2014-02-13 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob881cad1009d7ecbecef1908086f35de55fa0d6ca
1 2014-02-13  Richard Biener  <rguenther@suse.de>
3         PR bootstrap/59878
4         * doc/install.texi (ISL): Update recommended version to 0.12.2,
5         mention the possibility of an in-tree build.
6         (CLooG): Update recommended version to 0.18.1, mention the
7         possibility of an in-tree build and clarify that the ISL
8         bundled with CLooG does not work.
10 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
12         PR target/43546
13         * expr.c (compress_float_constant): If x is a hard register,
14         extend into a pseudo and then move to x.
16 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
18         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
19         caused by bad second argument to warning_at() with -mhotpatch and
20         nested functions (e.g. with gfortran).
22 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
24         * opts.c (option_name): Remove "enabled by default" rider.
26 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
28         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
30 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
31             Uros Bizjak  <ubizjak@gmail.com>
33         PR target/60151
34         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU
35         assembler.
36         * configure: Regenerated.
38 2014-02-12  Richard Biener  <rguenther@suse.de>
40         * vec.c (vec_prefix::calculate_allocation): Move as
41         inline variant to vec.h.
42         (vec_prefix::calculate_allocation_1): New out-of-line version.
43         * vec.h (vec_prefix::calculate_allocation_1): Declare.
44         (vec_prefix::m_has_auto_buf): Rename to ...
45         (vec_prefix::m_using_auto_storage): ... this.
46         (vec_prefix::calculate_allocation): Inline the easy cases
47         and dispatch to calculate_allocation_1 which doesn't need the
48         prefix address.
49         (va_heap::reserve): Use gcc_checking_assert.
50         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
51         m_using_auto_storage.
52         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
53         member and adjust.
54         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
55         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
56         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
58 2014-02-12  Richard Biener  <rguenther@suse.de>
60         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
61         when we found a dependence.
63 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
65         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
66         common code...
67         (maybe_fold_stmt): ... into this new function.
68         * omp-low.c (lower_omp): Update comment.
70         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
71         last use.
73         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
74         dereference.
76 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
78         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
79         identifiers in comments.
80         (cortexa53_extra_costs): Likewise.
81         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in
82         comments.
83         (cortexa7_extra_costs): Likewise.
84         (cortexa12_extra_costs): Likewise.
85         (cortexa15_extra_costs): Likewise.
86         (v7m_extra_costs): Likewise.
88 2014-02-12  Richard Biener  <rguenther@suse.de>
90         PR middle-end/60092
91         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
92         of posix_memalign being successful.
93         (lower_stmt): Restrict lowering of posix_memalign to when
94         -ftree-bit-ccp is enabled.
96 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
98         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
99         arg_loc.
100         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
102 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
104         PR rtl-optimization/60116
105         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
106         other_insn once the combination has been validated.
108 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
110         PR lto/59468
111         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
112         and wrapper.
113         * ipa-devirt.c: Include demangle.h
114         (odr_violation_reported): New static variable.
115         (add_type_duplicate): Update odr_violations.
116         (maybe_record_node): Add completep parameter; update it.
117         (record_target_from_binfo): Add COMPLETEP parameter;
118         update it as needed.
119         (possible_polymorphic_call_targets_1): Likewise.
120         (struct polymorphic_call_target_d): Add nonconstruction_targets;
121         rename FINAL to COMPLETE.
122         (record_targets_from_bases): Sanity check we found the binfo;
123         fix COMPLETEP updating.
124         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
125         parameter, fix computing of COMPLETEP.
126         (dump_possible_polymorphic_call_targets): Imrove readability of dump; at
127         LTO time do demangling.
128         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
129         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
130         parameter.
131         (gimple_get_virt_method_for_binfo): Likewise.
132         * gimple-fold.h (gimple_get_virt_method_for_binfo,
133         gimple_get_virt_method_for_vtable): Update prototypes.
135 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
137         PR target/49008
138         * genautomata.c (add_presence_absence): Fix typo with
139         {final_}presence_list.
141 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
143         PR target/60137
144         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
145         for VSX/Altivec vectors that land in GPR registers.
147 2014-02-11  Richard Henderson  <rth@redhat.com>
148             Jakub Jelinek  <jakub@redhat.com>
150         PR debug/59776
151         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
152         around drhs if type conversion to lacc->type is not useless.
154 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
156         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
157         tuning struct.
158         (cortex-a57.cortex-a53): Likewise.
159         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
161 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
163         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
164         arm_restrict_it.
166 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
168         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
169         add_options_for_arm_vfp3.
171 2014-02-11  Jeff Law  <law@redhat.com>
173         PR middle-end/54041
174         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
175         object with an undesirable mode.
177 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
179         PR libgomp/60107
180         * config/i386/sol2-9.h: New file.
181         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
182         *-*-solaris2.9*): Use it.
184 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
186         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
187         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
189 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
191         * config/microblaze/microblaze.c: Extend mcpu version format
193 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
195         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
197 2014-02-10  Richard Henderson <rth@redhat.com>
199         PR target/59927
200         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
201         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
202         ms-abi vs -mno-accumulate-outgoing-args.
