PR pch/60010
[official-gcc.git] / gcc / ChangeLog
blob3a322921d1bdb153382fc8e3b59c8e6b395f80f7
1 2014-02-14  Kyle McMartin  <kyle@redhat.com>
3         PR pch/60010
4         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
6 2014-02-14  Richard Biener  <rguenther@suse.de>
8         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
9         (get_frame_arg): Drop the assert with langhook types_compatible_p.
10         Do not strip INDIRECT_REFs.
12 2014-02-14  Richard Biener  <rguenther@suse.de>
14         PR lto/60179
15         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
16         DECL_FUNCTION_SPECIFIC_TARGET.
17         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
18         * tree-streamer-out.c (pack_ts_target_option): Remove.
19         (streamer_pack_tree_bitfields): Do not stream
20         TS_TARGET_OPTION.
21         (write_ts_function_decl_tree_pointers): Do not stream
22         DECL_FUNCTION_SPECIFIC_TARGET.
23         * tree-streamer-in.c (unpack_ts_target_option): Remove.
24         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
25         (lto_input_ts_function_decl_tree_pointers): Do not stream
26         DECL_FUNCTION_SPECIFIC_TARGET.
28 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
30         * tree-vect-loop.c (vect_is_slp_reduction): Don't set
31         use_stmt twice.
32         (get_initial_def_for_induction, vectorizable_induction): Ignore
33         debug stmts when looking for exit_phi.
34         (vectorizable_live_operation): Fix up condition.
36 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
38         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
39         nreverse() because it changes the content of original tree list.
41 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
43         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
44         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
46 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
48         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
49         GNU coding standards.
51 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
53         PR debug/60152
54         * dwarf2out.c (gen_subprogram_die): Don't call
55         add_calling_convention_attribute if subr_die is old_die.
57 2014-02-13  Sharad Singhai  <singhai@google.com>
59         * doc/optinfo.texi: Fix order of nodes.
61 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
63         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
64         operands[2], not operands[3].
66 2014-02-13  Richard Biener  <rguenther@suse.de>
68         PR bootstrap/59878
69         * doc/install.texi (ISL): Update recommended version to 0.12.2,
70         mention the possibility of an in-tree build.
71         (CLooG): Update recommended version to 0.18.1, mention the
72         possibility of an in-tree build and clarify that the ISL
73         bundled with CLooG does not work.
75 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
77         PR target/43546
78         * expr.c (compress_float_constant): If x is a hard register,
79         extend into a pseudo and then move to x.
81 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
83         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
84         caused by bad second argument to warning_at() with -mhotpatch and
85         nested functions (e.g. with gfortran).
87 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
89         * opts.c (option_name): Remove "enabled by default" rider.
91 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
93         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
95 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
96             Uros Bizjak  <ubizjak@gmail.com>
98         PR target/60151
99         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
100         * configure: Regenerated.
102 2014-02-12  Richard Biener  <rguenther@suse.de>
104         * vec.c (vec_prefix::calculate_allocation): Move as
105         inline variant to vec.h.
106         (vec_prefix::calculate_allocation_1): New out-of-line version.
107         * vec.h (vec_prefix::calculate_allocation_1): Declare.
108         (vec_prefix::m_has_auto_buf): Rename to ...
109         (vec_prefix::m_using_auto_storage): ... this.
110         (vec_prefix::calculate_allocation): Inline the easy cases
111         and dispatch to calculate_allocation_1 which doesn't need the
112         prefix address.
113         (va_heap::reserve): Use gcc_checking_assert.
114         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
115         m_using_auto_storage.
116         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
117         member and adjust.
118         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
119         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
120         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
122 2014-02-12  Richard Biener  <rguenther@suse.de>
124         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
125         when we found a dependence.
127 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
129         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
130         common code...
131         (maybe_fold_stmt): ... into this new function.
132         * omp-low.c (lower_omp): Update comment.
134         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
135         last use.
137         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
138         dereference.
140 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
142         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
143         identifiers in comments.
144         (cortexa53_extra_costs): Likewise.
145         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
146         (cortexa7_extra_costs): Likewise.
147         (cortexa12_extra_costs): Likewise.
148         (cortexa15_extra_costs): Likewise.
149         (v7m_extra_costs): Likewise.
151 2014-02-12  Richard Biener  <rguenther@suse.de>
153         PR middle-end/60092
154         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
155         of posix_memalign being successful.
156         (lower_stmt): Restrict lowering of posix_memalign to when
157         -ftree-bit-ccp is enabled.
159 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
161         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
162         arg_loc.
163         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
165 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
167         PR rtl-optimization/60116
168         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
169         other_insn once the combination has been validated.
171 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
173         PR lto/59468
174         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
175         and wrapper.
176         * ipa-devirt.c: Include demangle.h
177         (odr_violation_reported): New static variable.
178         (add_type_duplicate): Update odr_violations.
179         (maybe_record_node): Add completep parameter; update it.
180         (record_target_from_binfo): Add COMPLETEP parameter;
181         update it as needed.
182         (possible_polymorphic_call_targets_1): Likewise.
183         (struct polymorphic_call_target_d): Add nonconstruction_targets;
184         rename FINAL to COMPLETE.
185         (record_targets_from_bases): Sanity check we found the binfo;
186         fix COMPLETEP updating.
187         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
188         parameter, fix computing of COMPLETEP.
189         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
190         at LTO time do demangling.
191         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
192         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
193         parameter.
194         (gimple_get_virt_method_for_binfo): Likewise.
195         * gimple-fold.h (gimple_get_virt_method_for_binfo,
196         gimple_get_virt_method_for_vtable): Update prototypes.
198 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
200         PR target/49008
201         * genautomata.c (add_presence_absence): Fix typo with
202         {final_}presence_list.
204 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
206         PR target/60137
207         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
208         for VSX/Altivec vectors that land in GPR registers.
210 2014-02-11  Richard Henderson  <rth@redhat.com>
211             Jakub Jelinek  <jakub@redhat.com>
213         PR debug/59776
214         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
215         around drhs if type conversion to lacc->type is not useless.
217 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
219         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
220         tuning struct.
221         (cortex-a57.cortex-a53): Likewise.
222         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
224 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
226         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
227         arm_restrict_it.
229 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
231         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
232         add_options_for_arm_vfp3.
234 2014-02-11  Jeff Law  <law@redhat.com>
236         PR middle-end/54041
237         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
238         object with an undesirable mode.
240 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
242         PR libgomp/60107
243         * config/i386/sol2-9.h: New file.
244         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
245         *-*-solaris2.9*): Use it.
247 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
249         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
250         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
252 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
254         * config/microblaze/microblaze.c: Extend mcpu version format
256 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
258         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
260 2014-02-10  Richard Henderson <rth@redhat.com>
262         PR target/59927
263         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
264         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
265         ms-abi vs -mno-accumulate-outgoing-args.
266         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
267         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
268         respect to ms-abi.
270 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
272         PR middle-end/60080
273         * cfgexpand.c (expand_asm_operands): Attach source location to
274         ASM_INPUT rtx objects.
275         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
277 2014-02-10  Nick Clifton  <nickc@redhat.com>
279         * config/mn10300/mn10300.c (popcount): New function.
280         (mn10300_expand_prologue): Include saved registers in stack usage
281         count.
283 2014-02-10  Jeff Law  <law@redhat.com>
285         PR middle-end/52306
286         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
287         when changing the SET_DEST of a prior insn to avoid an input reload.
289 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
291         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
292         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
293         -mcall-openbsd, or -mcall-linux.
294         (CC1_ENDIAN_BIG_SPEC): Remove.
295         (CC1_ENDIAN_LITTLE_SPEC): Remove.
296         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
297         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
298         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
299         and %cc1_endian_default.
300         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
302 2014-02-10  Richard Biener  <rguenther@suse.de>
304         PR tree-optimization/60115
305         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
306         MEM_REF handling.  Properly verify that the accesses are not
307         out of the objects bound.
309 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
311         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
312         coretex to cortex.
314 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
316         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
317         proper constants and fix formatting.
318         (possible_polymorphic_call_targets): Fix formatting.
320 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
321             Ilya Tocar  <ilya.tocar@intel.com>
323         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
324         (_mm512_loadu_epi32): Renamed into...
325         (_mm512_loadu_si512): This.
326         (_mm512_storeu_epi32): Renamed into...
327         (_mm512_storeu_si512): This.
328         (_mm512_maskz_ceil_ps): Removed.
329         (_mm512_maskz_ceil_pd): Ditto.
330         (_mm512_maskz_floor_ps): Ditto.
331         (_mm512_maskz_floor_pd): Ditto.
332         (_mm512_floor_round_ps): Ditto.
333         (_mm512_floor_round_pd): Ditto.
334         (_mm512_ceil_round_ps): Ditto.
335         (_mm512_ceil_round_pd): Ditto.
336         (_mm512_mask_floor_round_ps): Ditto.
337         (_mm512_mask_floor_round_pd): Ditto.
338         (_mm512_mask_ceil_round_ps): Ditto.
339         (_mm512_mask_ceil_round_pd): Ditto.
340         (_mm512_maskz_floor_round_ps): Ditto.
341         (_mm512_maskz_floor_round_pd): Ditto.
342         (_mm512_maskz_ceil_round_ps): Ditto.
343         (_mm512_maskz_ceil_round_pd): Ditto.
344         (_mm512_expand_pd): Ditto.
345         (_mm512_expand_ps): Ditto.
346         * config/i386/i386.c (ix86_builtins): Remove
347         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
348         (bdesc_args): Ditto.
349         * config/i386/predicates.md (const1256_operand): New.
350         (const_1_to_2_operand): Ditto.
351         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
352         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
353         (*avx512pf_gatherpf<mode>sf): Ditto.
354         (avx512pf_gatherpf<mode>df): Ditto.
355         (*avx512pf_gatherpf<mode>df_mask): Ditto.
356         (*avx512pf_gatherpf<mode>df): Ditto.
357         (avx512pf_scatterpf<mode>sf): Ditto.
358         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
359         (*avx512pf_scatterpf<mode>sf): Ditto.
360         (avx512pf_scatterpf<mode>df): Ditto.
361         (*avx512pf_scatterpf<mode>df_mask): Ditto.
362         (*avx512pf_scatterpf<mode>df): Ditto.
363         (avx512f_expand<mode>): Removed.
364         (<shift_insn><mode>3<mask_name>): Change predicate type.
366 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
368         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
369         not at the end of datarefs vector use ordered_remove to avoid
370         reordering datarefs vector.
372         PR c/59984
373         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
374         mark local addressable non-static vars as GOVD_PRIVATE
375         instead of GOVD_LOCAL.
376         * omp-low.c (lower_omp_for): Move gimple_bind_vars
377         and BLOCK_VARS of gimple_bind_block to new_stmt rather
378         than copying them.
380         PR middle-end/60092
381         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
382         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
383         assume_aligned or alloc_align attributes.
384         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
385         arguments.  Handle also assume_aligned and alloc_align attributes.
386         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
387         calls to functions with assume_aligned or alloc_align attributes.
388         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
390 2014-02-08  Terry Guo  <terry.guo@arm.com>
392         * doc/invoke.texi: Document ARM -march=armv7e-m.
394 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
396         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
397         flag on __cilkrts_rethrow builtin.
399         PR ipa/60026
400         * ipa-cp.c (determine_versionability): Fail at -O0
401         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
402         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
404         Revert:
405         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
407         PR ipa/60026
408         * tree-inline.c (copy_forbidden): Fail for
409         __attribute__((optimize (0))) functions.
411 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
413         * varpool.c: Include pointer-set.h.
414         (varpool_remove_unreferenced_decls): Variables in other partitions
415         will not be output; be however careful to not lose information
416         about partitioning.
418 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
420         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
421         lookup in the vtable constructor.
423 2014-02-07  Jeff Law  <law@redhat.com>
425         PR target/40977
426         * config/m68k/m68k.md (ashldi_extsi): Turn into a
427         define_insn_and_split.
429         * ipa-inline.c (inline_small_functions): Fix typos.
431 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
433         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
434         (s390_can_use_return_insn): Declare.
435         * config/s390/s390.h (EPILOGUE_USES): Define.
436         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
437         instructions.
438         (s390_chunkify_start): Handle return JUMP_LABELs.
439         (s390_early_mach): Emit a main_pool instruction on the entry edge.
440         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
441         (s390_can_use_return_insn): New functions.
442         (s390_fix_long_loop_prediction): Handle conditional returns.
443         (TARGET_SET_UP_BY_PROLOGUE): Define.
