PR target/59794
[official-gcc.git] / gcc / ChangeLog
blob5000f33123be35d1791e13694a0ca97ffcf45f69
1 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
3         PR target/59794
4         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
5         only when -Wpsabi is enabled.
7 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
9          PR target/59799
10         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
11         passing arrays in registers are the same as for structs, so remove the
12         special case for them.
14 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
16         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
17         destination type, extract only the valid bits if the source type is not
18         integral and has a different mode.
20 2014-02-19  Richard Biener  <rguenther@suse.de>
22         PR ipa/60243
23         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
24         for all calls.
26 2014-02-19  Richard Biener  <rguenther@suse.de>
28         PR ipa/60243
29         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
30         (ipa_modify_call_arguments): Emit an argument load explicitely and
31         preserve virtual SSA form there and for the replacement call.
32         Do not update SSA form nor free dominance info.
34 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
36         * ipa.c (function_and_variable_visibility): Also clear WEAK
37         flag when disolving COMDAT_GROUP.
39 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
41         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
42         * ipa-prop.c (ipa_set_jf_known_type): Return early when
43         not devirtualizing.
44         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
45         do more sanity checks.
46         (detect_type_change): Return true when giving up early.
47         (compute_complex_assign_jump_func): Fix type parameter of
48         ipa_set_ancestor_jf.
49         (compute_complex_ancestor_jump_func): Likewise.
50         (update_jump_functions_after_inlining): Fix updating of
51         ancestor function.
52         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
54 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
56         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
57         inline clones when edge disappears.
59 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
61         PR target/60203
62         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
63         Split 64-bit moves into 2 patterns.  Do not allow the use of
64         direct move for TDmode in little endian, since the decimal value
65         has little endian bytes within a word, but the 64-bit pieces are
66         ordered in a big endian fashion, and normal subreg's of TDmode are
67         not allowed.
68         (mov<mode>_64bit_dm): Likewise.
69         (movtd_64bit_nodm): Likewise.
71 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
73         PR tree-optimization/60174
74         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
75         statement of an SSA_NAME that occurs in an abnormal PHI node.
77 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
79         PR sanitizer/60142
80         * final.c (SEEN_BB): Remove.
81         (SEEN_NOTE, SEEN_EMITTED): Renumber.
82         (final_scan_insn): Don't force_source_line on second
83         NOTE_INSN_BASIC_BLOCK.
85 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
87         PR target/60205
88         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
89         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
90         (type_natural_mode): Warn ABI change when %zmm register is not
91         available for AVX512F vector value passing.
93 2014-02-18  Kai Tietz  <ktietz@redhat.com>
95         PR target/60193
96         * config/i386/i386.c (ix86_expand_prologue): Use value in
97         rax register as displacement when restoring %r10 or %rax.
98         Fix wrong offset when restoring both registers.
100 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
102         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
103         assertion with conditional return.
105 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
106             Uros Bizjak  <ubizjak@gmail.com>
108         PR driver/60233
109         * config/i386/driver-i386.c (host_detect_local_cpu): If
110         YMM state is not saved by the OS, also clear has_f16c.  Move
111         CPUID 0x80000001 handling before YMM state saving checking.
113 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
115         PR rtl-optimization/58960
116         * haifa-sched.c (alloc_global_sched_pressure_data): New,
117         factored out from ...
118         (sched_init): ... here.
119         (free_global_sched_pressure_data): New, factored out from ...
120         (sched_finish): ... here.
121         * sched-int.h (free_global_sched_pressure_data): Declare.
122         * sched-rgn.c (nr_regions_initial): New static global.
123         (haifa_find_rgns): Initialize it.
124         (schedule_region): Disable sched-pressure for the newly
125         generated regions.
127 2014-02-17  Richard Biener  <rguenther@suse.de>
129         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
130         release SSA defs of pattern stmts.
132 2014-02-17  Richard Biener  <rguenther@suse.de>
134         * tree-inline.c (expand_call_inline): Release the virtual
135         operand defined by the call we are about to inline.
137 2014-02-17  Richard Biener  <rguenther@suse.de>
139         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
141 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
142             Ilya Tocar  <ilya.tocar@intel.com>
144         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
145         arguments order in builtin.
146         (_mm512_permutexvar_epi64): Ditto.
147         (_mm512_mask_permutexvar_epi64): Ditto
148         (_mm512_maskz_permutexvar_epi32): Ditto
149         (_mm512_permutexvar_epi32): Ditto
150         (_mm512_mask_permutexvar_epi32): Ditto
152 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
154         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
155         (p8_vmrgow): Likewise.
157 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
159         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
160         endian targets.
162 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
164         PR target/60203
165         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
166         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
167         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
168         using direct move instructions on ISA 2.07.  Also adjust
169         instruction length for 64-bit.
170         (mov<mode>_64bit, TFmode/TDmode): Likewise.
171         (mov<mode>_32bit, TFmode/TDmode): Likewise.
173 2014-02-15  Alan Modra  <amodra@gmail.com>
175         PR target/58675
176         PR target/57935
177         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
178         find_replacement on parts of insn rtl that might be reloaded.
180 2014-02-15  Richard Biener  <rguenther@suse.de>
182         PR tree-optimization/60183
183         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
184         (tree_ssa_phiprop): Calculate and free post-dominators.
186 2014-02-14  Jeff Law  <law@redhat.com>
188         PR rtl-optimization/60131
189         * ree.c (get_extended_src_reg): New function.
190         (combine_reaching_defs): Use it rather than assuming location of REG.
191         (find_and_remove_re): Verify first operand of extension is
192         a REG before adding the insns to the copy list.
194 2014-02-14  Roland McGrath  <mcgrathr@google.com>
196         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
197         * configure: Regenerated.
198         * config.in: Regenerated.
199         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
200         instead of ASM_SHORT.
202 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
203             Richard Earnshaw  <rearnsha@arm.com>
205         PR rtl-optimization/59535
206         * lra-constraints.c (process_alt_operands): Encourage alternative
207         when unassigned pseudo class is superset of the alternative class.
208         (inherit_reload_reg): Don't inherit when optimizing for code size.
209         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
210         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
211         modes not less than 4 for Thumb1.
213 2014-02-14  Kyle McMartin  <kyle@redhat.com>
215         PR pch/60010
216         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
218 2014-02-14  Richard Biener  <rguenther@suse.de>
220         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
221         (get_frame_arg): Drop the assert with langhook types_compatible_p.
222         Do not strip INDIRECT_REFs.
224 2014-02-14  Richard Biener  <rguenther@suse.de>
226         PR lto/60179
227         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
228         DECL_FUNCTION_SPECIFIC_TARGET.
229         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
230         * tree-streamer-out.c (pack_ts_target_option): Remove.
231         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
232         (write_ts_function_decl_tree_pointers): Do not stream
233         DECL_FUNCTION_SPECIFIC_TARGET.
234         * tree-streamer-in.c (unpack_ts_target_option): Remove.
235         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
236         (lto_input_ts_function_decl_tree_pointers): Do not stream
237         DECL_FUNCTION_SPECIFIC_TARGET.
239 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
241         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
242         (get_initial_def_for_induction, vectorizable_induction): Ignore
243         debug stmts when looking for exit_phi.
244         (vectorizable_live_operation): Fix up condition.
246 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
248         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
249         nreverse() because it changes the content of original tree list.
251 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
253         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
254         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
256 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
258         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
259         GNU coding standards.
261 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
263         PR debug/60152
264         * dwarf2out.c (gen_subprogram_die): Don't call
265         add_calling_convention_attribute if subr_die is old_die.
267 2014-02-13  Sharad Singhai  <singhai@google.com>
269         * doc/optinfo.texi: Fix order of nodes.
271 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
273         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
274         operands[2], not operands[3].
276 2014-02-13  Richard Biener  <rguenther@suse.de>
278         PR bootstrap/59878
279         * doc/install.texi (ISL): Update recommended version to 0.12.2,
280         mention the possibility of an in-tree build.
281         (CLooG): Update recommended version to 0.18.1, mention the
282         possibility of an in-tree build and clarify that the ISL
283         bundled with CLooG does not work.
285 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
287         PR target/43546
288         * expr.c (compress_float_constant): If x is a hard register,
289         extend into a pseudo and then move to x.
291 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
293         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
294         caused by bad second argument to warning_at() with -mhotpatch and
295         nested functions (e.g. with gfortran).
297 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
299         * opts.c (option_name): Remove "enabled by default" rider.
301 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
303         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
305 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
306             Uros Bizjak  <ubizjak@gmail.com>
308         PR target/60151
309         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
310         * configure: Regenerated.
312 2014-02-12  Richard Biener  <rguenther@suse.de>
314         * vec.c (vec_prefix::calculate_allocation): Move as
315         inline variant to vec.h.
316         (vec_prefix::calculate_allocation_1): New out-of-line version.
317         * vec.h (vec_prefix::calculate_allocation_1): Declare.
318         (vec_prefix::m_has_auto_buf): Rename to ...
319         (vec_prefix::m_using_auto_storage): ... this.
320         (vec_prefix::calculate_allocation): Inline the easy cases
321         and dispatch to calculate_allocation_1 which doesn't need the
322         prefix address.
323         (va_heap::reserve): Use gcc_checking_assert.
324         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
325         m_using_auto_storage.
326         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
327         member and adjust.
328         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
329         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
330         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
332 2014-02-12  Richard Biener  <rguenther@suse.de>
334         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
335         when we found a dependence.
337 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
339         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
340         common code...
341         (maybe_fold_stmt): ... into this new function.
342         * omp-low.c (lower_omp): Update comment.
344         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
345         last use.
347         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
348         dereference.
350 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
352         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
353         identifiers in comments.
354         (cortexa53_extra_costs): Likewise.
355         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
356         (cortexa7_extra_costs): Likewise.
357         (cortexa12_extra_costs): Likewise.
358         (cortexa15_extra_costs): Likewise.
359         (v7m_extra_costs): Likewise.
361 2014-02-12  Richard Biener  <rguenther@suse.de>
363         PR middle-end/60092
364         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
365         of posix_memalign being successful.
366         (lower_stmt): Restrict lowering of posix_memalign to when
367         -ftree-bit-ccp is enabled.
369 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
371         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
372         arg_loc.
373         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
375 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
377         PR rtl-optimization/60116
378         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
379         other_insn once the combination has been validated.
381 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
383         PR lto/59468
384         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
385         and wrapper.
386         * ipa-devirt.c: Include demangle.h
387         (odr_violation_reported): New static variable.
388         (add_type_duplicate): Update odr_violations.
389         (maybe_record_node): Add completep parameter; update it.
390         (record_target_from_binfo): Add COMPLETEP parameter;
391         update it as needed.
392         (possible_polymorphic_call_targets_1): Likewise.
393         (struct polymorphic_call_target_d): Add nonconstruction_targets;
394         rename FINAL to COMPLETE.
395         (record_targets_from_bases): Sanity check we found the binfo;
396         fix COMPLETEP updating.
397         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
398         parameter, fix computing of COMPLETEP.
399         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
400         at LTO time do demangling.
401         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
402         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
403         parameter.
404         (gimple_get_virt_method_for_binfo): Likewise.
405         * gimple-fold.h (gimple_get_virt_method_for_binfo,
406         gimple_get_virt_method_for_vtable): Update prototypes.
408 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
410         PR target/49008
411         * genautomata.c (add_presence_absence): Fix typo with
412         {final_}presence_list.
414 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
416         PR target/60137
417         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
418         for VSX/Altivec vectors that land in GPR registers.
420 2014-02-11  Richard Henderson  <rth@redhat.com>
421             Jakub Jelinek  <jakub@redhat.com>
423         PR debug/59776
424         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
425         around drhs if type conversion to lacc->type is not useless.
427 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
429         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
430         tuning struct.
431         (cortex-a57.cortex-a53): Likewise.
432         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
434 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
436         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
437         arm_restrict_it.
439 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
441         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
442         add_options_for_arm_vfp3.
444 2014-02-11  Jeff Law  <law@redhat.com>
446         PR middle-end/54041
447         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
448         object with an undesirable mode.
450 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
452         PR libgomp/60107
453         * config/i386/sol2-9.h: New file.
454         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
455         *-*-solaris2.9*): Use it.
457 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
459         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
460         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
462 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
464         * config/microblaze/microblaze.c: Extend mcpu version format
466 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
468         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
470 2014-02-10  Richard Henderson <rth@redhat.com>
472         PR target/59927
473         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
474         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
475         ms-abi vs -mno-accumulate-outgoing-args.
476         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
477         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
478         respect to ms-abi.
480 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
482         PR middle-end/60080
483         * cfgexpand.c (expand_asm_operands): Attach source location to
484         ASM_INPUT rtx objects.
485         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
487 2014-02-10  Nick Clifton  <nickc@redhat.com>
489         * config/mn10300/mn10300.c (popcount): New function.
490         (mn10300_expand_prologue): Include saved registers in stack usage
491         count.
493 2014-02-10  Jeff Law  <law@redhat.com>
495         PR middle-end/52306
496         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
497         when changing the SET_DEST of a prior insn to avoid an input reload.
499 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
501         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
502         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
503         -mcall-openbsd, or -mcall-linux.
504         (CC1_ENDIAN_BIG_SPEC): Remove.