203         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
204         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
205         respect to ms-abi.
207 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
209         PR middle-end/60080
210         * cfgexpand.c (expand_asm_operands): Attach source location to
211         ASM_INPUT rtx objects.
212         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
214 2014-02-10  Nick Clifton  <nickc@redhat.com>
216         * config/mn10300/mn10300.c (popcount): New function.
217         (mn10300_expand_prologue): Include saved registers in stack usage
218         count.
220 2014-02-10  Jeff Law  <law@redhat.com>
222         PR middle-end/52306
223         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
224         when changing the SET_DEST of a prior insn to avoid an input
225         reload.
227 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
229         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
230         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
231         -mcall-openbsd, or -mcall-linux.
232         (CC1_ENDIAN_BIG_SPEC): Remove.
233         (CC1_ENDIAN_LITTLE_SPEC): Remove.
234         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
235         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
236         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
237         and %cc1_endian_default.
238         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
240 2014-02-10  Richard Biener  <rguenther@suse.de>
242         PR tree-optimization/60115
243         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
244         MEM_REF handling.  Properly verify that the accesses are not
245         out of the objects bound.
247 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
249         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
250         coretex to cortex.
252 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
254         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
255         proper constants and fix formatting.
256         (possible_polymorphic_call_targets): Fix formatting.
258 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
259             Ilya Tocar  <ilya.tocar@intel.com>
261         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
262         (_mm512_loadu_epi32): Renamed into...
263         (_mm512_loadu_si512): This.
264         (_mm512_storeu_epi32): Renamed into...
265         (_mm512_storeu_si512): This.
266         (_mm512_maskz_ceil_ps): Removed.
267         (_mm512_maskz_ceil_pd): Ditto.
268         (_mm512_maskz_floor_ps): Ditto.
269         (_mm512_maskz_floor_pd): Ditto.
270         (_mm512_floor_round_ps): Ditto.
271         (_mm512_floor_round_pd): Ditto.
272         (_mm512_ceil_round_ps): Ditto.
273         (_mm512_ceil_round_pd): Ditto.
274         (_mm512_mask_floor_round_ps): Ditto.
275         (_mm512_mask_floor_round_pd): Ditto.
276         (_mm512_mask_ceil_round_ps): Ditto.
277         (_mm512_mask_ceil_round_pd): Ditto.
278         (_mm512_maskz_floor_round_ps): Ditto.
279         (_mm512_maskz_floor_round_pd): Ditto.
280         (_mm512_maskz_ceil_round_ps): Ditto.
281         (_mm512_maskz_ceil_round_pd): Ditto.
282         (_mm512_expand_pd): Ditto.
283         (_mm512_expand_ps): Ditto.
284         * config/i386/i386.c (ix86_builtins): Remove
285         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
286         (bdesc_args): Ditto.
287         * config/i386/predicates.md (const1256_operand): New.
288         (const_1_to_2_operand): Ditto.
289         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
290         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
291         (*avx512pf_gatherpf<mode>sf): Ditto.
292         (avx512pf_gatherpf<mode>df): Ditto.
293         (*avx512pf_gatherpf<mode>df_mask): Ditto.
294         (*avx512pf_gatherpf<mode>df): Ditto.
295         (avx512pf_scatterpf<mode>sf): Ditto.
296         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
297         (*avx512pf_scatterpf<mode>sf): Ditto.
298         (avx512pf_scatterpf<mode>df): Ditto.
299         (*avx512pf_scatterpf<mode>df_mask): Ditto.
300         (*avx512pf_scatterpf<mode>df): Ditto.
301         (avx512f_expand<mode>): Removed.
302         (<shift_insn><mode>3<mask_name>): Change predicate type.
304 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
306         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
307         not at the end of datarefs vector use ordered_remove to avoid
308         reordering datarefs vector.
310         PR c/59984
311         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
312         mark local addressable non-static vars as GOVD_PRIVATE
313         instead of GOVD_LOCAL.
314         * omp-low.c (lower_omp_for): Move gimple_bind_vars
315         and BLOCK_VARS of gimple_bind_block to new_stmt rather
316         than copying them.
318         PR middle-end/60092
319         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
320         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
321         assume_aligned or alloc_align attributes.
322         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
323         arguments.  Handle also assume_aligned and alloc_align attributes.
324         (evaluate_stmt): Adjust bit_value_assume_aligned caller.
325         Handle calls to functions with assume_aligned or alloc_align
326         attributes.
327         * doc/extend.texi: Document assume_aligned and alloc_align
328         attributes.
330 2014-02-08  Terry Guo  <terry.guo@arm.com>
332         * doc/invoke.texi: Document ARM -march=armv7e-m.
334 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
336         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
337         flag on __cilkrts_rethrow builtin.
339         PR ipa/60026
340         * ipa-cp.c (determine_versionability): Fail at -O0
341         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
342         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
344         Revert:
345         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
347         PR ipa/60026
348         * tree-inline.c (copy_forbidden): Fail for
349         __attribute__((optimize (0))) functions.
