng.zhang@arm.com>
100         * doc/invoke.texi (free): Document AArch64.
102 2014-04-14  Richard Biener  <rguenther@suse.de>
104         PR tree-optimization/60042
105         * tree-ssa-pre.c (inhibit_phi_insertion): Remove.
106         (insert_into_preds_of_block): Do not prevent PHI insertion
107         for REFERENCE exprs here ...
108         (eliminate_dom_walker::before_dom_children): ... but prevent
109         their use here under similar conditions when applied to the
110         IL after PRE optimizations.
112 2014-04-14  Richard Biener  <rguenther@suse.de>
114         * passes.def: Move early points-to after early SRA.
116 2014-04-14  Richard Biener  <rguenther@suse.de>
118         * tree-ssa-forwprop.c (simplify_gimple_switch): Enhance
119         check for which sign-changes we allow when forwarding
120         a converted value into a switch.
122 2014-04-14  Eric Botcazou  <ebotcazou@adacore.com>
124         * stor-layout.c (place_field): Finalize non-constant offset for the
125         field, if any.
127 2014-04-14  Richard Biener  <rguenther@suse.de>
129         * tree-switch-conversion.c (lshift_cheap_p): Get speed_p
130         as argument.
131         (expand_switch_using_bit_tests_p): Likewise.
132         (process_switch): Compute and pass on speed_p based on the
133         switch stmt.
134         * tree-ssa-math-opts.c (gimple_expand_builtin_pow): Use
135         optimize_bb_for_speed_p.
137 2014-04-14  Eric Botcazou  <ebotcazou@adacore.com>
139         * cfgloop.h (struct loop): Rename force_vect into force_vectorize.
140         * function.h (struct function): Rename has_force_vect_loops into
141         has_force_vectorize_loops.
142         * lto-streamer-in.c (input_cfg): Adjust for renaming.
143         (input_struct_function_base): Likewise.
144         * lto-streamer-out.c (output_cfg): Likewise.
145         (output_struct_function_base): Likewise.
146         * omp-low.c (expand_omp_simd): Likewise.
147         * tree-cfg.c (move_sese_region_to_fn): Likewise.
148         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
149         (version_loop_for_if_conversion): Likewise.
150         (tree_if_conversion): Likewise.
151         (main_tree_if_conversion): Likewise.
152         (gate_tree_if_conversion): Likewise.
153         * tree-inline.c (copy_loops): Likewise.
154         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Likewise.
155         * tree-ssa-loop.c (tree_loop_vectorize): Likewise.
156         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Likewise.
157         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Likewise.
158         * tree-vectorizer.c (vectorize_loops): Likewise.
159         * tree-vectorizer.h (unlimited_cost_model): Likewise.
161 2014-04-14  Richard Biener  <rguenther@suse.de>
163         PR lto/60720
164         * lto-streamer-out.c (wrap_refs): New function.
165         (lto_output): Wrap symbol references in global initializes in
166         type-preserving MEM_REFs.
168 2014-04-14  Christian Bruel  <christian.bruel@st.com>
170         * config/sh/sh-mem.cc (sh_expand_strlen): Unroll last word.
172 2014-04-14  Christian Bruel  <christian.bruel@st.com>
174         * config/sh/sh.md (setmemqi): New expand pattern.
175         * config/sh/sh.h (CLEAR_RATIO): Define.
176         * config/sh/sh-mem.cc (sh_expand_setmem): Define.
177         * config/sh/sh-protos.h (sh_expand_setmem): Declare.
179 2014-04-14  Richard Biener  <rguenther@suse.de>
181         PR middle-end/55022
182         * fold-const.c (negate_expr_p): Don't negate directional rounding
183         division.
184         (fold_negate_expr): Likewise.
186 2014-04-14  Richard Biener  <rguenther@suse.de>
188         PR tree-optimization/59817
189         PR tree-optimization/60453
190         * graphite-scop-detection.c (graphite_can_represent_scev): Complete
191         recursion to catch all CHRECs in the scalar evolution and restrict
192         the predicate for the remains appropriately.
194 2014-04-12  Catherine Moore  <clm@codesourcery.com>
196         * config/mips/constraints.md: Add new register constraint "kb".
197         * config/mips/mips.md (*mov<mode>_internal): Use constraint "kb".
198         (*movhi_internal): Likewise.
199         (*movqi_internal): Likewise.
200         * config/mips/mips.h (M16_STORE_REGS): New register class.
201         (REG_CLASS_NAMES): Add M16_STORE_REGS.
202         (REG_CLASS_CONTENTS): Likewise.
203         * config/mips/mips.c (mips_regno_to_class): Add M16_STORE_REGS.
205 2014-04-11  Tobias Burnus  <burnus@net-b.de>
207         PR c/60194
208         * doc/invoke.texi (-Wformat-signedness): Document it.
209         (Wformat=2): Mention that this enables -Wformat-signedness.
211 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
213         * common/config/epiphany/epiphany-common.c
214         (epiphany_option_optimization_table): Enable section anchors by
215         default at -O1 or higher.
216         * config/epiphany/epiphany.c (TARGET_MAX_ANCHOR_OFFSET): Define.
217         (TARGET_MIN_ANCHOR_OFFSET): Likewise.
218         (epiphany_rtx_costs) <SET>: For binary operators, the set as such
219         carries no extra cost.
220         (epiphany_legitimate_address_p): For BLKmode, apply SImode check.
221         * config/epiphany/epiphany.h (ASM_OUTPUT_DEF): Define.
222         * config/epiphany/predicates.md (memclob_operand): New predicate.
223         * config/epiphany/epiphany.md (stack_adjust_add, stack_adjust_str):
224         Use memclob_operand predicate and X constraint for operand 3.
226 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
228         * config/epiphany/epiphany.c (epiphany_rtx_cost): Compare
229         with CC_N_NE / CC_C_LTU / CC_C_GTU carries no extra cost for
230         its operands.
232 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
234         PR rtl-optimization/60651
235         * mode-switching.c (optimize_mode_switching): Make sure to emit
236         sets of a lower numbered entity before sets of a higher numbered
237         entity to a mode of the same or lower priority.
238         When creating a seginfo for a basic block that starts with a code
239         label, move the insertion point past the code label.
240         (new_seginfo): Document and enforce requirement that
241         NOTE_INSN_BASIC_BLOCK only appears for empty blocks.
242         * doc/tm.texi.in: Document ordering constraint for emitted mode sets.
243         * doc/tm.texi: Regenerate.
245 2014-01-11  Joern Rennecke  <joern.rennecke@embecosm.com>
247         PR target/60811
248         * config/arc/arc.c (arc_save_restore): Fix assert typo.
250 2013-04-11  Jakub Jelinek  <jakub@redhat.com>
252         * BASE-VER: Set to 4.10.0.
254 2014-04-11  Tobias Burnus  <burnus@net-b.de>
256         PR other/59055
257         * doc/bugreport.texi (Bugs): Remove nodes pointing to the nirvana.
258         * doc/gcc.texi (Service): Update description in the @menu
259         * doc/invoke.texi (Option Summary): Remove misplaced and
260         duplicated @menu.
262 2014-04-11  Steve Ellcey  <sellcey@mips.com>
263             Jakub Jelinek  <jakub@redhat.com>
265         PR middle-end/60556
266         * expr.c (convert_move): Use emit_store_flag_force instead of
267         emit_store_flag.  Pass lowpart_mode instead of VOIDmode as 5th
268         argument to it.
270 2014-04-11  Richard Biener  <rguenther@suse.de>
272         PR middle-end/60797
273         * varasm.c (assemble_alias): Avoid endless error reporting
274         recursion by setting TREE_ASM_WRITTEN.
276 2014-04-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
278         * config/s390/s390.md: Add a splitter for NOT rtx.
280 2014-04-11  Jakub Jelinek  <jakub@redhat.com>
282         PR rtl-optimization/60663
283         * cse.c (cse_insn): Set src_volatile on ASM_OPERANDS in PARALLEL.
285 2014-04-10  Jan Hubicka  <hubicka@ucw.cz>
286             Jakub Jelinek  <jakub@redhat.com>
288         PR lto/60567
289         * ipa.c (function_and_variable_visibility): Copy forced_by_abi
290         flag from decl_node to node.
292 2014-04-10  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
294         PR debug/60655
295         * config/arm/arm.c (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Define
296         (arm_const_not_ok_for_debug_p): Reject MINUS with SYM_REF's
297         ameliorating the cases where it can be.
299 2014-04-09  David Edelsohn  <dje.gcc@gmail.com>
301         Revert
302         2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
304         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
305         (loadsync_<mode>): Change mode.
306         (load_quadpti, store_quadpti): New.
307         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
308         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
309         * config/rs6000/predicates.md (quad_memory_operand): !TARGET_SYNC_TI.
311 2014-04-09  Cong Hou  <congh@google.com>
313         PR testsuite/60773
314         * doc/sourcebuild.texi (vect_widen_mult_si_to_di_pattern): Add
315         documentation.
317 2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
319         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
320         instead of vnor to exploit possible fusion opportunity in the
321         future.
322         (altivec_expand_vec_perm_const_le): Likewise.
324 2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
326         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
327         (loadsync_<mode>): Change mode.
328         (load_quadpti, store_quadpti): New.
329         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
330         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
332 2014-04-08  Richard Sandiford  <rdsandiford@googlemail.com>
334         PR target/60763
335         * config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
336         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
337         Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.
339 2014-04-08  Richard Biener  <rguenther@suse.de>
341         PR middle-end/60706
342         * tree-pretty-print.c (pp_double_int): For HWI32 hosts with
343         a 64bit widest int print double-int similar to on HWI64 hosts.
345 2014-04-08  Richard Biener  <rguenther@suse.de>
347         PR tree-optimization/60785
348         * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
349         default defs properly.
351 2014-04-08  Nathan Sidwell  <nathan@codesourcery.com>
353         * doc/invoke (Wnon-virtual-dtor): Update to match implementation.
354         (Weffc++): Likewise.
356 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
358         * ipa-devirt.c (maybe_record_node): When node is not recorded,
359         set completep to false rather than true.
361 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
363         PR target/60504
364         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
365         ARM_TARGET2_DWARF_FORMAT.
367 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
369         PR target/60609
370         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
371         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
372         ADDR_DIFF_VEC.
374 2014-04-07  Richard Biener  <rguenther@suse.de>
376         PR tree-optimization/60766
377         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
378         (may_eliminate_iv): Convert cand_value_at result to desired type.
380 2014-04-07  Jason Merrill  <jason@redhat.com>
382         PR c++/60731
383         * common.opt (-fno-gnu-unique): Add.
384         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
386 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
388         * haifa-sched.c: Fix outdated function reference and minor
389         grammar errors in introductory comment.
391 2014-04-07  Richard Biener  <rguenther@suse.de>
393         PR middle-end/60750
394         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
395         for noreturn calls.
396         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
398 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
400         PR debug/55794
401         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
402         size accounting for thunks.
403         (pa_asm_output_mi_thunk): Use final_start_function() and
404         final_end_function() to output function start and end directives.
406 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
408         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have
409         device specific ISA/ feature information. Remove short_sp and
410         errata_skip ds.  Add avr_device_specific_features enum to have device
411         specific info.
412         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
413         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
414         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
415         updated device specific info.
416         * config/avr/avr-mcus.def: Merge device specific details to
417         dev_attribute field.
418         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
419         errata_skip.
420         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
421         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
422         assembler if RMW isa supported by current device.
423         * config/avr/genmultilib.awk: Update as device info structure changed.
424         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
426 2014-04-04  Cong Hou  <congh@google.com>
428         PR tree-optimization/60656
429         * tree-vect-stmts.c (supportable_widening_operation):
430         Fix a bug that elements in a vector with vect_used_by_reduction
431         property are incorrectly reordered when the operation on it is not
432         consistant with the one in reduction operation.
434 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
436         PR rtl-optimization/60155
437         * gcse.c (record_set_data): New function.
438         (single_set_gcse): New function.
439         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
440         (hoist_code): Likewise.
441         (get_pressure_class_and_nregs): Likewise.
443 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
445         * explow.c (probe_stack_range): Emit a final optimization blockage.
447 2014-04-04  Anthony Green  <green@moxielogic.com>
449         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
450         typos.
452 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
454         PR ipa/59626
455         * lto-cgraph.c (input_overwrite_node): Check that partitioning
456         flags are set only during streaming.
457         * ipa.c (process_references, walk_polymorphic_call_targets,
458         symtab_remove_unreachable_nodes): Drop bodies of always inline
459         after early inlining.
460         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
462 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
463         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
465         PR debug/60655
466         * dwarf2out.c (const_ok_for_output_1): Reject expressions
467         containing a NOT.
469 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
471         PR bootstrap/60743
472         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
473         duration.
474         (cortex_a53_fdivd): Likewise.
476 2014-04-04  Martin Jambor  <mjambor@suse.cz>
478         PR ipa/60640
479         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
480         Adjust all callers.
481         * cgraph.c (clone_of_p): Also return true if thunks match.
482         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
483         cgraph_function_or_thunk_node and an obsolete comment.
484         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
485         file.
486         (build_function_decl_skip_args): Likewise.
487         (set_new_clone_decl_and_node_flags): New function.
488         (duplicate_thunk_for_node): Likewise.
489         (redirect_edge_duplicating_thunks): Likewise.
490         (cgraph_clone_node): New parameter args_to_skip, pass it to
491         redirect_edge_duplicating_thunks which is called instead of
492         cgraph_redirect_edge_callee.
493         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
494         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
496 2014-04-04  Jeff Law  <law@redhat.com>
498         PR target/60657
499         * config/arm/predicates.md (const_int_I_operand): New predicate.
500         (const_int_M_operand): Similarly.
501         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
502         const_int_operand.
503         (insv_t2, extv_reg, extzv_t2): Likewise.
504         (load_multiple_with_writeback): Similarly for const_int_I_operand.
505         (pop_multiple_with_writeback_and_return): Likewise.
506         (vfp_pop_multiple_with_writeback): Likewise
508 2014-04-04  Richard Biener  <rguenther@suse.de>
510         PR ipa/60746
511         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
512         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
513         non-GIMPLE_LABELs.
514         * gimplify.h (gimple_add_tmp_var_fn): Declare.
515         * gimplify.c (gimple_add_tmp_var_fn): New function.
516         * gimple-expr.h (create_tmp_reg_fn): Declare.
517         * gimple-expr.c (create_tmp_reg_fn): New function.
518         * gimple-low.c (record_vars_into): Don't change cfun.
519         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
520         code generation without cfun.
522 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
524         PR bootstrap/60719
525         * Makefile.in (install-driver): Fix shell scripting.
527 2014-04-03  Cong Hou  <congh@google.com>
529         PR tree-optimization/60505
530         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
531         threshold of number of iterations below which no vectorization
532         will be done.
533         * tree-vect-loop.c (new_loop_vec_info):
534         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
535         * tree-vect-loop.c (vect_analyze_loop_operations):
536         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
537         * tree-vect-loop.c (vect_transform_loop):
538         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
539         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
540         of iterations of the loop and see if we should build the epilogue.
542 2014-04-03  Richard Biener  <rguenther@suse.de>
544         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx member.
545         (streamer_tree_cache_create): Adjust.
546         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
547         to allow optional nodes array.
548         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
549         (streamer_tree_cache_append): Likewise.
550         (streamer_tree_cache_create): Create nodes array optionally
551         as specified by parameter.
552         * lto-streamer-out.c (create_output_block): Avoid maintaining
553         the node array in the writer cache.
554         (DFS_write_tree): Remove assertion.
555         (produce_asm_for_decls): Free the out decl state hash table early.
556         * lto-streamer-in.c (lto_data_in_create): Adjust for
557         streamer_tree_cache_create prototype change.
559 2014-04-03  Richard Biener  <rguenther@suse.de>
561         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
562         set TREE_CHAIN to NULL_TREE.
564 2014-04-03  Richard Biener  <rguenther@suse.de>
566         PR tree-optimization/60740
567         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
568         over all GIMPLE_COND operands.
570 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
572         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
573         (Weffc++): Remove Scott's numbering, merge lists and reference
574         Wnon-virtual-dtor.
576         c-family/
578         cp/
579         * class.c (accessible_nvdtor_p): New.
580         (check_bases): Don't check base destructor here ...
581         (check_bases_and_members): ... check them here.  Trigger on
582         Wnon-virtual-dtor flag.
583         (finish_struct_1): Use accessible_nvdtor_p.
585         testsuite/
586         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
587         * g++.dg/warn/Wnvdtor-2.C: New.
588         * g++.dg/warn/Wnvdtor-3.C: New.
589         * g++.dg/warn/Wnvdtor-4.C: New.
590         * g++.dg/warn/Weff1.C: Delete.
591         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
592         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
594 2014-04-03  Nick Clifton  <nickc@redhat.com>
596         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
597         properly.
599 2014-04-03  Martin Jambor  <mjambor@suse.cz>
601         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
602         mention gcc_unreachable before failing.
603         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
604         removed symbols.
606 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
608         PR ipa/60659
609         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type
610         inconsistent code and instead mark the context inconsistent.
611         (possible_polymorphic_call_targets): For inconsistent contexts
612         return empty complete list.
614 2014-04-02  Anthony Green  <green@moxielogic.com>
616         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
617         (extendqisi2, extendhisi2): Define.
618         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
619         (WCHAR_TYPE): Change to unsigned int.
621 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
623         PR tree-optimization/60733
624         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
625         insertion point for PHI candidates to be the end of the feeding
626         block for the PHI argument.
628 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
630         PR rtl-optimization/60650
631         * lra-constraints.c (process_alt_operands): Decrease reject for
632         earlyclobber matching.
634 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
636         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
638 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
640         * config/spu/spu.c (pad_bb): Do not crash when the last
641         insn is CODE_FOR_blockage.
643 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
645         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
646         lies outside the target mode.
648 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
650         PR target/60735
651         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
652         software floating point or no floating point registers, do not
653         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
654         in GPRs that occurs after we tested for GPRs that would never be
655         true.
657         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
658         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
659         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
660         specifically allow DDmode, since that does not use the SPE SIMD
661         instructions.
663 2014-04-02  Richard Biener  <rguenther@suse.de>
665         PR middle-end/60729
666         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
667         MODE_INTs.  Properly use negv_optab.
668         (expand_abs): Likewise.
670 2014-04-02  Richard Biener  <rguenther@suse.de>
672         PR bootstrap/60719
673         * Makefile.in (install-driver): Guard extra installs with special
674         names properly.
676 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
678         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
679         Document vec_vgbbd.
681 2014-04-01  Richard Henderson  <rth@redhat.com>
683         PR target/60704
684         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
685         alternative enabled before register allocation.
687 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
689         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
690         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
691         typo.
692         (nios2_large_got_address): Remove unneeded 'sym' parameter.
693         (nios2_got_address): Update nios2_large_got_address call site.
694         (nios2_delegitimize_address): New function.
695         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
696         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
697         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
699 2014-04-01  Martin Husemann  <martin@duskware.de>
701         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
702         for -mabi=32.
704 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
706         PR rtl-optimization/60604
707         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
708         check from register_operand.
709         (register_operand): Redefine in terms of general_operand.
710         (nonmemory_operand): Use register_operand for the non-constant cases.
712 2014-04-01  Richard Biener  <rguenther@suse.de>
714         * gimple.h (struct gimple_statement_base): Align subcode to 16 bits.
716 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
718         * doc/invoke.texi (mapp-regs): Clarify.
720 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
722         * config/i386/avx512fintrin.h (__v32hi): Define type.
723         (__v64qi): Likewise.
724         (_mm512_set1_epi8): Define.
725         (_mm512_set1_epi16): Define.
726         (_mm512_set4_epi32): Define.
727         (_mm512_set4_epi64): Define.
728         (_mm512_set4_pd): Define.
729         (_mm512_set4_ps): Define.
730         (_mm512_setr4_epi64): Define.
731         (_mm512_setr4_epi32): Define.
732         (_mm512_setr4_pd): Define.
733         (_mm512_setr4_ps): Define.
734         (_mm512_setzero_epi32): Define.
736 2014-03-31  Martin Jambor  <mjambor@suse.cz>
738         PR middle-end/60647
739         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
740         callsite_arguments_match_p.  Updated all callers.  Also check types of
741         corresponding formal parameters and actual arguments.
742         (not_all_callers_have_enough_arguments_p) Renamed to
743         some_callers_have_mismatched_arguments_p.
745 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
747         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
749 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
751         PR target/60034
752         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
753         section anchor.
755 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
757         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
758         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
759         Split out
760         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
761         Use FMAMODE_NOVF512 mode iterator.
762         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
763         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
764         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
765         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
766         Split out
767         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
768         Use VF_128_256 mode iterator.
769         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
770         Ditto.
772 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
774         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
775         static chain if needed.
777 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
779         PR target/60697
780         * lra-constraints.c (index_part_to_reg): New.
781         (process_address): Use it.
783 2014-03-27  Jeff Law  <law@redhat.com>
784             Jakub Jelinek  <jakub@redhat.com>
786         PR target/60648
787         * expr.c (do_tablejump): Use simplify_gen_binary rather than
788         gen_rtx_{PLUS,MULT} to build up the address expression.
790         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
791         creating non-canonical RTL.
793 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
795         PR ipa/60243
796         * ipa-inline.c (want_inline_small_function_p): Short circuit large
797         functions; reorganize to make cheap checks first.
798         (inline_small_functions): Do not estimate growth when dumping;
799         it is expensive.
800         * ipa-inline.h (inline_summary): Add min_size.
801         (growth_likely_positive): New function.
802         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
803         (set_cond_stmt_execution_predicate): Cleanup.
804         (estimate_edge_size_and_time): Compute min_size.
805         (estimate_calls_size_and_time): Likewise.
806         (estimate_node_size_and_time): Likewise.
807         (inline_update_overall_summary): Update min_size.
808         (do_estimate_edge_time): Likewise.
809         (do_estimate_edge_size): Update.
810         (do_estimate_edge_hints): Update.
811         (growth_likely_positive): New function.
813 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
815         PR target/60693
816         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
817         also if addr has VOIDmode.
819 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
821         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
822         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
823         Declare extern.
824         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
825         instructions as well as AdvancedSIMD loads.