505         (CC1_ENDIAN_LITTLE_SPEC): Remove.
506         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
507         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
508         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
509         and %cc1_endian_default.
510         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
512 2014-02-10  Richard Biener  <rguenther@suse.de>
514         PR tree-optimization/60115
515         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
516         MEM_REF handling.  Properly verify that the accesses are not
517         out of the objects bound.
519 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
521         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
522         coretex to cortex.
524 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
526         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
527         proper constants and fix formatting.
528         (possible_polymorphic_call_targets): Fix formatting.
530 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
531             Ilya Tocar  <ilya.tocar@intel.com>
533         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
534         (_mm512_loadu_epi32): Renamed into...
535         (_mm512_loadu_si512): This.
536         (_mm512_storeu_epi32): Renamed into...
537         (_mm512_storeu_si512): This.
538         (_mm512_maskz_ceil_ps): Removed.
539         (_mm512_maskz_ceil_pd): Ditto.
540         (_mm512_maskz_floor_ps): Ditto.
541         (_mm512_maskz_floor_pd): Ditto.
542         (_mm512_floor_round_ps): Ditto.
543         (_mm512_floor_round_pd): Ditto.
544         (_mm512_ceil_round_ps): Ditto.
545         (_mm512_ceil_round_pd): Ditto.
546         (_mm512_mask_floor_round_ps): Ditto.
547         (_mm512_mask_floor_round_pd): Ditto.
548         (_mm512_mask_ceil_round_ps): Ditto.
549         (_mm512_mask_ceil_round_pd): Ditto.
550         (_mm512_maskz_floor_round_ps): Ditto.
551         (_mm512_maskz_floor_round_pd): Ditto.
552         (_mm512_maskz_ceil_round_ps): Ditto.
553         (_mm512_maskz_ceil_round_pd): Ditto.
554         (_mm512_expand_pd): Ditto.
555         (_mm512_expand_ps): Ditto.
556         * config/i386/i386.c (ix86_builtins): Remove
557         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
558         (bdesc_args): Ditto.
559         * config/i386/predicates.md (const1256_operand): New.
560         (const_1_to_2_operand): Ditto.
561         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
562         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
563         (*avx512pf_gatherpf<mode>sf): Ditto.
564         (avx512pf_gatherpf<mode>df): Ditto.
565         (*avx512pf_gatherpf<mode>df_mask): Ditto.
566         (*avx512pf_gatherpf<mode>df): Ditto.
567         (avx512pf_scatterpf<mode>sf): Ditto.
568         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
569         (*avx512pf_scatterpf<mode>sf): Ditto.
570         (avx512pf_scatterpf<mode>df): Ditto.
571         (*avx512pf_scatterpf<mode>df_mask): Ditto.
572         (*avx512pf_scatterpf<mode>df): Ditto.
573         (avx512f_expand<mode>): Removed.
574         (<shift_insn><mode>3<mask_name>): Change predicate type.
576 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
578         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
579         not at the end of datarefs vector use ordered_remove to avoid
580         reordering datarefs vector.
582         PR c/59984
583         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
584         mark local addressable non-static vars as GOVD_PRIVATE
585         instead of GOVD_LOCAL.
586         * omp-low.c (lower_omp_for): Move gimple_bind_vars
587         and BLOCK_VARS of gimple_bind_block to new_stmt rather
588         than copying them.
590         PR middle-end/60092
591         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
592         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
593         assume_aligned or alloc_align attributes.
594         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
595         arguments.  Handle also assume_aligned and alloc_align attributes.
596         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
597         calls to functions with assume_aligned or alloc_align attributes.
598         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
600 2014-02-08  Terry Guo  <terry.guo@arm.com>
602         * doc/invoke.texi: Document ARM -march=armv7e-m.
604 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
606         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
607         flag on __cilkrts_rethrow builtin.
609         PR ipa/60026
610         * ipa-cp.c (determine_versionability): Fail at -O0
611         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
612         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
614         Revert:
615         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
617         PR ipa/60026
618         * tree-inline.c (copy_forbidden): Fail for
619         __attribute__((optimize (0))) functions.
621 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
623         * varpool.c: Include pointer-set.h.
624         (varpool_remove_unreferenced_decls): Variables in other partitions
625         will not be output; be however careful to not lose information
626         about partitioning.
628 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
630         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
631         lookup in the vtable constructor.
633 2014-02-07  Jeff Law  <law@redhat.com>
635         PR target/40977
636         * config/m68k/m68k.md (ashldi_extsi): Turn into a
637         define_insn_and_split.
639         * ipa-inline.c (inline_small_functions): Fix typos.
641 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
643         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
644         (s390_can_use_return_insn): Declare.
645         * config/s390/s390.h (EPILOGUE_USES): Define.
646         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
647         instructions.
648         (s390_chunkify_start): Handle return JUMP_LABELs.
649         (s390_early_mach): Emit a main_pool instruction on the entry edge.
650         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
651         (s390_can_use_return_insn): New functions.
652         (s390_fix_long_loop_prediction): Handle conditional returns.
653         (TARGET_SET_UP_BY_PROLOGUE): Define.
654         * config/s390/s390.md (ANY_RETURN): New code iterator.
655         (*creturn, *csimple_return, return, simple_return): New patterns.
657 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
659         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
660         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
661         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
662         REG_CFA_RESTORE list when deciding not to restore a register.
664 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
666         * config/s390/s390.c: Include tree-pass.h and context.h.
667         (s390_early_mach): New function, split out from...
668         (s390_emit_prologue): ...here.
669         (pass_data_s390_early_mach): New pass structure.
670         (pass_s390_early_mach): New class.
671         (s390_option_override): Create and register early_mach pass.
672         Move to end of file.
674 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
676         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
677         to match for the exit block.
679 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
681         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
682         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
683         Reject misaligned operands.
685 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
687         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
689 2014-02-07  Richard Biener  <rguenther@suse.de>
691         PR middle-end/60092
692         * gimple-low.c (lower_builtin_posix_memalign): New function.
693         (lower_stmt): Call it to lower posix_memalign in a way
694         to make alignment info accessible.
696 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
698         PR c++/60082
699         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
700         __builtin_setjmp_receiver.
702 2014-02-07  Richard Biener  <rguenther@suse.de>
704         PR middle-end/60092
705         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
706         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
707         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
708         Handle BUILT_IN_POSIX_MEMALIGN.
709         (find_func_clobbers): Likewise.
710         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
711         (call_may_clobber_ref_p_1): Likewise.
713 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
715         PR ipa/59918
716         * ipa-devirt.c (record_target_from_binfo): Remove overactive
717         sanity check.
719 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
721         PR ipa/59469
722         * lto-cgraph.c (lto_output_node): Use
723         symtab_get_symbol_partitioning_class.
724         (lto_output_varpool_node): likewise.
725         (symtab_get_symbol_partitioning_class): Move here from
726         lto/lto-partition.c
727         * cgraph.h (symbol_partitioning_class): Likewise.
728         (symtab_get_symbol_partitioning_class): Declare.
730 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
732         * ggc.h (ggc_internal_cleared_alloc): New macro.
733         * vec.h (vec_safe_copy): Handle memory stats.
734         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
735         * target-globals.c (save_target_globals): Likewise.
737 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
739         PR target/60077
740         * expr.c (emit_move_resolve_push): Export; be bit more selective
741         on when to clear alias set.
742         * expr.h (emit_move_resolve_push): Declare.
743         * function.h (struct function): Add tail_call_marked.
744         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
745         * config/i386/i386-protos.h (ix86_expand_push): Remove.
746         * config/i386/i386.md (TImode move expander): De not call
747         ix86_expand_push.
748         (FP push expanders): Preserve memory attributes.
749         * config/i386/sse.md (push<mode>1): Remove.
750         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
751         (ix86_expand_push): Remove.
752         * config/i386/mmx.md (push<mode>1): Remove.
754 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
756         PR rtl-optimization/60030
757         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
758         lopart with paradoxical subreg before shifting it up by hprec.
760 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
762         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
763         Remove extra newline at end of file.
764         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
765         (arm_issue_rate): Handle cortexa57.
766         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
767         (cortex-a57.cortex-a53): Likewise.
769 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
771         PR target/59575
772         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
773         don't record in REG_FRAME_RELATED_EXPR registers not set in that
774         bitmask.
775         (arm_expand_prologue): Adjust all callers.
776         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
777         info, registers also at the lowest numbered registers side.  Use
778         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
779         XEXP.
781         PR debug/59992
782         * var-tracking.c (adjust_mems): Before adding a SET to
783         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
785 2014-02-06  Alan Modra  <amodra@gmail.com>
787         PR target/60032
788         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
789         change SDmode to DDmode when lra_in_progress.
791 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
793         PR middle-end/59150
794         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
795         free_data_ref on the dr first, and before goto again also set dr
796         to the next dr.  For simd_lane_access, free old datarefs[i] before
797         overwriting it.  For get_vectype_for_scalar_type failure, don't
798         free_data_ref if simd_lane_access.
800         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
802         PR target/60062
803         * tree.h (opts_for_fn): New inline function.
804         (opt_for_fn): Define.
805         * config/i386/i386.c (ix86_function_regparm): Use
806         opt_for_fn (decl, optimize) instead of optimize.
808 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
810         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
811         for SYMBOL_REF in large memory model.
813 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
815         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
816         and crypto support.
817         (cortex-a57): Likewise.
818         (cortex-a57.cortex-a53): Likewise.
820 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
821             Kugan Vivekanandarajah  <kuganv@linaro.org>
823         * config/arm/arm.c (arm_vector_alignment_reachable): Check
824         unaligned_access.
825         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
827 2014-02-06  Richard Biener  <rguenther@suse.de>
829         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
830         set_loop_copy and initialize_original_copy_tables.
832 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
834         * config/aarch64/aarch64-simd.md
835         (aarch64_ashr_simddi): Change QI to SI.
837 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
838             Jakub Jelinek  <jakub@redhat.com>
840         PR middle-end/60013
841         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
842         of the dataflow.
844 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
846         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
847         CODE_FOR_altivec_vpku[hw]um to
848         CODE_FOR_altivec_vpku[hw]um_direct.
849         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
850         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
851         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
852         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
854 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
856         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
857         generation for -maltivec=be.
858         (altivec_vsumsws): Simplify redundant test.
860 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
862         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
863         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
864         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
865         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
866         gen_altivec_vpkuwum.
867         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
868         BYTES_BIG_ENDIAN.
869         (altivec_vpks<VI_char>ss): Likewise.
870         (altivec_vpks<VI_char>us): Likewise.
871         (altivec_vpku<VI_char>us): Likewise.
872         (altivec_vpku<VI_char>um): Likewise.
873         (altivec_vpku<VI_char>um_direct): New (copy of
874         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
875         internal use).
876         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
877         target is little endian and -maltivec=be is not specified.
878         (*altivec_vupkhs<VU_char>_direct): New (copy of
879         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
880         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
881         target is little endian and -maltivec=be is not specified.
882         (*altivec_vupkls<VU_char>_direct): New (copy of
883         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
884         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
885         little endian and -maltivec=be is not specified.
886         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
887         little endian and -maltivec=be is not specified.
889 2014-02-05  Richard Henderson <rth@redhat.com>
891         PR debug/52727
892         * combine-stack-adj.c: Revert r206943.
893         * sched-int.h (struct deps_desc): Add last_args_size.
894         * sched-deps.c (init_deps): Initialize it.
895         (sched_analyze_insn): Add OUTPUT dependencies between insns that
896         contain REG_ARGS_SIZE notes.
898 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
900         * lto-cgraph.c (asm_nodes_output): Make global.
901         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
902         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
903         (driver_handle_option): Handle OPT_fwpa.
905 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
907         PR ipa/59947
908         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
909         a comment typo and formatting issue.  If odr_hash hasn't been
910         created, return vNULL and set *completep to false.
912         PR middle-end/57499
913         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
914         bb with no successors.
916 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
918         PR target/59718
919         * doc/invoke.texi (-march): Clarify documentation for ARM.
920         (-mtune): Likewise.
921         (-mcpu): Likewise.
923 2014-02-05  Richard Biener  <rguenther@suse.de>
925         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
926         when not vectorizing because of too many alias checks.
927         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
928         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
930 2014-02-05  Nick Clifton  <nickc@redhat.com>
932         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
933         accept extended registers in any mode when compiling for the MN10300.
935 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
937         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
938         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
939         sanitization attributes.
940         (can_inline_edge_p): Likewise.
941         (sanitize_attrs_match_for_inline_p): New function.
943 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
945         * ipa-prop.c (detect_type_change): Shor circuit testing of
946         type changes on THIS pointer.
948 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
950         PR target/59777
951         * config/pa/pa.c (legitimize_tls_address): Return original address
952         if not passed a SYMBOL_REF rtx.
953         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
954         addresses.
955         (pa_emit_move_sequence): Simplify TLS source operands.
956         (pa_legitimate_constant_p): Reject all TLS constants.
957         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
958         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
960 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
962         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
963         groups when we know they are controlled by LTO.
964         * varasm.c (default_binds_local_p_1): If object is in other partition,
965         it will be resolved locally.
967 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
969         * config/host-linux.c (linux_gt_pch_use_address): Don't
970         use SSIZE_MAX because it is not always defined.