351 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
353         * varpool.c: Include pointer-set.h.
354         (varpool_remove_unreferenced_decls): Variables in other partitions
355         will not be output; be however careful to not lose information
356         about partitioning.
358 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
360         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
361         lookup in the vtable constructor.
363 2014-02-07  Jeff Law  <law@redhat.com>
365         PR target/40977
366         * config/m68k/m68k.md (ashldi_extsi): Turn into a
367         define_insn_and_split.
369         * ipa-inline.c (inline_small_functions): Fix typos.
371 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
373         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
374         (s390_can_use_return_insn): Declare.
375         * config/s390/s390.h (EPILOGUE_USES): Define.
376         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
377         instructions.
378         (s390_chunkify_start): Handle return JUMP_LABELs.
379         (s390_early_mach): Emit a main_pool instruction on the entry edge.
380         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
381         (s390_can_use_return_insn): New functions.
382         (s390_fix_long_loop_prediction): Handle conditional returns.
383         (TARGET_SET_UP_BY_PROLOGUE): Define.
384         * config/s390/s390.md (ANY_RETURN): New code iterator.
385         (*creturn, *csimple_return, return, simple_return): New patterns.
387 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
389         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
390         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
391         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
392         REG_CFA_RESTORE list when deciding not to restore a register.
394 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
396         * config/s390/s390.c: Include tree-pass.h and context.h.
397         (s390_early_mach): New function, split out from...
398         (s390_emit_prologue): ...here.
399         (pass_data_s390_early_mach): New pass structure.
400         (pass_s390_early_mach): New class.
401         (s390_option_override): Create and register early_mach pass.
402         Move to end of file.
404 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
406         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
407         to match for the exit block.
409 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
411         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
412         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
413         Reject misaligned operands.
415 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
417         * optabs.c (expand_atomic_compare_and_swap): Allow expander to
418         fail.
420 2014-02-07  Richard Biener  <rguenther@suse.de>
422         PR middle-end/60092
423         * gimple-low.c (lower_builtin_posix_memalign): New function.
424         (lower_stmt): Call it to lower posix_memalign in a way
425         to make alignment info accessible.
427 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
429         PR c++/60082
430         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
431         __builtin_setjmp_receiver.
433 2014-02-07  Richard Biener  <rguenther@suse.de>
435         PR middle-end/60092
436         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
437         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
438         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
439         Handle BUILT_IN_POSIX_MEMALIGN.
440         (find_func_clobbers): Likewise.
441         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
442         (call_may_clobber_ref_p_1): Likewise.
444 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
446         PR ipa/59918
447         * ipa-devirt.c (record_target_from_binfo): Remove overactive sanity check.
449 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
451         PR ipa/59469
452         * lto-cgraph.c (lto_output_node): Use
453         symtab_get_symbol_partitioning_class.
454         (lto_output_varpool_node): likewise.
455         (symtab_get_symbol_partitioning_class): Move here from
456         lto/lto-partition.c
457         * cgraph.h (symbol_partitioning_class): Likewise.
458         (symtab_get_symbol_partitioning_class): Declare.
460 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
462         * ggc.h (ggc_internal_cleared_alloc): New macro.
463         * vec.h (vec_safe_copy): Handle memory stats.
464         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
465         * target-globals.c (save_target_globals): Likewise.
467 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
469         PR target/60077
470         * expr.c (emit_move_resolve_push): Export; be bit more selective
471         on when to clear alias set.
472         * expr.h (emit_move_resolve_push): Declare.
473         * function.h (struct function): Add tail_call_marked.
474         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
475         * config/i386/i386-protos.h (ix86_expand_push): Remove.
476         * config/i386/i386.md (TImode move expander): De not call
477         ix86_expand_push.
478         (FP push expanders): Preserve memory attributes.
479         * config/i386/sse.md (push<mode>1): Remove.
480         * config/i386/i386.c (ix86_expand_vector_move): Handle push
481         operation.
482         (ix86_expand_push): Remove.
483         * config/i386/mmx.md (push<mode>1): Remove.
485 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
487         PR rtl-optimization/60030
488         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
489         lopart with paradoxical subreg before shifting it up by hprec.
491 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
493         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
494         Remove extra newline at end of file.
495         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
496         (arm_issue_rate): Handle cortexa57.
497         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
498         (cortex-a57.cortex-a53): Likewise.
500 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
502         PR target/59575
503         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
504         don't record in REG_FRAME_RELATED_EXPR registers not set in that
505         bitmask.
506         (arm_expand_prologue): Adjust all callers.
507         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
508         info, registers also at the lowest numbered registers side.  Use
509         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
510         XEXP.
512         PR debug/59992
513         * var-tracking.c (adjust_mems): Before adding a SET
514         to amd->side_effects, adjust it's SET_SRC using
515         simplify_replace_fn_rtx.
517 2014-02-06  Alan Modra  <amodra@gmail.com>
519         PR target/60032
520         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
521         change SDmode to DDmode when lra_in_progress.