827 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
829         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
830         Use crypto_aese type.
831         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
832         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
833         crypto_aese, crypto_aesmc.  Move to types.md.
834         * config/arm/types.md (crypto_aes): Split into crypto_aese,
835         crypto_aesmc.
836         * config/arm/iterators.md (crypto_type): Likewise.
838 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
840         * cgraph.c: Include expr.h and tree-dfa.h.
841         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
842         remove LHS.
844 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
846         PR target/60675
847         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
848         regs from checking multi-reg pseudos.
850 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
852         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
854 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
856         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
857         if it would clobber the stack pointer, even temporarily.
859 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
861         * mode-switching.c: Make small adjustments to the top comment.
863 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
865         * config/rs6000/constraints.md (wD constraint): New constraint to
866         match the constant integer to get the top DImode/DFmode out of a
867         vector in a VSX register.
869         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
870         match the constant integer to get the top DImode/DFmode out of a
871         vector in a VSX register.
873         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
874         for ISA 2.07.
876         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
877         vbpermq builtins.
879         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
880         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
882         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
883         Optimize vec_extract of 64-bit values, where the value being
884         extracted is in the top word, where we can use scalar
885         instructions.  Add direct move and store support.  Combine the big
886         endian/little endian vector select load support into a single insn.
887         (vsx_extract_<mode>_internal1): Likewise.
888         (vsx_extract_<mode>_internal2): Likewise.
889         (vsx_extract_<mode>_load): Likewise.
890         (vsx_extract_<mode>_store): Likewise.
891         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
892         combined into vsx_extract_<mode>_load.
893         (vsx_extract_<mode>_one_le): Likewise.
895         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
896         define the top 64-bit vector element.
898         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
899         constraint.
901         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
902         Document vec_vbpermq builtin.
904         PR target/60672
905         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
906         enable use of xxsldwi and xxpermdi builtin functions.
907         (vec_xxpermdi): Likewise.
909         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
910         Document use of vec_xxsldwi and vec_xxpermdi builtins.
912 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
914         PR rtl-optimization/60650
915         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
916         first_p.  Use it.
917         (find_spills_for): New.
918         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
919         Spill all pseudos on the second iteration.
921 2014-03-27  Marek Polacek  <polacek@redhat.com>
923         PR c/50347
924         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
925         types.
927 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
929         * config/s390/s390.c (s390_can_use_return_insn): Check for
930         call-saved FPRs on 31 bit.
932 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
934         PR middle-end/60682
935         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
936         if they need regimplification, just drop them instead of
937         calling gimple_regimplify_operands on them.
939 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
941         PR target/60580
942         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
943         (aarch64_frame_pointer_required): Adjust logic.
944         (aarch64_can_eliminate): Adjust logic.
945         (aarch64_override_options_after_change): Adjust logic.
947 2014-03-27  Dehao Chen  <dehao@google.com>
949         * ipa-inline.c (early_inliner): Update node's inline info.
951 2014-03-26  Dehao Chen  <dehao@google.com>
953         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
954         compiler inserted conditional jumps for NAN float check.
956 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
958         * ubsan.h (ubsan_create_data): Change second argument's type
959         to const location_t *.
960         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
961         _("<unknown>").
962         (ubsan_create_data): Change second argument to const location_t *PLOC.
963         Create Loc field whenever PLOC is non-NULL.
964         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
965         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
966         callers.
968         PR other/59545
969         * real.c (real_to_integer2): Change type of low to UHWI.
971 2014-03-26  Tobias Burnus  <burnus@net-b.de>
973         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
974         (CILK_SELF_SPECS): New define.
975         (driver_self_specs): Use it.
977 2014-03-26  Richard Biener  <rguenther@suse.de>
979         * tree-pretty-print.c (percent_K_format): Implement special
980         case for LTO and its stripped down BLOCK tree.
982 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
984         PR sanitizer/60636
985         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
987         * tree-vrp.c (simplify_internal_call_using_ranges): If only
988         one range is range_int_cst_p, but not both, at least optimize
989         addition/subtraction of 0 and multiplication by 0 or 1.
990         * gimple-fold.c (gimple_fold_call): Fold
991         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
992         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
993         INTEGER_CSTs, try to fold at least x * 0 and y - y.
995 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
997         PR rtl-optimization/60452
998         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
999         <case REG>: Return 1 for invalid offsets from the frame pointer.
1001 2014-03-26  Marek Polacek  <polacek@redhat.com>
1003         PR c/37428
1004         * doc/extend.texi (C Extensions): Mention variable-length arrays in
1005         a structure/union.
1007 2014-03-26  Marek Polacek  <polacek@redhat.com>
1009         PR c/39525
1010         * doc/extend.texi (Designated Inits): Describe what happens to omitted
1011         field members.
1013 2014-03-26  Marek Polacek  <polacek@redhat.com>
1015         PR other/59545
1016         * ira-color.c (update_conflict_hard_regno_costs): Perform the
1017         multiplication in unsigned type.
1019 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
1021         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
1023 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
1025         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
1027 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
1029         PR ipa/60315
1030         * cif-code.def (UNREACHABLE) New code.
1031         * ipa-inline.c (inline_small_functions): Skip edges to
1032         __builtlin_unreachable.
1033         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
1034         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
1035         predicate to __bulitin_unreachable.
1036         (set_cond_stmt_execution_predicate): Fix issue when
1037         invert_tree_comparison returns ERROR_MARK.
1038         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
1039         propagate to inline clones.
1040         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
1041         to unreachable.
1042         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
1043         * cgraphclones.c (cgraph_clone_node): If call destination is already
1044         ureachable, do not redirect it back.
1045         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
1046         unreachable.
1048 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
1050         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
1051         Do not modify inline clones.
1053 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
1055         * config/i386/i386.md (general_sext_operand): New mode attr.
1056         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
1057         don't generate (sign_extend (const_int)).
1058         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
1059         operands[2].  Use We constraint instead of <i> and
1060         <general_sext_operand> predicate instead of <general_operand>.
1061         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
1062         * config/i386/constraints.md (We): New constraint.
1063         * config/i386/predicates.md (x86_64_sext_operand,
1064         sext_operand): New predicates.
1066 2014-03-25  Martin Jambor  <mjambor@suse.cz>
1068         PR ipa/60600
1069         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
1070         inconsistent devirtualizations to __builtin_unreachable.
1072 2014-03-25  Marek Polacek  <polacek@redhat.com>
1074         PR c/35449
1075         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
1077 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
1079         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
1080         order of elements for big-endian.
1082 2014-03-25  Richard Biener  <rguenther@suse.de>
1084         PR middle-end/60635
1085         * gimplify-me.c (gimple_regimplify_operands): Update the
1086         re-gimplifed stmt.
1088 2014-03-25  Martin Jambor  <mjambor@suse.cz>
1090         PR ipa/59176
1091         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
1092         (lto_output_varpool_node): Likewise.
1093         (input_overwrite_node): Likewise.
1094         (input_varpool_node): Likewise.
1096 2014-03-25  Richard Biener  <rguenther@suse.de>
1098         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
1099         (run_gcc): Likewise.
1101 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
1103         * combine.c (simplify_compare_const): Add MODE argument.
1104         Handle mode_width 0 as very large mode_width.
1105         (try_combine, simplify_comparison): Adjust callers.
1107         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
1108         type to avoid signed integer overflow.
1109         * explow.c (plus_constant): Likewise.
1111 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1113         * doc/generic.texi: Correct typos.
1115 2014-03-24  Tobias Burnus  <burnus@net-b.de>
1117         * doc/invoke.texi (-flto): Expand section about
1118         using static libraries with LTO.
1120 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1122         PR rtl-optimization/60501
1123         * optabs.def (addptr3_optab): New optab.
1124         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
1125         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
1126         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
1128         * lra.c (emit_add3_insn): Use the addptr pattern if available.
1130         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
1132 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
1134         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
1135         _mm512_set1_pd.
1137         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
1138         (_mm256_undefined_ps): Define.
1139         (_mm256_undefined_pd): Define.
1140         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
1141         (_mm_undefined_pd): Define.
1142         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
1143         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
1144         (_mm512_undefined_ps): Define.
1145         (_mm512_undefined_pd): Define.
1146         Use _mm*_undefined_*.
1147         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
1149 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
1151         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
1152         (lshr_simd): DI mode added.
1153         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
1154         (aarch64_ushr_simddi): Likewise.
1155         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
1156         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
1157         (vshrd_n_u64): Likewise.
1159 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1161         * Makefile.in (s-macro_list): Depend on cc1.
1163 2014-03-23  Teresa Johnson  <tejohnson@google.com>
1165         * ipa-utils.c (ipa_print_order): Use specified dump file.
1167 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
1169         PR rtl-optimization/60601
1170         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
1172         * gcc.c (eval_spec_function): Initialize save_growing_value.
1174 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
1176         PR sanitizer/60613
1177         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
1178         code == MINUS_EXPR, never swap op0 with op1.
1180         * toplev.c (init_local_tick): Avoid signed integer multiplication
1181         overflow.
1182         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
1183         shift by first operand's bitsize.
1185 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
1187         PR target/60610
1188         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
1189         redefine to 1 or 0.
1190         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
1191         TARGET_ISA_64BIT_P(x).
1193 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1195         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
1196         pattern for vector nor instead of subtract from splat(-1).
1197         (altivec_expand_vec_perm_const_le): Likewise.
1199 2014-03-21  Richard Henderson  <rth@twiddle.net>
1201         PR target/60598
1202         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
1203         related insns after epilogue_completed.
1205 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1207         PR ipa/59176
1208         * cgraph.h (symtab_node): New flag body_removed.
1209         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
1210         when removing bodies.
1211         * symtab.c (dump_symtab_base): Dump body_removed flag.
1212         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
1213         had their bodies removed.
1215 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1217         PR ipa/60419
1218         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
1219         in the border.
1221 2014-03-21  Richard Biener  <rguenther@suse.de>
1223         PR tree-optimization/60577
1224         * tree-core.h (struct tree_base): Document nothrow_flag use
1225         in VAR_DECL_NONALIASED.
1226         * tree.h (VAR_DECL_NONALIASED): New.
1227         (may_be_aliased): Adjust.
1228         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
1230 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1232         * expr.c (expand_expr_real_1): Remove outdated comment.
1234 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
1236         PR middle-end/60597
1237         * ira.c (adjust_cleared_regs): Call copy_rtx on
1238         *reg_equiv[REGNO (loc)].src_p before passing it to
1239         simplify_replace_fn_rtx.
1241         PR target/60568
1242         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
1243         into CONST, put pic register as first operand of PLUS.  Use
1244         gen_const_mem for both 32-bit and 64-bit PIC got loads.
1246 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1248         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
1250 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1252         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
1253         around for store forwarding issue in the FPU on the UT699.
1254         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
1255         loads and operations if -mfix-ut699 is specified.
1256         (divtf3_hq): Tweak attribute.
1257         (sqrttf2_hq): Likewise.
1259 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1261         * calls.c (store_one_arg): Remove incorrect const qualification on the
1262         type of the temporary.
1263         * cfgexpand.c (expand_return): Likewise.
1264         * expr.c (expand_constructor): Likewise.
1265         (expand_expr_real_1): Likewise.
1267 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1269         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
1270         of parts.
1272 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
1274         PR target/60039
1275         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
1277 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
1279         * config/arm/aarch-common-protos.h
1280         (alu_cost_table): Fix spelling of "extend".
1281         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
1283 2014-03-19  Richard Biener  <rguenther@suse.de>
1285         PR middle-end/60553
1286         * tree-core.h (tree_type_common): Re-order pointer members
1287         to reduce recursion depth during GC walks.
1289 2014-03-19  Marek Polacek  <polacek@redhat.com>
1291         PR sanitizer/60569
1292         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
1293         before accessing it.
1295 2014-03-19  Richard Biener  <rguenther@suse.de>
1297         PR lto/59543
1298         * lto-streamer-in.c (input_function): In WPA stage do not drop
1299         debug stmts.
1301 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
1303         PR tree-optimization/60559
1304         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
1305         with build_zero_cst assignment.
1307 2014-03-18  Kai Tietz  <ktietz@redhat.com>
1309         PR rtl-optimization/56356
1310         * sdbout.c (sdbout_parms): Verify that parms'
1311         incoming argument is valid.
1312         (sdbout_reg_parms): Likewise.
1314 2014-03-18  Richard Henderson  <rth@redhat.com>
1316         PR target/60562
1317         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
1318         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
1319         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
1321 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
1323         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
1324         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
1325         Italicize plugin event names in description.  Explain that
1326         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
1327         Remind that no GCC functions should be called after PLUGIN_FINISH.
1328         Explain what pragmas with expansion are.
1330 2014-03-18  Martin Liska  <mliska@suse.cz>
1332         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
1333         gimple call statement is update.
1334         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
1335         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
1337 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1339         PR sanitizer/60557
1340         * ubsan.c (ubsan_instrument_unreachable): Call
1341         initialize_sanitizer_builtins.
1342         (ubsan_pass): Likewise.
1344         PR sanitizer/60535
1345         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1346         varpool_finalize_decl instead of rest_of_decl_compilation.
1348 2014-03-18  Richard Biener  <rguenther@suse.de>
1350         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1351         by using bitmap_and_compl instead of bitmap_and_compl_into.
1352         (df_rd_transfer_function): Likewise.
1354 2014-03-18  Richard Biener  <rguenther@suse.de>
1356         * doc/lto.texi (fresolution): Fix typo.
1358 2014-03-18  Richard Biener  <rguenther@suse.de>
1360         * doc/invoke.texi (flto): Update for changes in 4.9.
1362 2014-03-18  Richard Biener  <rguenther@suse.de>
1364         * doc/loop.texi: Remove section on the removed lambda framework.
1365         Update loop docs with recent changes in preserving loop structure.
1367 2014-03-18  Richard Biener  <rguenther@suse.de>
1369         * doc/lto.texi (-fresolution): Document.
1371 2014-03-18  Richard Biener  <rguenther@suse.de>
1373         * doc/contrib.texi: Adjust my name.
1375 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1377         PR ipa/58721
1378         * internal-fn.c: Include diagnostic-core.h.
1379         (expand_BUILTIN_EXPECT): New function.
1380         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1381         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1382         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1383         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1384         IFN_BUILTIN_EXPECT.
1385         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1386         Revert 3 argument __builtin_expect code.
1387         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1388         * gimple-fold.c (gimple_fold_call): Likewise.
1389         * tree.h (fold_builtin_expect): New prototype.
1390         * builtins.c (build_builtin_expect_predicate): Add predictor
1391         argument, if non-NULL, create 3 argument __builtin_expect.
1392         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1393         pass it through to build_builtin_expect_predicate.
1394         (fold_builtin_2): Adjust caller.
1395         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1396         * internal-fn.def (BUILTIN_EXPECT): New.
1398 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1400         PR ipa/58721
1401         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1402         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1403         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1405 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1407         PR ipa/58721
1408         * predict.c (combine_predictions_for_bb): Fix up formatting.
1409         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1410         fill what it points to if non-NULL.
1411         (tree_predict_by_opcode): Adjust caller, use the predictor.
1412         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1414 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1416         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1417         proper constant for the store mode.
1419 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1421         * symtab.c (change_decl_assembler_name): Fix transparent alias
1422         chain construction.
1424 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1426         * config/aarch64/aarch64.c: Correct the comments about the
1427         aarch64 stack layout.
1429 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1431         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1432         check for GF_OMP_FOR_KIND_FOR.
1434 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1436         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1437         ymm and zmm register names.
1439 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1441         PR target/60516
1442         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1443         note creation for the 2010-08-31 changes.
1445 2014-03-17  Marek Polacek  <polacek@redhat.com>
1447         PR middle-end/60534
1448         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1449         as -fno-tree-loop-vectorize.
1450         (expand_omp_simd): Likewise.
1452 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1454         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1455         (eligible_for_call_delay): New prototype.
1456         * config/sparc/sparc.c (tls_call_delay): Rename into...
1457         (eligible_for_call_delay): ...this.  Return false if the instruction
1458         cannot be put in the delay slot of a branch.
1459         (eligible_for_restore_insn): Simplify.
1460         (eligible_for_return_delay): Return false if the instruction cannot be
1461         put in the delay slot of a branch and simplify.
1462         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1463         put in the delay slot of a branch.
1464         * config/sparc/sparc.md (fix_ut699): New attribute.
1465         (tls_call_delay): Delete.
1466         (in_call_delay): Reimplement.
1467         (eligible_for_sibcall_delay): Rename into...
1468         (in_sibcall_delay): ...this.
1469         (eligible_for_return_delay): Rename into...
1470         (in_return_delay): ...this.
1471         (in_branch_delay): Reimplement.
1472         (in_uncond_branch_delay): Delete.
1473         (in_annul_branch_delay): Delete.
1475 2014-03-14  Richard Henderson  <rth@redhat.com>
1477         PR target/60525
1478         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1479         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1480         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1481         (floathi splitters): Remove.
1482         (float<SWI48x>xf2): New pattern.
1483         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1484         code that tried to handle DImode for 32-bit, but which was excluded
1485         by the pattern's condition.  Drop allocation of stack temporary.
1486         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1487         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1488         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1489         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1490         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1491         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1492         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1493         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1494         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1495         (*float<SWI48x><X87MODEF>2_i387): Remove.
1496         (all float _with_temp splitters): Remove.
1497         (*float<SWI48x><MODEF>2_i387): New pattern.
1498         (*float<SWI48><MODEF>2_sse): New pattern.
1499         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1500         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1502 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1503             Marek Polacek  <polacek@redhat.com>
1505         PR middle-end/60484
1506         * common.opt (dump_base_name_prefixed): New Variable.
1507         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1508         if x_dump_base_name_prefixed is already set, set it at the end.
1510 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1512         PR rtl-optimization/60508
1513         * lra-constraints.c (get_reload_reg): Add new parameter
1514         in_subreg_p.
1515         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1516         Pass the new parameter values.
1518 2014-03-14  Richard Biener  <rguenther@suse.de>
1520         * common.opt: Revert unintented changes from r205065.
1521         * opts.c: Likewise.
1523 2014-03-14  Richard Biener  <rguenther@suse.de>
1525         PR middle-end/60518
1526         * cfghooks.c (split_block): Properly adjust all loops the
1527         block was a latch of.
1529 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1531         PR lto/60461
1532         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1533         and simplify it.
1535 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1537         PR target/59396
1538         * config/avr/avr.c (avr_set_current_function): Pass function name
1539         through default_strip_name_encoding before sanity checking instead
1540         of skipping the first char of the assembler name.
1542 2014-03-13  Richard Henderson  <rth@redhat.com>
1544         PR debug/60438
1545         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1546         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1547         * config/i386/i386-protos.h: Likewise.
1548         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1549         in the expander instead of a splitter.
1550         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1551         any possibility of requiring a memory.
1552         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1553         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1554         (fp branch splitters): Update for ix86_split_fp_branch.
1555         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1556         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1557         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1558         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1559         (*fop_<MODEF>_3_i387): Likewise.
1560         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1561         (splitters for the fop_* register patterns): Remove.
1562         (fscalexf4_i387): Rename from *fscalexf4_i387.
1563         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1565 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1567         PR tree-optimization/59779
1568         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1569         type for bitsize and maxsize instead of HOST_WIDE_INT.
1571 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1573         PR rtl-optimization/57320
1574         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1575         the CFG after thread_prologue_and_epilogue_insns.
1577 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1579         PR rtl-optimization/57189
1580         * lra-constraints.c (process_alt_operands): Disfavor spilling
1581         vector pseudos.
1583 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1585         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1587 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1589         PR tree-optimization/59025
1590         PR middle-end/60418
1591         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1592         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1594 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1596         PR target/60486
1597         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1598         calls of avr_out_plus_1.
1600 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1602         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1603         BB's single pred and update the father loop's latch info later.
1605 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1607         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1608         (VEC_M): Likewise.
1609         (VEC_N): Likewise.
1610         (VEC_R): Likewise.
1611         (VEC_base): Likewise.
1612         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1613         registers, we need to swap double words in little endian mode.
1615         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1616         to be a container mode for 128-bit integer operations added in ISA
1617         2.07.  Unlike TImode and PTImode, the preferred register set is
1618         the Altivec/VMX registers for the 128-bit operations.
1620         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1621         declarations.
1622         (rs6000_split_128bit_ok_p): Likewise.
1624         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1625         macros for creating ISA 2.07 normal and overloaded builtin
1626         functions with 3 arguments.
1627         (BU_P8V_OVERLOAD_3): Likewise.
1628         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1629         for use as overloaded functions.
1630         (VPERM_1TI_UNS): Likewise.
1631         (VSEL_1TI): Likewise.
1632         (VSEL_1TI_UNS): Likewise.
1633         (ST_INTERNAL_1ti): Likewise.
1634         (LD_INTERNAL_1ti): Likewise.
1635         (XXSEL_1TI): Likewise.
1636         (XXSEL_1TI_UNS): Likewise.
1637         (VPERM_1TI): Likewise.
1638         (VPERM_1TI_UNS): Likewise.
1639         (XXPERMDI_1TI): Likewise.
1640         (SET_1TI): Likewise.
1641         (LXVD2X_V1TI): Likewise.
1642         (STXVD2X_V1TI): Likewise.
1643         (VEC_INIT_V1TI): Likewise.
1644         (VEC_SET_V1TI): Likewise.
1645         (VEC_EXT_V1TI): Likewise.
1646         (EQV_V1TI): Likewise.
1647         (NAND_V1TI): Likewise.
1648         (ORC_V1TI): Likewise.
1649         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1650         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1651         overloaded builtin.
1652         (VADDUQM): Likewise.
1653         (VSUBCUQ): Likewise.
1654         (VADDEUQM): Likewise.
1655         (VADDECUQ): Likewise.
1656         (VSUBEUQM): Likewise.
1657         (VSUBECUQ): Likewise.
1659         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1660         __int128_t and __uint128_t types.
1661         (__uint128_type): Likewise.
1662         (altivec_categorize_keyword): Add support for vector __int128_t,
1663         vector __uint128_t, vector __int128, and vector unsigned __int128
1664         as a container type for TImode operations that need to be done in
1665         VSX/Altivec registers.
1666         (rs6000_macro_to_expand): Likewise.