972 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
974         PR bootstrap/59913
975         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
976         threshold for pseudo splitting.
977         (update_ebb_live_info): Process call argument hard registers and
978         hard registers from insn definition too.
979         (max_small_class_regs_num): New constant.
980         (inherit_in_ebb): Update live hard regs through EBBs.  Update
981         reloads_num only for small register classes.  Don't split for
982         outputs of jumps.
984 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
986         PR ipa/60058
987         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
988         is non-null.
990 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
992         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
993         visibility is safe.
995 2014-02-04  Marek Polacek  <polacek@redhat.com>
997         * gdbinit.in (pel): Define.
999 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1001         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1002         behavior.
1004 2014-02-04  Richard Biener  <rguenther@suse.de>
1006         PR lto/59723
1007         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1008         in function context local.
1009         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1010         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1011         similar to LTO_imported_decl_ref.
1013 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1015         PR tree-optimization/60002
1016         * cgraphclones.c (build_function_decl_skip_args): Clear
1017         DECL_LANG_SPECIFIC.
1019         PR tree-optimization/60023
1020         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1021         false to gsi_replace.
1022         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1023         has been in some EH region and vec_stmt could throw, add
1024         vec_stmt into the same EH region.
1025         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1026         has no lhs, ignore it.
1027         * internal-fn.c (expand_MASK_LOAD): Likewise.
1029         PR ipa/60026
1030         * tree-inline.c (copy_forbidden): Fail for
1031         __attribute__((optimize (0))) functions.
1033         PR other/58712
1034         * omp-low.c (simd_clone_struct_copy): If from->inbranch
1035         is set, copy one less argument.
1036         (expand_simd_clones): Don't subtract clone_info->inbranch
1037         from simd_clone_struct_alloc argument.
1039         PR rtl-optimization/57915
1040         * recog.c (simplify_while_replacing): If all unary/binary/relational
1041         operation arguments are constant, attempt to simplify those.
1043         PR middle-end/59261
1044         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1045         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1047 2014-02-04  Richard Biener  <rguenther@suse.de>
1049         PR tree-optimization/60012
1050         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1051         TBAA disambiguation to all DDRs.
1053 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1055         PR target/59788
1056         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1057         (LINK_SPEC): Use it for -shared, -shared-libgcc.
1059 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1061         PR ipa/59882
1062         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1064 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1066         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1067         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1069 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1071         PR ipa/59831
1072         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1073         to figure out targets of polymorphic calls with known decl.
1074         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1075         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1076         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1077         (get_polymorphic_call_info): ... here.
1078         (get_polymorphic_call_info_from_invariant): New function.
1080 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1082         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1083         lookup via vtable pointer; check for type consistency
1084         and turn inconsitent facts into UNREACHABLE.
1085         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1086         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1087         type inconsistent querries; return UNREACHABLE instead.
1089 2014-02-03  Richard Henderson  <rth@twiddle.net>
1091         PR tree-opt/59924
1092         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1093         already processed this node.
1094         (normalize_one_pred_1): Pass along mark_set.
1095         (normalize_one_pred): Create and destroy a pointer_set_t.
1096         (normalize_one_pred_chain): Likewise.
1098 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
1100         PR gcov-profile/58602
1101         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1103 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1105         PR ipa/59831
1106         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1107         -fno-devirtualize; try to devirtualize by the knowledge of
1108         virtual table pointer given by aggregate propagation.
1109         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1110         (ipa_print_node_jump_functions): Dump also offset that
1111         is relevant for polymorphic calls.
1112         (determine_known_aggregate_parts): Add arg_type parameter; use it
1113         instead of determining the type from pointer type.
1114         (ipa_compute_jump_functions_for_edge): Update call of
1115         determine_known_aggregate_parts.
1116         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1117         (gimple_get_virt_method_for_binfo): ... here; simplify using
1118         vtable_pointer_value_to_vtable.
1119         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1120         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1121         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1122         (vtable_pointer_value_to_vtable): Break out from ...; handle also
1123         POINTER_PLUS_EXPR.
1124         (vtable_pointer_value_to_binfo): ... here.
1125         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1127 2014-02-03  Teresa Johnson  <tejohnson@google.com>
1129         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1130         redef of outer loop index variable.
1132 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
1134         PR c++/53017
1135         PR c++/59211
1136         * doc/extend.texi (Function Attributes): Typo.
1138 2014-02-03  Cong Hou  <congh@google.com>
1140         PR tree-optimization/60000
1141         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1142         if the vectorized statement is a store.  A store statement can only
1143         appear at the end of pattern statements.
1145 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1147         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1148         (ix86_option_override_internal): Default long double to 64-bit for
1149         32-bit Bionic and to 128-bit for 64-bit Bionic.
1151         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1152         TARGET_LONG_DOUBLE_128 is true.
1153         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1155         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1156         (mlong-double-64): Negate -mlong-double-128.
1157         (mlong-double-128): New option.
1159         * config/i386/i386-c.c (ix86_target_macros): Define
1160         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1162         * doc/invoke.texi: Document -mlong-double-128.
1164 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1166         PR rtl-optimization/60024
1167         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1169 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
1171         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1173 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1175         PR rtl-optimization/57662
1176         * sel-sched.c (code_motion_path_driver): Do not mark already not
1177         existing blocks in the visiting bitmap.
1179 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1181         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1182         on the insn being emitted.
1184 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
1185             Will Deacon  <will.deacon@arm.com>
1187         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1189 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1191         * config/arm/arm-tables.opt: Regenerate.
1193 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1195         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1196         for vector types other than V16QImode.
1197         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1198         define_expand, and call altivec_expand_vec_perm_le when producing
1199         code with little endian element order.
1200         (*altivec_vperm_<mode>_internal): New insn having previous
1201         behavior of altivec_vperm_<mode>.
1202         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1203         altivec_expand_vec_perm_le when producing code with little endian
1204         element order.
1205         (*altivec_vperm_<mode>_uns_internal): New insn having previous
1206         behavior of altivec_vperm_<mode>_uns.
1208 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1210         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1211         (altivec_vsumsws): Add handling for -maltivec=be with a little
1212         endian target.
1213         (altivec_vsumsws_direct): New.
1214         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1215         gen_altivec_vsumsws.
1217 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1219         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1220         vtable_pointer_value_to_binfo): New functions.
1221         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1222         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1224 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
1226         * config/nios2/nios2.md (load_got_register): Initialize GOT
1227         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1228         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1230 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1232         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1233         preserverd by passthrough, do not propagate the type.
1235 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1237         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1238         (mips_atomic_assign_expand_fenv): New function.
1239         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1241 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1243         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1244         (__builtin_mips_set_fcsr): Likewise.
1245         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1246         MIPS_USI_FTYPE_VOID.
1247         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1248         (mips16_expand_set_fcsr): Likewise.
1249         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1250         (mips16_set_fcsr_stub): Likewise.
1251         (mips16_get_fcsr_one_only_stub): New class.
1252         (mips16_set_fcsr_one_only_stub): Likewise.
1253         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1254         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1255         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1256         (hard_float): New availability predicate.
1257         (mips_builtins): Add get_fcsr and set_fcsr.
1258         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1259         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1260         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1261         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1262         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1263         patterns.
1265 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1267         * config/mips/mips.c (mips_one_only_stub): New class.
1268         (mips_need_mips16_rdhwr_p): Replace with...
1269         (mips16_rdhwr_stub): ...this new variable.
1270         (mips16_stub_call_address): New function.
1271         (mips16_rdhwr_one_only_stub): New class.
1272         (mips_expand_thread_pointer): Use mips16_stub_call_address.
1273         (mips_output_mips16_rdhwr): Delete.
1274         (mips_finish_stub): New function.
1275         (mips_code_end): Use it to handle rdhwr stubs.
1277 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
1279         PR target/60017
1280         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1281         when calculating size of integer atomic types.
1283 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
1285         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1287 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
1289         PR tree-optimization/60003
1290         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1291         * profile.c (branch_prob): Use gimple_call_builtin_p
1292         to check for BUILT_IN_SETJMP_RECEIVER.
1293         * tree-inline.c (copy_bb): Call notice_special_calls.
1295 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1297         PR bootstrap/59985
1298         * lra-constraints.c (process_alt_operands): Update reload_sum only
1299         on the first pass.
1301 2014-01-31  Richard Henderson  <rth@redhat.com>
1303         PR middle-end/60004
1304         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1305         until after else_eh is processed.
1307 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1309         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1310         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1311         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1312         in smmintrin.h, remove them.
1313         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1314         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1315         * config/i386/i386.md (ROUND_SAE): Fix value.
1316         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1317         (const48_operand): New.
1318         * config/i386/subst.md (round), (round_expand): Use
1319         const_4_or_8_to_11_operand.
1320         (round_saeonly), (round_saeonly_expand): Use const48_operand.
1322 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1324         * config/i386/constraints.md (Yk): Swap meaning with k.
1325         * config/i386/i386.md (movhi_internal): Change Yk to k.
1326         (movqi_internal): Ditto.
1327         (*k<logic><mode>): Ditto.
1328         (*andhi_1): Ditto.
1329         (*andqi_1): Ditto.
1330         (kandn<mode>): Ditto.
1331         (*<code>hi_1): Ditto.
1332         (*<code>qi_1): Ditto.
1333         (kxnor<mode>): Ditto.
1334         (kortestzhi): Ditto.
1335         (kortestchi): Ditto.
1336         (kunpckhi): Ditto.
1337         (*one_cmplhi2_1): Ditto.
1338         (*one_cmplqi2_1): Ditto.
1339         * config/i386/sse.md (): Change k to Yk.
1340         (avx512f_load<mode>_mask): Ditto.
1341         (avx512f_blendm<mode>): Ditto.
1342         (avx512f_store<mode>_mask): Ditto.
1343         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1344         (avx512f_storedqu<mode>_mask): Ditto.
1345         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1346         Ditto.
1347         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1348         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1349         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1350         (avx512f_maskcmp<mode>3): Ditto.
1351         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1352         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1353         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1354         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1355         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1356         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1357         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1358         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1359         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1360         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1361         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1362         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1363         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1364         (vec_extract_lo_<mode>_maskm): Ditto.
1365         (vec_extract_hi_<mode>_maskm): Ditto.
1366         (avx512f_vternlog<mode>_mask): Ditto.
1367         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1368         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1369         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1370         (avx512f_<code>v8div16qi2_mask): Ditto.
1371         (avx512f_<code>v8div16qi2_mask_store): Ditto.
1372         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1373         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1374         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1375         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1376         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1377         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1378         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1379         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1380         (avx512cd_maskb_vec_dupv8di): Ditto.
1381         (avx512cd_maskw_vec_dupv16si): Ditto.
1382         (avx512f_vpermi2var<mode>3_maskz): Ditto.
1383         (avx512f_vpermi2var<mode>3_mask): Ditto.
1384         (avx512f_vpermi2var<mode>3_mask): Ditto.
1385         (avx512f_vpermt2var<mode>3_maskz): Ditto.
1386         (*avx512f_gathersi<mode>): Ditto.
1387         (*avx512f_gathersi<mode>_2): Ditto.
1388         (*avx512f_gatherdi<mode>): Ditto.
1389         (*avx512f_gatherdi<mode>_2): Ditto.
1390         (*avx512f_scattersi<mode>): Ditto.
1391         (*avx512f_scatterdi<mode>): Ditto.
1392         (avx512f_compress<mode>_mask): Ditto.
1393         (avx512f_compressstore<mode>_mask): Ditto.
1394         (avx512f_expand<mode>_mask): Ditto.
1395         * config/i386/subst.md (mask): Change k to Yk.
1396         (mask_scalar_merge): Ditto.
1397         (sd): Ditto.
1399 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
1401         * doc/extend.texi (Vector Extensions): Document ?: in C++.
1403 2014-01-31  Richard Biener  <rguenther@suse.de>
1405         PR middle-end/59990
1406         * builtins.c (fold_builtin_memory_op): Make sure to not
1407         use a floating-point mode or a boolean or enumeral type for
1408         the copy operation.
1410 2014-01-30  DJ Delorie  <dj@redhat.com>
1412         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1413         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1414         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1415         whenever main() has an epilogue.
1417 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1419         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1420         unused variable "field".
1421         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1422         (vsx_mergeh_<mode>): Likewise.
1423         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1424         (altivec_vmrghh): Likewise.
1425         (altivec_vmrghw): Likewise.
1426         (altivec_vmrglb): Likewise.
1427         (altivec_vmrglh): Likewise.
1428         (altivec_vmrglw): Likewise.
1429         (altivec_vspltb): Add missing uses.
1430         (altivec_vsplth): Likewise.
1431         (altivec_vspltw): Likewise.
1432         (altivec_vspltsf): Likewise.
1434 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1436         PR target/59923
1437         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1438         frame related instructions.
1440 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1442         PR rtl-optimization/59959
1443         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1444         any reload of register whose subreg is invalid.
1446 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1448         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1449         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1450         Add missing return type - void.
1452 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1454         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1455         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1456         remove element index adjustment for endian (now handled in vsx.md
1457         and altivec.md).
1458         (altivec_expand_vec_perm_const): Use
1459         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1460         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1461         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1462         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1463         define_expand and a new define_insn *altivec_vspltb_internal;
1464         adjust for -maltivec=be on a little endian target.