444         * config/s390/s390.md (ANY_RETURN): New code iterator.
445         (*creturn, *csimple_return, return, simple_return): New patterns.
447 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
449         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
450         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
451         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
452         REG_CFA_RESTORE list when deciding not to restore a register.
454 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
456         * config/s390/s390.c: Include tree-pass.h and context.h.
457         (s390_early_mach): New function, split out from...
458         (s390_emit_prologue): ...here.
459         (pass_data_s390_early_mach): New pass structure.
460         (pass_s390_early_mach): New class.
461         (s390_option_override): Create and register early_mach pass.
462         Move to end of file.
464 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
466         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
467         to match for the exit block.
469 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
471         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
472         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
473         Reject misaligned operands.
475 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
477         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
479 2014-02-07  Richard Biener  <rguenther@suse.de>
481         PR middle-end/60092
482         * gimple-low.c (lower_builtin_posix_memalign): New function.
483         (lower_stmt): Call it to lower posix_memalign in a way
484         to make alignment info accessible.
486 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
488         PR c++/60082
489         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
490         __builtin_setjmp_receiver.
492 2014-02-07  Richard Biener  <rguenther@suse.de>
494         PR middle-end/60092
495         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
496         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
497         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
498         Handle BUILT_IN_POSIX_MEMALIGN.
499         (find_func_clobbers): Likewise.
500         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
501         (call_may_clobber_ref_p_1): Likewise.
503 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
505         PR ipa/59918
506         * ipa-devirt.c (record_target_from_binfo): Remove overactive
507         sanity check.
509 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
511         PR ipa/59469
512         * lto-cgraph.c (lto_output_node): Use
513         symtab_get_symbol_partitioning_class.
514         (lto_output_varpool_node): likewise.
515         (symtab_get_symbol_partitioning_class): Move here from
516         lto/lto-partition.c
517         * cgraph.h (symbol_partitioning_class): Likewise.
518         (symtab_get_symbol_partitioning_class): Declare.
520 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
522         * ggc.h (ggc_internal_cleared_alloc): New macro.
523         * vec.h (vec_safe_copy): Handle memory stats.
524         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
525         * target-globals.c (save_target_globals): Likewise.
527 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
529         PR target/60077
530         * expr.c (emit_move_resolve_push): Export; be bit more selective
531         on when to clear alias set.
532         * expr.h (emit_move_resolve_push): Declare.
533         * function.h (struct function): Add tail_call_marked.
534         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
535         * config/i386/i386-protos.h (ix86_expand_push): Remove.
536         * config/i386/i386.md (TImode move expander): De not call
537         ix86_expand_push.
538         (FP push expanders): Preserve memory attributes.
539         * config/i386/sse.md (push<mode>1): Remove.
540         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
541         (ix86_expand_push): Remove.
542         * config/i386/mmx.md (push<mode>1): Remove.
544 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
546         PR rtl-optimization/60030
547         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
548         lopart with paradoxical subreg before shifting it up by hprec.
550 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
552         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
553         Remove extra newline at end of file.
554         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
555         (arm_issue_rate): Handle cortexa57.
556         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
557         (cortex-a57.cortex-a53): Likewise.
559 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
561         PR target/59575
562         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
563         don't record in REG_FRAME_RELATED_EXPR registers not set in that
564         bitmask.
565         (arm_expand_prologue): Adjust all callers.
566         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
567         info, registers also at the lowest numbered registers side.  Use
568         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
569         XEXP.
571         PR debug/59992
572         * var-tracking.c (adjust_mems): Before adding a SET to
573         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
575 2014-02-06  Alan Modra  <amodra@gmail.com>
577         PR target/60032
578         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
579         change SDmode to DDmode when lra_in_progress.
581 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
583         PR middle-end/59150
584         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
585         free_data_ref on the dr first, and before goto again also set dr
586         to the next dr.  For simd_lane_access, free old datarefs[i] before
587         overwriting it.  For get_vectype_for_scalar_type failure, don't
588         free_data_ref if simd_lane_access.
590         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
592         PR target/60062
593         * tree.h (opts_for_fn): New inline function.
594         (opt_for_fn): Define.
595         * config/i386/i386.c (ix86_function_regparm): Use
596         opt_for_fn (decl, optimize) instead of optimize.
598 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
600         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
601         for SYMBOL_REF in large memory model.
603 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
605         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
606         and crypto support.
607         (cortex-a57): Likewise.
608         (cortex-a57.cortex-a53): Likewise.
610 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
611             Kugan Vivekanandarajah  <kuganv@linaro.org>
613         * config/arm/arm.c (arm_vector_alignment_reachable): Check
614         unaligned_access.
615         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
617 2014-02-06  Richard Biener  <rguenther@suse.de>
619         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
620         set_loop_copy and initialize_original_copy_tables.
622 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
624         * config/aarch64/aarch64-simd.md
625         (aarch64_ashr_simddi): Change QI to SI.
627 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
628             Jakub Jelinek  <jakub@redhat.com>
630         PR middle-end/60013
631         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
632         of the dataflow.
634 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
636         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
637         CODE_FOR_altivec_vpku[hw]um to
638         CODE_FOR_altivec_vpku[hw]um_direct.
639         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
640         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
641         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
642         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
644 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
646         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
647         generation for -maltivec=be.
648         (altivec_vsumsws): Simplify redundant test.
650 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
652         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
653         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
654         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
655         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
656         gen_altivec_vpkuwum.
657         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
658         BYTES_BIG_ENDIAN.
659         (altivec_vpks<VI_char>ss): Likewise.
660         (altivec_vpks<VI_char>us): Likewise.
661         (altivec_vpku<VI_char>us): Likewise.
662         (altivec_vpku<VI_char>um): Likewise.
663         (altivec_vpku<VI_char>um_direct): New (copy of
664         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
665         internal use).
666         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
667         target is little endian and -maltivec=be is not specified.
668         (*altivec_vupkhs<VU_char>_direct): New (copy of
669         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
670         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
671         target is little endian and -maltivec=be is not specified.
672         (*altivec_vupkls<VU_char>_direct): New (copy of
673         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
674         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
675         little endian and -maltivec=be is not specified.
676         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
677         little endian and -maltivec=be is not specified.
679 2014-02-05  Richard Henderson <rth@redhat.com>
681         PR debug/52727
682         * combine-stack-adj.c: Revert r206943.
683         * sched-int.h (struct deps_desc): Add last_args_size.
684         * sched-deps.c (init_deps): Initialize it.
685         (sched_analyze_insn): Add OUTPUT dependencies between insns that
686         contain REG_ARGS_SIZE notes.
688 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
690         * lto-cgraph.c (asm_nodes_output): Make global.
691         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
692         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
693         (driver_handle_option): Handle OPT_fwpa.
695 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
697         PR ipa/59947
698         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
699         a comment typo and formatting issue.  If odr_hash hasn't been
700         created, return vNULL and set *completep to false.
702         PR middle-end/57499
703         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
704         bb with no successors.
706 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
708         PR target/59718
709         * doc/invoke.texi (-march): Clarify documentation for ARM.
710         (-mtune): Likewise.
711         (-mcpu): Likewise.
713 2014-02-05  Richard Biener  <rguenther@suse.de>
715         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
716         when not vectorizing because of too many alias checks.
717         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
718         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
720 2014-02-05  Nick Clifton  <nickc@redhat.com>
722         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
723         accept extended registers in any mode when compiling for the MN10300.
725 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
727         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
728         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
729         sanitization attributes.
730         (can_inline_edge_p): Likewise.
731         (sanitize_attrs_match_for_inline_p): New function.
733 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
735         * ipa-prop.c (detect_type_change): Shor circuit testing of
736         type changes on THIS pointer.
738 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
740         PR target/59777
741         * config/pa/pa.c (legitimize_tls_address): Return original address
742         if not passed a SYMBOL_REF rtx.
743         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
744         addresses.
745         (pa_emit_move_sequence): Simplify TLS source operands.
746         (pa_legitimate_constant_p): Reject all TLS constants.
747         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
748         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
750 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
752         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
753         groups when we know they are controlled by LTO.
754         * varasm.c (default_binds_local_p_1): If object is in other partition,
755         it will be resolved locally.
757 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
759         * config/host-linux.c (linux_gt_pch_use_address): Don't
760         use SSIZE_MAX because it is not always defined.
762 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
764         PR bootstrap/59913
765         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
766         threshold for pseudo splitting.
767         (update_ebb_live_info): Process call argument hard registers and
768         hard registers from insn definition too.
769         (max_small_class_regs_num): New constant.
770         (inherit_in_ebb): Update live hard regs through EBBs.  Update
771         reloads_num only for small register classes.  Don't split for
772         outputs of jumps.
774 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
776         PR ipa/60058
777         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
778         is non-null.
780 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
782         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
783         visibility is safe.
785 2014-02-04  Marek Polacek  <polacek@redhat.com>
787         * gdbinit.in (pel): Define.
789 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
791         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
792         behavior.
794 2014-02-04  Richard Biener  <rguenther@suse.de>
796         PR lto/59723
797         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
798         in function context local.
799         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
800         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
801         similar to LTO_imported_decl_ref.
803 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
805         PR tree-optimization/60002
806         * cgraphclones.c (build_function_decl_skip_args): Clear
807         DECL_LANG_SPECIFIC.
809         PR tree-optimization/60023
810         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
811         false to gsi_replace.
812         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
813         has been in some EH region and vec_stmt could throw, add
814         vec_stmt into the same EH region.
815         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
816         has no lhs, ignore it.
817         * internal-fn.c (expand_MASK_LOAD): Likewise.
819         PR ipa/60026
820         * tree-inline.c (copy_forbidden): Fail for
821         __attribute__((optimize (0))) functions.
823         PR other/58712
824         * omp-low.c (simd_clone_struct_copy): If from->inbranch
825         is set, copy one less argument.
826         (expand_simd_clones): Don't subtract clone_info->inbranch
827         from simd_clone_struct_alloc argument.
829         PR rtl-optimization/57915
830         * recog.c (simplify_while_replacing): If all unary/binary/relational
831         operation arguments are constant, attempt to simplify those.
833         PR middle-end/59261
834         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
835         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
837 2014-02-04  Richard Biener  <rguenther@suse.de>
839         PR tree-optimization/60012
840         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
841         TBAA disambiguation to all DDRs.
843 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
845         PR target/59788
846         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
847         (LINK_SPEC): Use it for -shared, -shared-libgcc.
849 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
851         PR ipa/59882
852         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
854 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
856         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
857         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
859 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
861         PR ipa/59831
862         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
863         to figure out targets of polymorphic calls with known decl.
864         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
865         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
866         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
867         (get_polymorphic_call_info): ... here.
868         (get_polymorphic_call_info_from_invariant): New function.
870 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
872         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
873         lookup via vtable pointer; check for type consistency
874         and turn inconsitent facts into UNREACHABLE.
875         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
876         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
877         type inconsistent querries; return UNREACHABLE instead.
879 2014-02-03  Richard Henderson  <rth@twiddle.net>
881         PR tree-opt/59924
882         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
883         already processed this node.
884         (normalize_one_pred_1): Pass along mark_set.
885         (normalize_one_pred): Create and destroy a pointer_set_t.
886         (normalize_one_pred_chain): Likewise.
888 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
890         PR gcov-profile/58602
891         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
893 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
895         PR ipa/59831
896         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
897         -fno-devirtualize; try to devirtualize by the knowledge of
898         virtual table pointer given by aggregate propagation.
899         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
900         (ipa_print_node_jump_functions): Dump also offset that
901         is relevant for polymorphic calls.
902         (determine_known_aggregate_parts): Add arg_type parameter; use it
903         instead of determining the type from pointer type.
904         (ipa_compute_jump_functions_for_edge): Update call of
905         determine_known_aggregate_parts.
906         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
907         (gimple_get_virt_method_for_binfo): ... here; simplify using
908         vtable_pointer_value_to_vtable.
909         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
910         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
911         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
912         (vtable_pointer_value_to_vtable): Break out from ...; handle also
913         POINTER_PLUS_EXPR.
914         (vtable_pointer_value_to_binfo): ... here.
915         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
917 2014-02-03  Teresa Johnson  <tejohnson@google.com>
919         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
920         redef of outer loop index variable.
922 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
924         PR c++/53017
925         PR c++/59211
926         * doc/extend.texi (Function Attributes): Typo.
928 2014-02-03  Cong Hou  <congh@google.com>
930         PR tree-optimization/60000
931         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
932         if the vectorized statement is a store.  A store statement can only
933         appear at the end of pattern statements.