1667         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1668         to support 128-bit integer instructions vaddcuq, vadduqm,
1669         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1670         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1672         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1673         for V1TImode, and set up preferences to use VSX/Altivec registers.
1674         Setup VSX reload handlers.
1675         (rs6000_debug_reg_global): Likewise.
1676         (rs6000_init_hard_regno_mode_ok): Likewise.
1677         (rs6000_preferred_simd_mode): Likewise.
1678         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1679         (easy_altivec_constant): Likewise.
1680         (output_vec_const_move): Likewise.
1681         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1682         simple move.
1683         (rs6000_expand_vector_extract): Likewise.
1684         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1685         addressing.
1686         (rs6000_const_vec): Add support for V1TImode.
1687         (rs6000_emit_le_vsx_load): Swap double words when loading or
1688         storing TImode/V1TImode.
1689         (rs6000_emit_le_vsx_store): Likewise.
1690         (rs6000_emit_le_vsx_move): Likewise.
1691         (rs6000_emit_move): Add support for V1TImode.
1692         (altivec_expand_ld_builtin): Likewise.
1693         (altivec_expand_st_builtin): Likewise.
1694         (altivec_expand_vec_init_builtin): Likewise.
1695         (altivec_expand_builtin): Likewise.
1696         (rs6000_init_builtins): Add support for V1TImode type.  Add
1697         support for ISA 2.07 128-bit integer builtins.  Define type names
1698         for the VSX/Altivec vector types.
1699         (altivec_init_builtins): Add support for overloaded vector
1700         functions with V1TImode type.
1701         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1702         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1703         external function.
1704         (rs6000_split_128bit_ok_p): Likewise.
1705         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1706         __int128_t and vector __uint128_t.
1708         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1709         and mode attributes.
1710         (VSX_M): Likewise.
1711         (VSX_M2): Likewise.
1712         (VSm): Likewise.
1713         (VSs): Likewise.
1714         (VSr): Likewise.
1715         (VSv): Likewise.
1716         (VS_scalar): Likewise.
1717         (VS_double): Likewise.
1718         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1720         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1721         we support the ISA 2.07 128-bit integer arithmetic instructions.
1722         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1723         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1724         and TImode types for use with the builtin functions.
1725         (V1TI_type_node): Likewise.
1726         (unsigned_V1TI_type_node): Likewise.
1727         (intTI_type_internal_node): Likewise.
1728         (uintTI_type_internal_node): Likewise.
1730         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1731         128-bit builtin functions.
1732         (UNSPEC_VADDEUQM): Likewise.
1733         (UNSPEC_VADDECUQ): Likewise.
1734         (UNSPEC_VSUBCUQ): Likewise.
1735         (UNSPEC_VSUBEUQM): Likewise.
1736         (UNSPEC_VSUBECUQ): Likewise.
1737         (VM): Add V1TImode to vector mode iterators.
1738         (VM2): Likewise.
1739         (VI_unit): Likewise.
1740         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1741         (altivec_vaddcuq): Likewise.
1742         (altivec_vsubuqm): Likewise.
1743         (altivec_vsubcuq): Likewise.
1744         (altivec_vaddeuqm): Likewise.
1745         (altivec_vaddecuq): Likewise.
1746         (altivec_vsubeuqm): Likewise.
1747         (altivec_vsubecuq): Likewise.
1749         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1750         mode iterators.
1751         (BOOL_128): Likewise.
1752         (BOOL_REGS_OUTPUT): Likewise.
1753         (BOOL_REGS_OP1): Likewise.
1754         (BOOL_REGS_OP2): Likewise.
1755         (BOOL_REGS_UNARY): Likewise.
1756         (BOOL_REGS_AND_CR0): Likewise.
1758         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1759         128-bit integer builtin support.
1760         (vec_vadduqm): Likewise.
1761         (vec_vaddecuq): Likewise.
1762         (vec_vaddeuqm): Likewise.
1763         (vec_vsubecuq): Likewise.
1764         (vec_vsubeuqm): Likewise.
1765         (vec_vsubcuq): Likewise.
1766         (vec_vsubuqm): Likewise.
1768         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1769         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1770         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1771         128-bit integer add/subtract to ISA 2.07.
1773 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1775         * config/arc/arc.c (arc_predicate_delay_insns):
1776         Fix third argument passed to conditionalize_nonjump.
1778 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1780         * config/aarch64/aarch64-builtins.c
1781         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1782         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1783         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1784         instead of __builtin_lfloor.
1785         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1787 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1789         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1790         (tree_ssa_ifcombine_bb_1): New function.
1791         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1792         is an empty forwarder block to then_bb or vice versa and then_bb
1793         and else_bb are effectively swapped.
1795 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1797         PR target/60264
1798         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1799         REG_CFA_DEF_CFA note.
1800         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1801         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1803 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1805         PR tree-optimization/60454
1806         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1808 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1810         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1811         Do not define target_cpu_default2 to generic.
1812         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1813         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1814         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1816 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1817             Marc Glisse  <marc.glisse@inria.fr>
1819         PR tree-optimization/60502
1820         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1821         instead of build_low_bits_mask.
1823 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1825         PR middle-end/60482
1826         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1827         if there are multiple uses, but op doesn't live on E edge.
1828         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1829         clobber stmts before __builtin_unreachable.
1831 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1833         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1834         hard_frame_pointer_rtx.
1835         * cse.c (cse_insn): Remove volatile check.
1836         * cselib.c (cselib_process_insn): Likewise.
1837         * dse.c (scan_insn): Likewise.
1839 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1841         * config/arc/arc.c (conditionalize_nonjump): New function,
1842         broken out of ...
1843         (arc_ifcvt): ... this.
1844         (arc_predicate_delay_insns): Use it.
1846 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1848         * config/arc/predicates.md (extend_operand): During/after reload,
1849         allow const_int_operand.
1850         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1851         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1852         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1853         to "i".
1854         (umulsi3_highpart_i): Likewise.
1856 2014-03-11  Richard Biener  <rguenther@suse.de>
1858         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1859         Add asserts to guard possible wrong-code bugs.
1861 2014-03-11  Richard Biener  <rguenther@suse.de>
1863         PR tree-optimization/60429
1864         PR tree-optimization/60485
1865         * tree-ssa-structalias.c (set_union_with_increment): Properly
1866         take into account all fields that overlap the shifted vars.
1867         (do_sd_constraint): Likewise.
1868         (do_ds_constraint): Likewise.
1869         (get_constraint_for_ptr_offset): Likewise.
1871 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1873         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1874         (nios2_compute_frame_layout):
1875         Add calculation of cfun->machine->fp_save_offset.
1876         (nios2_expand_prologue): Correct setting of frame pointer register
1877         in prologue.
1878         (nios2_expand_epilogue): Update recovery of stack pointer from
1879         frame pointer accordingly.
1880         (nios2_initial_elimination_offset): Update calculation of offset
1881         for eliminating to HARD_FRAME_POINTER_REGNUM.
1883 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1885         PR ipa/60457
1886         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1887         cgraph_get_create_node on VAR_DECLs.
1889 2014-03-10  Richard Biener  <rguenther@suse.de>
1891         PR middle-end/60474
1892         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1894 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1896         * config/vms/vms.opt (vms_float_format): New variable.
1898 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1900         * doc/invoke.texi (-fcilkplus): Update implementation status.
1902 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1903             Richard Biener  <rguenther@suse.de>
1905         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1906         consistently accross all TUs.
1907         (run_gcc): Enable -fshort-double automatically at link at link-time
1908         and disallow override.
1910 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1912         PR target/58271
1913         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1914         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1915         if they can't be used.
1917 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1919         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1920         for Solaris 11/x86 ld.
1921         * configure: Regenerate.
1923 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1925         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1926         (LIB_TLS_SPEC): Save as ld_tls_libs.
1927         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1928         (HAVE_AS_IX86_TLSLDM): New test.
1929         * configure, config.in: Regenerate.
1930         * config/i386/i386.c (legitimize_tls_address): Fall back to
1931         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1932         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1933         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1934         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1936 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1938         * common.opt (fira-loop-pressure): Mark as optimization.
1940 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1942         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1943         an OpenMP mappable type.
1945 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1947         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1948         MULTILIB_OSDIRNAMES is not defined.
1950 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1951             Meador Inge  <meadori@codesourcery.com>
1953         PR target/58595
1954         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1955         (arm_legitimize_address): Call legitimize_tls_address for any
1956         arm_tls_referenced_p expression, handle constant addend.  Call it
1957         before testing for !TARGET_ARM.
1958         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1960 2014-03-06  Richard Biener  <rguenther@suse.de>
1962         PR middle-end/60445
1963         PR lto/60424
1964         PR lto/60427
1965         Revert
1966         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1968         * tree-streamer.c (record_common_node): Assert we don't record
1969         nodes with type double.
1970         (preload_common_node): Skip type double, complex double and double
1971         pointer since it is now frontend dependent due to fshort-double option.
1973 2014-03-06  Richard Biener  <rguenther@suse.de>
1975         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1976         or -fno-lto is specified and the linker has full plugin support.
1977         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1978         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1979         * lto-wrapper.c (merge_and_complain): Merge compile-time
1980         optimization levels.
1981         (run_gcc): And pass it through to the link options.
1983 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1985         PR debug/60381
1986         Revert:
1987         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1988         PR debug/59992
1989         * cselib.c (remove_useless_values): Skip to avoid quadratic
1990         behavior if the condition moved from...
1991         (cselib_process_insn): ... here holds.
1993 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1995         PR plugins/59335
1996         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1997         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1999         PR plugins/59335
2000         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
2001         (TM_H): Add x86-tune.def.
2003 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2005         * config/aarch64/aarch64.c (generic_tunings):
2006         Use cortexa57_extra_costs.
2008 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
2010         PR lto/60404
2011         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
2012         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
2013         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
2014         cost for in_lto_p.
2016 2014-03-04  Heiher  <r@hev.cc>
2018         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
2019         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
2021 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
2023         * config/i386/predicates.md (const2356_operand): Change to ...
2024         (const2367_operand): ... this.
2025         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
2026         const2367_operand.
2027         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2028         (*avx512pf_scatterpf<mode>sf): Ditto.
2029         (avx512pf_scatterpf<mode>df): Ditto.
2030         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2031         (*avx512pf_scatterpf<mode>df): Ditto.
2032         * config/i386/i386.c (ix86_expand_builtin): Update
2033         incorrect hint operand error message.
2035 2014-03-04  Richard Biener  <rguenther@suse.de>
2037         * lto-section-in.c (lto_get_section_data): Fix const cast.
2039 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
2041         * tree-streamer.c (record_common_node): Assert we don't record
2042         nodes with type double.
2043         (preload_common_node): Skip type double, complex double and double
2044         pointer since it is now frontend dependent due to fshort-double option.
2046 2014-03-04  Richard Biener  <rguenther@suse.de>
2048         PR lto/60405
2049         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
2050         (lto_input_toplevel_asms): Likewise.
2051         * lto-section-in.c (lto_get_section_data): Instead do it here
2052         for every section.
2054 2014-03-04  Richard Biener  <rguenther@suse.de>
2056         PR tree-optimization/60382
2057         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
2058         dead PHIs a reduction.
2060 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
2062         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
2063         hint value.
2064         (_mm_prefetch): Move out of GCC target("sse") pragma.
2065         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
2066         GCC target("prfchw") pragma.
2067         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
2068         for locality <= 2.
2069         * config/i386/i386.c (ix86_option_override_internal): Enable
2070         -mprfchw with -mprefetchwt1.
2072 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
2074         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
2075         Mark as varying.
2077 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
2079         * opts.h (CL_PCH_IGNORE): Define.
2080         * targhooks.c (option_affects_pch_p):
2081         Return false for options that have CL_PCH_IGNORE set.
2082         * opt-functions.awk: Process PchIgnore.
2083         * doc/options.texi: Document PchIgnore.
2085         * config/arc/arc.opt (misize): Add PchIgnore property.
2087 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2089         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
2090         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
2091         constraint on constants to permit them being loaded into
2092         GENERAL_REGS or BASE_REGS.
2094 2014-03-03  Nick Clifton  <nickc@redhat.com>
2096         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
2097         anti-cacnonical alternatives.
2098         (negandhi3_real): New pattern.
2099         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
2101 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2103         * config/avr/avr-mcus.def: Remove atxmega16x1.
2104         * config/avr/avr-tables.opt: Regenerate.
2105         * config/avr/t-multilib: Regenerate.
2106         * doc/avr-mmcu.texi: Regenerate.
2108 2014-03-03  Tobias Grosser  <tobias@grosser.es>
2109             Mircea Namolaru  <mircea.namolaru@inria.fr>
2111         PR tree-optimization/58028
2112         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
2113         scalar dimensions.
2115 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
2117         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
2118         not handled by recognizers.
2120 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
2122         PR middle-end/60175
2123         * function.c (expand_function_end): Don't emit
2124         clobber_return_register sequence if clobber_after is a BARRIER.
2125         * cfgexpand.c (construct_exit_block): Append instructions before
2126         return_label to prev_bb.
2128 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2130         * config/rs6000/constraints.md: Document reserved use of "wc".
2132 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
2134         PR ipa/60150
2135         * ipa.c (function_and_variable_visibility): When dissolving comdat
2136         group, also set all symbols to local.
2138 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
2140         PR ipa/60306
2142         Revert:
2143         2013-12-14   Jan Hubicka  <jh@suse.cz>
2144         PR middle-end/58477
2145         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
2147 2014-03-02  Jon Beniston  <jon@beniston.com>
2149         PR bootstrap/48230
2150         PR bootstrap/50927
2151         PR bootstrap/52466
2152         PR target/46898
2153         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
2154           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
2155         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
2156         (simple_return, *simple_return): New patterns
2157         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
2158         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
2160 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
2162         * dwarf2out.c (gen_subprogram_die): Tidy.
2164 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
2166         PR target/60071
2167         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
2168         (*mov_t_msb_neg_negc): ... this new insn.
2170 2014-02-28  Jason Merrill  <jason@redhat.com>
2172         PR c++/58678
2173         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
2174         function.
2176 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
2178         PR c++/60314
2179         * dwarf2out.c (decltype_auto_die): New static.
2180         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
2181         (gen_type_die_with_usage): Handle 'decltype(auto)'.
2182         (is_cxx_auto): Likewise.
2184 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
2186         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
2187         we are not using general regs only.
2189 2014-02-28  Richard Biener  <rguenther@suse.de>
2191         PR target/60280
2192         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
2193         previous fix and only allow to remove trivial pre-headers
2194         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
2195         (remove_forwarder_block): Properly update the latch of a loop.
2197 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2199         PR debug/59992
2200         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
2201         (cselib_preserved_hash_table): New.
2202         (preserve_constants_and_equivs): Move preserved vals to it.
2203         (cselib_find_slot): Look it up first.
2204         (cselib_init): Initialize it.
2205         (cselib_finish): Release it.
2206         (dump_cselib_table): Dump it.
2208 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2210         PR debug/59992
2211         * cselib.c (remove_useless_values): Skip to avoid quadratic
2212         behavior if the condition moved from...
2213         (cselib_process_insn): ... here holds.
2215 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2217         PR debug/57232
2218         * var-tracking.c (vt_initialize): Apply the same condition to
2219         preserve the CFA base value.
2221 2014-02-28  Joey Ye  <joey.ye@arm.com>
2223         PR target/PR60169
2224         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
2225         if reload in progress or completed.
2227 2014-02-28  Tobias Burnus  <burnus@net-b.de>
2229         PR middle-end/60147
2230         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
2231         NAMELIST_DECL.
2233 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
2235         * doc/tm.texi.in (Condition Code Status): Update documention for
2236         relative locations of cc0-setter and cc0-user.
2238 2014-02-27  Jeff Law  <law@redhat.com>
2240         PR rtl-optimization/52714
2241         * combine.c (try_combine): When splitting an unrecognized PARALLEL
2242         into two independent simple sets, if I3 is a jump, ensure the
2243         pattern we place into I3 is a (set (pc) ...).
2245 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
2246             Jeff Law  <law@redhat.com>
2248         PR rtl-optimization/49847
2249         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
2250         are in different blocks.
2251         * doc/tm.texi (Condition Code Status): Update documention for
2252         relative locations of cc0-setter and cc0-user.
2254 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
2256         PR target/59222
2257         * lra.c (lra_emit_add): Check SUBREG too.
2259 2014-02-27  Andreas Schwab  <schwab@suse.de>
2261         * config/m68k/m68k.c (m68k_option_override): Disable
2262         -flive-range-shrinkage for classic m68k.
2263         (m68k_override_options_after_change): Likewise.
2265 2014-02-27  Marek Polacek  <polacek@redhat.com>
2267         PR middle-end/59223
2268         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
2269         -Wmaybe-uninitialized.
2271 2014-02-27  Alan Modra  <amodra@gmail.com>
2273         PR target/57936
2274         * reload1.c (emit_input_reload_insns): When reload_override_in,
2275         set old to rl->in_reg when rl->in_reg is a subreg.
2277 2014-02-26  Richard Biener  <rguenther@suse.de>
2279         PR bootstrap/60343
2280         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
2282 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2284         * common/config/i386/predicates.md (const1256_operand): Remove.
2285         (const2356_operand): New.
2286         (const_1_to_2_operand): Remove.
2287         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2288         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2289         (*avx512pf_gatherpf<mode>sf): Ditto.
2290         (avx512pf_gatherpf<mode>df): Ditto.
2291         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2292         (*avx512pf_gatherpf<mode>df): Ditto.
2293         (avx512pf_scatterpf<mode>sf): Ditto.
2294         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2295         (*avx512pf_scatterpf<mode>sf): Ditto.
2296         (avx512pf_scatterpf<mode>df): Ditto.
2297         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2298         (*avx512pf_scatterpf<mode>df): Ditto.
2299         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
2301 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
2303         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
2304         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
2305         (_mm512_mask_testn_epi64_mask): Move to ...
2306         * config/i386/avx512cdintrin.h: Here.
2307         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
2308         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
2309         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
2310         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
2311         TARGET_AVX512F from TARGET_AVX512CD.
2313 2014-02-26  Richard Biener  <rguenther@suse.de>
2315         PR ipa/60327
2316         * ipa.c (walk_polymorphic_call_targets): Properly guard
2317         call to inline_update_overall_summary.
2319 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
2321         PR target/60280
2322         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
2323         and latches only if requested.  Fix latch if it is removed.
2324         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
2325         LOOPS_HAVE_PREHEADERS.
2327 2014-02-25  Andrew Pinski  <apinski@cavium.com>
2329         * builtins.c (expand_builtin_thread_pointer): Create a new target
2330         when the target is NULL.
2332 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
2334         PR rtl-optimization/60317
2335         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2336         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2337         * lra-assigns.c: Include params.h.
2338         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
2339         other reload pseudos considerations.
2341 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2343         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
2344         to use canonical form for nor<mode>3.
2346 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2348         PR target/55426
2349         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2350         conversions.
2352 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2354         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2355         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2356         (ix86_handle_option): Handle OPT_mprefetchwt1.
2357         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2358         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2359         PREFETCHWT1 CPUID.
2360         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2361         OPTION_MASK_ISA_PREFETCHWT1.
2362         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2363         (PTA_PREFETCHWT1): New.
2364         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2365         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2366         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2367         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2368         (*prefetch_avx512pf_<mode>_: Change into ...
2369         (*prefetch_prefetchwt1_<mode>: This.
2370         * config/i386/i386.opt (mprefetchwt1): New.
2371         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2372         (_mm_prefetch): Handle intent to write.
2373         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2375 2014-02-25  Richard Biener  <rguenther@suse.de>
2377         PR middle-end/60291
2378         * emit-rtl.c (mem_attrs_htab): Remove.
2379         (mem_attrs_htab_hash): Likewise.
2380         (mem_attrs_htab_eq): Likewise.
2381         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2382         (init_emit_once): Do not allocate mem_attrs_htab.
2384 2014-02-25  Richard Biener  <rguenther@suse.de>
2386         PR lto/60319
2387         * lto-opts.c (lto_write_options): Output non-explicit conservative
2388         -fwrapv, -fno-trapv and -fno-strict-overflow.
2389         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2390         (run_gcc): And pass them through.
2392 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2394         * sel-sched.c (calculate_new_fences): New parameter ptime.
2395         Calculate it as a maximum over all fence cycles.
2396         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2397         Print the final schedule timing when sched_verbose.
2399 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2401         PR rtl-optimization/60292
2402         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2403         bit fot the fence instruction.
2405 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2407         * calls.h: Fix typo in comment.
2409 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2411         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2412         adjusting offsetable addresses.
2414 2014-02-24  Guozhi Wei  <carrot@google.com>
2416         * sparseset.h (sparseset_pop): Fix the wrong index.
2418 2014-02-24  Walter Lee  <walt@tilera.com>
2420         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2421         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2422         triplet.
2423         * common/config/tilegx/tilegx-common.c
2424         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2425         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2426         (LINK_SPEC): Ditto.
2427         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2428         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2429         (tilegx_gimplify_va_arg_expr): Handle big endian.
2430         (tilegx_expand_unaligned_load): Ditto.
2431         (tilegx_expand_unaligned_store): Ditto.
2432         (TARGET_RETURN_IN_MSB): New.
2433         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2434         (TARGET_ENDIAN_DEFAULT): New.
2435         (TARGET_BIG_ENDIAN): Handle big endian.
2436         (BYTES_BIG_ENDIAN): Ditto.
2437         (WORDS_BIG_ENDIAN): Ditto.
2438         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2439         (ENDIAN_SPEC): New.
2440         (EXTRA_SPECS): New.
2441         * config/tilegx/tilegx.md (extv): Handle big endian.
2442         (extzv): Ditto.
2443         (insn_st<n>): Ditto.
2444         (insn_st<n>_add<bitsuffix>): Ditto.
2445         (insn_stnt<n>): Ditto.
2446         (insn_stnt<n>_add<bitsuffix>):Ditto.
2447         (vec_interleave_highv8qi): Handle big endian.
2448         (vec_interleave_highv8qi_be): New.
2449         (vec_interleave_highv8qi_le): New.
2450         (insn_v1int_h): Handle big endian.
2451         (vec_interleave_lowv8qi): Handle big endian.