1465         (altivec_vspltb_direct): New.
1466         (altivec_vsplth): Divide into a define_expand and a new
1467         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1468         little endian target.
1469         (altivec_vsplth_direct): New.
1470         (altivec_vspltw): Divide into a define_expand and a new
1471         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1472         little endian target.
1473         (altivec_vspltw_direct): New.
1474         (altivec_vspltsf): Divide into a define_expand and a new
1475         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1476         a little endian target.
1478 2014-01-30  Richard Biener  <rguenther@suse.de>
1480         PR tree-optimization/59993
1481         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1482         can propagate form the earlier stmt and avoid the transform
1483         when the intermediate result is needed.
1485 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1487         * README.Portability: Fix typo.
1489 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1491         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1492         comparison_operator with ordered_comparison_operator.
1494 2014-01-30  Nick Clifton  <nickc@redhat.com>
1496         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1497         Rename to mn10300_store_multiple_regs.
1498         * config/mn10300/mn10300.c: Likewise.
1499         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1500         store_multiple_regs.
1501         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1502         Call mn10300_store_multiple_regs.
1504 2014-01-30  Nick Clifton  <nickc@redhat.com>
1505             DJ Delorie  <dj@redhat.com>
1507         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1508         %fp 2 to keep registers after it properly word-aligned.
1509         (rl78_alloc_physical_registers_umul): Handle the case where both
1510         input operands are the same.
1512 2014-01-30  Richard Biener  <rguenther@suse.de>
1514         PR tree-optimization/59903
1515         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1516         check properly.
1518 2014-01-30  Jason Merrill  <jason@redhat.com>
1520         PR c++/59633
1521         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1523         PR c++/59645
1524         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1526 2014-01-30  Richard Biener  <rguenther@suse.de>
1528         PR tree-optimization/59951
1529         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1531 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1533         PR target/59784
1534         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1535         SFmode to DFmode case.
1537 2014-01-29  DJ Delorie  <dj@redhat.com>
1539         * config/msp430/msp430.opt (-minrt): New.
1540         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1541         if -minrt given.
1542         (ENDFILE_SPEC): Likewise.
1544 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1546         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1547         (estimate_function_body_sizes): Use it.
1549 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1551         PR c++/58561
1552         * dwarf2out.c (is_cxx_auto): New.
1553         (is_base_type): Use it.
1554         (gen_type_die_with_usage): Likewise.
1556 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1558         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1559         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1560         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1561         -maltivec=be with LE targets.
1562         (vsx_mergeh_<mode>): Likewise.
1563         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1564         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1565         (altivec_vmrghb): Replace with define_expand and new
1566         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1567         (altivec_vmrghb_direct): New define_insn.
1568         (altivec_vmrghh): Replace with define_expand and new
1569         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1570         (altivec_vmrghh_direct): New define_insn.
1571         (altivec_vmrghw): Replace with define_expand and new
1572         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1573         (altivec_vmrghw_direct): New define_insn.
1574         (*altivec_vmrghsf): Adjust for endianness.
1575         (altivec_vmrglb): Replace with define_expand and new
1576         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1577         (altivec_vmrglb_direct): New define_insn.
1578         (altivec_vmrglh): Replace with define_expand and new
1579         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1580         (altivec_vmrglh_direct): New define_insn.
1581         (altivec_vmrglw): Replace with define_expand and new
1582         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1583         (altivec_vmrglw_direct): New define_insn.
1584         (*altivec_vmrglsf): Adjust for endianness.
1585         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1586         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1587         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1588         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1589         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1590         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1591         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1592         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1594 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1596         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1597         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1598         whitespace.
1600 2014-01-29  Richard Biener  <rguenther@suse.de>
1602         PR tree-optimization/58742
1603         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1604         associate_pointerplus_align.
1605         (associate_pointerplus_diff): New function.
1606         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
1607         and associate_pointerplus_diff.
1609 2014-01-29  Richard Biener  <rguenther@suse.de>
1611         * lto-streamer.h (LTO_major_version): Bump to 3.
1612         (LTO_minor_version): Reset to 0.
1614 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
1616         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1617         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1618         (arm_file_start): Generate correct asm header for armv7ve.
1619         * config/arm/bpabi.h: Add multilib support for armv7ve.
1620         * config/arm/driver-arm.c: Change the architectures of cortex-a7
1621         and cortex-a15 to armv7ve.
1622         * config/arm/t-aprofile: Add multilib support for armv7ve.
1623         * doc/invoke.texi: Document -march=armv7ve.
1625 2014-01-29  Richard Biener  <rguenther@suse.de>
1627         PR tree-optimization/58742
1628         * tree-ssa-forwprop.c (associate_plusminus): Return true
1629         if we changed sth, defer EH cleanup to ...
1630         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
1631         (simplify_mult): New function.
1633 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1635         PR middle-end/59917
1636         PR tree-optimization/59920
1637         * tree.c (build_common_builtin_nodes): Remove
1638         __builtin_setjmp_dispatcher initialization.
1639         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1640         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1641         instead of gsi_after_labels + manually skipping debug stmts.
1642         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1643         ignore bbs with IFN_ABNORMAL_DISPATCHER.
1644         * tree-inline.c (copy_edges_for_bb): Remove
1645         can_make_abnormal_goto argument, instead add abnormal_goto_dest
1646         argument.  Ignore computed_goto_p stmts.  Don't call
1647         make_abnormal_goto_edges.  If a call might need abnormal edges
1648         for non-local gotos, see if it already has an edge to
1649         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1650         with true argument, don't do anything then, otherwise add
1651         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1652         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1653         caller.
1654         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1655         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1656         (lower_stmt): Don't set data->calls_builtin_setjmp.
1657         (lower_builtin_setjmp): Adjust comment.
1658         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1659         * tree-cfg.c (found_computed_goto): Remove.
1660         (factor_computed_gotos): Remove.
1661         (make_goto_expr_edges): Return bool, true for computed gotos.
1662         Don't call make_abnormal_goto_edges.
1663         (build_gimple_cfg): Don't set found_computed_goto, don't call
1664         factor_computed_gotos.
1665         (computed_goto_p): No longer static.
1666         (make_blocks): Don't set found_computed_goto.
1667         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1668         (make_edges): If make_goto_expr_edges returns true, push bb
1669         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1670         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1671         vector.  Record mapping between bbs and OpenMP regions if there
1672         are any, adjust make_gimple_omp_edges caller.  Call
1673         handle_abnormal_edges.
1674         (make_abnormal_goto_edges): Remove.
1675         * tree-cfg.h (make_abnormal_goto_edges): Remove.
1676         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1677         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1678         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1679         * internal-fn.def (ABNORMAL_DISPATCHER): New.
1680         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1681         filling *region also set *region_idx to (*region)->entry->index.
1683         PR other/58712
1684         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1685         For REGs set ORIGINAL_REGNO.
1687 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
1689         * doc/md.texi: Mention that a target shouldn't implement
1690         vec_widen_(s|u)mul_even/odd pair if it is less efficient
1691         than hi/lo pair.
1693 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1695         PR tree-optimization/59594
1696         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1697         a copy of the datarefs vector rather than the vector itself.
1699 2014-01-28  Jason Merrill  <jason@redhat.com>
1701         PR c++/53756
1702         * dwarf2out.c (auto_die): New static.
1703         (gen_type_die_with_usage): Handle C++1y 'auto'.
1704         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1705         on definition.
1707 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1709         PR target/59672
1710         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1711         (SPEC_X32): Likewise.
1712         (SPEC_64): Likewise.
1713         * config/i386/i386.c (ix86_option_override_internal): Turn off
1714         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1715         for TARGET_16BIT.
1716         (x86_file_start): Output .code16gcc for TARGET_16BIT.
1717         * config/i386/i386.h (TARGET_16BIT): New macro.
1718         (TARGET_16BIT_P): Likewise.
1719         * config/i386/i386.opt: Add m16.
1720         * doc/invoke.texi: Document -m16.
1722 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
1724         PR preprocessor/59935
1725         * input.c (location_get_source_line): Bail out on when line number
1726         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1728 2014-01-28  Richard Biener  <rguenther@suse.de>
1730         PR tree-optimization/58742
1731         * tree-ssa-forwprop.c (associate_plusminus): Handle
1732         pointer subtraction of the form (T)(P + A) - (T)P.
1734 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1736         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1737         at const_int_cost.
1739 2014-01-28  Richard Biener  <rguenther@suse.de>
1741         Revert
1742         2014-01-28  Richard Biener  <rguenther@suse.de>
1744         PR rtl-optimization/45364
1745         PR rtl-optimization/59890
1746         * var-tracking.c (local_get_addr_clear_given_value): Handle
1747         already cleared slot.
1748         (val_reset): Handle not allocated local_get_addr_cache.
1749         (vt_find_locations): Use post-order on the inverted CFG.
1751 2014-01-28  Richard Biener  <rguenther@suse.de>
1753         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1755 2014-01-28  Richard Biener  <rguenther@suse.de>
1757         PR rtl-optimization/45364
1758         PR rtl-optimization/59890
1759         * var-tracking.c (local_get_addr_clear_given_value): Handle
1760         already cleared slot.
1761         (val_reset): Handle not allocated local_get_addr_cache.
1762         (vt_find_locations): Use post-order on the inverted CFG.
1764 2014-01-28  Alan Modra  <amodra@gmail.com>
1766         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1767         * configure.ac <recursive call for build != host>: Define
1768         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1769         and LD_FOR_BUILD too.
1770         * configure: Regenerate.
1772 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
1774         * config/i386/i386.c (get_builtin_code_for_version): Separate
1775         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1776         Broadwell from Haswell.
1778 2014-01-27  Steve Ellcey  <sellcey@mips.com>
1780         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1781         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1782         * config/mips/mips.c (mips_option_override): Change setting
1783         of TARGET_DSP.
1784         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1785         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1786         Change from Mask to Var.
1788 2014-01-27  Jeff Law  <law@redhat.com>
1790         * ipa-inline.c (inline_small_functions): Fix typo.
1792 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
1794         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1795         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1796         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1797         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1798         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1799         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1800         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1801         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1802         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1803         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1804         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1805         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1806         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1807         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1808         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1809         (_mm512_storeu_epi64): Ditto.
1810         (_mm512_cmpge_epi32_mask): Ditto.
1811         (_mm512_cmpge_epu32_mask): Ditto.
1812         (_mm512_cmpge_epi64_mask): Ditto.
1813         (_mm512_cmpge_epu64_mask): Ditto.
1814         (_mm512_cmple_epi32_mask): Ditto.
1815         (_mm512_cmple_epu32_mask): Ditto.
1816         (_mm512_cmple_epi64_mask): Ditto.
1817         (_mm512_cmple_epu64_mask): Ditto.
1818         (_mm512_cmplt_epi32_mask): Ditto.
1819         (_mm512_cmplt_epu32_mask): Ditto.
1820         (_mm512_cmplt_epi64_mask): Ditto.
1821         (_mm512_cmplt_epu64_mask): Ditto.
1822         (_mm512_cmpneq_epi32_mask): Ditto.
1823         (_mm512_cmpneq_epu32_mask): Ditto.
1824         (_mm512_cmpneq_epi64_mask): Ditto.
1825         (_mm512_cmpneq_epu64_mask): Ditto.
1826         (_mm512_expand_pd): Ditto.
1827         (_mm512_expand_ps): Ditto.
1828         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1829         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1830         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1831         * config/i386/i386.c (ix86_builtins): Add
1832         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1833         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1834         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1835         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1836         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1837         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1838         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1839         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1840         IX86_BUILTIN_PMOVUSQW512_MEM.
1841         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1842         __builtin_ia32_pmovsqd512mem_mask,
1843         __builtin_ia32_pmovqd512mem_mask,
1844         __builtin_ia32_pmovusqw512mem_mask,
1845         __builtin_ia32_pmovsqw512mem_mask,
1846         __builtin_ia32_pmovqw512mem_mask,
1847         __builtin_ia32_pmovusdw512mem_mask,
1848         __builtin_ia32_pmovsdw512mem_mask,
1849         __builtin_ia32_pmovdw512mem_mask,
1850         __builtin_ia32_pmovqb512mem_mask,
1851         __builtin_ia32_pmovusqb512mem_mask,
1852         __builtin_ia32_pmovsqb512mem_mask,
1853         __builtin_ia32_pmovusdb512mem_mask,
1854         __builtin_ia32_pmovsdb512mem_mask,
1855         __builtin_ia32_pmovdb512mem_mask.
1856         (bdesc_args): Add __builtin_ia32_expanddf512,
1857         __builtin_ia32_expandsf512.
1858         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1859         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1860         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1861         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1862         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1863         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1864         (avx512f_<code>v8div16qi2_mask_store): This.
1865         (avx512f_expand<mode>): New.
1867 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
1869         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
1870         New.
1871         (_mm512_mask_prefetch_i64gather_pd): Ditto.
1872         (_mm512_prefetch_i32scatter_pd): Ditto.
1873         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
1874         (_mm512_prefetch_i64scatter_pd): Ditto.
1875         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
1876         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
1877         (_mm512_mask_prefetch_i64gather_ps): Ditto.
1878         (_mm512_prefetch_i32scatter_ps): Ditto.