935 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
937         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
938         (ix86_option_override_internal): Default long double to 64-bit for
939         32-bit Bionic and to 128-bit for 64-bit Bionic.
941         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
942         TARGET_LONG_DOUBLE_128 is true.
943         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
945         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
946         (mlong-double-64): Negate -mlong-double-128.
947         (mlong-double-128): New option.
949         * config/i386/i386-c.c (ix86_target_macros): Define
950         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
952         * doc/invoke.texi: Document -mlong-double-128.
954 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
956         PR rtl-optimization/60024
957         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
959 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
961         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
963 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
965         PR rtl-optimization/57662
966         * sel-sched.c (code_motion_path_driver): Do not mark already not
967         existing blocks in the visiting bitmap.
969 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
971         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
972         on the insn being emitted.
974 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
975             Will Deacon  <will.deacon@arm.com>
977         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
979 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
981         * config/arm/arm-tables.opt: Regenerate.
983 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
985         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
986         for vector types other than V16QImode.
987         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
988         define_expand, and call altivec_expand_vec_perm_le when producing
989         code with little endian element order.
990         (*altivec_vperm_<mode>_internal): New insn having previous
991         behavior of altivec_vperm_<mode>.
992         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
993         altivec_expand_vec_perm_le when producing code with little endian
994         element order.
995         (*altivec_vperm_<mode>_uns_internal): New insn having previous
996         behavior of altivec_vperm_<mode>_uns.
998 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1000         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1001         (altivec_vsumsws): Add handling for -maltivec=be with a little
1002         endian target.
1003         (altivec_vsumsws_direct): New.
1004         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1005         gen_altivec_vsumsws.
1007 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1009         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1010         vtable_pointer_value_to_binfo): New functions.
1011         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1012         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1014 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
1016         * config/nios2/nios2.md (load_got_register): Initialize GOT
1017         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1018         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1020 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1022         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1023         preserverd by passthrough, do not propagate the type.
1025 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1027         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1028         (mips_atomic_assign_expand_fenv): New function.
1029         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1031 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1033         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1034         (__builtin_mips_set_fcsr): Likewise.
1035         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1036         MIPS_USI_FTYPE_VOID.
1037         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1038         (mips16_expand_set_fcsr): Likewise.
1039         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1040         (mips16_set_fcsr_stub): Likewise.
1041         (mips16_get_fcsr_one_only_stub): New class.
1042         (mips16_set_fcsr_one_only_stub): Likewise.
1043         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1044         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1045         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1046         (hard_float): New availability predicate.
1047         (mips_builtins): Add get_fcsr and set_fcsr.
1048         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1049         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1050         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1051         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1052         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1053         patterns.
1055 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1057         * config/mips/mips.c (mips_one_only_stub): New class.
1058         (mips_need_mips16_rdhwr_p): Replace with...
1059         (mips16_rdhwr_stub): ...this new variable.
1060         (mips16_stub_call_address): New function.
1061         (mips16_rdhwr_one_only_stub): New class.
1062         (mips_expand_thread_pointer): Use mips16_stub_call_address.
1063         (mips_output_mips16_rdhwr): Delete.
1064         (mips_finish_stub): New function.
1065         (mips_code_end): Use it to handle rdhwr stubs.
1067 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
1069         PR target/60017
1070         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1071         when calculating size of integer atomic types.
1073 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
1075         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1077 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
1079         PR tree-optimization/60003
1080         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1081         * profile.c (branch_prob): Use gimple_call_builtin_p
1082         to check for BUILT_IN_SETJMP_RECEIVER.
1083         * tree-inline.c (copy_bb): Call notice_special_calls.
1085 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1087         PR bootstrap/59985
1088         * lra-constraints.c (process_alt_operands): Update reload_sum only
1089         on the first pass.
1091 2014-01-31  Richard Henderson  <rth@redhat.com>
1093         PR middle-end/60004
1094         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1095         until after else_eh is processed.
1097 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1099         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1100         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1101         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1102         in smmintrin.h, remove them.
1103         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1104         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1105         * config/i386/i386.md (ROUND_SAE): Fix value.
1106         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1107         (const48_operand): New.
1108         * config/i386/subst.md (round), (round_expand): Use
1109         const_4_or_8_to_11_operand.
1110         (round_saeonly), (round_saeonly_expand): Use const48_operand.
1112 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1114         * config/i386/constraints.md (Yk): Swap meaning with k.
1115         * config/i386/i386.md (movhi_internal): Change Yk to k.
1116         (movqi_internal): Ditto.
1117         (*k<logic><mode>): Ditto.
1118         (*andhi_1): Ditto.
1119         (*andqi_1): Ditto.
1120         (kandn<mode>): Ditto.
1121         (*<code>hi_1): Ditto.
1122         (*<code>qi_1): Ditto.
1123         (kxnor<mode>): Ditto.
1124         (kortestzhi): Ditto.
1125         (kortestchi): Ditto.
1126         (kunpckhi): Ditto.
1127         (*one_cmplhi2_1): Ditto.
1128         (*one_cmplqi2_1): Ditto.
1129         * config/i386/sse.md (): Change k to Yk.
1130         (avx512f_load<mode>_mask): Ditto.
1131         (avx512f_blendm<mode>): Ditto.
1132         (avx512f_store<mode>_mask): Ditto.
1133         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1134         (avx512f_storedqu<mode>_mask): Ditto.
1135         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1136         Ditto.
1137         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1138         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1139         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1140         (avx512f_maskcmp<mode>3): Ditto.
1141         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1142         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1143         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1144         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1145         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1146         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1147         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1148         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1149         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1150         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1151         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1152         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1153         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1154         (vec_extract_lo_<mode>_maskm): Ditto.
1155         (vec_extract_hi_<mode>_maskm): Ditto.
1156         (avx512f_vternlog<mode>_mask): Ditto.
1157         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1158         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1159         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1160         (avx512f_<code>v8div16qi2_mask): Ditto.
1161         (avx512f_<code>v8div16qi2_mask_store): Ditto.
1162         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1163         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1164         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1165         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1166         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1167         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1168         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1169         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1170         (avx512cd_maskb_vec_dupv8di): Ditto.
1171         (avx512cd_maskw_vec_dupv16si): Ditto.
1172         (avx512f_vpermi2var<mode>3_maskz): Ditto.
1173         (avx512f_vpermi2var<mode>3_mask): Ditto.
1174         (avx512f_vpermi2var<mode>3_mask): Ditto.
1175         (avx512f_vpermt2var<mode>3_maskz): Ditto.
1176         (*avx512f_gathersi<mode>): Ditto.
1177         (*avx512f_gathersi<mode>_2): Ditto.
1178         (*avx512f_gatherdi<mode>): Ditto.
1179         (*avx512f_gatherdi<mode>_2): Ditto.
1180         (*avx512f_scattersi<mode>): Ditto.
1181         (*avx512f_scatterdi<mode>): Ditto.
1182         (avx512f_compress<mode>_mask): Ditto.
1183         (avx512f_compressstore<mode>_mask): Ditto.
1184         (avx512f_expand<mode>_mask): Ditto.
1185         * config/i386/subst.md (mask): Change k to Yk.
1186         (mask_scalar_merge): Ditto.
1187         (sd): Ditto.
1189 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
1191         * doc/extend.texi (Vector Extensions): Document ?: in C++.
1193 2014-01-31  Richard Biener  <rguenther@suse.de>
1195         PR middle-end/59990
1196         * builtins.c (fold_builtin_memory_op): Make sure to not
1197         use a floating-point mode or a boolean or enumeral type for
1198         the copy operation.
1200 2014-01-30  DJ Delorie  <dj@redhat.com>
1202         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1203         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1204         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1205         whenever main() has an epilogue.
1207 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1209         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1210         unused variable "field".
1211         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1212         (vsx_mergeh_<mode>): Likewise.
1213         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1214         (altivec_vmrghh): Likewise.
1215         (altivec_vmrghw): Likewise.
1216         (altivec_vmrglb): Likewise.
1217         (altivec_vmrglh): Likewise.
1218         (altivec_vmrglw): Likewise.
1219         (altivec_vspltb): Add missing uses.
1220         (altivec_vsplth): Likewise.
1221         (altivec_vspltw): Likewise.
1222         (altivec_vspltsf): Likewise.
1224 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1226         PR target/59923
1227         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1228         frame related instructions.
1230 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1232         PR rtl-optimization/59959
1233         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1234         any reload of register whose subreg is invalid.
1236 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1238         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1239         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1240         Add missing return type - void.
1242 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1244         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1245         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1246         remove element index adjustment for endian (now handled in vsx.md
1247         and altivec.md).
1248         (altivec_expand_vec_perm_const): Use
1249         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1250         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1251         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1252         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1253         define_expand and a new define_insn *altivec_vspltb_internal;
1254         adjust for -maltivec=be on a little endian target.
1255         (altivec_vspltb_direct): New.
1256         (altivec_vsplth): Divide into a define_expand and a new
1257         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1258         little endian target.
1259         (altivec_vsplth_direct): New.
1260         (altivec_vspltw): Divide into a define_expand and a new
1261         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1262         little endian target.
1263         (altivec_vspltw_direct): New.
1264         (altivec_vspltsf): Divide into a define_expand and a new
1265         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1266         a little endian target.
1268 2014-01-30  Richard Biener  <rguenther@suse.de>
1270         PR tree-optimization/59993
1271         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1272         can propagate form the earlier stmt and avoid the transform
1273         when the intermediate result is needed.
1275 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1277         * README.Portability: Fix typo.
1279 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1281         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1282         comparison_operator with ordered_comparison_operator.
1284 2014-01-30  Nick Clifton  <nickc@redhat.com>
1286         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1287         Rename to mn10300_store_multiple_regs.
1288         * config/mn10300/mn10300.c: Likewise.
1289         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1290         store_multiple_regs.
1291         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1292         Call mn10300_store_multiple_regs.
1294 2014-01-30  Nick Clifton  <nickc@redhat.com>
1295             DJ Delorie  <dj@redhat.com>
1297         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1298         %fp 2 to keep registers after it properly word-aligned.
1299         (rl78_alloc_physical_registers_umul): Handle the case where both
1300         input operands are the same.
1302 2014-01-30  Richard Biener  <rguenther@suse.de>
1304         PR tree-optimization/59903
1305         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1306         check properly.
1308 2014-01-30  Jason Merrill  <jason@redhat.com>
1310         PR c++/59633
1311         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1313         PR c++/59645
1314         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1316 2014-01-30  Richard Biener  <rguenther@suse.de>
1318         PR tree-optimization/59951
1319         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1321 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1323         PR target/59784
1324         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1325         SFmode to DFmode case.
1327 2014-01-29  DJ Delorie  <dj@redhat.com>
1329         * config/msp430/msp430.opt (-minrt): New.
1330         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1331         if -minrt given.
1332         (ENDFILE_SPEC): Likewise.
1334 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1336         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1337         (estimate_function_body_sizes): Use it.
1339 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1341         PR c++/58561
1342         * dwarf2out.c (is_cxx_auto): New.
1343         (is_base_type): Use it.
1344         (gen_type_die_with_usage): Likewise.
1346 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1348         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1349         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1350         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1351         -maltivec=be with LE targets.
1352         (vsx_mergeh_<mode>): Likewise.
1353         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1354         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1355         (altivec_vmrghb): Replace with define_expand and new
1356         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1357         (altivec_vmrghb_direct): New define_insn.
1358         (altivec_vmrghh): Replace with define_expand and new
1359         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1360         (altivec_vmrghh_direct): New define_insn.
1361         (altivec_vmrghw): Replace with define_expand and new
1362         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1363         (altivec_vmrghw_direct): New define_insn.
1364         (*altivec_vmrghsf): Adjust for endianness.
1365         (altivec_vmrglb): Replace with define_expand and new
1366         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1367         (altivec_vmrglb_direct): New define_insn.
1368         (altivec_vmrglh): Replace with define_expand and new
1369         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1370         (altivec_vmrglh_direct): New define_insn.
1371         (altivec_vmrglw): Replace with define_expand and new
1372         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1373         (altivec_vmrglw_direct): New define_insn.
1374         (*altivec_vmrglsf): Adjust for endianness.
1375         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1376         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1377         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1378         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1379         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1380         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1381         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1382         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1384 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1386         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1387         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1388         whitespace.
1390 2014-01-29  Richard Biener  <rguenther@suse.de>
1392         PR tree-optimization/58742
1393         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1394         associate_pointerplus_align.
1395         (associate_pointerplus_diff): New function.
1396         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
1397         and associate_pointerplus_diff.