523 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
525         PR middle-end/59150
526         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
527         free_data_ref on the dr first, and before goto again also set dr
528         to the next dr.  For simd_lane_access, free old datarefs[i] before
529         overwriting it.  For get_vectype_for_scalar_type failure, don't
530         free_data_ref if simd_lane_access.
532         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
534         PR target/60062
535         * tree.h (opts_for_fn): New inline function.
536         (opt_for_fn): Define.
537         * config/i386/i386.c (ix86_function_regparm): Use
538         opt_for_fn (decl, optimize) instead of optimize.
540 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
542         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
543         for SYMBOL_REF in large memory model.
545 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
547         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
548         and crypto support.
549         (cortex-a57): Likewise.
550         (cortex-a57.cortex-a53): Likewise.
552 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
553             Kugan Vivekanandarajah  <kuganv@linaro.org>
555         * config/arm/arm.c (arm_vector_alignment_reachable): Check
556         unaligned_access.
557         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
559 2014-02-06  Richard Biener  <rguenther@suse.de>
561         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
562         set_loop_copy and initialize_original_copy_tables.
564 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
566         * config/aarch64/aarch64-simd.md
567         (aarch64_ashr_simddi): Change QI to SI.
569 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
570             Jakub Jelinek  <jakub@redhat.com>
572         PR middle-end/60013
573         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
574         of the dataflow.
576 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
578         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
579         CODE_FOR_altivec_vpku[hw]um to
580         CODE_FOR_altivec_vpku[hw]um_direct.
581         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
582         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
583         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
584         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
586 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
588         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
589         generation for -maltivec=be.
590         (altivec_vsumsws): Simplify redundant test.
592 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
594         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
595         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
596         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
597         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
598         gen_altivec_vpkuwum.
599         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
600         BYTES_BIG_ENDIAN.
601         (altivec_vpks<VI_char>ss): Likewise.
602         (altivec_vpks<VI_char>us): Likewise.
603         (altivec_vpku<VI_char>us): Likewise.
604         (altivec_vpku<VI_char>um): Likewise.
605         (altivec_vpku<VI_char>um_direct): New (copy of
606         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
607         internal use).
608         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
609         target is little endian and -maltivec=be is not specified.
610         (*altivec_vupkhs<VU_char>_direct): New (copy of
611         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal
612         use).
613         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
614         target is little endian and -maltivec=be is not specified.
615         (*altivec_vupkls<VU_char>_direct): New (copy of
616         altivec_vupkls<VU_char> that always emits vupkls*, for internal
617         use).
618         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
619         little endian and -maltivec=be is not specified.
620         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
621         little endian and -maltivec=be is not specified.
623 2014-02-05  Richard Henderson <rth@redhat.com>
625         PR debug/52727
626         * combine-stack-adj.c: Revert r206943.
627         * sched-int.h (struct deps_desc): Add last_args_size.
628         * sched-deps.c (init_deps): Initialize it.
629         (sched_analyze_insn): Add OUTPUT dependencies between insns that
630         contain REG_ARGS_SIZE notes.
632 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
634         * lto-cgraph.c (asm_nodes_output): Make global.
635         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
636         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA
637         parameter
638         (driver_handle_option): Handle OPT_fwpa.
640 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
642         PR ipa/59947
643         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
644         a comment typo and formatting issue.  If odr_hash hasn't been
645         created, return vNULL and set *completep to false.
647         PR middle-end/57499
648         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
649         bb with no successors.
651 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
653         PR target/59718
654         * doc/invoke.texi (-march): Clarify documentation for ARM.
655         (-mtune): Likewise.
656         (-mcpu): Likewise.
658 2014-02-05  Richard Biener  <rguenther@suse.de>
660         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
661         when not vectorizing because of too many alias checks.
662         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
663         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
665 2014-02-05  Nick Clifton  <nickc@redhat.com>
667         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
668         accept extended registers in any mode when compiling for the
669         MN10300.
671 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
673         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
674         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
675         sanitization attributes.
676         (can_inline_edge_p): Likewise.
677         (sanitize_attrs_match_for_inline_p): New function.
679 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
681         * ipa-prop.c (detect_type_change): Shor circuit testing of
682         type changes on THIS pointer.
684 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
686         PR target/59777
687         * config/pa/pa.c (legitimize_tls_address): Return original address
688         if not passed a SYMBOL_REF rtx.
689         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
690         addresses.
691         (pa_emit_move_sequence): Simplify TLS source operands.
692         (pa_legitimate_constant_p): Reject all TLS constants.
693         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
694         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
696 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
698         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
699         groups when we know they are controlled by LTO.
700         * varasm.c (default_binds_local_p_1): If object is in other partition,
701         it will be resolved locally.
703 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
705         * config/host-linux.c (linux_gt_pch_use_address): Don't
706         use SSIZE_MAX because it is not always defined. 
708 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
710         PR bootstrap/59913
711         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
712         threshold for pseudo splitting.
713         (update_ebb_live_info): Process call argument hard registers and
714         hard registers from insn definition too.
715         (max_small_class_regs_num): New constant.