1879         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
1880         (_mm512_prefetch_i64scatter_ps): Ditto.
1881         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
1882         * config/i386/i386-builtin-types.def: Define
1883         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
1884         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
1885         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
1886         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
1887         IX86_BUILTIN_SCATTERPFQPD.
1888         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
1889         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
1890         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
1891         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
1892         __builtin_ia32_scatterpfqps.
1893         (ix86_expand_builtin): Expand new built-ins.
1894         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
1895         fix memory access data type.
1896         (*avx512pf_gatherpf<mode>_mask): Ditto.
1897         (*avx512pf_gatherpf<mode>): Ditto.
1898         (avx512pf_scatterpf<mode>): Ditto.
1899         (*avx512pf_scatterpf<mode>_mask): Ditto.
1900         (*avx512pf_scatterpf<mode>): Ditto.
1901         (GATHER_SCATTER_SF_MEM_MODE): New.
1902         (avx512pf_gatherpf<mode>df): Ditto.
1903         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1904         (*avx512pf_scatterpf<mode>df): Ditto.
1906 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
1908         PR bootstrap/59934
1909         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
1910         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
1911         reached.
1913 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1915         * common/config/arm/arm-common.c
1916         (arm_rewrite_mcpu): Handle multiple names.
1917         * config/arm/arm.h
1918         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1920 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1922         * gimple-builder.h (create_gimple_tmp): Delete.
1924 2014-01-27  Christian Bruel  <christian.bruel@st.com>
1926         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
1927         words comparisons.
1929 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
1931         * config/pa/pa.md (call): Generate indirect long calls to non-local
1932         functions when outputing 32-bit code.
1933         (call_value): Likewise except for special call to buggy powf function.
1935         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
1936         portable runtime and PIC indirect calls.
1937         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
1938         and PIC call sequences.  Use ldo instead of blr to set return register
1939         in PIC call sequence.
1941 2014-01-25  Walter Lee  <walt@tilera.com>
1943         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
1944         avoid clobbering a live register.
1946 2014-01-25  Walter Lee  <walt@tilera.com>
1948         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
1949         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
1950         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
1951         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
1953 2014-01-25  Walter Lee  <walt@tilera.com>
1955         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
1956         arguments on even registers.
1957         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
1958         STACK_BOUNDARY.
1959         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
1960         (BIGGEST_ALIGNMENT): Ditto.
1961         (BIGGEST_FIELD_ALIGNMENT): Ditto.
1963 2014-01-25  Walter Lee  <walt@tilera.com>
1965         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
1966         insns before bundling.
1967         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
1969 2014-01-25  Walter Lee  <walt@tilera.com>
1971         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
1972         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
1973         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
1975 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1977         * config/mips/constraints.md (kl): Delete.
1978         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
1979         define expands, using...
1980         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
1981         instructions for MIPS16.
1982         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
1983         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
1985 2014-01-25  Walter Lee  <walt@tilera.com>
1987         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
1988         (clzdi2): Ditto.
1989         (ffsdi2): Ditto.
1991 2014-01-25  Walter Lee  <walt@tilera.com>
1993         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
1994         (TARGET_EXPAND_TO_RTL_HOOK): Define.
1996 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1998         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
1999         Handle XOR.
2001 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
2003         * print-rtl.c (in_call_function_usage): New var.
2004         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2005         EXPR_LIST mode as mode and not as reg note name.
2007         PR middle-end/59561
2008         * cfgloopmanip.c (copy_loop_info): If
2009         loop->warned_aggressive_loop_optimizations, make sure
2010         the flag is set in target loop too.
2012 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
2014         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2015         flag_cilkplus.
2016         * builtins.def: Likewise.
2017         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2018         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2019         * ira.c (ira_setup_eliminable_regset): Likewise.
2020         * omp-low.c (gate_expand_omp): Likewise.
2021         (execute_lower_omp): Likewise.
2022         (diagnose_sb_0): Likewise.
2023         (gate_diagnose_omp_blocks): Likewise.
2024         (simd_clone_clauses_extract): Likewise.
2025         (gate): Likewise.
2027 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2029         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2030         correction for little endian...
2031         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2032         here.
2034 2014-01-24  Jeff Law  <law@redhat.com>
2036         PR tree-optimization/59919
2037         * tree-vrp.c (find_assert_locations_1): Do not register asserts
2038         for non-returning calls.
2040 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
2042         * common/config/aarch64/aarch64-common.c
2043         (aarch64_rewrite_mcpu): Handle multiple names.
2044         * config/aarch64/aarch64.h
2045         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2047 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
2049         * input.c (add_file_to_cache_tab): Handle the case where fopen
2050         returns NULL.
2052 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
2054         PR target/59929
2055         * config/i386/i386.md (pushsf splitter): Get stack adjustment
2056         from push operand if code of push isn't PRE_DEC.
2058 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
2060         PR target/59909
2061         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2062         -mquad-memory-atomic.  Update -mquad-memory documentation to say
2063         it is only used for non-atomic loads/stores.
2065         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2066         -mquad-memory or -mquad-memory-atomic switches.
2068         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2069         -mquad-memory-atomic to ISA 2.07 support.
2071         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2072         to separate support of normal quad word memory operations (ldq, stq)
2073         from the atomic quad word memory operations.
2075         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2076         support to separate non-atomic quad word operations from atomic
2077         quad word operations.  Disable non-atomic quad word operations in
2078         little endian mode so that we don't have to swap words after the
2079         load and before the store.
2080         (quad_load_store_p): Add comment about atomic quad word support.
2081         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2082         options printed with -mdebug=reg.
2084         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2085         -mquad-memory-atomic as the test for whether we have quad word
2086         atomic instructions.
2087         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2088         or -mp8-vector are used, allow byte/half-word atomic operations.
2090         * config/rs6000/sync.md (load_lockedti): Insure that the address
2091         is a proper indexed or indirect address for the lqarx instruction.
2092         On little endian systems, swap the hi/lo registers after the lqarx
2093         instruction.
2094         (load_lockedpti): Use indexed_or_indirect_operand predicate to
2095         insure the address is valid for the lqarx instruction.
2096         (store_conditionalti): Insure that the address is a proper indexed
2097         or indirect address for the stqcrx. instruction.  On little endian
2098         systems, swap the hi/lo registers before doing the stqcrx.
2099         instruction.
2100         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2101         insure the address is valid for the stqcrx. instruction.
2103         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2104         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2105         type of quad memory support is available.
2107 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
2109         PR regression/59915
2110         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2111         there is a danger of looping.
2113 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
2115         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2116         force flag_ira_loop_pressure if set via command line.
2118 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2120         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2121         (ashr_simd): New builtin handling DI mode.
2122         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2123         (aarch64_sshr_simddi): New match pattern.
2124         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2125         (vshrd_n_s64): Likewise.
2126         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2128 2014-01-23  Nick Clifton  <nickc@redhat.com>
2130         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2131         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2132         favour of mcu specific scripts.
2133         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2134         430x multilibs.
2136 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2137             Alex Velenko  <Alex.Velenko@arm.com>
2139         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2140         (vaddv_s16): Likewise.
2141         (vaddv_s32): Likewise.
2142         (vaddv_u8): Likewise.
2143         (vaddv_u16): Likewise.
2144         (vaddv_u32): Likewise.
2145         (vaddvq_s8): Likewise.
2146         (vaddvq_s16): Likewise.
2147         (vaddvq_s32): Likewise.
2148         (vaddvq_s64): Likewise.
2149         (vaddvq_u8): Likewise.
2150         (vaddvq_u16): Likewise.
2151         (vaddvq_u32): Likewise.
2152         (vaddvq_u64): Likewise.
2153         (vaddv_f32): Likewise.
2154         (vaddvq_f32): Likewise.
2155         (vaddvq_f64): Likewise.
2156         (vmaxv_f32): Likewise.
2157         (vmaxv_s8): Likewise.
2158         (vmaxv_s16): Likewise.
2159         (vmaxv_s32): Likewise.
2160         (vmaxv_u8): Likewise.
2161         (vmaxv_u16): Likewise.
2162         (vmaxv_u32): Likewise.
2163         (vmaxvq_f32): Likewise.
2164         (vmaxvq_f64): Likewise.
2165         (vmaxvq_s8): Likewise.
2166         (vmaxvq_s16): Likewise.
2167         (vmaxvq_s32): Likewise.
2168         (vmaxvq_u8): Likewise.
2169         (vmaxvq_u16): Likewise.
2170         (vmaxvq_u32): Likewise.
2171         (vmaxnmv_f32): Likewise.
2172         (vmaxnmvq_f32): Likewise.
2173         (vmaxnmvq_f64): Likewise.
2174         (vminv_f32): Likewise.
2175         (vminv_s8): Likewise.
2176         (vminv_s16): Likewise.
2177         (vminv_s32): Likewise.
2178         (vminv_u8): Likewise.
2179         (vminv_u16): Likewise.
2180         (vminv_u32): Likewise.
2181         (vminvq_f32): Likewise.
2182         (vminvq_f64): Likewise.
2183         (vminvq_s8): Likewise.
2184         (vminvq_s16): Likewise.
2185         (vminvq_s32): Likewise.
2186         (vminvq_u8): Likewise.
2187         (vminvq_u16): Likewise.
2188         (vminvq_u32): Likewise.
2189         (vminnmv_f32): Likewise.
2190         (vminnmvq_f32): Likewise.
2191         (vminnmvq_f64): Likewise.
2193 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2195         * config/aarch64/aarch64-simd.md
2196         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2197         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2198         (*aarch64_mul3_elt<mode>): Likewise.
2199         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2200         (*aarch64_mul3_elt_to_64v2df): Likewise.
2201         (*aarch64_mla_elt<mode>): Likewise.
2202         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2203         (*aarch64_mls_elt<mode>): Likewise.
2204         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2205         (*aarch64_fma4_elt<mode>): Likewise.
2206         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2207         (*aarch64_fma4_elt_to_64v2df): Likewise.
2208         (*aarch64_fnma4_elt<mode>): Likewise.
2209         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2210         (*aarch64_fnma4_elt_to_64v2df): Likewise.
2211         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2212         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2213         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2214         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2215         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2216         (aarch64_sqdmull_lane<mode>_internal): Likewise.
2217         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2219 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2221         * config/aarch64/aarch64-simd.md
2222         (aarch64_be_checked_get_lane<mode>): New define_expand.
2223         * config/aarch64/aarch64-simd-builtins.def
2224         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2225         New builtin definition.
2226         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2227         Use new safe be builtin.
2229 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2231         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2232         New define_insn.
2233         (aarch64_be_st1<mode>): Likewise.
2234         (aarch_ld1<VALL:mode>): Define_expand modified.
2235         (aarch_st1<VALL:mode>): Likewise.
2236         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2237         (UNSPEC_ST1): Likewise.
2239 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
2241         * config/microblaze/microblaze.md: Add trap insn and attribute
2243 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
2245         PR preprocessor/58580
2246         * input.h (location_get_source_line): Take an additional line_size
2247         parameter.
2248         (void diagnostics_file_cache_fini): Declare new function.
2249         * input.c (struct fcache): New type.
2250         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2251         New static constants.
2252         (diagnostic_file_cache_init, total_lines_num)
2253         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2254         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2255         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2256         (get_next_line, read_next_line, goto_next_line, read_line_num):
2257         New static function definitions.
2258         (diagnostic_file_cache_fini): New function.
2259         (location_get_source_line): Take an additional output line_len
2260         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
2261         read_line_num.
2262         * diagnostic.c (diagnostic_finish): Call
2263         diagnostic_file_cache_fini.
2264         (adjust_line): Take an additional input parameter for the length
2265         of the line, rather than calculating it with strlen.
2266         (diagnostic_show_locus): Adjust the use of
2267         location_get_source_line and adjust_line with respect to their new
2268         signature.  While displaying a line now, do not stop at the first
2269         null byte.  Rather, display the zero byte as a space and keep
2270         going until we reach the size of the line.
2271         * Makefile.in: Add vec.o to OBJS-libcommon
2273 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2274             Ilya Tocar     <ilya.tocar@intel.com>
2276         * config/i386/avx512fintrin.h (_mm512_kmov): New.
2277         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2278         (__builtin_ia32_kmov16): Ditto.
2279         * config/i386/i386.md (UNSPEC_KMOV): New.
2280         (kmovw): Ditto.
2282 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2284         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2285         (_mm512_storeu_si512): Ditto.
2287 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
2289         PR target/52125
2290         * rtl.h (get_referenced_operands): Declare.
2291         * recog.c (get_referenced_operands): New function.
2292         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2293         operands have been referenced when recording LO_SUM references.
2295 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
2297         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2299 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2301         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2302         Enable for generic and recent AMD targets.
2304 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2306         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2307         ARG_SIZE note when adjustment was eliminated.
2309 2014-01-22  Jeff Law  <law@redhat.com>
2311         PR tree-optimization/59597
2312         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2313         in file.  Accept new argument REGISTERING and use it to modify
2314         dump output appropriately.
2315         (register_jump_thread): Corresponding changes.
2316         (mark_threaded_blocks): Reinstate code to cancel unprofitable
2317         thread paths involving joiner blocks.  Add code to dump cancelled
2318         jump threading paths.