1399 2014-01-29  Richard Biener  <rguenther@suse.de>
1401         * lto-streamer.h (LTO_major_version): Bump to 3.
1402         (LTO_minor_version): Reset to 0.
1404 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
1406         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1407         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1408         (arm_file_start): Generate correct asm header for armv7ve.
1409         * config/arm/bpabi.h: Add multilib support for armv7ve.
1410         * config/arm/driver-arm.c: Change the architectures of cortex-a7
1411         and cortex-a15 to armv7ve.
1412         * config/arm/t-aprofile: Add multilib support for armv7ve.
1413         * doc/invoke.texi: Document -march=armv7ve.
1415 2014-01-29  Richard Biener  <rguenther@suse.de>
1417         PR tree-optimization/58742
1418         * tree-ssa-forwprop.c (associate_plusminus): Return true
1419         if we changed sth, defer EH cleanup to ...
1420         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
1421         (simplify_mult): New function.
1423 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1425         PR middle-end/59917
1426         PR tree-optimization/59920
1427         * tree.c (build_common_builtin_nodes): Remove
1428         __builtin_setjmp_dispatcher initialization.
1429         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1430         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1431         instead of gsi_after_labels + manually skipping debug stmts.
1432         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1433         ignore bbs with IFN_ABNORMAL_DISPATCHER.
1434         * tree-inline.c (copy_edges_for_bb): Remove
1435         can_make_abnormal_goto argument, instead add abnormal_goto_dest
1436         argument.  Ignore computed_goto_p stmts.  Don't call
1437         make_abnormal_goto_edges.  If a call might need abnormal edges
1438         for non-local gotos, see if it already has an edge to
1439         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1440         with true argument, don't do anything then, otherwise add
1441         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1442         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1443         caller.
1444         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1445         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1446         (lower_stmt): Don't set data->calls_builtin_setjmp.
1447         (lower_builtin_setjmp): Adjust comment.
1448         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1449         * tree-cfg.c (found_computed_goto): Remove.
1450         (factor_computed_gotos): Remove.
1451         (make_goto_expr_edges): Return bool, true for computed gotos.
1452         Don't call make_abnormal_goto_edges.
1453         (build_gimple_cfg): Don't set found_computed_goto, don't call
1454         factor_computed_gotos.
1455         (computed_goto_p): No longer static.
1456         (make_blocks): Don't set found_computed_goto.
1457         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1458         (make_edges): If make_goto_expr_edges returns true, push bb
1459         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1460         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1461         vector.  Record mapping between bbs and OpenMP regions if there
1462         are any, adjust make_gimple_omp_edges caller.  Call
1463         handle_abnormal_edges.
1464         (make_abnormal_goto_edges): Remove.
1465         * tree-cfg.h (make_abnormal_goto_edges): Remove.
1466         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1467         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1468         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1469         * internal-fn.def (ABNORMAL_DISPATCHER): New.
1470         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1471         filling *region also set *region_idx to (*region)->entry->index.
1473         PR other/58712
1474         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1475         For REGs set ORIGINAL_REGNO.
1477 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
1479         * doc/md.texi: Mention that a target shouldn't implement
1480         vec_widen_(s|u)mul_even/odd pair if it is less efficient
1481         than hi/lo pair.
1483 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1485         PR tree-optimization/59594
1486         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1487         a copy of the datarefs vector rather than the vector itself.
1489 2014-01-28  Jason Merrill  <jason@redhat.com>
1491         PR c++/53756
1492         * dwarf2out.c (auto_die): New static.
1493         (gen_type_die_with_usage): Handle C++1y 'auto'.
1494         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1495         on definition.
1497 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1499         PR target/59672
1500         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1501         (SPEC_X32): Likewise.
1502         (SPEC_64): Likewise.
1503         * config/i386/i386.c (ix86_option_override_internal): Turn off
1504         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1505         for TARGET_16BIT.
1506         (x86_file_start): Output .code16gcc for TARGET_16BIT.
1507         * config/i386/i386.h (TARGET_16BIT): New macro.
1508         (TARGET_16BIT_P): Likewise.
1509         * config/i386/i386.opt: Add m16.
1510         * doc/invoke.texi: Document -m16.
1512 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
1514         PR preprocessor/59935
1515         * input.c (location_get_source_line): Bail out on when line number
1516         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1518 2014-01-28  Richard Biener  <rguenther@suse.de>
1520         PR tree-optimization/58742
1521         * tree-ssa-forwprop.c (associate_plusminus): Handle
1522         pointer subtraction of the form (T)(P + A) - (T)P.
1524 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1526         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1527         at const_int_cost.
1529 2014-01-28  Richard Biener  <rguenther@suse.de>
1531         Revert
1532         2014-01-28  Richard Biener  <rguenther@suse.de>
1534         PR rtl-optimization/45364
1535         PR rtl-optimization/59890
1536         * var-tracking.c (local_get_addr_clear_given_value): Handle
1537         already cleared slot.
1538         (val_reset): Handle not allocated local_get_addr_cache.
1539         (vt_find_locations): Use post-order on the inverted CFG.
1541 2014-01-28  Richard Biener  <rguenther@suse.de>
1543         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1545 2014-01-28  Richard Biener  <rguenther@suse.de>
1547         PR rtl-optimization/45364
1548         PR rtl-optimization/59890
1549         * var-tracking.c (local_get_addr_clear_given_value): Handle
1550         already cleared slot.
1551         (val_reset): Handle not allocated local_get_addr_cache.
1552         (vt_find_locations): Use post-order on the inverted CFG.
1554 2014-01-28  Alan Modra  <amodra@gmail.com>
1556         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1557         * configure.ac <recursive call for build != host>: Define
1558         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1559         and LD_FOR_BUILD too.
1560         * configure: Regenerate.
1562 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
1564         * config/i386/i386.c (get_builtin_code_for_version): Separate
1565         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1566         Broadwell from Haswell.
1568 2014-01-27  Steve Ellcey  <sellcey@mips.com>
1570         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1571         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1572         * config/mips/mips.c (mips_option_override): Change setting
1573         of TARGET_DSP.
1574         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1575         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1576         Change from Mask to Var.
1578 2014-01-27  Jeff Law  <law@redhat.com>
1580         * ipa-inline.c (inline_small_functions): Fix typo.
1582 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
1584         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1585         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1586         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1587         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1588         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1589         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1590         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1591         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1592         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1593         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1594         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1595         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1596         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1597         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1598         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1599         (_mm512_storeu_epi64): Ditto.
1600         (_mm512_cmpge_epi32_mask): Ditto.
1601         (_mm512_cmpge_epu32_mask): Ditto.
1602         (_mm512_cmpge_epi64_mask): Ditto.
1603         (_mm512_cmpge_epu64_mask): Ditto.
1604         (_mm512_cmple_epi32_mask): Ditto.
1605         (_mm512_cmple_epu32_mask): Ditto.
1606         (_mm512_cmple_epi64_mask): Ditto.
1607         (_mm512_cmple_epu64_mask): Ditto.
1608         (_mm512_cmplt_epi32_mask): Ditto.
1609         (_mm512_cmplt_epu32_mask): Ditto.
1610         (_mm512_cmplt_epi64_mask): Ditto.
1611         (_mm512_cmplt_epu64_mask): Ditto.
1612         (_mm512_cmpneq_epi32_mask): Ditto.
1613         (_mm512_cmpneq_epu32_mask): Ditto.
1614         (_mm512_cmpneq_epi64_mask): Ditto.
1615         (_mm512_cmpneq_epu64_mask): Ditto.
1616         (_mm512_expand_pd): Ditto.
1617         (_mm512_expand_ps): Ditto.
1618         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1619         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1620         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1621         * config/i386/i386.c (ix86_builtins): Add
1622         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1623         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1624         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1625         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1626         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1627         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1628         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1629         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1630         IX86_BUILTIN_PMOVUSQW512_MEM.
1631         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1632         __builtin_ia32_pmovsqd512mem_mask,
1633         __builtin_ia32_pmovqd512mem_mask,
1634         __builtin_ia32_pmovusqw512mem_mask,
1635         __builtin_ia32_pmovsqw512mem_mask,
1636         __builtin_ia32_pmovqw512mem_mask,
1637         __builtin_ia32_pmovusdw512mem_mask,
1638         __builtin_ia32_pmovsdw512mem_mask,
1639         __builtin_ia32_pmovdw512mem_mask,
1640         __builtin_ia32_pmovqb512mem_mask,
1641         __builtin_ia32_pmovusqb512mem_mask,
1642         __builtin_ia32_pmovsqb512mem_mask,
1643         __builtin_ia32_pmovusdb512mem_mask,
1644         __builtin_ia32_pmovsdb512mem_mask,
1645         __builtin_ia32_pmovdb512mem_mask.
1646         (bdesc_args): Add __builtin_ia32_expanddf512,
1647         __builtin_ia32_expandsf512.
1648         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1649         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1650         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1651         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1652         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1653         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1654         (avx512f_<code>v8div16qi2_mask_store): This.
1655         (avx512f_expand<mode>): New.
1657 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
1659         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
1660         New.
1661         (_mm512_mask_prefetch_i64gather_pd): Ditto.
1662         (_mm512_prefetch_i32scatter_pd): Ditto.
1663         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
1664         (_mm512_prefetch_i64scatter_pd): Ditto.
1665         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
1666         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
1667         (_mm512_mask_prefetch_i64gather_ps): Ditto.
1668         (_mm512_prefetch_i32scatter_ps): Ditto.
1669         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
1670         (_mm512_prefetch_i64scatter_ps): Ditto.
1671         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
1672         * config/i386/i386-builtin-types.def: Define
1673         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
1674         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
1675         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
1676         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
1677         IX86_BUILTIN_SCATTERPFQPD.
1678         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
1679         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
1680         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
1681         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
1682         __builtin_ia32_scatterpfqps.
1683         (ix86_expand_builtin): Expand new built-ins.
1684         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
1685         fix memory access data type.
1686         (*avx512pf_gatherpf<mode>_mask): Ditto.
1687         (*avx512pf_gatherpf<mode>): Ditto.
1688         (avx512pf_scatterpf<mode>): Ditto.
1689         (*avx512pf_scatterpf<mode>_mask): Ditto.
1690         (*avx512pf_scatterpf<mode>): Ditto.
1691         (GATHER_SCATTER_SF_MEM_MODE): New.
1692         (avx512pf_gatherpf<mode>df): Ditto.
1693         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1694         (*avx512pf_scatterpf<mode>df): Ditto.
1696 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
1698         PR bootstrap/59934
1699         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
1700         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
1701         reached.
1703 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1705         * common/config/arm/arm-common.c
1706         (arm_rewrite_mcpu): Handle multiple names.
1707         * config/arm/arm.h
1708         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1710 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1712         * gimple-builder.h (create_gimple_tmp): Delete.
1714 2014-01-27  Christian Bruel  <christian.bruel@st.com>
1716         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
1717         words comparisons.
1719 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
1721         * config/pa/pa.md (call): Generate indirect long calls to non-local
1722         functions when outputing 32-bit code.
1723         (call_value): Likewise except for special call to buggy powf function.
1725         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
1726         portable runtime and PIC indirect calls.
1727         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
1728         and PIC call sequences.  Use ldo instead of blr to set return register
1729         in PIC call sequence.
1731 2014-01-25  Walter Lee  <walt@tilera.com>
1733         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
1734         avoid clobbering a live register.
1736 2014-01-25  Walter Lee  <walt@tilera.com>
1738         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
1739         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
1740         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
1741         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
1743 2014-01-25  Walter Lee  <walt@tilera.com>
1745         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
1746         arguments on even registers.
1747         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
1748         STACK_BOUNDARY.
1749         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
1750         (BIGGEST_ALIGNMENT): Ditto.
1751         (BIGGEST_FIELD_ALIGNMENT): Ditto.
1753 2014-01-25  Walter Lee  <walt@tilera.com>
1755         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
1756         insns before bundling.
1757         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
1759 2014-01-25  Walter Lee  <walt@tilera.com>
1761         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
1762         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
1763         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
1765 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1767         * config/mips/constraints.md (kl): Delete.
1768         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
1769         define expands, using...
1770         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
1771         instructions for MIPS16.
1772         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
1773         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
1775 2014-01-25  Walter Lee  <walt@tilera.com>
1777         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
1778         (clzdi2): Ditto.
1779         (ffsdi2): Ditto.
1781 2014-01-25  Walter Lee  <walt@tilera.com>
1783         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
1784         (TARGET_EXPAND_TO_RTL_HOOK): Define.