716         (inherit_in_ebb): Update live hard regs through EBBs.  Update
717         reloads_num only for small register classes.  Don't split for
718         outputs of jumps.
720 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
722         PR ipa/60058
723         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
724         is non-null.
726 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
728         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default visibility is safe.
730 2014-02-04  Marek Polacek  <polacek@redhat.com>
732         * gdbinit.in (pel): Define.
734 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
736         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
737         behavior.
739 2014-02-04  Richard Biener  <rguenther@suse.de>
741         PR lto/59723
742         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
743         in function context local.
744         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
745         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
746         similar to LTO_imported_decl_ref.
748 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
750         PR tree-optimization/60002
751         * cgraphclones.c (build_function_decl_skip_args): Clear
752         DECL_LANG_SPECIFIC.
754         PR tree-optimization/60023
755         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
756         false to gsi_replace.
757         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
758         has been in some EH region and vec_stmt could throw, add
759         vec_stmt into the same EH region.
760         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
761         has no lhs, ignore it.
762         * internal-fn.c (expand_MASK_LOAD): Likewise.
764         PR ipa/60026
765         * tree-inline.c (copy_forbidden): Fail for
766         __attribute__((optimize (0))) functions.
768         PR other/58712
769         * omp-low.c (simd_clone_struct_copy): If from->inbranch
770         is set, copy one less argument.
771         (expand_simd_clones): Don't subtract clone_info->inbranch
772         from simd_clone_struct_alloc argument.
774         PR rtl-optimization/57915
775         * recog.c (simplify_while_replacing): If all unary/binary/relational
776         operation arguments are constant, attempt to simplify those.
778         PR middle-end/59261
779         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
780         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
782 2014-02-04  Richard Biener  <rguenther@suse.de>
784         PR tree-optimization/60012
785         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
786         TBAA disambiguation to all DDRs.
788 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
790         PR target/59788
791         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
792         (LINK_SPEC): Use it for -shared, -shared-libgcc.
794 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
796         PR ipa/59882
797         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
799 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
801         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
802         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
804 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
806         PR ipa/59831
807         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
808         to figure out targets of polymorphic calls with known decl.
809         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
810         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
811         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
812         (get_polymorphic_call_info): ... here.
813         (get_polymorphic_call_info_from_invariant): New function.
815 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
817         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
818         lookup via vtable pointer; check for type consistency
819         and turn inconsitent facts into UNREACHABLE.
820         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
821         * gimple-fold.c (gimple_get_virt_method_for_vtable): 
822         Do not ICE on type inconsistent querries; return UNREACHABLE
823         instead.
825 2014-02-03  Richard Henderson  <rth@twiddle.net>
827         PR tree-opt/59924
828         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
829         already processed this node.
830         (normalize_one_pred_1): Pass along mark_set.
831         (normalize_one_pred): Create and destroy a pointer_set_t.
832         (normalize_one_pred_chain): Likewise.
834 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
836         PR gcov-profile/58602
837         * gcc/gcov-io.c (gcov_open): Open with truncation when mode<0
839 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
841         PR ipa/59831
842         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on -fno-devirtualize;
843         Try to devirtualize by the knowledge of virtual table pointer given by
844         aggregate propagation.
845         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
846         ipa_print_node_jump_functions): Dump also offset that
847         is relevant for polymorphic calls.
848         (determine_known_aggregate_parts): Add arg_type parameter; use it
849         instead of determining the type from pointer type.
850         (ipa_compute_jump_functions_for_edge): Update call of
851         determine_known_aggregate_parts.
852         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
853         (gimple_get_virt_method_for_binfo): ... here; simplify using
854         vtable_pointer_value_to_vtable.
855         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
856         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
857         to unsigned HOST_WIDE_INT; Use vtable_pointer_value_to_vtable.
858         (vtable_pointer_value_to_vtable): Break out from ...; handle also
859         POINTER_PLUS_EXPR.
860         (vtable_pointer_value_to_binfo): ... here.
861         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
863 2014-02-03  Teresa Johnson  <tejohnson@google.com>
865         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
866         redef of outer loop index variable.
868 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
870         PR c++/53017
871         PR c++/59211
872         * doc/extend.texi (Function Attributes): Typo.
874 2014-02-03  Cong Hou  <congh@google.com>
876         PR tree-optimization/60000
877         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
878         if the vectorized statement is a store.  A store statement can only
879         appear at the end of pattern statements.
881 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
883         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
884         (ix86_option_override_internal): Default long double to 64-bit for
885         32-bit Bionic and to 128-bit for 64-bit Bionic.
887         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
888         TARGET_LONG_DOUBLE_128 is true.
889         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
891         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
892         (mlong-double-64): Negate -mlong-double-128.
893         (mlong-double-128): New option.
895         * config/i386/i386-c.c (ix86_target_macros): Define
896         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
898         * doc/invoke.texi: Document -mlong-double-128.
900 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
902         PR rtl-optimization/60024
903         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
905 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
907         * doc/invoke.texi: (fprofile-reorder-functions): Fix typo.
909 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
911         PR rtl-optimization/57662
912         * sel-sched.c (code_motion_path_driver): Do not mark already not
913         existing blocks in the visiting bitmap.