2320 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
2322         PR rtl-optimization/59477
2323         * lra-constraints.c (inherit_in_ebb): Process call for living hard
2324         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
2326 2014-01-22  Tom Tromey  <tromey@redhat.com>
2328         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2329         PARAMS.
2330         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2332 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2334         PR rtl-optimization/59896
2335         * lra-constraints.c (process_alt_operands): Check unused note for
2336         matched operands of insn with no output reloads.
2338 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
2340         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2341         (mips_move_from_gpr_cost): Likewise.
2343 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2345         PR rtl-optimization/59858
2346         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2347         ira_class_hard_regs_num.
2348         (process_alt_operands): Increase reject for dying matched operand.
2350 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
2352         PR target/59003
2353         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2354         smaller than size, perform several stores or loads and stores
2355         at dst + count - size to store or copy all of size bytes, rather
2356         than just last modesize bytes.
2358 2014-01-20  DJ Delorie  <dj@redhat.com>
2360         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2361         that CLOBBERs are REGs before propogating their values.
2363 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2365         PR middle-end/59789
2366         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2367         (cgraph_inline_failed_type): New function.
2368         * cgraph.h (DEFCIFCODE): Add type.
2369         (cgraph_inline_failed_type_t): New enum.
2370         (cgraph_inline_failed_type): New prototype.
2371         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2372         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2373         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2374         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2375         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2376         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2377         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2378         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2379         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2380         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2381         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2382         OPTIMIZATION_MISMATCH.
2383         * tree-inline.c (expand_call_inline): Emit errors during
2384         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2386 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2388         PR target/59685
2389         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2390         mode attribute in insn output.
2392 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
2394         * output.h (output_constant): Delete.
2395         * varasm.c (output_constant): Make private.
2397 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
2399         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2401 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2403         PR middle-end/59860
2404         * tree.h (fold_builtin_strcat): New prototype.
2405         * builtins.c (fold_builtin_strcat): No longer static.  Add len
2406         argument, if non-NULL, don't call c_strlen.  Optimize
2407         directly into __builtin_memcpy instead of __builtin_strcpy.
2408         (fold_builtin_2): Adjust fold_builtin_strcat caller.
2409         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2411 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2413         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2414         for SImode_address_operand operands, having only a REG argument.
2416 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2418         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2419         loader name using mbig-endian.
2420         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2422 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2424         * doc/invoke.texi (-march): Clarify documentation for AArch64.
2425         (-mtune): Likewise.
2426         (-mcpu): Likewise.
2428 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
2430         * config/aarch64/aarch64-protos.h
2431         (aarch64_cannot_change_mode_class_ptr): Declare.
2432         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2433         aarch64_cannot_change_mode_class_ptr): New.
2434         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2435         backend hook aarch64_cannot_change_mode_class.
2437 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2439         * common/config/aarch64/aarch64-common.c
2440         (aarch64_handle_option): Don't handle any option order logic here.
2441         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2442         selected_cpu, warn on architecture version mismatch.
2443         (aarch64_override_options): Fix parsing order for option strings.
2445 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2446             Iain Sandoe  <iain@codesourcery.com>
2448         PR bootstrap/59496
2449         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2450         warning.  Amend comment to reflect current functionality.
2452 2014-01-20  Richard Biener  <rguenther@suse.de>
2454         PR middle-end/59860
2455         * builtins.c (fold_builtin_strcat): Remove case better handled
2456         by tree-ssa-strlen.c.
2458 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2460         * config/aarch64/aarch64.opt
2461         (mcpu, march, mtune): Make case-insensitive.
2463 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2465         PR target/59880
2466         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2467         if operands[1] is a REG or ZERO_EXTEND of a REG.
2469 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2471         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2473 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2475         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2476         long non-pic millicode calls.
2478 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2480         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2482 2014-01-19  Kito Cheng  <kito@0xlab.org>
2484         * builtins.c (expand_movstr): Check movstr expand done or fail.
2486 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2487             H.J. Lu  <hongjiu.lu@intel.com>
2489         PR target/59379
2490         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2491         to DImode for zero-extended addresses.
2493 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2495         PR rtl-optimization/57763
2496         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2497         on the new indirect jump_insn and increment LABEL_NUSES (label).
2499 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2501         PR bootstrap/59580
2502         PR bootstrap/59583
2503         * config.gcc (x86_archs): New variable.
2504         (x86_64_archs): Likewise.
2505         (x86_cpus): Likewise.
2506         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2507         --with-arch/--with-cpu= options.
2508         Support --with-arch=/--with-cpu={nehalem,westmere,
2509         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2511 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2513         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2514         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2516 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2518         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2520 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2522         PR target/58944
2523         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2524         clear cpp_get_options (parse_in)->warn_unused_macros for
2525         ix86_target_macros_internal with cpp_define.
2527 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2529         * jump.c (delete_related_insns): Keep (use (insn))s.
2530         * reorg.c (redundant_insn): Check for barriers too.
2532 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2534         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2536 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2538         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2539         call to $$dyncall when TARGET_LONG_CALLS is true.
2541 2014-01-17  Jeff Law  <law@redhat.com>
2543         * ree.c (combine_set_extension): Temporarily disable test for
2544         changing number of hard registers.
2546 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2548         PR middle-end/58125
2549         * ipa-inline-analysis.c (inline_free_summary):
2550         Do not free summary of aliases.
2552 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2554         PR middle-end/59706
2555         * gimplify.c (gimplify_expr): Use create_tmp_var
2556         instead of create_tmp_var_raw.  If cond doesn't have
2557         integral type, don't add the IFN_ANNOTATE builtin at all.
2559 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2561         PR ipa/59736
2562         * ipa-cp.c (prev_edge_clone): New variable.
2563         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2564         Also resize prev_edge_clone vector.
2565         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2566         (ipcp_edge_removal_hook): New function.
2567         (ipcp_driver): Register ipcp_edge_removal_hook.
2569 2014-01-17  Andrew Pinski <apinski@cavium.com>
2570             Steve Ellcey  <sellcey@mips.com>
2572         PR target/59462
2573         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2574         of operator mode.
2576 2014-01-17  Jeff Law  <law@redhat.com>
2578         PR middle-end/57904
2579         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2580         so that pass_ccp runs first.
2582 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2584         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2585         (ix86_adjust_cost): Use !TARGET_XXX.
2586         (do_reorder_for_imul): Likewise.
2587         (swap_top_of_ready_list): Likewise.
2588         (ix86_sched_reorder): Likewise.
2590 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2592         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2593         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2594         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2595         (intel_memset): New.  Duplicate slm_memset.
2596         (intel_cost): New.  Duplicate slm_cost.
2597         (m_INTEL): New macro.
2598         (processor_target_table): Add "intel".
2599         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2600         with PROCESSOR_INTEL for "intel".
2601         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2602         PROCESSOR_SILVERMONT.
2603         (ix86_issue_rate): Likewise.
2604         (ix86_adjust_cost): Likewise.
2605         (ia32_multipass_dfa_lookahead): Likewise.
2606         (swap_top_of_ready_list): Likewise.
2607         (ix86_sched_reorder): Likewise.
2608         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2609         instead of TARGET_OPT_AGU.
2610         * config/i386/i386.h (TARGET_INTEL): New.
2611         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2612         (processor_type): Add PROCESSOR_INTEL.
2613         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2614         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2616 2014-01-17  Marek Polacek  <polacek@redhat.com>
2618         PR c/58346
2619         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2620         size is zero.
2622 2014-01-17  Richard Biener  <rguenther@suse.de>
2624         PR tree-optimization/46590
2625         * opts.c (default_options_table): Add entries for
2626         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2627         all enabled at -O1 but not for -Og.
2628         * common.opt (fbranch-count-reg): Remove Init(1).
2629         (fmove-loop-invariants): Likewise.
2630         (ftree-pta): Likewise.
2632 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2634         * config/i386/i386.c (ix86_data_alignment): For compatibility with
2635         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2636         decls to at least the GCC 4.8 used alignments.
2638         PR fortran/59440
2639         * tree-nested.c (convert_nonlocal_reference_stmt,
2640         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2641         of GIMPLE_BIND stmts, adjust associated decls.
2643 2014-01-17  Richard Biener  <rguenther@suse.de>
2645         PR tree-optimization/46590
2646         * vec.h (vec<>::bseach): New member function implementing
2647         binary search according to C89 bsearch.
2648         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2649         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2650         bitmap pointer again.  Make accesses_in_loop a flat array.
2651         (mem_ref_obstack): New global.
2652         (outermost_indep_loop): Adjust for mem_ref->stored changes.
2653         (mark_ref_stored): Likewise.
2654         (ref_indep_loop_p_2): Likewise.
2655         (set_ref_stored_in_loop): New helper function.
2656         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2657         (memref_free): Adjust.
2658         (record_mem_ref_loc): Simplify.
2659         (gather_mem_refs_stmt): Adjust.
2660         (sort_locs_in_loop_postorder_cmp): New function.
2661         (analyze_memory_references): Sort accesses_in_loop after
2662         loop postorder number.
2663         (find_ref_loc_in_loop_cmp): New function.
2664         (for_all_locs_in_loop): Find relevant cluster of locs in
2665         accesses_in_loop and iterate without recursion.
2666         (execute_sm): Avoid uninit warning.
2667         (struct ref_always_accessed): Simplify.
2668         (ref_always_accessed::operator ()): Likewise.
2669         (ref_always_accessed_p): Likewise.
2670         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2671         loop postorder numbers here.
2672         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2673         numbers.
2675 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2677         PR c++/57945
2678         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2679         on decls for which assemble_alias has been called.
2681 2014-01-17  Nick Clifton  <nickc@redhat.com>
2683         * config/msp430/msp430.opt: (mcpu): New option.
2684         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2685         (msp430_option_override): Parse target_cpu.  If the MCU name
2686         matches a generic string, clear target_mcu.
2687         (msp430_attr): Allow numeric interrupt values up to 63.
2688         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2689         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2690         option.
2691         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2692         Add mcpu matches.
2693         * config/msp430/msp430.md (popm): Use %J rather than %I.
2694         (addsi3): Use msp430_nonimmediate_operand for operand 2.
2695         (addhi_cy_i): Use immediate_operand for operand 2.
2696         * doc/invoke.texi: Document -mcpu option.
2698 2014-01-17  Richard Biener  <rguenther@suse.de>
2700         PR rtl-optimization/38518
2701         * df.h (df_analyze_loop): Declare.
2702         * df-core.c: Include cfgloop.h.
2703         (df_analyze_1): Split out main part of df_analyze.
2704         (df_analyze): Adjust.
2705         (loop_inverted_post_order_compute): New function.
2706         (loop_post_order_compute): Likewise.
2707         (df_analyze_loop): New function avoiding whole-function
2708         postorder computes.
2709         * loop-invariant.c (find_defs): Use df_analyze_loop.
2710         (find_invariants): Adjust.
2711         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2713 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
2715         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2716         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2718 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
2720         * ipa-ref.c (ipa_remove_stmt_references): Fix references
2721         traversal when removing references.
2723 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
2725         PR ipa/59775
2726         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2728 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
2730         PR middle-end/56791
2731         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2732         pushing a reload for an autoinc when we had previously reloaded an
2733         inner part of the address.
2735 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
2737         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2738         field.
2739         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2740         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2741         when not giving up or versioning for alias only because of
2742         loop->safelen.
2743         (vect_analyze_data_ref_dependences): Set to true.
2744         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2745         is a GIMPLE_PHI.
2746         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2747         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2748         to the condition.
2750         PR middle-end/58344
2751         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2753         PR target/59839
2754         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2755         operand 0 predicate for gathers, use a new pseudo as subtarget.
2757 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2759         PR middle-end/59609
2760         * lra-constraints.c (process_alt_operands): Add printing debug info.
2761         Check absence of input/output reloads for matched operands too.
2763 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2765         PR rtl-optimization/59835
2766         * ira.c (ira_init_register_move_cost): Increase cost for
2767         impossible modes.
2769 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
2771         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2773 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
2775         PR target/59780
2776         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2777         non-register objects.  Use gen_(high/low)part more consistently.
2778         Fix assertions.
2780 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
2782         PR target/59844
2783         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2784         endian support, remove tests for WORDS_BIG_ENDIAN.
2785         (p8_mfvsrd_3_<mode>): Likewise.
2786         (reload_gpr_from_vsx<mode>): Likewise.
2787         (reload_gpr_from_vsxsf): Likewise.
2788         (p8_mfvsrd_4_disf): Likewise.
2790 2014-01-16  Richard Biener  <rguenther@suse.de>
2792         PR rtl-optimization/46590
2793         * lcm.c (compute_antinout_edge): Use postorder iteration.
2794         (compute_laterin): Use inverted postorder iteration.
2796 2014-01-16  Nick Clifton  <nickc@redhat.com>
2798         PR middle-end/28865
2799         * varasm.c (output_constant): Return the number of bytes actually
2800         emitted.
2801         (output_constructor_array_range): Update the field size with the
2802         number of bytes emitted by output_constant.
2803         (output_constructor_regular_field): Likewise.  Also do not
2804         complain if the total number of bytes emitted is now greater
2805         than the expected fieldpos.
2806         * output.h (output_constant): Update prototype and descriptive comment.
2808 2014-01-16  Marek Polacek  <polacek@redhat.com>
2810         PR middle-end/59827
2811         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2812         it is error_mark_node.