2452         (vec_interleave_lowv8qi_be): New.
2453         (vec_interleave_lowv8qi_le): New.
2454         (insn_v1int_l): Handle big endian.
2455         (vec_interleave_highv4hi): Handle big endian.
2456         (vec_interleave_highv4hi_be): New.
2457         (vec_interleave_highv4hi_le): New.
2458         (insn_v2int_h): Handle big endian.
2459         (vec_interleave_lowv4hi): Handle big endian.
2460         (vec_interleave_lowv4hi_be): New.
2461         (vec_interleave_lowv4hi_le): New.
2462         (insn_v2int_l): Handle big endian.
2463         (vec_interleave_highv2si): Handle big endian.
2464         (vec_interleave_highv2si_be): New.
2465         (vec_interleave_highv2si_le): New.
2466         (insn_v4int_h): Handle big endian.
2467         (vec_interleave_lowv2si): Handle big endian.
2468         (vec_interleave_lowv2si_be): New.
2469         (vec_interleave_lowv2si_le): New.
2470         (insn_v4int_l): Handle big endian.
2471         * config/tilegx/tilegx.opt (mbig-endian): New option.
2472         (mlittle-endian): New option.
2473         * doc/install.texi: Document tilegxbe-linux.
2474         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2476 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2478         PR ipa/60266
2479         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2480         there are no parameter descriptors.
2482 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2484         PR rtl-optimization/60268
2485         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2486         initialization to ...
2487         (sched_rgn_init): ... here.
2488         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2490 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2492         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2493         names.
2495 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2497         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2498         definition.
2500 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2502         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2503         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2505 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2507         * config/microblaze/predicates.md: Add cmp_op predicate.
2508         * config/microblaze/microblaze.md: Add branch_compare instruction
2509         which uses cmp_op predicate and emits cmp insn before branch.
2510         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2511         to microblaze_expand_conditional_branch and consolidate logic.
2512         (microblaze_expand_conditional_branch): emit branch_compare
2513         insn instead of handling cmp op separate from branch insn.
2515 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2517         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2518         to permit subregs.
2520 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2522         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2523         define_insn with define_expand and new define_insn
2524         *altivec_lve<VI_char>x_internal.
2525         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2526         and new define_insn *altivec_stve<VI_char>x_internal.
2527         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2528         prototype.
2529         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2530         lve*x built-ins.
2531         (altivec_expand_stvex_be): New function.
2533 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2535         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2536         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2537         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2538         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2540 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2542         PR target/60298
2543         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2544         instead of emit_move_insn.
2546 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2548         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2549         vspltw with vsldoi.
2550         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2551         gen_altivec_vsumsws.
2553 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2555         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2556         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2557         (altivec_lvxl_<mode>): New define_expand incorporating
2558         -maltivec=be semantics where needed.
2559         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2560         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2561         semantics where needed.
2562         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2563         (altivec_stvx_<mode>): New define_expand incorporating
2564         -maltivec=be semantics where needed.
2565         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2566         VM2 iterator instead of V4SI.
2567         (altivec_stvxl_<mode>): New define_expand incorporating
2568         -maltivec=be semantics where needed.
2569         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2570         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2571         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2572         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2573         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2574         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2575         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2576         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2577         ALTIVEC_BUILTIN_STVXL.
2578         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2579         (altivec_expand_stvx_be): Likewise.
2580         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2581         (altivec_expand_lvx_be): Likewise.
2582         (altivec_expand_stvx_be): Likewise.
2583         (altivec_expand_builtin): Add cases for
2584         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2585         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2586         (altivec_init_builtins): Add definitions for
2587         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2588         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2590 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2592         * doc/invoke.texi (mvirt, mno-virt): Document.
2593         * config/mips/mips.opt (mvirt): New option.
2594         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2596 2014-02-21  Richard Biener  <rguenther@suse.de>
2598         PR tree-optimization/60276
2599         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2600         (STMT_VINFO_MIN_NEG_DIST): New macro.
2601         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2602         STMT_VINFO_MIN_NEG_DIST.
2603         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2604         made for negative dependence distances still hold.
2606 2014-02-21  Richard Biener  <rguenther@suse.de>
2608         PR middle-end/60291
2609         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2610         DECL_INITIAL for globals not in the current function context.
2612 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2614         PR tree-optimization/56490
2615         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2616         * tree-ssa-uninit.c: Include params.h.
2617         (compute_control_dep_chain): Add num_calls argument, return false
2618         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2619         num_calls to recursive call.
2620         (find_predicates): Change dep_chain into normal array,
2621         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2622         variable and adjust compute_control_dep_chain caller.
2623         (find_def_preds): Likewise.
2625 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2627         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2628         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2630 2014-02-21  Nick Clifton  <nickc@redhat.com>
2632         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2633         (pushhi1): Likewise.
2634         (popqi1): Add mode to pre_dec.
2635         (pophi1): Likewise.
2637 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2639         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2640         mode for mask of V8SFmode permutation.
2642 2014-02-20  Richard Henderson  <rth@redhat.com>
2644         PR c++/60272
2645         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2646         a new pseudo for OLDVAL.
2648 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2650         PR target/57896
2651         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2652         gen_reg_rtx if d->testing_p.
2653         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2654         if d->testing_p and we will certainly return true.
2655         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2656         if d->testing_p.
2658 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2660         * emit-rtl.c (gen_reg_rtx): Assert that
2661         crtl->emit.regno_pointer_align_length is non-zero.
2663 2014-02-20  Richard Henderson  <rth@redhat.com>
2665         PR c++/60272
2666         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2667         on failure the store back into EXPECT.
2669 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2670             Sandra Loosemore  <sandra@codesourcery.com>
2672         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2673         * config/nios2/nios2.c (nios2_function_profiler): Add
2674         -fPIC (flag_pic == 2) support.
2675         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2676         (nios2_large_offset_p): New function.
2677         (nios2_unspec_reloc_p): Move up position, update to use
2678         nios2_large_offset_p.
2679         (nios2_unspec_address): Remove function.
2680         (nios2_unspec_offset): New function.
2681         (nios2_large_got_address): New function.
2682         (nios2_got_address): Add large offset support.
2683         (nios2_legitimize_tls_address): Update usage of removed and new
2684         functions.
2685         (nios2_symbol_binds_local_p): New function.
2686         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2687         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2688         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2689         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2690         processing for (const (unspec ...)).
2691         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2693 2014-02-20  Richard Biener  <rguenther@suse.de>
2695         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2696         doing the substitution.
2697         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2699 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2701         PR ipa/55260
2702         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2703         info when checking whether lattices are bottom.
2705 2014-02-20  Richard Biener  <rguenther@suse.de>
2707         PR middle-end/60221
2708         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2709         regions at -O0.
2711 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2713         PR ipa/58555
2714         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2715         parameter specifying the scaling.
2716         (inline_call): Update.
2717         (want_inline_recursively): Guard division by zero.
2718         (recursive_inlining): Update.
2719         * ipa-inline.h (clone_inlined_nodes): Update.
2721 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2723         PR target/60204
2724         * config/i386/i386.c (classify_argument): Pass structures of size
2725         64 bytes or less in register.
2727 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2728             Kirill Yukhin  <kirill.yukhin@intel.com>
2730         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2731         (_mm_rcp28_round_ss): Ditto.
2732         (_mm_rsqrt28_round_sd): Ditto.
2733         (_mm_rsqrt28_round_ss): Ditto.
2734         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2735         (_mm_rcp14_round_ss): Ditto.
2736         (_mm_rsqrt14_round_sd): Ditto.
2737         (_mm_rsqrt14_round_ss): Ditto.
2738         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2739         the first input operand, get rid of match_dup.
2740         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2741         attribute to sse.
2742         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2743         Ditto.
2744         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2745         operand as the first input operand, set type attribute.
2746         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2747         Set type attribute.
2748         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2749         operand as the first input operand, set type attribute.
2751 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2753         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2754         bit of zero.
2756 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2758         PR target/60207
2759         * config/i386/i386.c (construct_container): Remove TFmode check
2760         for X86_64_INTEGER_CLASS.
2762 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2764         PR target/59794
2765         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2766         only when -Wpsabi is enabled.
2768 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2770          PR target/59799
2771         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2772         passing arrays in registers are the same as for structs, so remove the
2773         special case for them.
2775 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2777         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2778         destination type, extract only the valid bits if the source type is not
2779         integral and has a different mode.
2781 2014-02-19  Richard Biener  <rguenther@suse.de>
2783         PR ipa/60243
2784         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2785         for all calls.
2787 2014-02-19  Richard Biener  <rguenther@suse.de>
2789         PR ipa/60243
2790         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2791         (ipa_modify_call_arguments): Emit an argument load explicitely and
2792         preserve virtual SSA form there and for the replacement call.
2793         Do not update SSA form nor free dominance info.
2795 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2797         * ipa.c (function_and_variable_visibility): Also clear WEAK
2798         flag when disolving COMDAT_GROUP.
2800 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2802         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2803         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2804         not devirtualizing.
2805         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2806         do more sanity checks.
2807         (detect_type_change): Return true when giving up early.
2808         (compute_complex_assign_jump_func): Fix type parameter of
2809         ipa_set_ancestor_jf.
2810         (compute_complex_ancestor_jump_func): Likewise.
2811         (update_jump_functions_after_inlining): Fix updating of
2812         ancestor function.
2813         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2815 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2817         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2818         inline clones when edge disappears.
2820 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2822         PR target/60203
2823         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2824         Split 64-bit moves into 2 patterns.  Do not allow the use of
2825         direct move for TDmode in little endian, since the decimal value
2826         has little endian bytes within a word, but the 64-bit pieces are
2827         ordered in a big endian fashion, and normal subreg's of TDmode are
2828         not allowed.
2829         (mov<mode>_64bit_dm): Likewise.
2830         (movtd_64bit_nodm): Likewise.
2832 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2834         PR tree-optimization/60174
2835         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2836         statement of an SSA_NAME that occurs in an abnormal PHI node.
2838 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2840         PR sanitizer/60142
2841         * final.c (SEEN_BB): Remove.
2842         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2843         (final_scan_insn): Don't force_source_line on second
2844         NOTE_INSN_BASIC_BLOCK.
2846 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2848         PR target/60205
2849         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2850         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2851         (type_natural_mode): Warn ABI change when %zmm register is not
2852         available for AVX512F vector value passing.
2854 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2856         PR target/60193
2857         * config/i386/i386.c (ix86_expand_prologue): Use value in
2858         rax register as displacement when restoring %r10 or %rax.
2859         Fix wrong offset when restoring both registers.
2861 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2863         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2864         assertion with conditional return.
2866 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2867             Uros Bizjak  <ubizjak@gmail.com>
2869         PR driver/60233
2870         * config/i386/driver-i386.c (host_detect_local_cpu): If
2871         YMM state is not saved by the OS, also clear has_f16c.  Move
2872         CPUID 0x80000001 handling before YMM state saving checking.
2874 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2876         PR rtl-optimization/58960
2877         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2878         factored out from ...
2879         (sched_init): ... here.
2880         (free_global_sched_pressure_data): New, factored out from ...
2881         (sched_finish): ... here.
2882         * sched-int.h (free_global_sched_pressure_data): Declare.
2883         * sched-rgn.c (nr_regions_initial): New static global.
2884         (haifa_find_rgns): Initialize it.
2885         (schedule_region): Disable sched-pressure for the newly
2886         generated regions.
2888 2014-02-17  Richard Biener  <rguenther@suse.de>
2890         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2891         release SSA defs of pattern stmts.
2893 2014-02-17  Richard Biener  <rguenther@suse.de>
2895         * tree-inline.c (expand_call_inline): Release the virtual
2896         operand defined by the call we are about to inline.
2898 2014-02-17  Richard Biener  <rguenther@suse.de>
2900         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2902 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2903             Ilya Tocar  <ilya.tocar@intel.com>
2905         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2906         arguments order in builtin.
2907         (_mm512_permutexvar_epi64): Ditto.
2908         (_mm512_mask_permutexvar_epi64): Ditto
2909         (_mm512_maskz_permutexvar_epi32): Ditto
2910         (_mm512_permutexvar_epi32): Ditto
2911         (_mm512_mask_permutexvar_epi32): Ditto
2913 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2915         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2916         (p8_vmrgow): Likewise.
2918 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2920         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2921         endian targets.
2923 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2925         PR target/60203
2926         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2927         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2928         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2929         using direct move instructions on ISA 2.07.  Also adjust
2930         instruction length for 64-bit.
2931         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2932         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2934 2014-02-15  Alan Modra  <amodra@gmail.com>
2936         PR target/58675
2937         PR target/57935
2938         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2939         find_replacement on parts of insn rtl that might be reloaded.
2941 2014-02-15  Richard Biener  <rguenther@suse.de>
2943         PR tree-optimization/60183
2944         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2945         (tree_ssa_phiprop): Calculate and free post-dominators.
2947 2014-02-14  Jeff Law  <law@redhat.com>
2949         PR rtl-optimization/60131
2950         * ree.c (get_extended_src_reg): New function.
2951         (combine_reaching_defs): Use it rather than assuming location of REG.
2952         (find_and_remove_re): Verify first operand of extension is
2953         a REG before adding the insns to the copy list.
2955 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2957         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2958         * configure: Regenerated.
2959         * config.in: Regenerated.
2960         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2961         instead of ASM_SHORT.
2963 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2964             Richard Earnshaw  <rearnsha@arm.com>
2966         PR rtl-optimization/59535
2967         * lra-constraints.c (process_alt_operands): Encourage alternative
2968         when unassigned pseudo class is superset of the alternative class.
2969         (inherit_reload_reg): Don't inherit when optimizing for code size.
2970         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2971         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2972         modes not less than 4 for Thumb1.
2974 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2976         PR pch/60010
2977         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2979 2014-02-14  Richard Biener  <rguenther@suse.de>
2981         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2982         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2983         Do not strip INDIRECT_REFs.
2985 2014-02-14  Richard Biener  <rguenther@suse.de>
2987         PR lto/60179
2988         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2989         DECL_FUNCTION_SPECIFIC_TARGET.
2990         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2991         * tree-streamer-out.c (pack_ts_target_option): Remove.
2992         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2993         (write_ts_function_decl_tree_pointers): Do not stream
2994         DECL_FUNCTION_SPECIFIC_TARGET.
2995         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2996         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2997         (lto_input_ts_function_decl_tree_pointers): Do not stream
2998         DECL_FUNCTION_SPECIFIC_TARGET.
3000 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
3002         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
3003         (get_initial_def_for_induction, vectorizable_induction): Ignore
3004         debug stmts when looking for exit_phi.
3005         (vectorizable_live_operation): Fix up condition.
3007 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3009         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
3010         nreverse() because it changes the content of original tree list.
3012 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3014         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
3015         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
3017 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
3019         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
3020         GNU coding standards.
3022 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
3024         PR debug/60152
3025         * dwarf2out.c (gen_subprogram_die): Don't call
3026         add_calling_convention_attribute if subr_die is old_die.
3028 2014-02-13  Sharad Singhai  <singhai@google.com>
3030         * doc/optinfo.texi: Fix order of nodes.
3032 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
3034         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
3035         operands[2], not operands[3].
3037 2014-02-13  Richard Biener  <rguenther@suse.de>
3039         PR bootstrap/59878
3040         * doc/install.texi (ISL): Update recommended version to 0.12.2,
3041         mention the possibility of an in-tree build.
3042         (CLooG): Update recommended version to 0.18.1, mention the
3043         possibility of an in-tree build and clarify that the ISL
3044         bundled with CLooG does not work.
3046 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
3048         PR target/43546
3049         * expr.c (compress_float_constant): If x is a hard register,
3050         extend into a pseudo and then move to x.
3052 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
3054         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
3055         caused by bad second argument to warning_at() with -mhotpatch and
3056         nested functions (e.g. with gfortran).
3058 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
3060         * opts.c (option_name): Remove "enabled by default" rider.
3062 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
3064         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
3066 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
3067             Uros Bizjak  <ubizjak@gmail.com>
3069         PR target/60151
3070         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
3071         * configure: Regenerated.
3073 2014-02-12  Richard Biener  <rguenther@suse.de>
3075         * vec.c (vec_prefix::calculate_allocation): Move as
3076         inline variant to vec.h.
3077         (vec_prefix::calculate_allocation_1): New out-of-line version.
3078         * vec.h (vec_prefix::calculate_allocation_1): Declare.
3079         (vec_prefix::m_has_auto_buf): Rename to ...
3080         (vec_prefix::m_using_auto_storage): ... this.
3081         (vec_prefix::calculate_allocation): Inline the easy cases
3082         and dispatch to calculate_allocation_1 which doesn't need the
3083         prefix address.
3084         (va_heap::reserve): Use gcc_checking_assert.
3085         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
3086         m_using_auto_storage.
3087         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
3088         member and adjust.
3089         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
3090         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
3091         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
3093 2014-02-12  Richard Biener  <rguenther@suse.de>
3095         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
3096         when we found a dependence.
3098 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
3100         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
3101         common code...
3102         (maybe_fold_stmt): ... into this new function.
3103         * omp-low.c (lower_omp): Update comment.
3105         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
3106         last use.
3108         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
3109         dereference.
3111 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
3113         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
3114         identifiers in comments.
3115         (cortexa53_extra_costs): Likewise.
3116         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
3117         (cortexa7_extra_costs): Likewise.
3118         (cortexa12_extra_costs): Likewise.
3119         (cortexa15_extra_costs): Likewise.
3120         (v7m_extra_costs): Likewise.
3122 2014-02-12  Richard Biener  <rguenther@suse.de>
3124         PR middle-end/60092
3125         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
3126         of posix_memalign being successful.
3127         (lower_stmt): Restrict lowering of posix_memalign to when
3128         -ftree-bit-ccp is enabled.
3130 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
3132         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
3133         arg_loc.
3134         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
3136 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
3138         PR rtl-optimization/60116
3139         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
3140         other_insn once the combination has been validated.
3142 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
3144         PR lto/59468
3145         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
3146         and wrapper.
3147         * ipa-devirt.c: Include demangle.h
3148         (odr_violation_reported): New static variable.
3149         (add_type_duplicate): Update odr_violations.
3150         (maybe_record_node): Add completep parameter; update it.
3151         (record_target_from_binfo): Add COMPLETEP parameter;
3152         update it as needed.
3153         (possible_polymorphic_call_targets_1): Likewise.
3154         (struct polymorphic_call_target_d): Add nonconstruction_targets;
3155         rename FINAL to COMPLETE.
3156         (record_targets_from_bases): Sanity check we found the binfo;
3157         fix COMPLETEP updating.
3158         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
3159         parameter, fix computing of COMPLETEP.
3160         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
3161         at LTO time do demangling.
3162         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
3163         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
3164         parameter.
3165         (gimple_get_virt_method_for_binfo): Likewise.
3166         * gimple-fold.h (gimple_get_virt_method_for_binfo,
3167         gimple_get_virt_method_for_vtable): Update prototypes.
3169 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
3171         PR target/49008
3172         * genautomata.c (add_presence_absence): Fix typo with
3173         {final_}presence_list.
3175 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
3177         PR target/60137
3178         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
3179         for VSX/Altivec vectors that land in GPR registers.
3181 2014-02-11  Richard Henderson  <rth@redhat.com>
3182             Jakub Jelinek  <jakub@redhat.com>
3184         PR debug/59776
3185         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
3186         around drhs if type conversion to lacc->type is not useless.
3188 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3190         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
3191         tuning struct.
3192         (cortex-a57.cortex-a53): Likewise.
3193         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
3195 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3197         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
3198         arm_restrict_it.
3200 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
3202         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
3203         add_options_for_arm_vfp3.
3205 2014-02-11  Jeff Law  <law@redhat.com>
3207         PR middle-end/54041
3208         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
3209         object with an undesirable mode.
3211 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3213         PR libgomp/60107
3214         * config/i386/sol2-9.h: New file.
3215         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
3216         *-*-solaris2.9*): Use it.
3218 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3220         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
3221         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
3223 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3225         * config/microblaze/microblaze.c: Extend mcpu version format
3227 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
3229         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
3231 2014-02-10  Richard Henderson  <rth@redhat.com>
3233         PR target/59927
3234         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
3235         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
3236         ms-abi vs -mno-accumulate-outgoing-args.
3237         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
3238         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
3239         respect to ms-abi.
3241 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3243         PR middle-end/60080
3244         * cfgexpand.c (expand_asm_operands): Attach source location to
3245         ASM_INPUT rtx objects.
3246         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
3248 2014-02-10  Nick Clifton  <nickc@redhat.com>
3250         * config/mn10300/mn10300.c (popcount): New function.
3251         (mn10300_expand_prologue): Include saved registers in stack usage
3252         count.
3254 2014-02-10  Jeff Law  <law@redhat.com>
3256         PR middle-end/52306
3257         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
3258         when changing the SET_DEST of a prior insn to avoid an input reload.
3260 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
3262         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
3263         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
3264         -mcall-openbsd, or -mcall-linux.
3265         (CC1_ENDIAN_BIG_SPEC): Remove.
3266         (CC1_ENDIAN_LITTLE_SPEC): Remove.
3267         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3268         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
3269         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
3270         and %cc1_endian_default.
3271         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3273 2014-02-10  Richard Biener  <rguenther@suse.de>
3275         PR tree-optimization/60115
3276         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3277         MEM_REF handling.  Properly verify that the accesses are not
3278         out of the objects bound.
3280 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3282         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
3283         coretex to cortex.
3285 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
3287         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
3288         proper constants and fix formatting.
3289         (possible_polymorphic_call_targets): Fix formatting.
3291 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
3292             Ilya Tocar  <ilya.tocar@intel.com>
3294         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
3295         (_mm512_loadu_epi32): Renamed into...
3296         (_mm512_loadu_si512): This.
3297         (_mm512_storeu_epi32): Renamed into...
3298         (_mm512_storeu_si512): This.
3299         (_mm512_maskz_ceil_ps): Removed.
3300         (_mm512_maskz_ceil_pd): Ditto.
3301         (_mm512_maskz_floor_ps): Ditto.
3302         (_mm512_maskz_floor_pd): Ditto.
3303         (_mm512_floor_round_ps): Ditto.