915 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
917         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
918         on the insn being emitted.
920 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
921             Will Deacon  <will.deacon@arm.com>
923         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
925 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
927         * config/arm/arm-tables.opt: Regenerate.
929 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
931         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
932         for vector types other than V16QImode.
933         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
934         define_expand, and call altivec_expand_vec_perm_le when producing
935         code with little endian element order.
936         (*altivec_vperm_<mode>_internal): New insn having previous
937         behavior of altivec_vperm_<mode>.
938         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
939         altivec_expand_vec_perm_le when producing code with little endian
940         element order.
941         (*altivec_vperm_<mode>_uns_internal): New insn having previous
942         behavior of altivec_vperm_<mode>_uns.
944 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
946         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
947         (altivec_vsumsws): Add handling for -maltivec=be with a little
948         endian target.
949         (altivec_vsumsws_direct): New.
950         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
951         gen_altivec_vsumsws.
953 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
955         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
956         vtable_pointer_value_to_binfo): New functions.
957         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
958         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
960 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
962         * config/nios2/nios2.md (load_got_register): Initialize GOT
963         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
964         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
966 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
968         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
969         preserverd by passthrough, do not propagate the type.
971 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
973         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
974         (mips_atomic_assign_expand_fenv): New function.
975         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
977 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
979         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
980         (__builtin_mips_set_fcsr): Likewise.
981         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
982         MIPS_USI_FTYPE_VOID.
983         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
984         (mips16_expand_set_fcsr): Likewise.
985         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
986         (mips16_set_fcsr_stub): Likewise.
987         (mips16_get_fcsr_one_only_stub): New class.
988         (mips16_set_fcsr_one_only_stub): Likewise.
989         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
990         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
991         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
992         (hard_float): New availability predicate.
993         (mips_builtins): Add get_fcsr and set_fcsr.
994         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
995         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
996         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
997         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
998         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
999         patterns.
1001 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1003         * config/mips/mips.c (mips_one_only_stub): New class.
1004         (mips_need_mips16_rdhwr_p): Replace with...
1005         (mips16_rdhwr_stub): ...this new variable.
1006         (mips16_stub_call_address): New function.
1007         (mips16_rdhwr_one_only_stub): New class.
1008         (mips_expand_thread_pointer): Use mips16_stub_call_address.
1009         (mips_output_mips16_rdhwr): Delete.
1010         (mips_finish_stub): New function.
1011         (mips_code_end): Use it to handle rdhwr stubs.
1013 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
1015         PR target/60017
1016         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1017         when calculating size of integer atomic types.
1019 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
1021         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1023 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
1025         PR tree-optimization/60003
1026         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1027         * profile.c (branch_prob): Use gimple_call_builtin_p
1028         to check for BUILT_IN_SETJMP_RECEIVER.
1029         * tree-inline.c (copy_bb): Call notice_special_calls.
1031 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1033         PR bootstrap/59985
1034         * lra-constraints.c (process_alt_operands): Update reload_sum only
1035         on the first pass.
1037 2014-01-31  Richard Henderson  <rth@redhat.com>
1039         PR middle-end/60004
1040         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1041         until after else_eh is processed.
1043 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1045         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1046         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1047         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1048         in smmintrin.h, remove them.
1049         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1050         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1051         * config/i386/i386.md (ROUND_SAE): Fix value.
1052         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1053         (const48_operand): New.
1054         * config/i386/subst.md (round), (round_expand): Use
1055         const_4_or_8_to_11_operand.
1056         (round_saeonly), (round_saeonly_expand): Use const48_operand.
1058 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1060         * config/i386/constraints.md (Yk): Swap meaning with k.
1061         * config/i386/i386.md (movhi_internal): Change Yk to k.
1062         (movqi_internal): Ditto.
1063         (*k<logic><mode>): Ditto.
1064         (*andhi_1): Ditto.
1065         (*andqi_1): Ditto.
1066         (kandn<mode>): Ditto.
1067         (*<code>hi_1): Ditto.
1068         (*<code>qi_1): Ditto.
1069         (kxnor<mode>): Ditto.
1070         (kortestzhi): Ditto.
1071         (kortestchi): Ditto.
1072         (kunpckhi): Ditto.
1073         (*one_cmplhi2_1): Ditto.
1074         (*one_cmplqi2_1): Ditto.
1075         * config/i386/sse.md (): Change k to Yk.
1076         (avx512f_load<mode>_mask): Ditto.
1077         (avx512f_blendm<mode>): Ditto.
1078         (avx512f_store<mode>_mask): Ditto.
1079         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1080         (avx512f_storedqu<mode>_mask): Ditto.
1081         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1082         Ditto.
1083         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1084         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1085         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1086         (avx512f_maskcmp<mode>3): Ditto.
1087         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1088         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1089         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1090         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1091         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1092         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1093         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1094         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1095         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1096         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1097         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1098         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1099         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1100         (vec_extract_lo_<mode>_maskm): Ditto.
1101         (vec_extract_hi_<mode>_maskm): Ditto.