2814 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
2816         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2817         VALID_AVX256_REG_OR_OI_MODE.
2819 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
2821         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2822         current procedure should be profiled.
2824 2014-01-15  Andrew Pinski  <apinski@cavium.com>
2826         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2827         of moving from/to the STACK_REG register class.
2829 2014-01-15  Richard Henderson <rth@redhat.com>
2831         PR debug/54694
2832         * reginfo.c (global_regs_decl): Globalize.
2833         * rtl.h (global_regs_decl): Declare.
2834         * ira.c (do_reload): Diagnose frame_pointer_needed and it
2835         reserved via global_regs.
2837 2014-01-15  Teresa Johnson  <tejohnson@google.com>
2839         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2841 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
2843         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2844         and vmulosh rather than call gen_vec_widen_smult_*.
2845         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2846         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2847         (vec_widen_smult_even_v16qi): Likewise.
2848         (vec_widen_umult_even_v8hi): Likewise.
2849         (vec_widen_smult_even_v8hi): Likewise.
2850         (vec_widen_umult_odd_v16qi): Likewise.
2851         (vec_widen_smult_odd_v16qi): Likewise.
2852         (vec_widen_umult_odd_v8hi): Likewise.
2853         (vec_widen_smult_odd_v8hi): Likewise.
2854         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2855         vmuloub rather than call gen_vec_widen_umult_*.
2856         (vec_widen_umult_lo_v16qi): Likewise.
2857         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2858         vmulosb rather than call gen_vec_widen_smult_*.
2859         (vec_widen_smult_lo_v16qi): Likewise.
2860         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2861         rather than call gen_vec_widen_umult_*.
2862         (vec_widen_umult_lo_v8hi): Likewise.
2863         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2864         rather than call gen_vec_widen_smult_*.
2865         (vec_widen_smult_lo_v8hi): Likewise.
2867 2014-01-15  Jeff Law  <law@redhat.com>
2869         PR tree-optimization/59747
2870         * ree.c (find_and_remove_re): Properly handle case where a second
2871         eliminated extension requires widening a copy created for elimination
2872         of a prior extension.
2873         (combine_set_extension): Ensure that the number of hard regs needed
2874         for a destination register does not change when we widen it.
2876 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2878         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
2879         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
2880         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
2881         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
2882         (avr-*-rtems*): Likewise.
2883         (bfin*-rtems*): Likewise.
2884         (moxie-*-rtems*): Likewise.
2885         (h8300-*-rtems*): Likewise.
2886         (i[34567]86-*-rtems*): Likewise.
2887         (lm32-*-rtems*): Likewise.
2888         (m32r-*-rtems*): Likewise.
2889         (m68k-*-rtems*): Likewise.
2890         (microblaze*-*-rtems*): Likewise.
2891         (mips*-*-rtems*): Likewise.
2892         (powerpc-*-rtems*): Likewise.
2893         (sh-*-rtems*): Likewise.
2894         (sparc-*-rtems*): Likewise.
2895         (sparc64-*-rtems*): Likewise.
2896         (v850-*-rtems*): Likewise.
2897         (m32c-*-rtems*): Likewise.
2899 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
2901         PR rtl-optimization/59511
2902         * ira.c (ira_init_register_move_cost): Use memory costs for some
2903         cases of register move cost calculations.
2904         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
2905         instead of BB frequency.
2906         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
2907         * lra-assigns.c (find_hard_regno_for): Ditto.
2909 2014-01-15  Richard Biener  <rguenther@suse.de>
2911         PR tree-optimization/59822
2912         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
2913         (vectorizable_load): Use it to hoist defs of uses of invariant
2914         loads out of the loop.
2916 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
2917             Kugan Vivekanandarajah  <kuganv@linaro.org>
2919         PR target/59695
2920         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
2921         truncation.
2923 2014-01-15  Richard Biener  <rguenther@suse.de>
2925         PR rtl-optimization/59802
2926         * lcm.c (compute_available): Use inverted postorder to seed
2927         the initial worklist.
2929 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2931         PR target/59803
2932         * config/s390/s390.c (s390_preferred_reload_class): Don't return
2933         ADDR_REGS for invalid symrefs in non-PIC code.
2935 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
2937         PR other/58712
2938         * builtins.c (determine_block_size): Initialize *probable_max_size
2939         even if len_rtx is CONST_INT.
2941 2014-01-14  Andrew Pinski  <apinski@cavium.com>
2943         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
2944         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
2945         (cortexa53_tunings): Likewise.
2946         (aarch64_sched_issue_rate): New function.
2947         (TARGET_SCHED_ISSUE_RATE): Define.
2949 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2951         * ira-costs.c (find_costs_and_classes): Add missed
2952         ira_init_register_move_cost_if_necessary.
2954 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2956         PR target/59787
2957         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
2959 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
2961         PR target/59794
2962         * config/i386/i386.c (type_natural_mode): Add a bool parameter
2963         to indicate if type is used for function return value.  Warn ABI
2964         change if the vector mode isn't available for function return value.
2965         (ix86_function_arg_advance): Pass false to type_natural_mode.
2966         (ix86_function_arg): Likewise.
2967         (ix86_gimplify_va_arg): Likewise.
2968         (function_arg_32): Don't warn ABI change.
2969         (ix86_function_value): Pass true to type_natural_mode.
2970         (ix86_return_in_memory): Likewise.
2971         (ix86_struct_value_rtx): Removed.
2972         (TARGET_STRUCT_VALUE_RTX): Likewise.
2974 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2976         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
2977         converting a conditional jump into a conditional return.
2979 2014-01-14  Richard Biener  <rguenther@suse.de>
2981         PR tree-optimization/58921
2982         PR tree-optimization/59006
2983         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
2984         hoisting invariant stmts.
2985         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
2986         invariant loads on the preheader edge if possible.
2988 2014-01-14  Joey Ye  <joey.ye@arm.com>
2990         * doc/plugin.texi (Building GCC plugins): Update to C++.
2992 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
2994         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
2995         (_mm_rcp28_round_ss): Ditto.
2996         (_mm_rsqrt28_round_sd): Ditto.
2997         (_mm_rsqrt28_round_ss): Ditto.
2998         (_mm_rcp28_sd): Ditto.
2999         (_mm_rcp28_ss): Ditto.
3000         (_mm_rsqrt28_sd): Ditto.
3001         (_mm_rsqrt28_ss): Ditto.
3002         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3003         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3004         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3005         (IX86_BUILTIN_RCP28SD): Ditto.
3006         (IX86_BUILTIN_RCP28SS): Ditto.
3007         (IX86_BUILTIN_RSQRT28SD): Ditto.
3008         (IX86_BUILTIN_RSQRT28SS): Ditto.
3009         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3010         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3011         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3012         (ix86_expand_special_args_builtin): Expand new FTYPE.
3013         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3014         (srcp14<mode>): Make insn unary.
3015         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3016         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3017         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3018         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3019         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3020         Fix rounding: make it SAE only.
3021         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3022         Ditto.
3023         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3024         Ditto.
3025         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3026         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3027         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3028         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3029         (round_saeonly_mask_scalar_operand4): Ditto.
3030         (round_saeonly_mask_scalar_op3): Ditto.
3031         (round_saeonly_mask_scalar_op4): Ditto.
3033 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3035         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3036         Implement -maltivec=be for vec_insert and vec_extract.
3038 2014-01-10  DJ Delorie  <dj@redhat.com>
3040         * config/msp430/msp430.md (call_internal): Don't allow memory
3041         references with SP as the base register.
3042         (call_value_internal): Likewise.
3043         * config/msp430/constraints.md (Yc): New.  For memory references
3044         that don't use SP as a base register.
3046         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3047         "an integer without a # prefix"
3048         * config/msp430/msp430.md (epilogue_helper): Use it.
3050 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3052         PR target/59617
3053         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3054         AVX512F gather builtins.
3055         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3056         on gather decls with INTEGER_TYPE masktype.
3057         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3058         directly into the builtin rather than hoisting it before loop.
3060         PR tree-optimization/59387
3061         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3062         (scev_const_prop): If folded_casts and type has undefined overflow,
3063         use force_gimple_operand instead of force_gimple_operand_gsi and
3064         for each added stmt if it is assign with
3065         arith_code_with_undefined_signed_overflow, call
3066         rewrite_to_defined_overflow.
3067         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3068         gimple-fold.h instead.
3069         (arith_code_with_undefined_signed_overflow,
3070         rewrite_to_defined_overflow): Moved to ...
3071         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3072         rewrite_to_defined_overflow): ... here.  No longer static.
3073         Include gimplify-me.h.
3074         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3075         rewrite_to_defined_overflow): New prototypes.
3077 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3079         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3081 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
3083         * builtins.c (get_object_alignment_2): Minor tweak.
3084         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3086 2014-01-13  Christian Bruel  <christian.bruel@st.com>
3088         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3089         optimized non constant lengths.
3091 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3093         PR libgomp/59194
3094         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3095         load as __atomic_load_N if possible.
3097 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3099         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3100         target parameter.
3101         (rs6000_expand_builtin): Adjust call.
3103 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3105         PR target/58115
3106         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3107         * config/rs6000/rs6000.c: Include target-globals.h.
3108         (rs6000_set_current_function): Instead of doing target_reinit
3109         unconditionally, use save_target_globals_default_opts and
3110         restore_target_globals.
3112         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3113         FPSCR.
3114         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3115         (rs6000_expand_builtin): Handle mffs and mtfsf.
3116         (rs6000_init_builtins): Define mffs and mtfsf.
3117         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3118         (rs6000_mffs): New pattern.
3119         (rs6000_mtfsf): New pattern.
3121 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
3123         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3124         Start narrowing with START.  Apply candidate-use pair
3125         and check overall cost in narrowing.
3126         (iv_ca_prune): Pass new argument.
3128 2014-01-10  Jeff Law  <law@redhat.com>
3130         PR middle-end/59743
3131         * ree.c (combine_reaching_defs): Ensure the defining statement
3132         occurs before the extension when optimizing extensions with
3133         different source and destination hard registers.
3135 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3137         PR ipa/58585
3138         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3139         vtables into the type inheritance graph.
3141 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3143         PR rtl-optimization/59754
3144         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3145         modes in the REGNO != REGNO case.
3147 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3149         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3151 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3153         PR tree-optimization/59745
3154         * tree-predcom.c (tree_predictive_commoning_loop): Call
3155         free_affine_expand_cache if giving up because components is NULL.
3157         * target-globals.c (save_target_globals): Allocate < 4KB structs using
3158         GC in payload of target_globals struct instead of allocating them on
3159         the heap and the larger structs separately using GC.
3160         * target-globals.h (struct target_globals): Make regs, hard_regs,
3161         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3162         of GTY((skip)) and change type to void *.
3163         (reset_target_globals): Cast loads from those fields to corresponding
3164         types.
3166 2014-01-10  Steve Ellcey  <sellcey@mips.com>
3168         PR plugins/59335
3169         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3170         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3171         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3173 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
3175         PR target/59744
3176         * aarch64-modes.def (CC_Zmode): New flags mode.
3177         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3178         represents an equality.
3179         (aarch64_get_condition_code): Handle CC_Zmode.
3180         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3182 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3184         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3185         extraction in good case.
3187 2014-01-10  Richard Biener  <rguenther@suse.de>
3189         PR tree-optimization/59374
3190         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3191         checking after SLP discovery.  Mark stmts not participating
3192         in any SLP instance properly.
3194 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3196         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3197         when handling a SET rtx.
3199 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3201         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3202         (cortex-a57): Likewise.
3203         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3205 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3207         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3208         non-iwmmxt builtins.
3210 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3212         PR ipa/58252
3213         PR ipa/59226
3214         * ipa-devirt.c record_target_from_binfo): Take as argument
3215         stack of binfos and lookup matching one for virtual inheritance.
3216         (possible_polymorphic_call_targets_1): Update.
3218 2014-01-10  Huacai Chen  <chenhc@lemote.com>
3220         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3221         kernel strings for Loongson-2E/2F/3A.
3223 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3225         PR middle-end/59670
3226         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3227         is_gimple_call before calling gimple_call_internal_p.
3229 2014-01-09  Steve Ellcey  <sellcey@mips.com>
3231         * Makefile.in (TREE_FLOW_H): Remove.
3232         (TREE_SSA_H): Add file names from tree-flow.h.
3233         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3234         * tree.h: Remove tree-flow.h reference.
3235         * hash-table.h: Remove tree-flow.h reference.
3236         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3237         reference with tree-ssa-loop.h.
3239 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3241         * doc/invoke.texi: Add -maltivec={be,le} options, and document
3242         default element-order behavior for -maltivec.
3243         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3244         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3245         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3246         when targeting big endian, at least for now.
3247         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3249 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3251         PR middle-end/47735
3252         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3253         var satisfies use_register_for_decl, just take into account type
3254         alignment, rather than decl alignment.
3256         PR tree-optimization/59622
3257         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
3258         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3259         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3260         Don't devirtualize for inplace at all.  For targets.length () == 1,
3261         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3263 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3265         * config/i386/i386.md (cpu): Remove the unused btver1.
3267 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3269         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3271 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3273         PR target/58115
3274         * tree-core.h (struct target_globals): New forward declaration.