3304         (_mm512_floor_round_pd): Ditto.
3305         (_mm512_ceil_round_ps): Ditto.
3306         (_mm512_ceil_round_pd): Ditto.
3307         (_mm512_mask_floor_round_ps): Ditto.
3308         (_mm512_mask_floor_round_pd): Ditto.
3309         (_mm512_mask_ceil_round_ps): Ditto.
3310         (_mm512_mask_ceil_round_pd): Ditto.
3311         (_mm512_maskz_floor_round_ps): Ditto.
3312         (_mm512_maskz_floor_round_pd): Ditto.
3313         (_mm512_maskz_ceil_round_ps): Ditto.
3314         (_mm512_maskz_ceil_round_pd): Ditto.
3315         (_mm512_expand_pd): Ditto.
3316         (_mm512_expand_ps): Ditto.
3317         * config/i386/i386.c (ix86_builtins): Remove
3318         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
3319         (bdesc_args): Ditto.
3320         * config/i386/predicates.md (const1256_operand): New.
3321         (const_1_to_2_operand): Ditto.
3322         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
3323         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3324         (*avx512pf_gatherpf<mode>sf): Ditto.
3325         (avx512pf_gatherpf<mode>df): Ditto.
3326         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3327         (*avx512pf_gatherpf<mode>df): Ditto.
3328         (avx512pf_scatterpf<mode>sf): Ditto.
3329         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3330         (*avx512pf_scatterpf<mode>sf): Ditto.
3331         (avx512pf_scatterpf<mode>df): Ditto.
3332         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3333         (*avx512pf_scatterpf<mode>df): Ditto.
3334         (avx512f_expand<mode>): Removed.
3335         (<shift_insn><mode>3<mask_name>): Change predicate type.
3337 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3339         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
3340         not at the end of datarefs vector use ordered_remove to avoid
3341         reordering datarefs vector.
3343         PR c/59984
3344         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
3345         mark local addressable non-static vars as GOVD_PRIVATE
3346         instead of GOVD_LOCAL.
3347         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3348         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3349         than copying them.
3351         PR middle-end/60092
3352         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3353         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3354         assume_aligned or alloc_align attributes.
3355         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3356         arguments.  Handle also assume_aligned and alloc_align attributes.
3357         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3358         calls to functions with assume_aligned or alloc_align attributes.
3359         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3361 2014-02-08  Terry Guo  <terry.guo@arm.com>
3363         * doc/invoke.texi: Document ARM -march=armv7e-m.
3365 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3367         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3368         flag on __cilkrts_rethrow builtin.
3370         PR ipa/60026
3371         * ipa-cp.c (determine_versionability): Fail at -O0
3372         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3373         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3375         Revert:
3376         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3378         PR ipa/60026
3379         * tree-inline.c (copy_forbidden): Fail for
3380         __attribute__((optimize (0))) functions.
3382 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3384         * varpool.c: Include pointer-set.h.
3385         (varpool_remove_unreferenced_decls): Variables in other partitions
3386         will not be output; be however careful to not lose information
3387         about partitioning.
3389 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3391         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3392         lookup in the vtable constructor.
3394 2014-02-07  Jeff Law  <law@redhat.com>
3396         PR target/40977
3397         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3398         define_insn_and_split.
3400         * ipa-inline.c (inline_small_functions): Fix typos.
3402 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3404         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3405         (s390_can_use_return_insn): Declare.
3406         * config/s390/s390.h (EPILOGUE_USES): Define.
3407         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3408         instructions.
3409         (s390_chunkify_start): Handle return JUMP_LABELs.
3410         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3411         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3412         (s390_can_use_return_insn): New functions.
3413         (s390_fix_long_loop_prediction): Handle conditional returns.
3414         (TARGET_SET_UP_BY_PROLOGUE): Define.
3415         * config/s390/s390.md (ANY_RETURN): New code iterator.
3416         (*creturn, *csimple_return, return, simple_return): New patterns.
3418 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3420         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3421         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3422         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3423         REG_CFA_RESTORE list when deciding not to restore a register.
3425 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3427         * config/s390/s390.c: Include tree-pass.h and context.h.
3428         (s390_early_mach): New function, split out from...
3429         (s390_emit_prologue): ...here.
3430         (pass_data_s390_early_mach): New pass structure.
3431         (pass_s390_early_mach): New class.
3432         (s390_option_override): Create and register early_mach pass.
3433         Move to end of file.
3435 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3437         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3438         to match for the exit block.
3440 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3442         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3443         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3444         Reject misaligned operands.
3446 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3448         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3450 2014-02-07  Richard Biener  <rguenther@suse.de>
3452         PR middle-end/60092
3453         * gimple-low.c (lower_builtin_posix_memalign): New function.
3454         (lower_stmt): Call it to lower posix_memalign in a way
3455         to make alignment info accessible.
3457 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3459         PR c++/60082
3460         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3461         __builtin_setjmp_receiver.
3463 2014-02-07  Richard Biener  <rguenther@suse.de>
3465         PR middle-end/60092
3466         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3467         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3468         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3469         Handle BUILT_IN_POSIX_MEMALIGN.
3470         (find_func_clobbers): Likewise.
3471         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3472         (call_may_clobber_ref_p_1): Likewise.
3474 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3476         PR ipa/59918
3477         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3478         sanity check.
3480 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3482         PR ipa/59469
3483         * lto-cgraph.c (lto_output_node): Use
3484         symtab_get_symbol_partitioning_class.
3485         (lto_output_varpool_node): likewise.
3486         (symtab_get_symbol_partitioning_class): Move here from
3487         lto/lto-partition.c
3488         * cgraph.h (symbol_partitioning_class): Likewise.
3489         (symtab_get_symbol_partitioning_class): Declare.
3491 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3493         * ggc.h (ggc_internal_cleared_alloc): New macro.
3494         * vec.h (vec_safe_copy): Handle memory stats.
3495         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3496         * target-globals.c (save_target_globals): Likewise.
3498 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3500         PR target/60077
3501         * expr.c (emit_move_resolve_push): Export; be bit more selective
3502         on when to clear alias set.
3503         * expr.h (emit_move_resolve_push): Declare.
3504         * function.h (struct function): Add tail_call_marked.
3505         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3506         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3507         * config/i386/i386.md (TImode move expander): De not call
3508         ix86_expand_push.
3509         (FP push expanders): Preserve memory attributes.
3510         * config/i386/sse.md (push<mode>1): Remove.
3511         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3512         (ix86_expand_push): Remove.
3513         * config/i386/mmx.md (push<mode>1): Remove.
3515 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3517         PR rtl-optimization/60030
3518         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3519         lopart with paradoxical subreg before shifting it up by hprec.
3521 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3523         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3524         Remove extra newline at end of file.
3525         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3526         (arm_issue_rate): Handle cortexa57.
3527         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3528         (cortex-a57.cortex-a53): Likewise.
3530 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3532         PR target/59575
3533         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3534         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3535         bitmask.
3536         (arm_expand_prologue): Adjust all callers.
3537         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3538         info, registers also at the lowest numbered registers side.  Use
3539         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3540         XEXP.
3542         PR debug/59992
3543         * var-tracking.c (adjust_mems): Before adding a SET to
3544         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3546 2014-02-06  Alan Modra  <amodra@gmail.com>
3548         PR target/60032
3549         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3550         change SDmode to DDmode when lra_in_progress.
3552 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3554         PR middle-end/59150
3555         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3556         free_data_ref on the dr first, and before goto again also set dr
3557         to the next dr.  For simd_lane_access, free old datarefs[i] before
3558         overwriting it.  For get_vectype_for_scalar_type failure, don't
3559         free_data_ref if simd_lane_access.
3561         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3563         PR target/60062
3564         * tree.h (opts_for_fn): New inline function.
3565         (opt_for_fn): Define.
3566         * config/i386/i386.c (ix86_function_regparm): Use
3567         opt_for_fn (decl, optimize) instead of optimize.
3569 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3571         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3572         for SYMBOL_REF in large memory model.
3574 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3576         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3577         and crypto support.
3578         (cortex-a57): Likewise.
3579         (cortex-a57.cortex-a53): Likewise.
3581 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3582             Kugan Vivekanandarajah  <kuganv@linaro.org>
3584         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3585         unaligned_access.
3586         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3588 2014-02-06  Richard Biener  <rguenther@suse.de>
3590         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3591         set_loop_copy and initialize_original_copy_tables.
3593 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3595         * config/aarch64/aarch64-simd.md
3596         (aarch64_ashr_simddi): Change QI to SI.
3598 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3599             Jakub Jelinek  <jakub@redhat.com>
3601         PR middle-end/60013
3602         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3603         of the dataflow.
3605 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3607         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3608         CODE_FOR_altivec_vpku[hw]um to
3609         CODE_FOR_altivec_vpku[hw]um_direct.
3610         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3611         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3612         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3613         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3615 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3617         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3618         generation for -maltivec=be.
3619         (altivec_vsumsws): Simplify redundant test.
3621 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3623         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3624         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3625         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3626         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3627         gen_altivec_vpkuwum.
3628         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3629         BYTES_BIG_ENDIAN.
3630         (altivec_vpks<VI_char>ss): Likewise.
3631         (altivec_vpks<VI_char>us): Likewise.
3632         (altivec_vpku<VI_char>us): Likewise.
3633         (altivec_vpku<VI_char>um): Likewise.
3634         (altivec_vpku<VI_char>um_direct): New (copy of
3635         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3636         internal use).
3637         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3638         target is little endian and -maltivec=be is not specified.
3639         (*altivec_vupkhs<VU_char>_direct): New (copy of
3640         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3641         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3642         target is little endian and -maltivec=be is not specified.
3643         (*altivec_vupkls<VU_char>_direct): New (copy of
3644         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3645         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3646         little endian and -maltivec=be is not specified.
3647         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3648         little endian and -maltivec=be is not specified.
3650 2014-02-05  Richard Henderson  <rth@redhat.com>
3652         PR debug/52727
3653         * combine-stack-adj.c: Revert r206943.
3654         * sched-int.h (struct deps_desc): Add last_args_size.
3655         * sched-deps.c (init_deps): Initialize it.
3656         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3657         contain REG_ARGS_SIZE notes.
3659 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3661         * lto-cgraph.c (asm_nodes_output): Make global.
3662         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3663         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3664         (driver_handle_option): Handle OPT_fwpa.
3666 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3668         PR ipa/59947
3669         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3670         a comment typo and formatting issue.  If odr_hash hasn't been
3671         created, return vNULL and set *completep to false.
3673         PR middle-end/57499
3674         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3675         bb with no successors.
3677 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3679         PR target/59718
3680         * doc/invoke.texi (-march): Clarify documentation for ARM.
3681         (-mtune): Likewise.
3682         (-mcpu): Likewise.
3684 2014-02-05  Richard Biener  <rguenther@suse.de>
3686         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3687         when not vectorizing because of too many alias checks.
3688         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3689         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3691 2014-02-05  Nick Clifton  <nickc@redhat.com>
3693         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3694         accept extended registers in any mode when compiling for the MN10300.
3696 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3698         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3699         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3700         sanitization attributes.
3701         (can_inline_edge_p): Likewise.
3702         (sanitize_attrs_match_for_inline_p): New function.
3704 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3706         * ipa-prop.c (detect_type_change): Shor circuit testing of
3707         type changes on THIS pointer.
3709 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3711         PR target/59777
3712         * config/pa/pa.c (legitimize_tls_address): Return original address
3713         if not passed a SYMBOL_REF rtx.
3714         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3715         addresses.
3716         (pa_emit_move_sequence): Simplify TLS source operands.
3717         (pa_legitimate_constant_p): Reject all TLS constants.
3718         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3719         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3721 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3723         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3724         groups when we know they are controlled by LTO.
3725         * varasm.c (default_binds_local_p_1): If object is in other partition,
3726         it will be resolved locally.
3728 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3730         * config/host-linux.c (linux_gt_pch_use_address): Don't
3731         use SSIZE_MAX because it is not always defined.
3733 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3735         PR bootstrap/59913
3736         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3737         threshold for pseudo splitting.
3738         (update_ebb_live_info): Process call argument hard registers and
3739         hard registers from insn definition too.
3740         (max_small_class_regs_num): New constant.
3741         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3742         reloads_num only for small register classes.  Don't split for
3743         outputs of jumps.
3745 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3747         PR ipa/60058
3748         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3749         is non-null.
3751 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3753         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3754         visibility is safe.
3756 2014-02-04  Marek Polacek  <polacek@redhat.com>
3758         * gdbinit.in (pel): Define.
3760 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3762         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3763         behavior.
3765 2014-02-04  Richard Biener  <rguenther@suse.de>
3767         PR lto/59723
3768         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3769         in function context local.
3770         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3771         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3772         similar to LTO_imported_decl_ref.
3774 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3776         PR tree-optimization/60002
3777         * cgraphclones.c (build_function_decl_skip_args): Clear
3778         DECL_LANG_SPECIFIC.
3780         PR tree-optimization/60023
3781         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3782         false to gsi_replace.
3783         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3784         has been in some EH region and vec_stmt could throw, add
3785         vec_stmt into the same EH region.
3786         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3787         has no lhs, ignore it.
3788         * internal-fn.c (expand_MASK_LOAD): Likewise.
3790         PR ipa/60026
3791         * tree-inline.c (copy_forbidden): Fail for
3792         __attribute__((optimize (0))) functions.
3794         PR other/58712
3795         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3796         is set, copy one less argument.
3797         (expand_simd_clones): Don't subtract clone_info->inbranch
3798         from simd_clone_struct_alloc argument.
3800         PR rtl-optimization/57915
3801         * recog.c (simplify_while_replacing): If all unary/binary/relational
3802         operation arguments are constant, attempt to simplify those.
3804         PR middle-end/59261
3805         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3806         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3808 2014-02-04  Richard Biener  <rguenther@suse.de>
3810         PR tree-optimization/60012
3811         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3812         TBAA disambiguation to all DDRs.
3814 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3816         PR target/59788
3817         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3818         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3820 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3822         PR ipa/59882
3823         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3825 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3827         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3828         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3830 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3832         PR ipa/59831
3833         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3834         to figure out targets of polymorphic calls with known decl.
3835         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3836         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3837         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3838         (get_polymorphic_call_info): ... here.
3839         (get_polymorphic_call_info_from_invariant): New function.
3841 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3843         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3844         lookup via vtable pointer; check for type consistency
3845         and turn inconsitent facts into UNREACHABLE.
3846         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3847         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3848         type inconsistent querries; return UNREACHABLE instead.
3850 2014-02-03  Richard Henderson  <rth@twiddle.net>
3852         PR tree-opt/59924
3853         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3854         already processed this node.
3855         (normalize_one_pred_1): Pass along mark_set.
3856         (normalize_one_pred): Create and destroy a pointer_set_t.
3857         (normalize_one_pred_chain): Likewise.
3859 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3861         PR gcov-profile/58602
3862         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3864 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3866         PR ipa/59831
3867         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3868         -fno-devirtualize; try to devirtualize by the knowledge of
3869         virtual table pointer given by aggregate propagation.
3870         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3871         (ipa_print_node_jump_functions): Dump also offset that
3872         is relevant for polymorphic calls.
3873         (determine_known_aggregate_parts): Add arg_type parameter; use it
3874         instead of determining the type from pointer type.
3875         (ipa_compute_jump_functions_for_edge): Update call of
3876         determine_known_aggregate_parts.
3877         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3878         (gimple_get_virt_method_for_binfo): ... here; simplify using
3879         vtable_pointer_value_to_vtable.
3880         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3881         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3882         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3883         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3884         POINTER_PLUS_EXPR.
3885         (vtable_pointer_value_to_binfo): ... here.
3886         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3888 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3890         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3891         redef of outer loop index variable.
3893 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3895         PR c++/53017
3896         PR c++/59211
3897         * doc/extend.texi (Function Attributes): Typo.
3899 2014-02-03  Cong Hou  <congh@google.com>
3901         PR tree-optimization/60000
3902         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3903         if the vectorized statement is a store.  A store statement can only
3904         appear at the end of pattern statements.
3906 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3908         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3909         (ix86_option_override_internal): Default long double to 64-bit for
3910         32-bit Bionic and to 128-bit for 64-bit Bionic.
3912         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3913         TARGET_LONG_DOUBLE_128 is true.
3914         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3916         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3917         (mlong-double-64): Negate -mlong-double-128.
3918         (mlong-double-128): New option.
3920         * config/i386/i386-c.c (ix86_target_macros): Define
3921         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3923         * doc/invoke.texi: Document -mlong-double-128.
3925 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3927         PR rtl-optimization/60024
3928         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3930 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3932         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3934 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3936         PR rtl-optimization/57662
3937         * sel-sched.c (code_motion_path_driver): Do not mark already not
3938         existing blocks in the visiting bitmap.
3940 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3942         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3943         on the insn being emitted.
3945 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3946             Will Deacon  <will.deacon@arm.com>
3948         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3950 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3952         * config/arm/arm-tables.opt: Regenerate.
3954 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3956         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3957         for vector types other than V16QImode.
3958         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3959         define_expand, and call altivec_expand_vec_perm_le when producing
3960         code with little endian element order.
3961         (*altivec_vperm_<mode>_internal): New insn having previous
3962         behavior of altivec_vperm_<mode>.
3963         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3964         altivec_expand_vec_perm_le when producing code with little endian
3965         element order.
3966         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3967         behavior of altivec_vperm_<mode>_uns.
3969 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3971         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3972         (altivec_vsumsws): Add handling for -maltivec=be with a little
3973         endian target.
3974         (altivec_vsumsws_direct): New.
3975         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3976         gen_altivec_vsumsws.
3978 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3980         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3981         vtable_pointer_value_to_binfo): New functions.
3982         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3983         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3985 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3987         * config/nios2/nios2.md (load_got_register): Initialize GOT
3988         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3989         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3991 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3993         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3994         preserverd by passthrough, do not propagate the type.
3996 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3998         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3999         (mips_atomic_assign_expand_fenv): New function.
4000         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
4002 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
4004         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
4005         (__builtin_mips_set_fcsr): Likewise.
4006         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
4007         MIPS_USI_FTYPE_VOID.
4008         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
4009         (mips16_expand_set_fcsr): Likewise.
4010         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
4011         (mips16_set_fcsr_stub): Likewise.
4012         (mips16_get_fcsr_one_only_stub): New class.
4013         (mips16_set_fcsr_one_only_stub): Likewise.
4014         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
4015         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
4016         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
4017         (hard_float): New availability predicate.
4018         (mips_builtins): Add get_fcsr and set_fcsr.
4019         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
4020         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
4021         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
4022         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
4023         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
4024         patterns.
4026 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
4028         * config/mips/mips.c (mips_one_only_stub): New class.
4029         (mips_need_mips16_rdhwr_p): Replace with...
4030         (mips16_rdhwr_stub): ...this new variable.
4031         (mips16_stub_call_address): New function.
4032         (mips16_rdhwr_one_only_stub): New class.
4033         (mips_expand_thread_pointer): Use mips16_stub_call_address.
4034         (mips_output_mips16_rdhwr): Delete.
4035         (mips_finish_stub): New function.
4036         (mips_code_end): Use it to handle rdhwr stubs.
4038 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
4040         PR target/60017
4041         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
4042         when calculating size of integer atomic types.
4044 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
4046         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
4048 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
4050         PR tree-optimization/60003
4051         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
4052         * profile.c (branch_prob): Use gimple_call_builtin_p
4053         to check for BUILT_IN_SETJMP_RECEIVER.
4054         * tree-inline.c (copy_bb): Call notice_special_calls.
4056 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
4058         PR bootstrap/59985
4059         * lra-constraints.c (process_alt_operands): Update reload_sum only
4060         on the first pass.
4062 2014-01-31  Richard Henderson  <rth@redhat.com>
4064         PR middle-end/60004
4065         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
4066         until after else_eh is processed.
4068 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
4070         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
4071         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
4072         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
4073         in smmintrin.h, remove them.
4074         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
4075         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
4076         * config/i386/i386.md (ROUND_SAE): Fix value.
4077         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
4078         (const48_operand): New.
4079         * config/i386/subst.md (round), (round_expand): Use
4080         const_4_or_8_to_11_operand.
4081         (round_saeonly), (round_saeonly_expand): Use const48_operand.
4083 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
4085         * config/i386/constraints.md (Yk): Swap meaning with k.
4086         * config/i386/i386.md (movhi_internal): Change Yk to k.
4087         (movqi_internal): Ditto.
4088         (*k<logic><mode>): Ditto.
4089         (*andhi_1): Ditto.
4090         (*andqi_1): Ditto.
4091         (kandn<mode>): Ditto.
4092         (*<code>hi_1): Ditto.
4093         (*<code>qi_1): Ditto.
4094         (kxnor<mode>): Ditto.
4095         (kortestzhi): Ditto.
4096         (kortestchi): Ditto.
4097         (kunpckhi): Ditto.
4098         (*one_cmplhi2_1): Ditto.
4099         (*one_cmplqi2_1): Ditto.
4100         * config/i386/sse.md (): Change k to Yk.
4101         (avx512f_load<mode>_mask): Ditto.
4102         (avx512f_blendm<mode>): Ditto.
4103         (avx512f_store<mode>_mask): Ditto.
4104         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
4105         (avx512f_storedqu<mode>_mask): Ditto.
4106         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
4107         Ditto.
4108         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
4109         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
4110         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
4111         (avx512f_maskcmp<mode>3): Ditto.
4112         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
4113         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
4114         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
4115         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
4116         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
4117         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
4118         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
4119         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
4120         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
4121         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
4122         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
4123         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
4124         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
4125         (vec_extract_lo_<mode>_maskm): Ditto.
4126         (vec_extract_hi_<mode>_maskm): Ditto.
4127         (avx512f_vternlog<mode>_mask): Ditto.
4128         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
4129         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
4130         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
4131         (avx512f_<code>v8div16qi2_mask): Ditto.
4132         (avx512f_<code>v8div16qi2_mask_store): Ditto.
4133         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
4134         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
4135         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
4136         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
4137         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
4138         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4139         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
4140         (*avx512pf_scatterpf<mode>df_mask): Ditto.
4141         (avx512cd_maskb_vec_dupv8di): Ditto.
4142         (avx512cd_maskw_vec_dupv16si): Ditto.