1102         (avx512f_vternlog<mode>_mask): Ditto.
1103         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1104         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1105         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1106         (avx512f_<code>v8div16qi2_mask): Ditto.
1107         (avx512f_<code>v8div16qi2_mask_store): Ditto.
1108         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1109         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1110         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1111         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1112         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1113         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1114         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1115         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1116         (avx512cd_maskb_vec_dupv8di): Ditto.
1117         (avx512cd_maskw_vec_dupv16si): Ditto.
1118         (avx512f_vpermi2var<mode>3_maskz): Ditto.
1119         (avx512f_vpermi2var<mode>3_mask): Ditto.
1120         (avx512f_vpermi2var<mode>3_mask): Ditto.
1121         (avx512f_vpermt2var<mode>3_maskz): Ditto.
1122         (*avx512f_gathersi<mode>): Ditto.
1123         (*avx512f_gathersi<mode>_2): Ditto.
1124         (*avx512f_gatherdi<mode>): Ditto.
1125         (*avx512f_gatherdi<mode>_2): Ditto.
1126         (*avx512f_scattersi<mode>): Ditto.
1127         (*avx512f_scatterdi<mode>): Ditto.
1128         (avx512f_compress<mode>_mask): Ditto.
1129         (avx512f_compressstore<mode>_mask): Ditto.
1130         (avx512f_expand<mode>_mask): Ditto.
1131         * config/i386/subst.md (mask): Change k to Yk.
1132         (mask_scalar_merge): Ditto.
1133         (sd): Ditto.
1135 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
1137         * doc/extend.texi (Vector Extensions): Document ?: in C++.
1139 2014-01-31  Richard Biener  <rguenther@suse.de>
1141         PR middle-end/59990
1142         * builtins.c (fold_builtin_memory_op): Make sure to not
1143         use a floating-point mode or a boolean or enumeral type for
1144         the copy operation.
1146 2014-01-30  DJ Delorie  <dj@redhat.com>
1148         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1149         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1150         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1151         whenever main() has an epilogue.
1153 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1155         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1156         unused variable "field".
1157         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1158         (vsx_mergeh_<mode>): Likewise.
1159         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1160         (altivec_vmrghh): Likewise.
1161         (altivec_vmrghw): Likewise.
1162         (altivec_vmrglb): Likewise.
1163         (altivec_vmrglh): Likewise.
1164         (altivec_vmrglw): Likewise.
1165         (altivec_vspltb): Add missing uses.
1166         (altivec_vsplth): Likewise.
1167         (altivec_vspltw): Likewise.
1168         (altivec_vspltsf): Likewise.
1170 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1172         PR target/59923
1173         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1174         frame related instructions.
1176 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1178         PR rtl-optimization/59959
1179         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1180         any reload of register whose subreg is invalid.
1182 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1184         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1185         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1186         Add missing return type - void.
1188 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1190         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1191         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1192         remove element index adjustment for endian (now handled in vsx.md
1193         and altivec.md).
1194         (altivec_expand_vec_perm_const): Use
1195         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1196         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1197         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1198         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1199         define_expand and a new define_insn *altivec_vspltb_internal;
1200         adjust for -maltivec=be on a little endian target.
1201         (altivec_vspltb_direct): New.
1202         (altivec_vsplth): Divide into a define_expand and a new
1203         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1204         little endian target.
1205         (altivec_vsplth_direct): New.
1206         (altivec_vspltw): Divide into a define_expand and a new
1207         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1208         little endian target.
1209         (altivec_vspltw_direct): New.
1210         (altivec_vspltsf): Divide into a define_expand and a new
1211         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1212         a little endian target.
1214 2014-01-30  Richard Biener  <rguenther@suse.de>
1216         PR tree-optimization/59993
1217         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1218         can propagate form the earlier stmt and avoid the transform
1219         when the intermediate result is needed.
1221 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1223         * README.Portability: Fix typo.
1225 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1227         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1228         comparison_operator with ordered_comparison_operator.
1230 2014-01-30  Nick Clifton  <nickc@redhat.com>
1232         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1233         Rename to mn10300_store_multiple_regs.
1234         * config/mn10300/mn10300.c: Likewise.
1235         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1236         store_multiple_regs.
1237         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1238         Call mn10300_store_multiple_regs.
1240 2014-01-30  Nick Clifton  <nickc@redhat.com>
1241             DJ Delorie  <dj@redhat.com>
1243         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1244         %fp 2 to keep registers after it properly word-aligned.
1245         (rl78_alloc_physical_registers_umul): Handle the case where both
1246         input operands are the same.
1248 2014-01-30  Richard Biener  <rguenther@suse.de>
1250         PR tree-optimization/59903
1251         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1252         check properly.
1254 2014-01-30  Jason Merrill  <jason@redhat.com>
1256         PR c++/59633
1257         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1259         PR c++/59645
1260         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1262 2014-01-30  Richard Biener  <rguenther@suse.de>
1264         PR tree-optimization/59951
1265         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1267 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1269         PR target/59784
1270         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1271         SFmode to DFmode case.