3275         (struct tree_target_option): Add globals field.
3276         * tree.h (TREE_TARGET_GLOBALS): Define.
3277         (prepare_target_option_nodes_for_pch): New prototype.
3278         * target-globals.h (struct target_globals): Define even if
3279         !SWITCHABLE_TARGET.
3280         * tree.c (prepare_target_option_node_for_pch,
3281         prepare_target_option_nodes_for_pch): New functions.
3282         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3283         * config/i386/i386.c: Include target-globals.h.
3284         (ix86_set_current_function): Instead of doing target_reinit
3285         unconditionally, use save_target_globals_default_opts and
3286         restore_target_globals.
3288 2014-01-09  Richard Biener  <rguenther@suse.de>
3290         PR tree-optimization/59715
3291         * tree-cfg.h (split_critical_edges): Declare.
3292         * tree-cfg.c (split_critical_edges): Export.
3293         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3295 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
3297         * cfgexpand.c (expand_stack_vars): Optionally disable
3298         asan stack protection.
3299         (expand_used_vars): Likewise.
3300         (partition_stack_vars): Likewise.
3301         * asan.c (asan_emit_stack_protection): Optionally disable
3302         after return stack usage.
3303         (instrument_derefs): Optionally disable memory access instrumentation.
3304         (instrument_builtin_call): Likewise.
3305         (instrument_strlen_call): Likewise.
3306         (asan_protect_global): Optionally disable global variables protection.
3307         * doc/invoke.texi: Added doc for new options.
3308         * params.def: Added new options.
3309         * params.h: Likewise.
3311 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3313         PR rtl-optimization/59724
3314         * ifcvt.c (cond_exec_process_if_block): Don't call
3315         flow_find_head_matching_sequence with 0 longest_match.
3316         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3317         non-active insns if !stop_after.
3318         (try_head_merge_bb): Revert 2014-01-07 changes.
3320 2014-01-08  Jeff Law  <law@redhat.com>
3322         * ree.c (get_sub_rtx): New function, extracted from...
3323         (merge_def_and_ext): Here.
3324         (combine_reaching_defs): Use get_sub_rtx.
3326 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
3328         * cgraph.h (varpool_variable_node): Do not choke on null node.
3330 2014-01-08  Catherine Moore  <clm@codesourcery.com>
3332         * config/mips/mips.md (simple_return): Attempt to use JRC
3333         for microMIPS.
3334         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3336 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3338         PR rtl-optimization/59137
3339         * reorg.c (steal_delay_list_from_target): Call update_block for
3340         elided insns.
3341         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3343 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3345         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3346         two duplicate entries.
3348 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3350         Revert:
3351         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
3353         * config/mips/mips.c (mips_truncated_op_cost): New function.
3354         (mips_rtx_costs): Adjust test for BADDU.
3355         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3357         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
3359         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3360         (*baddu_si): ...this new pattern.
3362 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3364         PR ipa/59722
3365         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3367 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3369         PR middle-end/57748
3370         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3371         inner_reference_p.
3372         (expand_expr, expand_normal): Adjust.
3373         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3374         inner_reference_p. Use inner_reference_p to expand inner references.
3375         (store_expr): Adjust.
3376         * cfgexpand.c (expand_call_stmt): Adjust.
3378 2014-01-08  Rong Xu  <xur@google.com>
3380         * gcov-io.c (gcov_var): Move from gcov-io.h.
3381         (gcov_position): Ditto.
3382         (gcov_is_error): Ditto.
3383         (gcov_rewrite): Ditto.
3384         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3385         only part to libgcc/libgcov.h.
3387 2014-01-08  Marek Polacek  <polacek@redhat.com>
3389         PR middle-end/59669
3390         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3392 2014-01-08  Marek Polacek  <polacek@redhat.com>
3394         PR sanitizer/59667
3395         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3397 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3399         PR rtl-optimization/59649
3400         * stor-layout.c (get_mode_bounds): For BImode return
3401         0 and STORE_FLAG_VALUE.
3403 2014-01-08  Richard Biener  <rguenther@suse.de>
3405         PR middle-end/59630
3406         * gimple.h (is_gimple_builtin_call): Remove.
3407         (gimple_builtin_call_types_compatible_p): New.
3408         (gimple_call_builtin_p): New overload.
3409         * gimple.c (is_gimple_builtin_call): Remove.
3410         (validate_call): Rename to ...
3411         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
3412         check return types.
3413         (validate_type): New static function.
3414         (gimple_call_builtin_p): New overload and adjust.
3415         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3416         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
3417         (gimple_fold_stmt_to_constant_1): Likewise.
3418         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3420 2014-01-08  Richard Biener  <rguenther@suse.de>
3422         PR middle-end/59471
3423         * gimplify.c (gimplify_expr): Gimplify register-register type
3424         VIEW_CONVERT_EXPRs to separate stmts.
3426 2014-01-07  Jeff Law  <law@redhat.com>
3428         PR middle-end/53623
3429         * ree.c (combine_set_extension): Handle case where source
3430         and destination registers in an extension insn are different.
3431         (combine_reaching_defs): Allow source and destination registers
3432         in extension to be different under limited circumstances.
3433         (add_removable_extension): Remove restriction that the
3434         source and destination registers in the extension are the same.
3435         (find_and_remove_re): Emit a copy from the extension's
3436         destination to its source after the defining insn if
3437         the source and destination registers are different.
3439         PR middle-end/59285
3440         * ifcvt.c (merge_if_block): If we are merging a block with more than
3441         one successor with a block with no successors, remove any BARRIER
3442         after the second block.
3444 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3446         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3448 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3450         PR target/59652
3451         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3452         for 14-bit register offsets when INT14_OK_STRICT is false.
3454 2014-01-07  Roland Stigge  <stigge@antcom.de>
3455             Michael Meissner  <meissner@linux.vnet.ibm.com>
3457         PR 57386/target
3458         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3459         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3461 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3463         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3464         -mcpu.
3466 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3468         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3469         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3470         rtx is const0_rtx or not.
3472 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3474         PR target/58115
3475         * target-globals.c (save_target_globals): Remove this_fn_optab
3476         handling.
3477         * toplev.c: Include optabs.h.
3478         (target_reinit): Temporarily restore the global options if another
3479         set of options are in force.
3481 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3483         PR rtl-optimization/58668
3484         * cfgcleanup.c (flow_find_cross_jump): Don't count
3485         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3486         to determine what is counted.
3487         (flow_find_head_matching_sequence): Use active_insn_p to determine
3488         what is counted.
3489         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3490         counting change.
3491         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3492         determine what is counted.
3494         PR tree-optimization/59643
3495         * tree-predcom.c (split_data_refs_to_components): If one dr is
3496         read and one write, determine_offset fails and the write isn't
3497         in the bad component, just put the read into the bad component.
3499 2014-01-07  Mike Stump  <mikestump@comcast.net>
3500             Jakub Jelinek  <jakub@redhat.com>
3502         PR pch/59436
3503         * tree-core.h (struct tree_optimization_option): Change optabs
3504         type from unsigned char * to void *.
3505         * optabs.c (init_tree_optimization_optabs): Adjust
3506         TREE_OPTIMIZATION_OPTABS initialization.
3508 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3510         PR target/59644
3511         * config/i386/i386.h (struct machine_function): Add
3512         no_drap_save_restore field.
3513         * config/i386/i386.c (ix86_save_reg): Use
3514         !cfun->machine->no_drap_save_restore instead of
3515         crtl->stack_realign_needed.
3516         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3517         this function clears frame_pointer_needed.  Set
3518         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3519         and DRAP reg is needed.
3521 2014-01-06  Marek Polacek  <polacek@redhat.com>
3523         PR c/57773
3524         * doc/implement-c.texi: Mention that other integer types are
3525         permitted as bit-field types in strictly conforming mode.
3527 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3529         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3530         is newly allocated.
3532 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3534         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3536 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3538         PR ipa/59008
3539         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3540         to int.
3541         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3543 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3545         PR debug/59350
3546         PR debug/59510
3547         * var-tracking.c (add_stores): Preserve the value of the source even if
3548         we don't record the store.
3550 2014-01-06  Terry Guo  <terry.guo@arm.com>
3552         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3554 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3556         PR bootstrap/59541
3557         * config/darwin.c (darwin_function_section): Adjust return values to
3558         correspond to optimisation changes made in r206070.
3560 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3562         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3563         from prefetch_block tune setting.
3564         (nocona_cost): Correct size of prefetch block to 64.
3566 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3568         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3569         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3570         used to save the static chain register in the computation of the offset
3571         from which the FP registers need to be restored.
3573 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3575         PR tree-optimization/59519
3576         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3577         ICE if get_current_def (current_new_name) is already non-NULL, as long
3578         as it is a phi result of some other phi in *new_exit_bb that has
3579         the same argument.
3581         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3582         or vmovdqu* for misaligned_operand.
3583         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3584         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3585         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3586         aligned_mem for AVX512F masked aligned load and store builtins and for
3587         non-temporal moves.
3589 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3591         PR tree-optimization/59651
3592         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3593         Address range for negative step should be added by TYPE_SIZE_UNIT.
3595 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3597         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3598         overlapping registers also for registers other than the stack pointer.
3600 2014-01-03  Marek Polacek  <polacek@redhat.com>
3602         PR other/59661
3603         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3604         __builtin_FILE.
3606 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
3608         PR target/59625
3609         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3610         asm goto as jump.
3612         * config/i386/i386.md (MODE_SIZE): New mode attribute.
3613         (push splitter): Use <P:MODE_SIZE> instead of
3614         GET_MODE_SIZE (<P:MODE>mode).
3615         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3616         (mov -1, reg peephole2): Likewise.
3617         * config/i386/sse.md (*mov<mode>_internal,
3618         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3619         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3620         *<code><mode>3, *andnot<mode>3<mask_name>,
3621         <mask_codefor><code><mode>3<mask_name>): Likewise.
3622         * config/i386/subst.md (mask_mode512bit_condition,
3623         sd_mask_mode512bit_condition): Likewise.
3625 2014-01-02  Xinliang David Li  <davidxl@google.com>
3627         PR tree-optimization/59303
3628         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3629         (dump_predicates): Better output format.
3630         (pred_equal_p): New function.
3631         (is_neq_relop_p): Ditto.
3632         (is_neq_zero_form_p): Ditto.
3633         (pred_expr_equal_p): Ditto.
3634         (pred_neg_p): Ditto.
3635         (simplify_pred): Ditto.
3636         (simplify_preds_2): Ditto.
3637         (simplify_preds_3): Ditto.
3638         (simplify_preds_4): Ditto.
3639         (simplify_preds): Ditto.
3640         (push_pred): Ditto.
3641         (push_to_worklist): Ditto.
3642         (get_pred_info_from_cmp): Ditto.
3643         (is_degenerated_phi): Ditto.
3644         (normalize_one_pred_1): Ditto.
3645         (normalize_one_pred): Ditto.
3646         (normalize_one_pred_chain): Ditto.
3647         (normalize_preds): Ditto.
3648         (normalize_cond_1): Remove function.
3649         (normalize_cond): Ditto.
3650         (is_gcond_subset_of): Ditto.
3651         (is_subset_of_any): Ditto.
3652         (is_or_set_subset_of): Ditto.
3653         (is_and_set_subset_of): Ditto.
3654         (is_norm_cond_subset_of): Ditto.
3655         (pred_chain_length_cmp): Ditto.
3656         (convert_control_dep_chain_into_preds): Type change.
3657         (find_predicates): Ditto.
3658         (find_def_preds): Ditto.
3659         (destroy_predicates_vecs): Ditto.
3660         (find_matching_predicates_in_rest_chains): Ditto.
3661         (use_pred_not_overlap_with_undef_path_pred): Ditto.
3662         (is_pred_expr_subset): Ditto.
3663         (is_pred_chain_subset_of): Ditto.
3664         (is_included_in): Ditto.
3665         (is_superset_of): Ditto.
3667 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3669         Update copyright years.
3671 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3673         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3674         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3675         config/arc/arc.md, config/arc/arc.opt,
3676         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3677         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3678         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3679         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3680         config/linux-protos.h, config/linux.c, config/winnt-c.c,
3681         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3682         vtable-verify.c, vtable-verify.h: Use the standard form for the
3683         copyright notice.
3685 2014-01-02  Tobias Burnus  <burnus@net-b.de>
3687         * gcc.c (process_command): Update copyright notice dates.
3688         * gcov-dump.c: Ditto.
3689         * gcov.c: Ditto.
3690         * doc/cpp.texi: Bump @copying's copyright year.
3691         * doc/cppinternals.texi: Ditto.
3692         * doc/gcc.texi: Ditto.
3693         * doc/gccint.texi: Ditto.
3694         * doc/gcov.texi: Ditto.
3695         * doc/install.texi: Ditto.
3696         * doc/invoke.texi: Ditto.
3698 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3700         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3702 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
3704         * config/i386/sse.md (*mov<mode>_internal): Guard
3705         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3707         PR rtl-optimization/59647
3708         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3709         new_rtx into UNSIGNED_FLOAT rtxes.
3711 Copyright (C) 2014 Free Software Foundation, Inc.
3713 Copying and distribution of this file, with or without modification,
3714 are permitted in any medium without royalty provided the copyright
3715 notice and this notice are preserved.