4143         (avx512f_vpermi2var<mode>3_maskz): Ditto.
4144         (avx512f_vpermi2var<mode>3_mask): Ditto.
4145         (avx512f_vpermi2var<mode>3_mask): Ditto.
4146         (avx512f_vpermt2var<mode>3_maskz): Ditto.
4147         (*avx512f_gathersi<mode>): Ditto.
4148         (*avx512f_gathersi<mode>_2): Ditto.
4149         (*avx512f_gatherdi<mode>): Ditto.
4150         (*avx512f_gatherdi<mode>_2): Ditto.
4151         (*avx512f_scattersi<mode>): Ditto.
4152         (*avx512f_scatterdi<mode>): Ditto.
4153         (avx512f_compress<mode>_mask): Ditto.
4154         (avx512f_compressstore<mode>_mask): Ditto.
4155         (avx512f_expand<mode>_mask): Ditto.
4156         * config/i386/subst.md (mask): Change k to Yk.
4157         (mask_scalar_merge): Ditto.
4158         (sd): Ditto.
4160 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
4162         * doc/extend.texi (Vector Extensions): Document ?: in C++.
4164 2014-01-31  Richard Biener  <rguenther@suse.de>
4166         PR middle-end/59990
4167         * builtins.c (fold_builtin_memory_op): Make sure to not
4168         use a floating-point mode or a boolean or enumeral type for
4169         the copy operation.
4171 2014-01-30  DJ Delorie  <dj@redhat.com>
4173         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
4174         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
4175         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
4176         whenever main() has an epilogue.
4178 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4180         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
4181         unused variable "field".
4182         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
4183         (vsx_mergeh_<mode>): Likewise.
4184         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
4185         (altivec_vmrghh): Likewise.
4186         (altivec_vmrghw): Likewise.
4187         (altivec_vmrglb): Likewise.
4188         (altivec_vmrglh): Likewise.
4189         (altivec_vmrglw): Likewise.
4190         (altivec_vspltb): Add missing uses.
4191         (altivec_vsplth): Likewise.
4192         (altivec_vspltw): Likewise.
4193         (altivec_vspltsf): Likewise.
4195 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4197         PR target/59923
4198         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
4199         frame related instructions.
4201 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
4203         PR rtl-optimization/59959
4204         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
4205         any reload of register whose subreg is invalid.
4207 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4209         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
4210         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
4211         Add missing return type - void.
4213 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4215         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
4216         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
4217         remove element index adjustment for endian (now handled in vsx.md
4218         and altivec.md).
4219         (altivec_expand_vec_perm_const): Use
4220         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
4221         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
4222         (vsx_xxspltw_<mode>): Adjust element index for little endian.
4223         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
4224         define_expand and a new define_insn *altivec_vspltb_internal;
4225         adjust for -maltivec=be on a little endian target.
4226         (altivec_vspltb_direct): New.
4227         (altivec_vsplth): Divide into a define_expand and a new
4228         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
4229         little endian target.
4230         (altivec_vsplth_direct): New.
4231         (altivec_vspltw): Divide into a define_expand and a new
4232         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
4233         little endian target.
4234         (altivec_vspltw_direct): New.
4235         (altivec_vspltsf): Divide into a define_expand and a new
4236         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
4237         a little endian target.
4239 2014-01-30  Richard Biener  <rguenther@suse.de>
4241         PR tree-optimization/59993
4242         * tree-ssa-forwprop.c (associate_pointerplus): Check we
4243         can propagate form the earlier stmt and avoid the transform
4244         when the intermediate result is needed.
4246 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
4248         * README.Portability: Fix typo.
4250 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
4252         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
4253         comparison_operator with ordered_comparison_operator.
4255 2014-01-30  Nick Clifton  <nickc@redhat.com>
4257         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
4258         Rename to mn10300_store_multiple_regs.
4259         * config/mn10300/mn10300.c: Likewise.
4260         * config/mn10300/mn10300.md (store_movm): Fix typo: call
4261         store_multiple_regs.
4262         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
4263         Call mn10300_store_multiple_regs.
4265 2014-01-30  Nick Clifton  <nickc@redhat.com>
4266             DJ Delorie  <dj@redhat.com>
4268         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
4269         %fp 2 to keep registers after it properly word-aligned.
4270         (rl78_alloc_physical_registers_umul): Handle the case where both
4271         input operands are the same.
4273 2014-01-30  Richard Biener  <rguenther@suse.de>
4275         PR tree-optimization/59903
4276         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4277         check properly.
4279 2014-01-30  Jason Merrill  <jason@redhat.com>
4281         PR c++/59633
4282         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
4284         PR c++/59645
4285         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
4287 2014-01-30  Richard Biener  <rguenther@suse.de>
4289         PR tree-optimization/59951
4290         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
4292 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
4294         PR target/59784
4295         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
4296         SFmode to DFmode case.
4298 2014-01-29  DJ Delorie  <dj@redhat.com>
4300         * config/msp430/msp430.opt (-minrt): New.
4301         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
4302         if -minrt given.
4303         (ENDFILE_SPEC): Likewise.
4305 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
4307         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
4308         (estimate_function_body_sizes): Use it.
4310 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
4312         PR c++/58561
4313         * dwarf2out.c (is_cxx_auto): New.
4314         (is_base_type): Use it.
4315         (gen_type_die_with_usage): Likewise.
4317 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4319         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
4320         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
4321         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
4322         -maltivec=be with LE targets.
4323         (vsx_mergeh_<mode>): Likewise.
4324         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
4325         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
4326         (altivec_vmrghb): Replace with define_expand and new
4327         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
4328         (altivec_vmrghb_direct): New define_insn.
4329         (altivec_vmrghh): Replace with define_expand and new
4330         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
4331         (altivec_vmrghh_direct): New define_insn.
4332         (altivec_vmrghw): Replace with define_expand and new
4333         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
4334         (altivec_vmrghw_direct): New define_insn.
4335         (*altivec_vmrghsf): Adjust for endianness.
4336         (altivec_vmrglb): Replace with define_expand and new
4337         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
4338         (altivec_vmrglb_direct): New define_insn.
4339         (altivec_vmrglh): Replace with define_expand and new
4340         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
4341         (altivec_vmrglh_direct): New define_insn.
4342         (altivec_vmrglw): Replace with define_expand and new
4343         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
4344         (altivec_vmrglw_direct): New define_insn.
4345         (*altivec_vmrglsf): Adjust for endianness.
4346         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4347         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4348         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4349         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4350         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4351         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4352         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4353         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4355 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4357         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4358         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4359         whitespace.
4361 2014-01-29  Richard Biener  <rguenther@suse.de>
4363         PR tree-optimization/58742
4364         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4365         associate_pointerplus_align.
4366         (associate_pointerplus_diff): New function.
4367         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4368         and associate_pointerplus_diff.
4370 2014-01-29  Richard Biener  <rguenther@suse.de>
4372         * lto-streamer.h (LTO_major_version): Bump to 3.
4373         (LTO_minor_version): Reset to 0.
4375 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4377         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4378         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4379         (arm_file_start): Generate correct asm header for armv7ve.
4380         * config/arm/bpabi.h: Add multilib support for armv7ve.
4381         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4382         and cortex-a15 to armv7ve.
4383         * config/arm/t-aprofile: Add multilib support for armv7ve.
4384         * doc/invoke.texi: Document -march=armv7ve.
4386 2014-01-29  Richard Biener  <rguenther@suse.de>
4388         PR tree-optimization/58742
4389         * tree-ssa-forwprop.c (associate_plusminus): Return true
4390         if we changed sth, defer EH cleanup to ...
4391         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4392         (simplify_mult): New function.
4394 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4396         PR middle-end/59917
4397         PR tree-optimization/59920
4398         * tree.c (build_common_builtin_nodes): Remove
4399         __builtin_setjmp_dispatcher initialization.
4400         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4401         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4402         instead of gsi_after_labels + manually skipping debug stmts.
4403         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4404         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4405         * tree-inline.c (copy_edges_for_bb): Remove
4406         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4407         argument.  Ignore computed_goto_p stmts.  Don't call
4408         make_abnormal_goto_edges.  If a call might need abnormal edges
4409         for non-local gotos, see if it already has an edge to
4410         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4411         with true argument, don't do anything then, otherwise add
4412         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4413         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4414         caller.
4415         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4416         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4417         (lower_stmt): Don't set data->calls_builtin_setjmp.
4418         (lower_builtin_setjmp): Adjust comment.
4419         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4420         * tree-cfg.c (found_computed_goto): Remove.
4421         (factor_computed_gotos): Remove.
4422         (make_goto_expr_edges): Return bool, true for computed gotos.
4423         Don't call make_abnormal_goto_edges.
4424         (build_gimple_cfg): Don't set found_computed_goto, don't call
4425         factor_computed_gotos.
4426         (computed_goto_p): No longer static.
4427         (make_blocks): Don't set found_computed_goto.
4428         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4429         (make_edges): If make_goto_expr_edges returns true, push bb
4430         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4431         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4432         vector.  Record mapping between bbs and OpenMP regions if there
4433         are any, adjust make_gimple_omp_edges caller.  Call
4434         handle_abnormal_edges.
4435         (make_abnormal_goto_edges): Remove.
4436         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4437         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4438         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4439         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4440         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4441         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4442         filling *region also set *region_idx to (*region)->entry->index.
4444         PR other/58712
4445         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4446         For REGs set ORIGINAL_REGNO.
4448 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4450         * doc/md.texi: Mention that a target shouldn't implement
4451         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4452         than hi/lo pair.
4454 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4456         PR tree-optimization/59594
4457         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4458         a copy of the datarefs vector rather than the vector itself.
4460 2014-01-28  Jason Merrill  <jason@redhat.com>
4462         PR c++/53756
4463         * dwarf2out.c (auto_die): New static.
4464         (gen_type_die_with_usage): Handle C++1y 'auto'.
4465         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4466         on definition.
4468 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4470         PR target/59672
4471         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4472         (SPEC_X32): Likewise.
4473         (SPEC_64): Likewise.
4474         * config/i386/i386.c (ix86_option_override_internal): Turn off
4475         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4476         for TARGET_16BIT.
4477         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4478         * config/i386/i386.h (TARGET_16BIT): New macro.
4479         (TARGET_16BIT_P): Likewise.
4480         * config/i386/i386.opt: Add m16.
4481         * doc/invoke.texi: Document -m16.
4483 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4485         PR preprocessor/59935
4486         * input.c (location_get_source_line): Bail out on when line number
4487         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4489 2014-01-28  Richard Biener  <rguenther@suse.de>
4491         PR tree-optimization/58742
4492         * tree-ssa-forwprop.c (associate_plusminus): Handle
4493         pointer subtraction of the form (T)(P + A) - (T)P.
4495 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4497         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4498         at const_int_cost.
4500 2014-01-28  Richard Biener  <rguenther@suse.de>
4502         Revert
4503         2014-01-28  Richard Biener  <rguenther@suse.de>
4505         PR rtl-optimization/45364
4506         PR rtl-optimization/59890
4507         * var-tracking.c (local_get_addr_clear_given_value): Handle
4508         already cleared slot.
4509         (val_reset): Handle not allocated local_get_addr_cache.
4510         (vt_find_locations): Use post-order on the inverted CFG.
4512 2014-01-28  Richard Biener  <rguenther@suse.de>
4514         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4516 2014-01-28  Richard Biener  <rguenther@suse.de>
4518         PR rtl-optimization/45364
4519         PR rtl-optimization/59890
4520         * var-tracking.c (local_get_addr_clear_given_value): Handle
4521         already cleared slot.
4522         (val_reset): Handle not allocated local_get_addr_cache.
4523         (vt_find_locations): Use post-order on the inverted CFG.
4525 2014-01-28  Alan Modra  <amodra@gmail.com>
4527         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4528         * configure.ac <recursive call for build != host>: Define
4529         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4530         and LD_FOR_BUILD too.
4531         * configure: Regenerate.
4533 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4535         * config/i386/i386.c (get_builtin_code_for_version): Separate
4536         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4537         Broadwell from Haswell.
4539 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4541         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4542         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4543         * config/mips/mips.c (mips_option_override): Change setting
4544         of TARGET_DSP.
4545         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4546         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4547         Change from Mask to Var.
4549 2014-01-27  Jeff Law  <law@redhat.com>
4551         * ipa-inline.c (inline_small_functions): Fix typo.
4553 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4555         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4556         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4557         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4558         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4559         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4560         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4561         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4562         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4563         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4564         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4565         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4566         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4567         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4568         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4569         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4570         (_mm512_storeu_epi64): Ditto.
4571         (_mm512_cmpge_epi32_mask): Ditto.
4572         (_mm512_cmpge_epu32_mask): Ditto.
4573         (_mm512_cmpge_epi64_mask): Ditto.
4574         (_mm512_cmpge_epu64_mask): Ditto.
4575         (_mm512_cmple_epi32_mask): Ditto.
4576         (_mm512_cmple_epu32_mask): Ditto.
4577         (_mm512_cmple_epi64_mask): Ditto.
4578         (_mm512_cmple_epu64_mask): Ditto.
4579         (_mm512_cmplt_epi32_mask): Ditto.
4580         (_mm512_cmplt_epu32_mask): Ditto.
4581         (_mm512_cmplt_epi64_mask): Ditto.
4582         (_mm512_cmplt_epu64_mask): Ditto.
4583         (_mm512_cmpneq_epi32_mask): Ditto.
4584         (_mm512_cmpneq_epu32_mask): Ditto.
4585         (_mm512_cmpneq_epi64_mask): Ditto.
4586         (_mm512_cmpneq_epu64_mask): Ditto.
4587         (_mm512_expand_pd): Ditto.
4588         (_mm512_expand_ps): Ditto.
4589         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4590         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4591         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4592         * config/i386/i386.c (ix86_builtins): Add
4593         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4594         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4595         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4596         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4597         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4598         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4599         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4600         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4601         IX86_BUILTIN_PMOVUSQW512_MEM.
4602         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4603         __builtin_ia32_pmovsqd512mem_mask,
4604         __builtin_ia32_pmovqd512mem_mask,
4605         __builtin_ia32_pmovusqw512mem_mask,
4606         __builtin_ia32_pmovsqw512mem_mask,
4607         __builtin_ia32_pmovqw512mem_mask,
4608         __builtin_ia32_pmovusdw512mem_mask,
4609         __builtin_ia32_pmovsdw512mem_mask,
4610         __builtin_ia32_pmovdw512mem_mask,
4611         __builtin_ia32_pmovqb512mem_mask,
4612         __builtin_ia32_pmovusqb512mem_mask,
4613         __builtin_ia32_pmovsqb512mem_mask,
4614         __builtin_ia32_pmovusdb512mem_mask,
4615         __builtin_ia32_pmovsdb512mem_mask,
4616         __builtin_ia32_pmovdb512mem_mask.
4617         (bdesc_args): Add __builtin_ia32_expanddf512,
4618         __builtin_ia32_expandsf512.
4619         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4620         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4621         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4622         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4623         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4624         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4625         (avx512f_<code>v8div16qi2_mask_store): This.
4626         (avx512f_expand<mode>): New.
4628 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4630         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4631         New.
4632         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4633         (_mm512_prefetch_i32scatter_pd): Ditto.
4634         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4635         (_mm512_prefetch_i64scatter_pd): Ditto.
4636         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4637         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4638         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4639         (_mm512_prefetch_i32scatter_ps): Ditto.
4640         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4641         (_mm512_prefetch_i64scatter_ps): Ditto.
4642         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4643         * config/i386/i386-builtin-types.def: Define
4644         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4645         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4646         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4647         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4648         IX86_BUILTIN_SCATTERPFQPD.
4649         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4650         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4651         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4652         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4653         __builtin_ia32_scatterpfqps.
4654         (ix86_expand_builtin): Expand new built-ins.
4655         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4656         fix memory access data type.
4657         (*avx512pf_gatherpf<mode>_mask): Ditto.
4658         (*avx512pf_gatherpf<mode>): Ditto.
4659         (avx512pf_scatterpf<mode>): Ditto.
4660         (*avx512pf_scatterpf<mode>_mask): Ditto.
4661         (*avx512pf_scatterpf<mode>): Ditto.
4662         (GATHER_SCATTER_SF_MEM_MODE): New.
4663         (avx512pf_gatherpf<mode>df): Ditto.
4664         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4665         (*avx512pf_scatterpf<mode>df): Ditto.
4667 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4669         PR bootstrap/59934
4670         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4671         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4672         reached.
4674 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4676         * common/config/arm/arm-common.c
4677         (arm_rewrite_mcpu): Handle multiple names.
4678         * config/arm/arm.h
4679         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4681 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4683         * gimple-builder.h (create_gimple_tmp): Delete.
4685 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4687         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4688         words comparisons.
4690 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4692         * config/pa/pa.md (call): Generate indirect long calls to non-local
4693         functions when outputing 32-bit code.
4694         (call_value): Likewise except for special call to buggy powf function.
4696         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4697         portable runtime and PIC indirect calls.
4698         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4699         and PIC call sequences.  Use ldo instead of blr to set return register
4700         in PIC call sequence.
4702 2014-01-25  Walter Lee  <walt@tilera.com>
4704         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4705         avoid clobbering a live register.
4707 2014-01-25  Walter Lee  <walt@tilera.com>
4709         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4710         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4711         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4712         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4714 2014-01-25  Walter Lee  <walt@tilera.com>
4716         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4717         arguments on even registers.
4718         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4719         STACK_BOUNDARY.
4720         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4721         (BIGGEST_ALIGNMENT): Ditto.
4722         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4724 2014-01-25  Walter Lee  <walt@tilera.com>
4726         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4727         insns before bundling.
4728         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4730 2014-01-25  Walter Lee  <walt@tilera.com>
4732         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4733         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4734         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4736 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4738         * config/mips/constraints.md (kl): Delete.
4739         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4740         define expands, using...
4741         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4742         instructions for MIPS16.
4743         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4744         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4746 2014-01-25  Walter Lee  <walt@tilera.com>
4748         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4749         (clzdi2): Ditto.
4750         (ffsdi2): Ditto.
4752 2014-01-25  Walter Lee  <walt@tilera.com>
4754         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4755         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4757 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4759         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4760         Handle XOR.
4762 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4764         * print-rtl.c (in_call_function_usage): New var.
4765         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4766         EXPR_LIST mode as mode and not as reg note name.
4768         PR middle-end/59561
4769         * cfgloopmanip.c (copy_loop_info): If
4770         loop->warned_aggressive_loop_optimizations, make sure
4771         the flag is set in target loop too.
4773 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4775         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4776         flag_cilkplus.
4777         * builtins.def: Likewise.
4778         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4779         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4780         * ira.c (ira_setup_eliminable_regset): Likewise.
4781         * omp-low.c (gate_expand_omp): Likewise.
4782         (execute_lower_omp): Likewise.
4783         (diagnose_sb_0): Likewise.
4784         (gate_diagnose_omp_blocks): Likewise.
4785         (simd_clone_clauses_extract): Likewise.
4786         (gate): Likewise.
4788 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4790         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4791         correction for little endian...
4792         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4793         here.
4795 2014-01-24  Jeff Law  <law@redhat.com>
4797         PR tree-optimization/59919
4798         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4799         for non-returning calls.
4801 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4803         * common/config/aarch64/aarch64-common.c
4804         (aarch64_rewrite_mcpu): Handle multiple names.
4805         * config/aarch64/aarch64.h
4806         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4808 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4810         * input.c (add_file_to_cache_tab): Handle the case where fopen
4811         returns NULL.
4813 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4815         PR target/59929
4816         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4817         from push operand if code of push isn't PRE_DEC.
4819 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4821         PR target/59909
4822         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4823         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4824         it is only used for non-atomic loads/stores.
4826         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4827         -mquad-memory or -mquad-memory-atomic switches.
4829         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4830         -mquad-memory-atomic to ISA 2.07 support.
4832         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4833         to separate support of normal quad word memory operations (ldq, stq)
4834         from the atomic quad word memory operations.
4836         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4837         support to separate non-atomic quad word operations from atomic
4838         quad word operations.  Disable non-atomic quad word operations in
4839         little endian mode so that we don't have to swap words after the
4840         load and before the store.
4841         (quad_load_store_p): Add comment about atomic quad word support.
4842         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4843         options printed with -mdebug=reg.
4845         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4846         -mquad-memory-atomic as the test for whether we have quad word
4847         atomic instructions.
4848         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4849         or -mp8-vector are used, allow byte/half-word atomic operations.
4851         * config/rs6000/sync.md (load_lockedti): Insure that the address
4852         is a proper indexed or indirect address for the lqarx instruction.
4853         On little endian systems, swap the hi/lo registers after the lqarx
4854         instruction.
4855         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4856         insure the address is valid for the lqarx instruction.
4857         (store_conditionalti): Insure that the address is a proper indexed
4858         or indirect address for the stqcrx. instruction.  On little endian
4859         systems, swap the hi/lo registers before doing the stqcrx.
4860         instruction.
4861         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4862         insure the address is valid for the stqcrx. instruction.
4864         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4865         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4866         type of quad memory support is available.
4868 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4870         PR regression/59915
4871         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4872         there is a danger of looping.
4874 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4876         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4877         force flag_ira_loop_pressure if set via command line.
4879 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4881         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4882         (ashr_simd): New builtin handling DI mode.
4883         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4884         (aarch64_sshr_simddi): New match pattern.
4885         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4886         (vshrd_n_s64): Likewise.
4887         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4889 2014-01-23  Nick Clifton  <nickc@redhat.com>
4891         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4892         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4893         favour of mcu specific scripts.
4894         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4895         430x multilibs.
4897 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4898             Alex Velenko  <Alex.Velenko@arm.com>
4900         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4901         (vaddv_s16): Likewise.
4902         (vaddv_s32): Likewise.
4903         (vaddv_u8): Likewise.
4904         (vaddv_u16): Likewise.
4905         (vaddv_u32): Likewise.
4906         (vaddvq_s8): Likewise.
4907         (vaddvq_s16): Likewise.
4908         (vaddvq_s32): Likewise.
4909         (vaddvq_s64): Likewise.
4910         (vaddvq_u8): Likewise.
4911         (vaddvq_u16): Likewise.