1786 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1788         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
1789         Handle XOR.
1791 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
1793         * print-rtl.c (in_call_function_usage): New var.
1794         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
1795         EXPR_LIST mode as mode and not as reg note name.
1797         PR middle-end/59561
1798         * cfgloopmanip.c (copy_loop_info): If
1799         loop->warned_aggressive_loop_optimizations, make sure
1800         the flag is set in target loop too.
1802 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
1804         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
1805         flag_cilkplus.
1806         * builtins.def: Likewise.
1807         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
1808         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
1809         * ira.c (ira_setup_eliminable_regset): Likewise.
1810         * omp-low.c (gate_expand_omp): Likewise.
1811         (execute_lower_omp): Likewise.
1812         (diagnose_sb_0): Likewise.
1813         (gate_diagnose_omp_blocks): Likewise.
1814         (simd_clone_clauses_extract): Likewise.
1815         (gate): Likewise.
1817 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1819         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
1820         correction for little endian...
1821         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
1822         here.
1824 2014-01-24  Jeff Law  <law@redhat.com>
1826         PR tree-optimization/59919
1827         * tree-vrp.c (find_assert_locations_1): Do not register asserts
1828         for non-returning calls.
1830 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
1832         * common/config/aarch64/aarch64-common.c
1833         (aarch64_rewrite_mcpu): Handle multiple names.
1834         * config/aarch64/aarch64.h
1835         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1837 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
1839         * input.c (add_file_to_cache_tab): Handle the case where fopen
1840         returns NULL.
1842 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
1844         PR target/59929
1845         * config/i386/i386.md (pushsf splitter): Get stack adjustment
1846         from push operand if code of push isn't PRE_DEC.
1848 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
1850         PR target/59909
1851         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1852         -mquad-memory-atomic.  Update -mquad-memory documentation to say
1853         it is only used for non-atomic loads/stores.
1855         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
1856         -mquad-memory or -mquad-memory-atomic switches.
1858         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
1859         -mquad-memory-atomic to ISA 2.07 support.
1861         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
1862         to separate support of normal quad word memory operations (ldq, stq)
1863         from the atomic quad word memory operations.
1865         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
1866         support to separate non-atomic quad word operations from atomic
1867         quad word operations.  Disable non-atomic quad word operations in
1868         little endian mode so that we don't have to swap words after the
1869         load and before the store.
1870         (quad_load_store_p): Add comment about atomic quad word support.
1871         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
1872         options printed with -mdebug=reg.
1874         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
1875         -mquad-memory-atomic as the test for whether we have quad word
1876         atomic instructions.
1877         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
1878         or -mp8-vector are used, allow byte/half-word atomic operations.
1880         * config/rs6000/sync.md (load_lockedti): Insure that the address
1881         is a proper indexed or indirect address for the lqarx instruction.
1882         On little endian systems, swap the hi/lo registers after the lqarx
1883         instruction.
1884         (load_lockedpti): Use indexed_or_indirect_operand predicate to
1885         insure the address is valid for the lqarx instruction.
1886         (store_conditionalti): Insure that the address is a proper indexed
1887         or indirect address for the stqcrx. instruction.  On little endian
1888         systems, swap the hi/lo registers before doing the stqcrx.
1889         instruction.
1890         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
1891         insure the address is valid for the stqcrx. instruction.
1893         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
1894         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
1895         type of quad memory support is available.
1897 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
1899         PR regression/59915
1900         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
1901         there is a danger of looping.
1903 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
1905         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1906         force flag_ira_loop_pressure if set via command line.
1908 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1910         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
1911         (ashr_simd): New builtin handling DI mode.
1912         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
1913         (aarch64_sshr_simddi): New match pattern.
1914         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
1915         (vshrd_n_s64): Likewise.
1916         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
1918 2014-01-23  Nick Clifton  <nickc@redhat.com>
1920         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
1921         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
1922         favour of mcu specific scripts.
1923         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
1924         430x multilibs.
1926 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1927             Alex Velenko  <Alex.Velenko@arm.com>
1929         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
1930         (vaddv_s16): Likewise.
1931         (vaddv_s32): Likewise.
1932         (vaddv_u8): Likewise.
1933         (vaddv_u16): Likewise.
1934         (vaddv_u32): Likewise.
1935         (vaddvq_s8): Likewise.
1936         (vaddvq_s16): Likewise.
1937         (vaddvq_s32): Likewise.
1938         (vaddvq_s64): Likewise.
1939         (vaddvq_u8): Likewise.
1940         (vaddvq_u16): Likewise.
1941         (vaddvq_u32): Likewise.
1942         (vaddvq_u64): Likewise.
1943         (vaddv_f32): Likewise.
1944         (vaddvq_f32): Likewise.
1945         (vaddvq_f64): Likewise.
1946         (vmaxv_f32): Likewise.
1947         (vmaxv_s8): Likewise.
1948         (vmaxv_s16): Likewise.
1949         (vmaxv_s32): Likewise.
1950         (vmaxv_u8): Likewise.
1951         (vmaxv_u16): Likewise.
1952         (vmaxv_u32): Likewise.
1953         (vmaxvq_f32): Likewise.
1954         (vmaxvq_f64): Likewise.
1955         (vmaxvq_s8): Likewise.
1956         (vmaxvq_s16): Likewise.
1957         (vmaxvq_s32): Likewise.
1958         (vmaxvq_u8): Likewise.
1959         (vmaxvq_u16): Likewise.
1960         (vmaxvq_u32): Likewise.
1961         (vmaxnmv_f32): Likewise.
1962         (vmaxnmvq_f32): Likewise.
1963         (vmaxnmvq_f64): Likewise.
1964         (vminv_f32): Likewise.
1965         (vminv_s8): Likewise.
1966         (vminv_s16): Likewise.
1967         (vminv_s32): Likewise.
1968         (vminv_u8): Likewise.
1969         (vminv_u16): Likewise.
1970         (vminv_u32): Likewise.
1971         (vminvq_f32): Likewise.
1972         (vminvq_f64): Likewise.
1973         (vminvq_s8): Likewise.
1974         (vminvq_s16): Likewise.
1975         (vminvq_s32): Likewise.
1976         (vminvq_u8): Likewise.
1977         (vminvq_u16): Likewise.
1978         (vminvq_u32): Likewise.
1979         (vminnmv_f32): Likewise.
1980         (vminnmvq_f32): Likewise.
1981         (vminnmvq_f64): Likewise.
1983 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1985         * config/aarch64/aarch64-simd.md
1986         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
1987         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
1988         (*aarch64_mul3_elt<mode>): Likewise.
1989         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
1990         (*aarch64_mul3_elt_to_64v2df): Likewise.
1991         (*aarch64_mla_elt<mode>): Likewise.
1992         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
1993         (*aarch64_mls_elt<mode>): Likewise.
1994         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
1995         (*aarch64_fma4_elt<mode>): Likewise.
1996         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
1997         (*aarch64_fma4_elt_to_64v2df): Likewise.
1998         (*aarch64_fnma4_elt<mode>): Likewise.
1999         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2000         (*aarch64_fnma4_elt_to_64v2df): Likewise.
2001         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2002         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2003         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2004         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2005         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2006         (aarch64_sqdmull_lane<mode>_internal): Likewise.
2007         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2009 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2011         * config/aarch64/aarch64-simd.md
2012         (aarch64_be_checked_get_lane<mode>): New define_expand.
2013         * config/aarch64/aarch64-simd-builtins.def
2014         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2015         New builtin definition.
2016         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2017         Use new safe be builtin.
2019 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2021         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2022         New define_insn.
2023         (aarch64_be_st1<mode>): Likewise.
2024         (aarch_ld1<VALL:mode>): Define_expand modified.
2025         (aarch_st1<VALL:mode>): Likewise.
2026         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2027         (UNSPEC_ST1): Likewise.
2029 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
2031         * config/microblaze/microblaze.md: Add trap insn and attribute
2033 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
2035         PR preprocessor/58580
2036         * input.h (location_get_source_line): Take an additional line_size
2037         parameter.
2038         (void diagnostics_file_cache_fini): Declare new function.
2039         * input.c (struct fcache): New type.
2040         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2041         New static constants.
2042         (diagnostic_file_cache_init, total_lines_num)
2043         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2044         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2045         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2046         (get_next_line, read_next_line, goto_next_line, read_line_num):
2047         New static function definitions.
2048         (diagnostic_file_cache_fini): New function.
2049         (location_get_source_line): Take an additional output line_len
2050         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
2051         read_line_num.
2052         * diagnostic.c (diagnostic_finish): Call
2053         diagnostic_file_cache_fini.
2054         (adjust_line): Take an additional input parameter for the length
2055         of the line, rather than calculating it with strlen.
2056         (diagnostic_show_locus): Adjust the use of
2057         location_get_source_line and adjust_line with respect to their new
2058         signature.  While displaying a line now, do not stop at the first
2059         null byte.  Rather, display the zero byte as a space and keep
2060         going until we reach the size of the line.
2061         * Makefile.in: Add vec.o to OBJS-libcommon
2063 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2064             Ilya Tocar     <ilya.tocar@intel.com>
2066         * config/i386/avx512fintrin.h (_mm512_kmov): New.
2067         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2068         (__builtin_ia32_kmov16): Ditto.
2069         * config/i386/i386.md (UNSPEC_KMOV): New.
2070         (kmovw): Ditto.
2072 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2074         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2075         (_mm512_storeu_si512): Ditto.
2077 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
2079         PR target/52125
2080         * rtl.h (get_referenced_operands): Declare.
2081         * recog.c (get_referenced_operands): New function.
2082         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2083         operands have been referenced when recording LO_SUM references.
2085 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
2087         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2089 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2091         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2092         Enable for generic and recent AMD targets.
2094 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2096         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2097         ARG_SIZE note when adjustment was eliminated.
2099 2014-01-22  Jeff Law  <law@redhat.com>
2101         PR tree-optimization/59597
2102         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2103         in file.  Accept new argument REGISTERING and use it to modify
2104         dump output appropriately.
2105         (register_jump_thread): Corresponding changes.
2106         (mark_threaded_blocks): Reinstate code to cancel unprofitable
2107         thread paths involving joiner blocks.  Add code to dump cancelled
2108         jump threading paths.
2110 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
2112         PR rtl-optimization/59477
2113         * lra-constraints.c (inherit_in_ebb): Process call for living hard
2114         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
2116 2014-01-22  Tom Tromey  <tromey@redhat.com>
2118         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2119         PARAMS.
2120         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2122 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2124         PR rtl-optimization/59896
2125         * lra-constraints.c (process_alt_operands): Check unused note for
2126         matched operands of insn with no output reloads.
2128 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
2130         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2131         (mips_move_from_gpr_cost): Likewise.
2133 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2135         PR rtl-optimization/59858
2136         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2137         ira_class_hard_regs_num.
2138         (process_alt_operands): Increase reject for dying matched operand.
2140 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
2142         PR target/59003
2143         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2144         smaller than size, perform several stores or loads and stores
2145         at dst + count - size to store or copy all of size bytes, rather
2146         than just last modesize bytes.
2148 2014-01-20  DJ Delorie  <dj@redhat.com>
2150         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2151         that CLOBBERs are REGs before propogating their values.
2153 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2155         PR middle-end/59789
2156         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2157         (cgraph_inline_failed_type): New function.
2158         * cgraph.h (DEFCIFCODE): Add type.
2159         (cgraph_inline_failed_type_t): New enum.
2160         (cgraph_inline_failed_type): New prototype.
2161         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2162         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2163         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2164         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2165         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2166         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2167         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2168         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2169         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2170         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2171         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2172         OPTIMIZATION_MISMATCH.
2173         * tree-inline.c (expand_call_inline): Emit errors during
2174         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2176 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2178         PR target/59685
2179         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2180         mode attribute in insn output.
2182 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
2184         * output.h (output_constant): Delete.
2185         * varasm.c (output_constant): Make private.
2187 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
2189         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2191 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2193         PR middle-end/59860
2194         * tree.h (fold_builtin_strcat): New prototype.
2195         * builtins.c (fold_builtin_strcat): No longer static.  Add len
2196         argument, if non-NULL, don't call c_strlen.  Optimize
2197         directly into __builtin_memcpy instead of __builtin_strcpy.
2198         (fold_builtin_2): Adjust fold_builtin_strcat caller.
2199         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2201 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2203         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2204         for SImode_address_operand operands, having only a REG argument.
2206 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2208         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2209         loader name using mbig-endian.
2210         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2212 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2214         * doc/invoke.texi (-march): Clarify documentation for AArch64.