4912         (vaddvq_u32): Likewise.
4913         (vaddvq_u64): Likewise.
4914         (vaddv_f32): Likewise.
4915         (vaddvq_f32): Likewise.
4916         (vaddvq_f64): Likewise.
4917         (vmaxv_f32): Likewise.
4918         (vmaxv_s8): Likewise.
4919         (vmaxv_s16): Likewise.
4920         (vmaxv_s32): Likewise.
4921         (vmaxv_u8): Likewise.
4922         (vmaxv_u16): Likewise.
4923         (vmaxv_u32): Likewise.
4924         (vmaxvq_f32): Likewise.
4925         (vmaxvq_f64): Likewise.
4926         (vmaxvq_s8): Likewise.
4927         (vmaxvq_s16): Likewise.
4928         (vmaxvq_s32): Likewise.
4929         (vmaxvq_u8): Likewise.
4930         (vmaxvq_u16): Likewise.
4931         (vmaxvq_u32): Likewise.
4932         (vmaxnmv_f32): Likewise.
4933         (vmaxnmvq_f32): Likewise.
4934         (vmaxnmvq_f64): Likewise.
4935         (vminv_f32): Likewise.
4936         (vminv_s8): Likewise.
4937         (vminv_s16): Likewise.
4938         (vminv_s32): Likewise.
4939         (vminv_u8): Likewise.
4940         (vminv_u16): Likewise.
4941         (vminv_u32): Likewise.
4942         (vminvq_f32): Likewise.
4943         (vminvq_f64): Likewise.
4944         (vminvq_s8): Likewise.
4945         (vminvq_s16): Likewise.
4946         (vminvq_s32): Likewise.
4947         (vminvq_u8): Likewise.
4948         (vminvq_u16): Likewise.
4949         (vminvq_u32): Likewise.
4950         (vminnmv_f32): Likewise.
4951         (vminnmvq_f32): Likewise.
4952         (vminnmvq_f64): Likewise.
4954 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4956         * config/aarch64/aarch64-simd.md
4957         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4958         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4959         (*aarch64_mul3_elt<mode>): Likewise.
4960         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4961         (*aarch64_mul3_elt_to_64v2df): Likewise.
4962         (*aarch64_mla_elt<mode>): Likewise.
4963         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4964         (*aarch64_mls_elt<mode>): Likewise.
4965         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4966         (*aarch64_fma4_elt<mode>): Likewise.
4967         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4968         (*aarch64_fma4_elt_to_64v2df): Likewise.
4969         (*aarch64_fnma4_elt<mode>): Likewise.
4970         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4971         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4972         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4973         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4974         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4975         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4976         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4977         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4978         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4980 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4982         * config/aarch64/aarch64-simd.md
4983         (aarch64_be_checked_get_lane<mode>): New define_expand.
4984         * config/aarch64/aarch64-simd-builtins.def
4985         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4986         New builtin definition.
4987         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4988         Use new safe be builtin.
4990 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4992         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4993         New define_insn.
4994         (aarch64_be_st1<mode>): Likewise.
4995         (aarch_ld1<VALL:mode>): Define_expand modified.
4996         (aarch_st1<VALL:mode>): Likewise.
4997         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4998         (UNSPEC_ST1): Likewise.
5000 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
5002         * config/microblaze/microblaze.md: Add trap insn and attribute
5004 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
5006         PR preprocessor/58580
5007         * input.h (location_get_source_line): Take an additional line_size
5008         parameter.
5009         (void diagnostics_file_cache_fini): Declare new function.
5010         * input.c (struct fcache): New type.
5011         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
5012         New static constants.
5013         (diagnostic_file_cache_init, total_lines_num)
5014         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
5015         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
5016         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
5017         (get_next_line, read_next_line, goto_next_line, read_line_num):
5018         New static function definitions.
5019         (diagnostic_file_cache_fini): New function.
5020         (location_get_source_line): Take an additional output line_len
5021         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
5022         read_line_num.
5023         * diagnostic.c (diagnostic_finish): Call
5024         diagnostic_file_cache_fini.
5025         (adjust_line): Take an additional input parameter for the length
5026         of the line, rather than calculating it with strlen.
5027         (diagnostic_show_locus): Adjust the use of
5028         location_get_source_line and adjust_line with respect to their new
5029         signature.  While displaying a line now, do not stop at the first
5030         null byte.  Rather, display the zero byte as a space and keep
5031         going until we reach the size of the line.
5032         * Makefile.in: Add vec.o to OBJS-libcommon
5034 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
5035             Ilya Tocar     <ilya.tocar@intel.com>
5037         * config/i386/avx512fintrin.h (_mm512_kmov): New.
5038         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
5039         (__builtin_ia32_kmov16): Ditto.
5040         * config/i386/i386.md (UNSPEC_KMOV): New.
5041         (kmovw): Ditto.
5043 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
5045         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
5046         (_mm512_storeu_si512): Ditto.
5048 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
5050         PR target/52125
5051         * rtl.h (get_referenced_operands): Declare.
5052         * recog.c (get_referenced_operands): New function.
5053         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
5054         operands have been referenced when recording LO_SUM references.
5056 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
5058         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
5060 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
5062         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
5063         Enable for generic and recent AMD targets.
5065 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
5067         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
5068         ARG_SIZE note when adjustment was eliminated.
5070 2014-01-22  Jeff Law  <law@redhat.com>
5072         PR tree-optimization/59597
5073         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
5074         in file.  Accept new argument REGISTERING and use it to modify
5075         dump output appropriately.
5076         (register_jump_thread): Corresponding changes.
5077         (mark_threaded_blocks): Reinstate code to cancel unprofitable
5078         thread paths involving joiner blocks.  Add code to dump cancelled
5079         jump threading paths.
5081 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
5083         PR rtl-optimization/59477
5084         * lra-constraints.c (inherit_in_ebb): Process call for living hard
5085         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
5087 2014-01-22  Tom Tromey  <tromey@redhat.com>
5089         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
5090         PARAMS.
5091         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
5093 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
5095         PR rtl-optimization/59896
5096         * lra-constraints.c (process_alt_operands): Check unused note for
5097         matched operands of insn with no output reloads.
5099 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
5101         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
5102         (mips_move_from_gpr_cost): Likewise.
5104 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
5106         PR rtl-optimization/59858
5107         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
5108         ira_class_hard_regs_num.
5109         (process_alt_operands): Increase reject for dying matched operand.
5111 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
5113         PR target/59003
5114         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
5115         smaller than size, perform several stores or loads and stores
5116         at dst + count - size to store or copy all of size bytes, rather
5117         than just last modesize bytes.
5119 2014-01-20  DJ Delorie  <dj@redhat.com>
5121         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
5122         that CLOBBERs are REGs before propogating their values.
5124 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
5126         PR middle-end/59789
5127         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
5128         (cgraph_inline_failed_type): New function.
5129         * cgraph.h (DEFCIFCODE): Add type.
5130         (cgraph_inline_failed_type_t): New enum.
5131         (cgraph_inline_failed_type): New prototype.
5132         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
5133         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
5134         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
5135         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
5136         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
5137         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
5138         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
5139         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
5140         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
5141         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
5142         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
5143         OPTIMIZATION_MISMATCH.
5144         * tree-inline.c (expand_call_inline): Emit errors during
5145         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
5147 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
5149         PR target/59685
5150         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
5151         mode attribute in insn output.
5153 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
5155         * output.h (output_constant): Delete.
5156         * varasm.c (output_constant): Make private.
5158 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
5160         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
5162 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5164         PR middle-end/59860
5165         * tree.h (fold_builtin_strcat): New prototype.
5166         * builtins.c (fold_builtin_strcat): No longer static.  Add len
5167         argument, if non-NULL, don't call c_strlen.  Optimize
5168         directly into __builtin_memcpy instead of __builtin_strcpy.
5169         (fold_builtin_2): Adjust fold_builtin_strcat caller.
5170         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
5172 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
5174         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5175         for SImode_address_operand operands, having only a REG argument.
5177 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5179         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
5180         loader name using mbig-endian.
5181         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
5183 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5185         * doc/invoke.texi (-march): Clarify documentation for AArch64.
5186         (-mtune): Likewise.
5187         (-mcpu): Likewise.
5189 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
5191         * config/aarch64/aarch64-protos.h
5192         (aarch64_cannot_change_mode_class_ptr): Declare.
5193         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
5194         aarch64_cannot_change_mode_class_ptr): New.
5195         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
5196         backend hook aarch64_cannot_change_mode_class.
5198 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5200         * common/config/aarch64/aarch64-common.c
5201         (aarch64_handle_option): Don't handle any option order logic here.
5202         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
5203         selected_cpu, warn on architecture version mismatch.
5204         (aarch64_override_options): Fix parsing order for option strings.
5206 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5207             Iain Sandoe  <iain@codesourcery.com>
5209         PR bootstrap/59496
5210         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
5211         warning.  Amend comment to reflect current functionality.
5213 2014-01-20  Richard Biener  <rguenther@suse.de>
5215         PR middle-end/59860
5216         * builtins.c (fold_builtin_strcat): Remove case better handled
5217         by tree-ssa-strlen.c.
5219 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
5221         * config/aarch64/aarch64.opt
5222         (mcpu, march, mtune): Make case-insensitive.
5224 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5226         PR target/59880
5227         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5228         if operands[1] is a REG or ZERO_EXTEND of a REG.
5230 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
5232         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
5234 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
5236         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
5237         long non-pic millicode calls.
5239 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5241         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
5243 2014-01-19  Kito Cheng  <kito@0xlab.org>
5245         * builtins.c (expand_movstr): Check movstr expand done or fail.
5247 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5248             H.J. Lu  <hongjiu.lu@intel.com>
5250         PR target/59379
5251         * config/i386/i386.md (*lea<mode>): Zero-extend return register
5252         to DImode for zero-extended addresses.
5254 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
5256         PR rtl-optimization/57763
5257         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
5258         on the new indirect jump_insn and increment LABEL_NUSES (label).
5260 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
5262         PR bootstrap/59580
5263         PR bootstrap/59583
5264         * config.gcc (x86_archs): New variable.
5265         (x86_64_archs): Likewise.
5266         (x86_cpus): Likewise.
5267         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
5268         --with-arch/--with-cpu= options.
5269         Support --with-arch=/--with-cpu={nehalem,westmere,
5270         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
5272 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5274         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
5275         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
5277 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5279         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
5281 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
5283         PR target/58944
5284         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
5285         clear cpp_get_options (parse_in)->warn_unused_macros for
5286         ix86_target_macros_internal with cpp_define.
5288 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
5290         * jump.c (delete_related_insns): Keep (use (insn))s.
5291         * reorg.c (redundant_insn): Check for barriers too.
5293 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5295         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
5297 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
5299         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
5300         call to $$dyncall when TARGET_LONG_CALLS is true.
5302 2014-01-17  Jeff Law  <law@redhat.com>
5304         * ree.c (combine_set_extension): Temporarily disable test for
5305         changing number of hard registers.
5307 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5309         PR middle-end/58125
5310         * ipa-inline-analysis.c (inline_free_summary):
5311         Do not free summary of aliases.
5313 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5315         PR middle-end/59706
5316         * gimplify.c (gimplify_expr): Use create_tmp_var
5317         instead of create_tmp_var_raw.  If cond doesn't have
5318         integral type, don't add the IFN_ANNOTATE builtin at all.
5320 2014-01-17  Martin Jambor  <mjambor@suse.cz>
5322         PR ipa/59736
5323         * ipa-cp.c (prev_edge_clone): New variable.
5324         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
5325         Also resize prev_edge_clone vector.
5326         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
5327         (ipcp_edge_removal_hook): New function.
5328         (ipcp_driver): Register ipcp_edge_removal_hook.
5330 2014-01-17  Andrew Pinski  <apinski@cavium.com>
5331             Steve Ellcey  <sellcey@mips.com>
5333         PR target/59462
5334         * config/mips/mips.c (mips_print_operand): Check operand mode instead
5335         of operator mode.
5337 2014-01-17  Jeff Law  <law@redhat.com>
5339         PR middle-end/57904
5340         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
5341         so that pass_ccp runs first.
5343 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5345         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5346         (ix86_adjust_cost): Use !TARGET_XXX.
5347         (do_reorder_for_imul): Likewise.
5348         (swap_top_of_ready_list): Likewise.
5349         (ix86_sched_reorder): Likewise.
5351 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5353         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5354         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5355         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5356         (intel_memset): New.  Duplicate slm_memset.
5357         (intel_cost): New.  Duplicate slm_cost.
5358         (m_INTEL): New macro.
5359         (processor_target_table): Add "intel".
5360         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5361         with PROCESSOR_INTEL for "intel".
5362         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5363         PROCESSOR_SILVERMONT.
5364         (ix86_issue_rate): Likewise.
5365         (ix86_adjust_cost): Likewise.
5366         (ia32_multipass_dfa_lookahead): Likewise.
5367         (swap_top_of_ready_list): Likewise.
5368         (ix86_sched_reorder): Likewise.
5369         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5370         instead of TARGET_OPT_AGU.
5371         * config/i386/i386.h (TARGET_INTEL): New.
5372         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5373         (processor_type): Add PROCESSOR_INTEL.
5374         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5375         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5377 2014-01-17  Marek Polacek  <polacek@redhat.com>
5379         PR c/58346
5380         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5381         size is zero.
5383 2014-01-17  Richard Biener  <rguenther@suse.de>
5385         PR tree-optimization/46590
5386         * opts.c (default_options_table): Add entries for
5387         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5388         all enabled at -O1 but not for -Og.
5389         * common.opt (fbranch-count-reg): Remove Init(1).
5390         (fmove-loop-invariants): Likewise.
5391         (ftree-pta): Likewise.
5393 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5395         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5396         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5397         decls to at least the GCC 4.8 used alignments.
5399         PR fortran/59440
5400         * tree-nested.c (convert_nonlocal_reference_stmt,
5401         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5402         of GIMPLE_BIND stmts, adjust associated decls.
5404 2014-01-17  Richard Biener  <rguenther@suse.de>
5406         PR tree-optimization/46590
5407         * vec.h (vec<>::bseach): New member function implementing
5408         binary search according to C89 bsearch.
5409         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5410         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5411         bitmap pointer again.  Make accesses_in_loop a flat array.
5412         (mem_ref_obstack): New global.
5413         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5414         (mark_ref_stored): Likewise.
5415         (ref_indep_loop_p_2): Likewise.
5416         (set_ref_stored_in_loop): New helper function.
5417         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5418         (memref_free): Adjust.
5419         (record_mem_ref_loc): Simplify.
5420         (gather_mem_refs_stmt): Adjust.
5421         (sort_locs_in_loop_postorder_cmp): New function.
5422         (analyze_memory_references): Sort accesses_in_loop after
5423         loop postorder number.
5424         (find_ref_loc_in_loop_cmp): New function.
5425         (for_all_locs_in_loop): Find relevant cluster of locs in
5426         accesses_in_loop and iterate without recursion.
5427         (execute_sm): Avoid uninit warning.
5428         (struct ref_always_accessed): Simplify.
5429         (ref_always_accessed::operator ()): Likewise.
5430         (ref_always_accessed_p): Likewise.
5431         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5432         loop postorder numbers here.
5433         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5434         numbers.
5436 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5438         PR c++/57945
5439         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5440         on decls for which assemble_alias has been called.
5442 2014-01-17  Nick Clifton  <nickc@redhat.com>
5444         * config/msp430/msp430.opt: (mcpu): New option.
5445         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5446         (msp430_option_override): Parse target_cpu.  If the MCU name
5447         matches a generic string, clear target_mcu.
5448         (msp430_attr): Allow numeric interrupt values up to 63.
5449         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5450         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5451         option.
5452         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5453         Add mcpu matches.
5454         * config/msp430/msp430.md (popm): Use %J rather than %I.
5455         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5456         (addhi_cy_i): Use immediate_operand for operand 2.
5457         * doc/invoke.texi: Document -mcpu option.
5459 2014-01-17  Richard Biener  <rguenther@suse.de>
5461         PR rtl-optimization/38518
5462         * df.h (df_analyze_loop): Declare.
5463         * df-core.c: Include cfgloop.h.
5464         (df_analyze_1): Split out main part of df_analyze.
5465         (df_analyze): Adjust.
5466         (loop_inverted_post_order_compute): New function.
5467         (loop_post_order_compute): Likewise.
5468         (df_analyze_loop): New function avoiding whole-function
5469         postorder computes.
5470         * loop-invariant.c (find_defs): Use df_analyze_loop.
5471         (find_invariants): Adjust.
5472         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5474 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5476         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5477         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5479 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5481         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5482         traversal when removing references.
5484 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5486         PR ipa/59775
5487         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5489 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5491         PR middle-end/56791
5492         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5493         pushing a reload for an autoinc when we had previously reloaded an
5494         inner part of the address.
5496 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5498         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5499         field.
5500         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5501         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5502         when not giving up or versioning for alias only because of
5503         loop->safelen.
5504         (vect_analyze_data_ref_dependences): Set to true.
5505         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5506         is a GIMPLE_PHI.
5507         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5508         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5509         to the condition.
5511         PR middle-end/58344
5512         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5514         PR target/59839
5515         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5516         operand 0 predicate for gathers, use a new pseudo as subtarget.
5518 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5520         PR middle-end/59609
5521         * lra-constraints.c (process_alt_operands): Add printing debug info.
5522         Check absence of input/output reloads for matched operands too.
5524 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5526         PR rtl-optimization/59835
5527         * ira.c (ira_init_register_move_cost): Increase cost for
5528         impossible modes.
5530 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5532         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5534 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5536         PR target/59780
5537         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5538         non-register objects.  Use gen_(high/low)part more consistently.
5539         Fix assertions.
5541 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5543         PR target/59844
5544         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5545         endian support, remove tests for WORDS_BIG_ENDIAN.
5546         (p8_mfvsrd_3_<mode>): Likewise.
5547         (reload_gpr_from_vsx<mode>): Likewise.
5548         (reload_gpr_from_vsxsf): Likewise.
5549         (p8_mfvsrd_4_disf): Likewise.
5551 2014-01-16  Richard Biener  <rguenther@suse.de>
5553         PR rtl-optimization/46590
5554         * lcm.c (compute_antinout_edge): Use postorder iteration.
5555         (compute_laterin): Use inverted postorder iteration.
5557 2014-01-16  Nick Clifton  <nickc@redhat.com>
5559         PR middle-end/28865
5560         * varasm.c (output_constant): Return the number of bytes actually
5561         emitted.
5562         (output_constructor_array_range): Update the field size with the
5563         number of bytes emitted by output_constant.
5564         (output_constructor_regular_field): Likewise.  Also do not
5565         complain if the total number of bytes emitted is now greater
5566         than the expected fieldpos.
5567         * output.h (output_constant): Update prototype and descriptive comment.
5569 2014-01-16  Marek Polacek  <polacek@redhat.com>
5571         PR middle-end/59827
5572         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5573         it is error_mark_node.
5575 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5577         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5578         VALID_AVX256_REG_OR_OI_MODE.
5580 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5582         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5583         current procedure should be profiled.
5585 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5587         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5588         of moving from/to the STACK_REG register class.
5590 2014-01-15  Richard Henderson  <rth@redhat.com>
5592         PR debug/54694
5593         * reginfo.c (global_regs_decl): Globalize.
5594         * rtl.h (global_regs_decl): Declare.
5595         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5596         reserved via global_regs.
5598 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5600         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5602 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5604         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5605         and vmulosh rather than call gen_vec_widen_smult_*.
5606         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5607         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5608         (vec_widen_smult_even_v16qi): Likewise.
5609         (vec_widen_umult_even_v8hi): Likewise.
5610         (vec_widen_smult_even_v8hi): Likewise.
5611         (vec_widen_umult_odd_v16qi): Likewise.
5612         (vec_widen_smult_odd_v16qi): Likewise.
5613         (vec_widen_umult_odd_v8hi): Likewise.
5614         (vec_widen_smult_odd_v8hi): Likewise.
5615         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5616         vmuloub rather than call gen_vec_widen_umult_*.
5617         (vec_widen_umult_lo_v16qi): Likewise.
5618         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5619         vmulosb rather than call gen_vec_widen_smult_*.
5620         (vec_widen_smult_lo_v16qi): Likewise.
5621         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5622         rather than call gen_vec_widen_umult_*.
5623         (vec_widen_umult_lo_v8hi): Likewise.
5624         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5625         rather than call gen_vec_widen_smult_*.
5626         (vec_widen_smult_lo_v8hi): Likewise.
5628 2014-01-15  Jeff Law  <law@redhat.com>
5630         PR tree-optimization/59747
5631         * ree.c (find_and_remove_re): Properly handle case where a second
5632         eliminated extension requires widening a copy created for elimination
5633         of a prior extension.
5634         (combine_set_extension): Ensure that the number of hard regs needed
5635         for a destination register does not change when we widen it.
5637 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5639         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5640         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5641         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5642         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5643         (avr-*-rtems*): Likewise.
5644         (bfin*-rtems*): Likewise.
5645         (moxie-*-rtems*): Likewise.
5646         (h8300-*-rtems*): Likewise.
5647         (i[34567]86-*-rtems*): Likewise.
5648         (lm32-*-rtems*): Likewise.
5649         (m32r-*-rtems*): Likewise.
5650         (m68k-*-rtems*): Likewise.
5651         (microblaze*-*-rtems*): Likewise.
5652         (mips*-*-rtems*): Likewise.
5653         (powerpc-*-rtems*): Likewise.
5654         (sh-*-rtems*): Likewise.
5655         (sparc-*-rtems*): Likewise.
5656         (sparc64-*-rtems*): Likewise.
5657         (v850-*-rtems*): Likewise.
5658         (m32c-*-rtems*): Likewise.
5660 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5662         PR rtl-optimization/59511
5663         * ira.c (ira_init_register_move_cost): Use memory costs for some
5664         cases of register move cost calculations.
5665         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5666         instead of BB frequency.
5667         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5668         * lra-assigns.c (find_hard_regno_for): Ditto.
5670 2014-01-15  Richard Biener  <rguenther@suse.de>
5672         PR tree-optimization/59822
5673         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5674         (vectorizable_load): Use it to hoist defs of uses of invariant
5675         loads out of the loop.