2215         (-mtune): Likewise.
2216         (-mcpu): Likewise.
2218 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
2220         * config/aarch64/aarch64-protos.h
2221         (aarch64_cannot_change_mode_class_ptr): Declare.
2222         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2223         aarch64_cannot_change_mode_class_ptr): New.
2224         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2225         backend hook aarch64_cannot_change_mode_class.
2227 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2229         * common/config/aarch64/aarch64-common.c
2230         (aarch64_handle_option): Don't handle any option order logic here.
2231         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2232         selected_cpu, warn on architecture version mismatch.
2233         (aarch64_override_options): Fix parsing order for option strings.
2235 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2236             Iain Sandoe  <iain@codesourcery.com>
2238         PR bootstrap/59496
2239         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2240         warning.  Amend comment to reflect current functionality.
2242 2014-01-20  Richard Biener  <rguenther@suse.de>
2244         PR middle-end/59860
2245         * builtins.c (fold_builtin_strcat): Remove case better handled
2246         by tree-ssa-strlen.c.
2248 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2250         * config/aarch64/aarch64.opt
2251         (mcpu, march, mtune): Make case-insensitive.
2253 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2255         PR target/59880
2256         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2257         if operands[1] is a REG or ZERO_EXTEND of a REG.
2259 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2261         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2263 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2265         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2266         long non-pic millicode calls.
2268 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2270         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2272 2014-01-19  Kito Cheng  <kito@0xlab.org>
2274         * builtins.c (expand_movstr): Check movstr expand done or fail.
2276 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2277             H.J. Lu  <hongjiu.lu@intel.com>
2279         PR target/59379
2280         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2281         to DImode for zero-extended addresses.
2283 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2285         PR rtl-optimization/57763
2286         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2287         on the new indirect jump_insn and increment LABEL_NUSES (label).
2289 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2291         PR bootstrap/59580
2292         PR bootstrap/59583
2293         * config.gcc (x86_archs): New variable.
2294         (x86_64_archs): Likewise.
2295         (x86_cpus): Likewise.
2296         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2297         --with-arch/--with-cpu= options.
2298         Support --with-arch=/--with-cpu={nehalem,westmere,
2299         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2301 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2303         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2304         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2306 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2308         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2310 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2312         PR target/58944
2313         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2314         clear cpp_get_options (parse_in)->warn_unused_macros for
2315         ix86_target_macros_internal with cpp_define.
2317 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2319         * jump.c (delete_related_insns): Keep (use (insn))s.
2320         * reorg.c (redundant_insn): Check for barriers too.
2322 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2324         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2326 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2328         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2329         call to $$dyncall when TARGET_LONG_CALLS is true.
2331 2014-01-17  Jeff Law  <law@redhat.com>
2333         * ree.c (combine_set_extension): Temporarily disable test for
2334         changing number of hard registers.
2336 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2338         PR middle-end/58125
2339         * ipa-inline-analysis.c (inline_free_summary):
2340         Do not free summary of aliases.
2342 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2344         PR middle-end/59706
2345         * gimplify.c (gimplify_expr): Use create_tmp_var
2346         instead of create_tmp_var_raw.  If cond doesn't have
2347         integral type, don't add the IFN_ANNOTATE builtin at all.
2349 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2351         PR ipa/59736
2352         * ipa-cp.c (prev_edge_clone): New variable.
2353         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2354         Also resize prev_edge_clone vector.
2355         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2356         (ipcp_edge_removal_hook): New function.
2357         (ipcp_driver): Register ipcp_edge_removal_hook.
2359 2014-01-17  Andrew Pinski <apinski@cavium.com>
2360             Steve Ellcey  <sellcey@mips.com>
2362         PR target/59462
2363         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2364         of operator mode.
2366 2014-01-17  Jeff Law  <law@redhat.com>
2368         PR middle-end/57904
2369         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2370         so that pass_ccp runs first.
2372 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2374         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2375         (ix86_adjust_cost): Use !TARGET_XXX.
2376         (do_reorder_for_imul): Likewise.
2377         (swap_top_of_ready_list): Likewise.
2378         (ix86_sched_reorder): Likewise.
2380 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2382         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2383         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2384         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2385         (intel_memset): New.  Duplicate slm_memset.
2386         (intel_cost): New.  Duplicate slm_cost.
2387         (m_INTEL): New macro.
2388         (processor_target_table): Add "intel".
2389         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2390         with PROCESSOR_INTEL for "intel".
2391         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2392         PROCESSOR_SILVERMONT.
2393         (ix86_issue_rate): Likewise.
2394         (ix86_adjust_cost): Likewise.
2395         (ia32_multipass_dfa_lookahead): Likewise.
2396         (swap_top_of_ready_list): Likewise.
2397         (ix86_sched_reorder): Likewise.
2398         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2399         instead of TARGET_OPT_AGU.
2400         * config/i386/i386.h (TARGET_INTEL): New.
2401         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2402         (processor_type): Add PROCESSOR_INTEL.
2403         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2404         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2406 2014-01-17  Marek Polacek  <polacek@redhat.com>
2408         PR c/58346
2409         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2410         size is zero.
2412 2014-01-17  Richard Biener  <rguenther@suse.de>
2414         PR tree-optimization/46590
2415         * opts.c (default_options_table): Add entries for
2416         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2417         all enabled at -O1 but not for -Og.
2418         * common.opt (fbranch-count-reg): Remove Init(1).
2419         (fmove-loop-invariants): Likewise.
2420         (ftree-pta): Likewise.
2422 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2424         * config/i386/i386.c (ix86_data_alignment): For compatibility with
2425         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2426         decls to at least the GCC 4.8 used alignments.
2428         PR fortran/59440
2429         * tree-nested.c (convert_nonlocal_reference_stmt,
2430         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2431         of GIMPLE_BIND stmts, adjust associated decls.
2433 2014-01-17  Richard Biener  <rguenther@suse.de>
2435         PR tree-optimization/46590
2436         * vec.h (vec<>::bseach): New member function implementing
2437         binary search according to C89 bsearch.
2438         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2439         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2440         bitmap pointer again.  Make accesses_in_loop a flat array.
2441         (mem_ref_obstack): New global.
2442         (outermost_indep_loop): Adjust for mem_ref->stored changes.
2443         (mark_ref_stored): Likewise.
2444         (ref_indep_loop_p_2): Likewise.
2445         (set_ref_stored_in_loop): New helper function.
2446         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2447         (memref_free): Adjust.
2448         (record_mem_ref_loc): Simplify.
2449         (gather_mem_refs_stmt): Adjust.
2450         (sort_locs_in_loop_postorder_cmp): New function.
2451         (analyze_memory_references): Sort accesses_in_loop after
2452         loop postorder number.
2453         (find_ref_loc_in_loop_cmp): New function.
2454         (for_all_locs_in_loop): Find relevant cluster of locs in
2455         accesses_in_loop and iterate without recursion.
2456         (execute_sm): Avoid uninit warning.
2457         (struct ref_always_accessed): Simplify.
2458         (ref_always_accessed::operator ()): Likewise.
2459         (ref_always_accessed_p): Likewise.
2460         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2461         loop postorder numbers here.
2462         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2463         numbers.
2465 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2467         PR c++/57945
2468         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2469         on decls for which assemble_alias has been called.
2471 2014-01-17  Nick Clifton  <nickc@redhat.com>
2473         * config/msp430/msp430.opt: (mcpu): New option.
2474         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2475         (msp430_option_override): Parse target_cpu.  If the MCU name
2476         matches a generic string, clear target_mcu.
2477         (msp430_attr): Allow numeric interrupt values up to 63.
2478         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2479         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2480         option.
2481         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2482         Add mcpu matches.
2483         * config/msp430/msp430.md (popm): Use %J rather than %I.
2484         (addsi3): Use msp430_nonimmediate_operand for operand 2.
2485         (addhi_cy_i): Use immediate_operand for operand 2.
2486         * doc/invoke.texi: Document -mcpu option.
2488 2014-01-17  Richard Biener  <rguenther@suse.de>
2490         PR rtl-optimization/38518
2491         * df.h (df_analyze_loop): Declare.
2492         * df-core.c: Include cfgloop.h.
2493         (df_analyze_1): Split out main part of df_analyze.
2494         (df_analyze): Adjust.
2495         (loop_inverted_post_order_compute): New function.
2496         (loop_post_order_compute): Likewise.
2497         (df_analyze_loop): New function avoiding whole-function
2498         postorder computes.
2499         * loop-invariant.c (find_defs): Use df_analyze_loop.
2500         (find_invariants): Adjust.
2501         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2503 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
2505         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2506         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2508 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
2510         * ipa-ref.c (ipa_remove_stmt_references): Fix references
2511         traversal when removing references.
2513 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
2515         PR ipa/59775
2516         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2518 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
2520         PR middle-end/56791
2521         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2522         pushing a reload for an autoinc when we had previously reloaded an
2523         inner part of the address.
2525 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
2527         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2528         field.
2529         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2530         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2531         when not giving up or versioning for alias only because of
2532         loop->safelen.
2533         (vect_analyze_data_ref_dependences): Set to true.
2534         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2535         is a GIMPLE_PHI.
2536         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2537         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2538         to the condition.
2540         PR middle-end/58344
2541         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2543         PR target/59839
2544         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2545         operand 0 predicate for gathers, use a new pseudo as subtarget.
2547 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2549         PR middle-end/59609
2550         * lra-constraints.c (process_alt_operands): Add printing debug info.
2551         Check absence of input/output reloads for matched operands too.
2553 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2555         PR rtl-optimization/59835
2556         * ira.c (ira_init_register_move_cost): Increase cost for
2557         impossible modes.
2559 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
2561         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2563 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
2565         PR target/59780
2566         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2567         non-register objects.  Use gen_(high/low)part more consistently.
2568         Fix assertions.
2570 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
2572         PR target/59844
2573         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2574         endian support, remove tests for WORDS_BIG_ENDIAN.
2575         (p8_mfvsrd_3_<mode>): Likewise.
2576         (reload_gpr_from_vsx<mode>): Likewise.
2577         (reload_gpr_from_vsxsf): Likewise.
2578         (p8_mfvsrd_4_disf): Likewise.
2580 2014-01-16  Richard Biener  <rguenther@suse.de>
2582         PR rtl-optimization/46590
2583         * lcm.c (compute_antinout_edge): Use postorder iteration.
2584         (compute_laterin): Use inverted postorder iteration.
2586 2014-01-16  Nick Clifton  <nickc@redhat.com>
2588         PR middle-end/28865
2589         * varasm.c (output_constant): Return the number of bytes actually
2590         emitted.
2591         (output_constructor_array_range): Update the field size with the
2592         number of bytes emitted by output_constant.
2593         (output_constructor_regular_field): Likewise.  Also do not
2594         complain if the total number of bytes emitted is now greater
2595         than the expected fieldpos.
2596         * output.h (output_constant): Update prototype and descriptive comment.
2598 2014-01-16  Marek Polacek  <polacek@redhat.com>
2600         PR middle-end/59827
2601         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2602         it is error_mark_node.
2604 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
2606         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2607         VALID_AVX256_REG_OR_OI_MODE.
2609 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
2611         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2612         current procedure should be profiled.
2614 2014-01-15  Andrew Pinski  <apinski@cavium.com>
2616         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2617         of moving from/to the STACK_REG register class.
2619 2014-01-15  Richard Henderson <rth@redhat.com>
2621         PR debug/54694
2622         * reginfo.c (global_regs_decl): Globalize.
2623         * rtl.h (global_regs_decl): Declare.
2624         * ira.c (do_reload): Diagnose frame_pointer_needed and it
2625         reserved via global_regs.
2627 2014-01-15  Teresa Johnson  <tejohnson@google.com>
2629         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2631 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
2633         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2634         and vmulosh rather than call gen_vec_widen_smult_*.
2635         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2636         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2637         (vec_widen_smult_even_v16qi): Likewise.
2638         (vec_widen_umult_even_v8hi): Likewise.
2639         (vec_widen_smult_even_v8hi): Likewise.
2640         (vec_widen_umult_odd_v16qi): Likewise.
2641         (vec_widen_smult_odd_v16qi): Likewise.
2642         (vec_widen_umult_odd_v8hi): Likewise.
2643         (vec_widen_smult_odd_v8hi): Likewise.
2644         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2645         vmuloub rather than call gen_vec_widen_umult_*.
2646         (vec_widen_umult_lo_v16qi): Likewise.
2647         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2648         vmulosb rather than call gen_vec_widen_smult_*.
2649         (vec_widen_smult_lo_v16qi): Likewise.
2650         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2651         rather than call gen_vec_widen_umult_*.
2652         (vec_widen_umult_lo_v8hi): Likewise.
2653         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2654         rather than call gen_vec_widen_smult_*.
2655         (vec_widen_smult_lo_v8hi): Likewise.
2657 2014-01-15  Jeff Law  <law@redhat.com>
2659         PR tree-optimization/59747
2660         * ree.c (find_and_remove_re): Properly handle case where a second
2661         eliminated extension requires widening a copy created for elimination
2662         of a prior extension.
2663         (combine_set_extension): Ensure that the number of hard regs needed
2664         for a destination register does not change when we widen it.
2666 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2668         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
2669         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
2670         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
2671         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
2672         (avr-*-rtems*): Likewise.
2673         (bfin*-rtems*): Likewise.
2674         (moxie-*-rtems*): Likewise.
2675         (h8300-*-rtems*): Likewise.
2676         (i[34567]86-*-rtems*): Likewise.
2677         (lm32-*-rtems*): Likewise.
2678         (m32r-*-rtems*): Likewise.
2679         (m68k-*-rtems*): Likewise.
2680         (microblaze*-*-rtems*): Likewise.
2681         (mips*-*-rtems*): Likewise.
2682         (powerpc-*-rtems*): Likewise.
2683         (sh-*-rtems*): Likewise.
2684         (sparc-*-rtems*): Likewise.
2685         (sparc64-*-rtems*): Likewise.
2686         (v850-*-rtems*): Likewise.
2687         (m32c-*-rtems*): Likewise.
2689 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
2691         PR rtl-optimization/59511
2692         * ira.c (ira_init_register_move_cost): Use memory costs for some
2693         cases of register move cost calculations.
2694         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
2695         instead of BB frequency.
2696         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
2697         * lra-assigns.c (find_hard_regno_for): Ditto.
2699 2014-01-15  Richard Biener  <rguenther@suse.de>
2701         PR tree-optimization/59822
2702         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
2703         (vectorizable_load): Use it to hoist defs of uses of invariant
2704         loads out of the loop.
2706 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
2707             Kugan Vivekanandarajah  <kuganv@linaro.org>
2709         PR target/59695
2710         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
2711         truncation.
2713 2014-01-15  Richard Biener  <rguenther@suse.de>
2715         PR rtl-optimization/59802
2716         * lcm.c (compute_available): Use inverted postorder to seed
2717         the initial worklist.
2719 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2721         PR target/59803
2722         * config/s390/s390.c (s390_preferred_reload_class): Don't return
2723         ADDR_REGS for invalid symrefs in non-PIC code.
2725 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
2727         PR other/58712
2728         * builtins.c (determine_block_size): Initialize *probable_max_size
2729         even if len_rtx is CONST_INT.
2731 2014-01-14  Andrew Pinski  <apinski@cavium.com>
2733         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
2734         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
2735         (cortexa53_tunings): Likewise.
2736         (aarch64_sched_issue_rate): New function.
2737         (TARGET_SCHED_ISSUE_RATE): Define.
2739 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2741         * ira-costs.c (find_costs_and_classes): Add missed
2742         ira_init_register_move_cost_if_necessary.
2744 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2746         PR target/59787
2747         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
2749 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
2751         PR target/59794
2752         * config/i386/i386.c (type_natural_mode): Add a bool parameter
2753         to indicate if type is used for function return value.  Warn ABI
2754         change if the vector mode isn't available for function return value.
2755         (ix86_function_arg_advance): Pass false to type_natural_mode.
2756         (ix86_function_arg): Likewise.
2757         (ix86_gimplify_va_arg): Likewise.
2758         (function_arg_32): Don't warn ABI change.
2759         (ix86_function_value): Pass true to type_natural_mode.
2760         (ix86_return_in_memory): Likewise.
2761         (ix86_struct_value_rtx): Removed.
2762         (TARGET_STRUCT_VALUE_RTX): Likewise.
2764 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2766         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
2767         converting a conditional jump into a conditional return.
2769 2014-01-14  Richard Biener  <rguenther@suse.de>
2771         PR tree-optimization/58921
2772         PR tree-optimization/59006
2773         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
2774         hoisting invariant stmts.
2775         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
2776         invariant loads on the preheader edge if possible.
2778 2014-01-14  Joey Ye  <joey.ye@arm.com>
2780         * doc/plugin.texi (Building GCC plugins): Update to C++.
2782 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
2784         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
2785         (_mm_rcp28_round_ss): Ditto.
2786         (_mm_rsqrt28_round_sd): Ditto.
2787         (_mm_rsqrt28_round_ss): Ditto.
2788         (_mm_rcp28_sd): Ditto.
2789         (_mm_rcp28_ss): Ditto.
2790         (_mm_rsqrt28_sd): Ditto.
2791         (_mm_rsqrt28_ss): Ditto.
2792         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
2793         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
2794         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
2795         (IX86_BUILTIN_RCP28SD): Ditto.
2796         (IX86_BUILTIN_RCP28SS): Ditto.
2797         (IX86_BUILTIN_RSQRT28SD): Ditto.
2798         (IX86_BUILTIN_RSQRT28SS): Ditto.
2799         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
2800         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
2801         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
2802         (ix86_expand_special_args_builtin): Expand new FTYPE.
2803         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
2804         (srcp14<mode>): Make insn unary.
2805         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
2806         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
2807         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2808         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
2809         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
2810         Fix rounding: make it SAE only.
2811         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2812         Ditto.
2813         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2814         Ditto.
2815         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
2816         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
2817         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2818         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
2819         (round_saeonly_mask_scalar_operand4): Ditto.
2820         (round_saeonly_mask_scalar_op3): Ditto.
2821         (round_saeonly_mask_scalar_op4): Ditto.
2823 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2825         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2826         Implement -maltivec=be for vec_insert and vec_extract.
2828 2014-01-10  DJ Delorie  <dj@redhat.com>
2830         * config/msp430/msp430.md (call_internal): Don't allow memory
2831         references with SP as the base register.
2832         (call_value_internal): Likewise.
2833         * config/msp430/constraints.md (Yc): New.  For memory references
2834         that don't use SP as a base register.
2836         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
2837         "an integer without a # prefix"
2838         * config/msp430/msp430.md (epilogue_helper): Use it.
2840 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2842         PR target/59617
2843         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
2844         AVX512F gather builtins.
2845         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
2846         on gather decls with INTEGER_TYPE masktype.
2847         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
2848         directly into the builtin rather than hoisting it before loop.
2850         PR tree-optimization/59387
2851         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
2852         (scev_const_prop): If folded_casts and type has undefined overflow,
2853         use force_gimple_operand instead of force_gimple_operand_gsi and
2854         for each added stmt if it is assign with
2855         arith_code_with_undefined_signed_overflow, call
2856         rewrite_to_defined_overflow.
2857         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
2858         gimple-fold.h instead.
2859         (arith_code_with_undefined_signed_overflow,
2860         rewrite_to_defined_overflow): Moved to ...
2861         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
2862         rewrite_to_defined_overflow): ... here.  No longer static.
2863         Include gimplify-me.h.
2864         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
2865         rewrite_to_defined_overflow): New prototypes.
2867 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2869         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
2871 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
2873         * builtins.c (get_object_alignment_2): Minor tweak.
2874         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
2876 2014-01-13  Christian Bruel  <christian.bruel@st.com>
2878         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
2879         optimized non constant lengths.
2881 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2883         PR libgomp/59194
2884         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
2885         load as __atomic_load_N if possible.
2887 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2889         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
2890         target parameter.
2891         (rs6000_expand_builtin): Adjust call.
2893 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2895         PR target/58115
2896         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
2897         * config/rs6000/rs6000.c: Include target-globals.h.
2898         (rs6000_set_current_function): Instead of doing target_reinit
2899         unconditionally, use save_target_globals_default_opts and
2900         restore_target_globals.
2902         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
2903         FPSCR.
2904         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
2905         (rs6000_expand_builtin): Handle mffs and mtfsf.
2906         (rs6000_init_builtins): Define mffs and mtfsf.
2907         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
2908         (rs6000_mffs): New pattern.
2909         (rs6000_mtfsf): New pattern.
2911 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
2913         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
2914         Start narrowing with START.  Apply candidate-use pair
2915         and check overall cost in narrowing.
2916         (iv_ca_prune): Pass new argument.
2918 2014-01-10  Jeff Law  <law@redhat.com>
2920         PR middle-end/59743
2921         * ree.c (combine_reaching_defs): Ensure the defining statement
2922         occurs before the extension when optimizing extensions with
2923         different source and destination hard registers.
2925 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
2927         PR ipa/58585
2928         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
2929         vtables into the type inheritance graph.
2931 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2933         PR rtl-optimization/59754
2934         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
2935         modes in the REGNO != REGNO case.
2937 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2939         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
2941 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2943         PR tree-optimization/59745
2944         * tree-predcom.c (tree_predictive_commoning_loop): Call
2945         free_affine_expand_cache if giving up because components is NULL.
2947         * target-globals.c (save_target_globals): Allocate < 4KB structs using
2948         GC in payload of target_globals struct instead of allocating them on
2949         the heap and the larger structs separately using GC.
2950         * target-globals.h (struct target_globals): Make regs, hard_regs,
2951         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
2952         of GTY((skip)) and change type to void *.
2953         (reset_target_globals): Cast loads from those fields to corresponding
2954         types.
2956 2014-01-10  Steve Ellcey  <sellcey@mips.com>
2958         PR plugins/59335
2959         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
2960         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
2961         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
2963 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
2965         PR target/59744
2966         * aarch64-modes.def (CC_Zmode): New flags mode.
2967         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
2968         represents an equality.
2969         (aarch64_get_condition_code): Handle CC_Zmode.
2970         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
2972 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2974         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
2975         extraction in good case.
2977 2014-01-10  Richard Biener  <rguenther@suse.de>
2979         PR tree-optimization/59374
2980         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
2981         checking after SLP discovery.  Mark stmts not participating
2982         in any SLP instance properly.
2984 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2986         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
2987         when handling a SET rtx.
2989 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2991         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
2992         (cortex-a57): Likewise.
2993         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
2995 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2997         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
2998         non-iwmmxt builtins.
3000 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3002         PR ipa/58252
3003         PR ipa/59226
3004         * ipa-devirt.c record_target_from_binfo): Take as argument
3005         stack of binfos and lookup matching one for virtual inheritance.
3006         (possible_polymorphic_call_targets_1): Update.
3008 2014-01-10  Huacai Chen  <chenhc@lemote.com>
3010         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3011         kernel strings for Loongson-2E/2F/3A.
3013 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3015         PR middle-end/59670
3016         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3017         is_gimple_call before calling gimple_call_internal_p.
3019 2014-01-09  Steve Ellcey  <sellcey@mips.com>
3021         * Makefile.in (TREE_FLOW_H): Remove.
3022         (TREE_SSA_H): Add file names from tree-flow.h.
3023         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3024         * tree.h: Remove tree-flow.h reference.
3025         * hash-table.h: Remove tree-flow.h reference.
3026         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3027         reference with tree-ssa-loop.h.
3029 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3031         * doc/invoke.texi: Add -maltivec={be,le} options, and document
3032         default element-order behavior for -maltivec.
3033         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3034         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3035         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3036         when targeting big endian, at least for now.
3037         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3039 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3041         PR middle-end/47735
3042         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3043         var satisfies use_register_for_decl, just take into account type
3044         alignment, rather than decl alignment.
3046         PR tree-optimization/59622
3047         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
3048         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3049         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3050         Don't devirtualize for inplace at all.  For targets.length () == 1,
3051         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3053 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3055         * config/i386/i386.md (cpu): Remove the unused btver1.
3057 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3059         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3061 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3063         PR target/58115
3064         * tree-core.h (struct target_globals): New forward declaration.
3065         (struct tree_target_option): Add globals field.
3066         * tree.h (TREE_TARGET_GLOBALS): Define.
3067         (prepare_target_option_nodes_for_pch): New prototype.
3068         * target-globals.h (struct target_globals): Define even if
3069         !SWITCHABLE_TARGET.
3070         * tree.c (prepare_target_option_node_for_pch,
3071         prepare_target_option_nodes_for_pch): New functions.
3072         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3073         * config/i386/i386.c: Include target-globals.h.
3074         (ix86_set_current_function): Instead of doing target_reinit
3075         unconditionally, use save_target_globals_default_opts and
3076         restore_target_globals.
3078 2014-01-09  Richard Biener  <rguenther@suse.de>
3080         PR tree-optimization/59715
3081         * tree-cfg.h (split_critical_edges): Declare.