5677 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5678             Kugan Vivekanandarajah  <kuganv@linaro.org>
5680         PR target/59695
5681         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5682         truncation.
5684 2014-01-15  Richard Biener  <rguenther@suse.de>
5686         PR rtl-optimization/59802
5687         * lcm.c (compute_available): Use inverted postorder to seed
5688         the initial worklist.
5690 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5692         PR target/59803
5693         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5694         ADDR_REGS for invalid symrefs in non-PIC code.
5696 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5698         PR other/58712
5699         * builtins.c (determine_block_size): Initialize *probable_max_size
5700         even if len_rtx is CONST_INT.
5702 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5704         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5705         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5706         (cortexa53_tunings): Likewise.
5707         (aarch64_sched_issue_rate): New function.
5708         (TARGET_SCHED_ISSUE_RATE): Define.
5710 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5712         * ira-costs.c (find_costs_and_classes): Add missed
5713         ira_init_register_move_cost_if_necessary.
5715 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5717         PR target/59787
5718         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5720 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5722         PR target/59794
5723         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5724         to indicate if type is used for function return value.  Warn ABI
5725         change if the vector mode isn't available for function return value.
5726         (ix86_function_arg_advance): Pass false to type_natural_mode.
5727         (ix86_function_arg): Likewise.
5728         (ix86_gimplify_va_arg): Likewise.
5729         (function_arg_32): Don't warn ABI change.
5730         (ix86_function_value): Pass true to type_natural_mode.
5731         (ix86_return_in_memory): Likewise.
5732         (ix86_struct_value_rtx): Removed.
5733         (TARGET_STRUCT_VALUE_RTX): Likewise.
5735 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5737         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5738         converting a conditional jump into a conditional return.
5740 2014-01-14  Richard Biener  <rguenther@suse.de>
5742         PR tree-optimization/58921
5743         PR tree-optimization/59006
5744         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5745         hoisting invariant stmts.
5746         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5747         invariant loads on the preheader edge if possible.
5749 2014-01-14  Joey Ye  <joey.ye@arm.com>
5751         * doc/plugin.texi (Building GCC plugins): Update to C++.
5753 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5755         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5756         (_mm_rcp28_round_ss): Ditto.
5757         (_mm_rsqrt28_round_sd): Ditto.
5758         (_mm_rsqrt28_round_ss): Ditto.
5759         (_mm_rcp28_sd): Ditto.
5760         (_mm_rcp28_ss): Ditto.
5761         (_mm_rsqrt28_sd): Ditto.
5762         (_mm_rsqrt28_ss): Ditto.
5763         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5764         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5765         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5766         (IX86_BUILTIN_RCP28SD): Ditto.
5767         (IX86_BUILTIN_RCP28SS): Ditto.
5768         (IX86_BUILTIN_RSQRT28SD): Ditto.
5769         (IX86_BUILTIN_RSQRT28SS): Ditto.
5770         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5771         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5772         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5773         (ix86_expand_special_args_builtin): Expand new FTYPE.
5774         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5775         (srcp14<mode>): Make insn unary.
5776         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5777         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5778         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5779         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5780         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5781         Fix rounding: make it SAE only.
5782         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5783         Ditto.
5784         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5785         Ditto.
5786         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5787         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5788         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5789         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5790         (round_saeonly_mask_scalar_operand4): Ditto.
5791         (round_saeonly_mask_scalar_op3): Ditto.
5792         (round_saeonly_mask_scalar_op4): Ditto.
5794 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5796         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5797         Implement -maltivec=be for vec_insert and vec_extract.
5799 2014-01-10  DJ Delorie  <dj@redhat.com>
5801         * config/msp430/msp430.md (call_internal): Don't allow memory
5802         references with SP as the base register.
5803         (call_value_internal): Likewise.
5804         * config/msp430/constraints.md (Yc): New.  For memory references
5805         that don't use SP as a base register.
5807         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5808         "an integer without a # prefix"
5809         * config/msp430/msp430.md (epilogue_helper): Use it.
5811 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5813         PR target/59617
5814         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5815         AVX512F gather builtins.
5816         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5817         on gather decls with INTEGER_TYPE masktype.
5818         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5819         directly into the builtin rather than hoisting it before loop.
5821         PR tree-optimization/59387
5822         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5823         (scev_const_prop): If folded_casts and type has undefined overflow,
5824         use force_gimple_operand instead of force_gimple_operand_gsi and
5825         for each added stmt if it is assign with
5826         arith_code_with_undefined_signed_overflow, call
5827         rewrite_to_defined_overflow.
5828         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5829         gimple-fold.h instead.
5830         (arith_code_with_undefined_signed_overflow,
5831         rewrite_to_defined_overflow): Moved to ...
5832         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5833         rewrite_to_defined_overflow): ... here.  No longer static.
5834         Include gimplify-me.h.
5835         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5836         rewrite_to_defined_overflow): New prototypes.
5838 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5840         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5842 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5844         * builtins.c (get_object_alignment_2): Minor tweak.
5845         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5847 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5849         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5850         optimized non constant lengths.
5852 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5854         PR libgomp/59194
5855         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5856         load as __atomic_load_N if possible.
5858 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5860         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5861         target parameter.
5862         (rs6000_expand_builtin): Adjust call.
5864 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5866         PR target/58115
5867         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5868         * config/rs6000/rs6000.c: Include target-globals.h.
5869         (rs6000_set_current_function): Instead of doing target_reinit
5870         unconditionally, use save_target_globals_default_opts and
5871         restore_target_globals.
5873         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5874         FPSCR.
5875         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5876         (rs6000_expand_builtin): Handle mffs and mtfsf.
5877         (rs6000_init_builtins): Define mffs and mtfsf.
5878         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5879         (rs6000_mffs): New pattern.
5880         (rs6000_mtfsf): New pattern.
5882 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5884         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5885         Start narrowing with START.  Apply candidate-use pair
5886         and check overall cost in narrowing.
5887         (iv_ca_prune): Pass new argument.
5889 2014-01-10  Jeff Law  <law@redhat.com>
5891         PR middle-end/59743
5892         * ree.c (combine_reaching_defs): Ensure the defining statement
5893         occurs before the extension when optimizing extensions with
5894         different source and destination hard registers.
5896 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5898         PR ipa/58585
5899         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5900         vtables into the type inheritance graph.
5902 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5904         PR rtl-optimization/59754
5905         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5906         modes in the REGNO != REGNO case.
5908 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5910         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5912 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5914         PR tree-optimization/59745
5915         * tree-predcom.c (tree_predictive_commoning_loop): Call
5916         free_affine_expand_cache if giving up because components is NULL.
5918         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5919         GC in payload of target_globals struct instead of allocating them on
5920         the heap and the larger structs separately using GC.
5921         * target-globals.h (struct target_globals): Make regs, hard_regs,
5922         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5923         of GTY((skip)) and change type to void *.
5924         (reset_target_globals): Cast loads from those fields to corresponding
5925         types.
5927 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5929         PR plugins/59335
5930         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5931         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5932         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5934 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5936         PR target/59744
5937         * aarch64-modes.def (CC_Zmode): New flags mode.
5938         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5939         represents an equality.
5940         (aarch64_get_condition_code): Handle CC_Zmode.
5941         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5943 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5945         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5946         extraction in good case.
5948 2014-01-10  Richard Biener  <rguenther@suse.de>
5950         PR tree-optimization/59374
5951         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5952         checking after SLP discovery.  Mark stmts not participating
5953         in any SLP instance properly.
5955 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5957         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5958         when handling a SET rtx.
5960 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5962         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5963         (cortex-a57): Likewise.
5964         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5966 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5968         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5969         non-iwmmxt builtins.
5971 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5973         PR ipa/58252
5974         PR ipa/59226
5975         * ipa-devirt.c record_target_from_binfo): Take as argument
5976         stack of binfos and lookup matching one for virtual inheritance.
5977         (possible_polymorphic_call_targets_1): Update.
5979 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5981         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5982         kernel strings for Loongson-2E/2F/3A.
5984 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5986         PR middle-end/59670
5987         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5988         is_gimple_call before calling gimple_call_internal_p.
5990 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5992         * Makefile.in (TREE_FLOW_H): Remove.
5993         (TREE_SSA_H): Add file names from tree-flow.h.
5994         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5995         * tree.h: Remove tree-flow.h reference.
5996         * hash-table.h: Remove tree-flow.h reference.
5997         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5998         reference with tree-ssa-loop.h.
6000 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6002         * doc/invoke.texi: Add -maltivec={be,le} options, and document
6003         default element-order behavior for -maltivec.
6004         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
6005         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
6006         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
6007         when targeting big endian, at least for now.
6008         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
6010 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6012         PR middle-end/47735
6013         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
6014         var satisfies use_register_for_decl, just take into account type
6015         alignment, rather than decl alignment.
6017         PR tree-optimization/59622
6018         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
6019         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
6020         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
6021         Don't devirtualize for inplace at all.  For targets.length () == 1,
6022         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
6024 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
6026         * config/i386/i386.md (cpu): Remove the unused btver1.
6028 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
6030         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
6032 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6034         PR target/58115
6035         * tree-core.h (struct target_globals): New forward declaration.
6036         (struct tree_target_option): Add globals field.
6037         * tree.h (TREE_TARGET_GLOBALS): Define.
6038         (prepare_target_option_nodes_for_pch): New prototype.
6039         * target-globals.h (struct target_globals): Define even if
6040         !SWITCHABLE_TARGET.
6041         * tree.c (prepare_target_option_node_for_pch,
6042         prepare_target_option_nodes_for_pch): New functions.
6043         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
6044         * config/i386/i386.c: Include target-globals.h.
6045         (ix86_set_current_function): Instead of doing target_reinit
6046         unconditionally, use save_target_globals_default_opts and
6047         restore_target_globals.
6049 2014-01-09  Richard Biener  <rguenther@suse.de>
6051         PR tree-optimization/59715
6052         * tree-cfg.h (split_critical_edges): Declare.
6053         * tree-cfg.c (split_critical_edges): Export.
6054         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
6056 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
6058         * cfgexpand.c (expand_stack_vars): Optionally disable
6059         asan stack protection.
6060         (expand_used_vars): Likewise.
6061         (partition_stack_vars): Likewise.
6062         * asan.c (asan_emit_stack_protection): Optionally disable
6063         after return stack usage.
6064         (instrument_derefs): Optionally disable memory access instrumentation.
6065         (instrument_builtin_call): Likewise.
6066         (instrument_strlen_call): Likewise.
6067         (asan_protect_global): Optionally disable global variables protection.
6068         * doc/invoke.texi: Added doc for new options.
6069         * params.def: Added new options.
6070         * params.h: Likewise.
6072 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
6074         PR rtl-optimization/59724
6075         * ifcvt.c (cond_exec_process_if_block): Don't call
6076         flow_find_head_matching_sequence with 0 longest_match.
6077         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
6078         non-active insns if !stop_after.
6079         (try_head_merge_bb): Revert 2014-01-07 changes.
6081 2014-01-08  Jeff Law  <law@redhat.com>
6083         * ree.c (get_sub_rtx): New function, extracted from...
6084         (merge_def_and_ext): Here.
6085         (combine_reaching_defs): Use get_sub_rtx.
6087 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
6089         * cgraph.h (varpool_variable_node): Do not choke on null node.
6091 2014-01-08  Catherine Moore  <clm@codesourcery.com>
6093         * config/mips/mips.md (simple_return): Attempt to use JRC
6094         for microMIPS.
6095         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
6097 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
6099         PR rtl-optimization/59137
6100         * reorg.c (steal_delay_list_from_target): Call update_block for
6101         elided insns.
6102         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
6104 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
6106         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6107         two duplicate entries.
6109 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
6111         Revert:
6112         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
6114         * config/mips/mips.c (mips_truncated_op_cost): New function.
6115         (mips_rtx_costs): Adjust test for BADDU.
6116         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
6118         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
6120         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
6121         (*baddu_si): ...this new pattern.
6123 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
6125         PR ipa/59722
6126         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
6128 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6130         PR middle-end/57748
6131         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
6132         inner_reference_p.
6133         (expand_expr, expand_normal): Adjust.
6134         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
6135         inner_reference_p. Use inner_reference_p to expand inner references.
6136         (store_expr): Adjust.
6137         * cfgexpand.c (expand_call_stmt): Adjust.
6139 2014-01-08  Rong Xu  <xur@google.com>
6141         * gcov-io.c (gcov_var): Move from gcov-io.h.
6142         (gcov_position): Ditto.
6143         (gcov_is_error): Ditto.
6144         (gcov_rewrite): Ditto.
6145         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
6146         only part to libgcc/libgcov.h.
6148 2014-01-08  Marek Polacek  <polacek@redhat.com>
6150         PR middle-end/59669
6151         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
6153 2014-01-08  Marek Polacek  <polacek@redhat.com>
6155         PR sanitizer/59667
6156         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
6158 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
6160         PR rtl-optimization/59649
6161         * stor-layout.c (get_mode_bounds): For BImode return
6162         0 and STORE_FLAG_VALUE.
6164 2014-01-08  Richard Biener  <rguenther@suse.de>
6166         PR middle-end/59630
6167         * gimple.h (is_gimple_builtin_call): Remove.
6168         (gimple_builtin_call_types_compatible_p): New.
6169         (gimple_call_builtin_p): New overload.
6170         * gimple.c (is_gimple_builtin_call): Remove.
6171         (validate_call): Rename to ...
6172         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
6173         check return types.
6174         (validate_type): New static function.
6175         (gimple_call_builtin_p): New overload and adjust.
6176         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
6177         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
6178         (gimple_fold_stmt_to_constant_1): Likewise.
6179         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
6181 2014-01-08  Richard Biener  <rguenther@suse.de>
6183         PR middle-end/59471
6184         * gimplify.c (gimplify_expr): Gimplify register-register type
6185         VIEW_CONVERT_EXPRs to separate stmts.
6187 2014-01-07  Jeff Law  <law@redhat.com>
6189         PR middle-end/53623
6190         * ree.c (combine_set_extension): Handle case where source
6191         and destination registers in an extension insn are different.
6192         (combine_reaching_defs): Allow source and destination registers
6193         in extension to be different under limited circumstances.
6194         (add_removable_extension): Remove restriction that the
6195         source and destination registers in the extension are the same.
6196         (find_and_remove_re): Emit a copy from the extension's
6197         destination to its source after the defining insn if
6198         the source and destination registers are different.
6200         PR middle-end/59285
6201         * ifcvt.c (merge_if_block): If we are merging a block with more than
6202         one successor with a block with no successors, remove any BARRIER
6203         after the second block.
6205 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
6207         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
6209 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
6211         PR target/59652
6212         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
6213         for 14-bit register offsets when INT14_OK_STRICT is false.
6215 2014-01-07  Roland Stigge  <stigge@antcom.de>
6216             Michael Meissner  <meissner@linux.vnet.ibm.com>
6218         PR 57386/target
6219         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
6220         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
6222 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
6224         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
6225         -mcpu.
6227 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
6229         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
6230         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
6231         rtx is const0_rtx or not.
6233 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
6235         PR target/58115
6236         * target-globals.c (save_target_globals): Remove this_fn_optab
6237         handling.
6238         * toplev.c: Include optabs.h.
6239         (target_reinit): Temporarily restore the global options if another
6240         set of options are in force.
6242 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
6244         PR rtl-optimization/58668
6245         * cfgcleanup.c (flow_find_cross_jump): Don't count
6246         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
6247         to determine what is counted.
6248         (flow_find_head_matching_sequence): Use active_insn_p to determine
6249         what is counted.
6250         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
6251         counting change.
6252         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
6253         determine what is counted.
6255         PR tree-optimization/59643
6256         * tree-predcom.c (split_data_refs_to_components): If one dr is
6257         read and one write, determine_offset fails and the write isn't
6258         in the bad component, just put the read into the bad component.
6260 2014-01-07  Mike Stump  <mikestump@comcast.net>
6261             Jakub Jelinek  <jakub@redhat.com>
6263         PR pch/59436
6264         * tree-core.h (struct tree_optimization_option): Change optabs
6265         type from unsigned char * to void *.
6266         * optabs.c (init_tree_optimization_optabs): Adjust
6267         TREE_OPTIMIZATION_OPTABS initialization.
6269 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
6271         PR target/59644
6272         * config/i386/i386.h (struct machine_function): Add
6273         no_drap_save_restore field.
6274         * config/i386/i386.c (ix86_save_reg): Use
6275         !cfun->machine->no_drap_save_restore instead of
6276         crtl->stack_realign_needed.
6277         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
6278         this function clears frame_pointer_needed.  Set
6279         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
6280         and DRAP reg is needed.
6282 2014-01-06  Marek Polacek  <polacek@redhat.com>
6284         PR c/57773
6285         * doc/implement-c.texi: Mention that other integer types are
6286         permitted as bit-field types in strictly conforming mode.
6288 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
6290         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
6291         is newly allocated.
6293 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
6295         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
6297 2014-01-06  Martin Jambor  <mjambor@suse.cz>
6299         PR ipa/59008
6300         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
6301         to int.
6302         * ipa-prop.c (ipa_print_node_params): Fix indentation.
6304 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
6306         PR debug/59350
6307         PR debug/59510
6308         * var-tracking.c (add_stores): Preserve the value of the source even if
6309         we don't record the store.
6311 2014-01-06  Terry Guo  <terry.guo@arm.com>
6313         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
6315 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
6317         PR bootstrap/59541
6318         * config/darwin.c (darwin_function_section): Adjust return values to
6319         correspond to optimisation changes made in r206070.
6321 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
6323         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
6324         from prefetch_block tune setting.
6325         (nocona_cost): Correct size of prefetch block to 64.
6327 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
6329         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
6330         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
6331         used to save the static chain register in the computation of the offset
6332         from which the FP registers need to be restored.
6334 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
6336         PR tree-optimization/59519
6337         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
6338         ICE if get_current_def (current_new_name) is already non-NULL, as long
6339         as it is a phi result of some other phi in *new_exit_bb that has
6340         the same argument.
6342         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
6343         or vmovdqu* for misaligned_operand.
6344         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
6345         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6346         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6347         aligned_mem for AVX512F masked aligned load and store builtins and for
6348         non-temporal moves.
6350 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6352         PR tree-optimization/59651
6353         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6354         Address range for negative step should be added by TYPE_SIZE_UNIT.
6356 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6358         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6359         overlapping registers also for registers other than the stack pointer.
6361 2014-01-03  Marek Polacek  <polacek@redhat.com>
6363         PR other/59661
6364         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6365         __builtin_FILE.
6367 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6369         PR target/59625
6370         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6371         asm goto as jump.
6373         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6374         (push splitter): Use <P:MODE_SIZE> instead of
6375         GET_MODE_SIZE (<P:MODE>mode).
6376         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6377         (mov -1, reg peephole2): Likewise.
6378         * config/i386/sse.md (*mov<mode>_internal,
6379         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6380         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6381         *<code><mode>3, *andnot<mode>3<mask_name>,
6382         <mask_codefor><code><mode>3<mask_name>): Likewise.
6383         * config/i386/subst.md (mask_mode512bit_condition,
6384         sd_mask_mode512bit_condition): Likewise.
6386 2014-01-02  Xinliang David Li  <davidxl@google.com>
6388         PR tree-optimization/59303
6389         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6390         (dump_predicates): Better output format.
6391         (pred_equal_p): New function.
6392         (is_neq_relop_p): Ditto.
6393         (is_neq_zero_form_p): Ditto.
6394         (pred_expr_equal_p): Ditto.
6395         (pred_neg_p): Ditto.
6396         (simplify_pred): Ditto.
6397         (simplify_preds_2): Ditto.
6398         (simplify_preds_3): Ditto.
6399         (simplify_preds_4): Ditto.
6400         (simplify_preds): Ditto.
6401         (push_pred): Ditto.
6402         (push_to_worklist): Ditto.
6403         (get_pred_info_from_cmp): Ditto.
6404         (is_degenerated_phi): Ditto.
6405         (normalize_one_pred_1): Ditto.
6406         (normalize_one_pred): Ditto.
6407         (normalize_one_pred_chain): Ditto.
6408         (normalize_preds): Ditto.
6409         (normalize_cond_1): Remove function.
6410         (normalize_cond): Ditto.
6411         (is_gcond_subset_of): Ditto.
6412         (is_subset_of_any): Ditto.
6413         (is_or_set_subset_of): Ditto.
6414         (is_and_set_subset_of): Ditto.
6415         (is_norm_cond_subset_of): Ditto.
6416         (pred_chain_length_cmp): Ditto.
6417         (convert_control_dep_chain_into_preds): Type change.
6418         (find_predicates): Ditto.
6419         (find_def_preds): Ditto.
6420         (destroy_predicates_vecs): Ditto.
6421         (find_matching_predicates_in_rest_chains): Ditto.
6422         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6423         (is_pred_expr_subset): Ditto.
6424         (is_pred_chain_subset_of): Ditto.
6425         (is_included_in): Ditto.
6426         (is_superset_of): Ditto.
6428 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6430         Update copyright years.
6432 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6434         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6435         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6436         config/arc/arc.md, config/arc/arc.opt,
6437         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6438         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6439         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6440         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6441         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6442         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6443         vtable-verify.c, vtable-verify.h: Use the standard form for the
6444         copyright notice.
6446 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6448         * gcc.c (process_command): Update copyright notice dates.
6449         * gcov-dump.c: Ditto.
6450         * gcov.c: Ditto.
6451         * doc/cpp.texi: Bump @copying's copyright year.
6452         * doc/cppinternals.texi: Ditto.
6453         * doc/gcc.texi: Ditto.
6454         * doc/gccint.texi: Ditto.
6455         * doc/gcov.texi: Ditto.
6456         * doc/install.texi: Ditto.
6457         * doc/invoke.texi: Ditto.
6459 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6461         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6463 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6465         * config/i386/sse.md (*mov<mode>_internal): Guard
6466         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6468         PR rtl-optimization/59647
6469         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6470         new_rtx into UNSIGNED_FLOAT rtxes.
6472 Copyright (C) 2014 Free Software Foundation, Inc.
6474 Copying and distribution of this file, with or without modification,
6475 are permitted in any medium without royalty provided the copyright
6476 notice and this notice are preserved.