3082         * tree-cfg.c (split_critical_edges): Export.
3083         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3085 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
3087         * cfgexpand.c (expand_stack_vars): Optionally disable
3088         asan stack protection.
3089         (expand_used_vars): Likewise.
3090         (partition_stack_vars): Likewise.
3091         * asan.c (asan_emit_stack_protection): Optionally disable
3092         after return stack usage.
3093         (instrument_derefs): Optionally disable memory access instrumentation.
3094         (instrument_builtin_call): Likewise.
3095         (instrument_strlen_call): Likewise.
3096         (asan_protect_global): Optionally disable global variables protection.
3097         * doc/invoke.texi: Added doc for new options.
3098         * params.def: Added new options.
3099         * params.h: Likewise.
3101 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3103         PR rtl-optimization/59724
3104         * ifcvt.c (cond_exec_process_if_block): Don't call
3105         flow_find_head_matching_sequence with 0 longest_match.
3106         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3107         non-active insns if !stop_after.
3108         (try_head_merge_bb): Revert 2014-01-07 changes.
3110 2014-01-08  Jeff Law  <law@redhat.com>
3112         * ree.c (get_sub_rtx): New function, extracted from...
3113         (merge_def_and_ext): Here.
3114         (combine_reaching_defs): Use get_sub_rtx.
3116 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
3118         * cgraph.h (varpool_variable_node): Do not choke on null node.
3120 2014-01-08  Catherine Moore  <clm@codesourcery.com>
3122         * config/mips/mips.md (simple_return): Attempt to use JRC
3123         for microMIPS.
3124         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3126 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3128         PR rtl-optimization/59137
3129         * reorg.c (steal_delay_list_from_target): Call update_block for
3130         elided insns.
3131         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3133 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3135         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3136         two duplicate entries.
3138 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3140         Revert:
3141         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
3143         * config/mips/mips.c (mips_truncated_op_cost): New function.
3144         (mips_rtx_costs): Adjust test for BADDU.
3145         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3147         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
3149         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3150         (*baddu_si): ...this new pattern.
3152 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3154         PR ipa/59722
3155         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3157 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3159         PR middle-end/57748
3160         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3161         inner_reference_p.
3162         (expand_expr, expand_normal): Adjust.
3163         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3164         inner_reference_p. Use inner_reference_p to expand inner references.
3165         (store_expr): Adjust.
3166         * cfgexpand.c (expand_call_stmt): Adjust.
3168 2014-01-08  Rong Xu  <xur@google.com>
3170         * gcov-io.c (gcov_var): Move from gcov-io.h.
3171         (gcov_position): Ditto.
3172         (gcov_is_error): Ditto.
3173         (gcov_rewrite): Ditto.
3174         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3175         only part to libgcc/libgcov.h.
3177 2014-01-08  Marek Polacek  <polacek@redhat.com>
3179         PR middle-end/59669
3180         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3182 2014-01-08  Marek Polacek  <polacek@redhat.com>
3184         PR sanitizer/59667
3185         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3187 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3189         PR rtl-optimization/59649
3190         * stor-layout.c (get_mode_bounds): For BImode return
3191         0 and STORE_FLAG_VALUE.
3193 2014-01-08  Richard Biener  <rguenther@suse.de>
3195         PR middle-end/59630
3196         * gimple.h (is_gimple_builtin_call): Remove.
3197         (gimple_builtin_call_types_compatible_p): New.
3198         (gimple_call_builtin_p): New overload.
3199         * gimple.c (is_gimple_builtin_call): Remove.
3200         (validate_call): Rename to ...
3201         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
3202         check return types.
3203         (validate_type): New static function.
3204         (gimple_call_builtin_p): New overload and adjust.
3205         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3206         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
3207         (gimple_fold_stmt_to_constant_1): Likewise.
3208         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3210 2014-01-08  Richard Biener  <rguenther@suse.de>
3212         PR middle-end/59471
3213         * gimplify.c (gimplify_expr): Gimplify register-register type
3214         VIEW_CONVERT_EXPRs to separate stmts.
3216 2014-01-07  Jeff Law  <law@redhat.com>
3218         PR middle-end/53623
3219         * ree.c (combine_set_extension): Handle case where source
3220         and destination registers in an extension insn are different.
3221         (combine_reaching_defs): Allow source and destination registers
3222         in extension to be different under limited circumstances.
3223         (add_removable_extension): Remove restriction that the
3224         source and destination registers in the extension are the same.
3225         (find_and_remove_re): Emit a copy from the extension's
3226         destination to its source after the defining insn if
3227         the source and destination registers are different.
3229         PR middle-end/59285
3230         * ifcvt.c (merge_if_block): If we are merging a block with more than
3231         one successor with a block with no successors, remove any BARRIER
3232         after the second block.
3234 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3236         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3238 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3240         PR target/59652
3241         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3242         for 14-bit register offsets when INT14_OK_STRICT is false.
3244 2014-01-07  Roland Stigge  <stigge@antcom.de>
3245             Michael Meissner  <meissner@linux.vnet.ibm.com>
3247         PR 57386/target
3248         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3249         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3251 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3253         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3254         -mcpu.
3256 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3258         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3259         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3260         rtx is const0_rtx or not.
3262 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3264         PR target/58115
3265         * target-globals.c (save_target_globals): Remove this_fn_optab
3266         handling.
3267         * toplev.c: Include optabs.h.
3268         (target_reinit): Temporarily restore the global options if another
3269         set of options are in force.
3271 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3273         PR rtl-optimization/58668
3274         * cfgcleanup.c (flow_find_cross_jump): Don't count
3275         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3276         to determine what is counted.
3277         (flow_find_head_matching_sequence): Use active_insn_p to determine
3278         what is counted.
3279         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3280         counting change.
3281         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3282         determine what is counted.
3284         PR tree-optimization/59643
3285         * tree-predcom.c (split_data_refs_to_components): If one dr is
3286         read and one write, determine_offset fails and the write isn't
3287         in the bad component, just put the read into the bad component.
3289 2014-01-07  Mike Stump  <mikestump@comcast.net>
3290             Jakub Jelinek  <jakub@redhat.com>
3292         PR pch/59436
3293         * tree-core.h (struct tree_optimization_option): Change optabs
3294         type from unsigned char * to void *.
3295         * optabs.c (init_tree_optimization_optabs): Adjust
3296         TREE_OPTIMIZATION_OPTABS initialization.
3298 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3300         PR target/59644
3301         * config/i386/i386.h (struct machine_function): Add
3302         no_drap_save_restore field.
3303         * config/i386/i386.c (ix86_save_reg): Use
3304         !cfun->machine->no_drap_save_restore instead of
3305         crtl->stack_realign_needed.
3306         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3307         this function clears frame_pointer_needed.  Set
3308         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3309         and DRAP reg is needed.
3311 2014-01-06  Marek Polacek  <polacek@redhat.com>
3313         PR c/57773
3314         * doc/implement-c.texi: Mention that other integer types are
3315         permitted as bit-field types in strictly conforming mode.
3317 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3319         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3320         is newly allocated.
3322 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3324         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3326 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3328         PR ipa/59008
3329         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3330         to int.
3331         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3333 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3335         PR debug/59350
3336         PR debug/59510
3337         * var-tracking.c (add_stores): Preserve the value of the source even if
3338         we don't record the store.
3340 2014-01-06  Terry Guo  <terry.guo@arm.com>
3342         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3344 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3346         PR bootstrap/59541
3347         * config/darwin.c (darwin_function_section): Adjust return values to
3348         correspond to optimisation changes made in r206070.
3350 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3352         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3353         from prefetch_block tune setting.
3354         (nocona_cost): Correct size of prefetch block to 64.
3356 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3358         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3359         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3360         used to save the static chain register in the computation of the offset
3361         from which the FP registers need to be restored.
3363 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3365         PR tree-optimization/59519
3366         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3367         ICE if get_current_def (current_new_name) is already non-NULL, as long
3368         as it is a phi result of some other phi in *new_exit_bb that has
3369         the same argument.
3371         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3372         or vmovdqu* for misaligned_operand.
3373         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3374         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3375         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3376         aligned_mem for AVX512F masked aligned load and store builtins and for
3377         non-temporal moves.
3379 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3381         PR tree-optimization/59651
3382         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3383         Address range for negative step should be added by TYPE_SIZE_UNIT.
3385 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3387         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3388         overlapping registers also for registers other than the stack pointer.
3390 2014-01-03  Marek Polacek  <polacek@redhat.com>
3392         PR other/59661
3393         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3394         __builtin_FILE.
3396 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
3398         PR target/59625
3399         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3400         asm goto as jump.
3402         * config/i386/i386.md (MODE_SIZE): New mode attribute.
3403         (push splitter): Use <P:MODE_SIZE> instead of
3404         GET_MODE_SIZE (<P:MODE>mode).
3405         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3406         (mov -1, reg peephole2): Likewise.
3407         * config/i386/sse.md (*mov<mode>_internal,
3408         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3409         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3410         *<code><mode>3, *andnot<mode>3<mask_name>,
3411         <mask_codefor><code><mode>3<mask_name>): Likewise.
3412         * config/i386/subst.md (mask_mode512bit_condition,
3413         sd_mask_mode512bit_condition): Likewise.
3415 2014-01-02  Xinliang David Li  <davidxl@google.com>
3417         PR tree-optimization/59303
3418         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3419         (dump_predicates): Better output format.
3420         (pred_equal_p): New function.
3421         (is_neq_relop_p): Ditto.
3422         (is_neq_zero_form_p): Ditto.
3423         (pred_expr_equal_p): Ditto.
3424         (pred_neg_p): Ditto.
3425         (simplify_pred): Ditto.
3426         (simplify_preds_2): Ditto.
3427         (simplify_preds_3): Ditto.
3428         (simplify_preds_4): Ditto.
3429         (simplify_preds): Ditto.
3430         (push_pred): Ditto.
3431         (push_to_worklist): Ditto.
3432         (get_pred_info_from_cmp): Ditto.
3433         (is_degenerated_phi): Ditto.
3434         (normalize_one_pred_1): Ditto.
3435         (normalize_one_pred): Ditto.
3436         (normalize_one_pred_chain): Ditto.
3437         (normalize_preds): Ditto.
3438         (normalize_cond_1): Remove function.
3439         (normalize_cond): Ditto.
3440         (is_gcond_subset_of): Ditto.
3441         (is_subset_of_any): Ditto.
3442         (is_or_set_subset_of): Ditto.
3443         (is_and_set_subset_of): Ditto.
3444         (is_norm_cond_subset_of): Ditto.
3445         (pred_chain_length_cmp): Ditto.
3446         (convert_control_dep_chain_into_preds): Type change.
3447         (find_predicates): Ditto.
3448         (find_def_preds): Ditto.
3449         (destroy_predicates_vecs): Ditto.
3450         (find_matching_predicates_in_rest_chains): Ditto.
3451         (use_pred_not_overlap_with_undef_path_pred): Ditto.
3452         (is_pred_expr_subset): Ditto.
3453         (is_pred_chain_subset_of): Ditto.
3454         (is_included_in): Ditto.
3455         (is_superset_of): Ditto.
3457 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3459         Update copyright years.
3461 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3463         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3464         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3465         config/arc/arc.md, config/arc/arc.opt,
3466         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3467         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3468         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3469         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3470         config/linux-protos.h, config/linux.c, config/winnt-c.c,
3471         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3472         vtable-verify.c, vtable-verify.h: Use the standard form for the
3473         copyright notice.
3475 2014-01-02  Tobias Burnus  <burnus@net-b.de>
3477         * gcc.c (process_command): Update copyright notice dates.
3478         * gcov-dump.c: Ditto.
3479         * gcov.c: Ditto.
3480         * doc/cpp.texi: Bump @copying's copyright year.
3481         * doc/cppinternals.texi: Ditto.
3482         * doc/gcc.texi: Ditto.
3483         * doc/gccint.texi: Ditto.
3484         * doc/gcov.texi: Ditto.
3485         * doc/install.texi: Ditto.
3486         * doc/invoke.texi: Ditto.
3488 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3490         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3492 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
3494         * config/i386/sse.md (*mov<mode>_internal): Guard
3495         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3497         PR rtl-optimization/59647
3498         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3499         new_rtx into UNSIGNED_FLOAT rtxes.
3501 Copyright (C) 2014 Free Software Foundation, Inc.
3503 Copying and distribution of this file, with or without modification,
3504 are permitted in any medium without royalty provided the copyright
3505 notice and this notice are preserved.