1273 2014-01-29  DJ Delorie  <dj@redhat.com>
1275         * config/msp430/msp430.opt (-minrt): New.
1276         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1277         if -minrt given.
1278         (ENDFILE_SPEC): Likewise.
1280 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1282         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1283         (estimate_function_body_sizes): Use it.
1285 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1287         PR c++/58561
1288         * dwarf2out.c (is_cxx_auto): New.
1289         (is_base_type): Use it.
1290         (gen_type_die_with_usage): Likewise.
1292 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1294         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1295         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1296         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1297         -maltivec=be with LE targets.
1298         (vsx_mergeh_<mode>): Likewise.
1299         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1300         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1301         (altivec_vmrghb): Replace with define_expand and new
1302         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1303         (altivec_vmrghb_direct): New define_insn.
1304         (altivec_vmrghh): Replace with define_expand and new
1305         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1306         (altivec_vmrghh_direct): New define_insn.
1307         (altivec_vmrghw): Replace with define_expand and new
1308         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1309         (altivec_vmrghw_direct): New define_insn.
1310         (*altivec_vmrghsf): Adjust for endianness.
1311         (altivec_vmrglb): Replace with define_expand and new
1312         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1313         (altivec_vmrglb_direct): New define_insn.
1314         (altivec_vmrglh): Replace with define_expand and new
1315         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1316         (altivec_vmrglh_direct): New define_insn.
1317         (altivec_vmrglw): Replace with define_expand and new
1318         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1319         (altivec_vmrglw_direct): New define_insn.
1320         (*altivec_vmrglsf): Adjust for endianness.
1321         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1322         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1323         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1324         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1325         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1326         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1327         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1328         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1330 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1332         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1333         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1334         whitespace.
1336 2014-01-29  Richard Biener  <rguenther@suse.de>
1338         PR tree-optimization/58742
1339         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1340         associate_pointerplus_align.
1341         (associate_pointerplus_diff): New function.
1342         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
1343         and associate_pointerplus_diff.
1345 2014-01-29  Richard Biener  <rguenther@suse.de>
1347         * lto-streamer.h (LTO_major_version): Bump to 3.
1348         (LTO_minor_version): Reset to 0.
1350 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
1352         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1353         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1354         (arm_file_start): Generate correct asm header for armv7ve.
1355         * config/arm/bpabi.h: Add multilib support for armv7ve.
1356         * config/arm/driver-arm.c: Change the architectures of cortex-a7
1357         and cortex-a15 to armv7ve.
1358         * config/arm/t-aprofile: Add multilib support for armv7ve.
1359         * doc/invoke.texi: Document -march=armv7ve.
1361 2014-01-29  Richard Biener  <rguenther@suse.de>
1363         PR tree-optimization/58742
1364         * tree-ssa-forwprop.c (associate_plusminus): Return true
1365         if we changed sth, defer EH cleanup to ...
1366         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
1367         (simplify_mult): New function.
1369 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1371         PR middle-end/59917
1372         PR tree-optimization/59920
1373         * tree.c (build_common_builtin_nodes): Remove
1374         __builtin_setjmp_dispatcher initialization.
1375         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1376         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1377         instead of gsi_after_labels + manually skipping debug stmts.
1378         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1379         ignore bbs with IFN_ABNORMAL_DISPATCHER.
1380         * tree-inline.c (copy_edges_for_bb): Remove
1381         can_make_abnormal_goto argument, instead add abnormal_goto_dest
1382         argument.  Ignore computed_goto_p stmts.  Don't call
1383         make_abnormal_goto_edges.  If a call might need abnormal edges
1384         for non-local gotos, see if it already has an edge to
1385         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1386         with true argument, don't do anything then, otherwise add
1387         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1388         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1389         caller.
1390         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1391         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1392         (lower_stmt): Don't set data->calls_builtin_setjmp.
1393         (lower_builtin_setjmp): Adjust comment.
1394         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1395         * tree-cfg.c (found_computed_goto): Remove.
1396         (factor_computed_gotos): Remove.
1397         (make_goto_expr_edges): Return bool, true for computed gotos.
1398         Don't call make_abnormal_goto_edges.
1399         (build_gimple_cfg): Don't set found_computed_goto, don't call
1400         factor_computed_gotos.
1401         (computed_goto_p): No longer static.
1402         (make_blocks): Don't set found_computed_goto.
1403         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1404         (make_edges): If make_goto_expr_edges returns true, push bb
1405         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1406         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1407         vector.  Record mapping between bbs and OpenMP regions if there
1408         are any, adjust make_gimple_omp_edges caller.  Call
1409         handle_abnormal_edges.
1410         (make_abnormal_goto_edges): Remove.
1411         * tree-cfg.h (make_abnormal_goto_edges): Remove.
1412         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1413         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1414         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1415         * internal-fn.def (ABNORMAL_DISPATCHER): New.
1416         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1417         filling *region also set *region_idx to (*region)->entry->index.
1419         PR other/58712
1420         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1421         For REGs set ORIGINAL_REGNO.
1423 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
1425         * doc/md.texi: Mention that a target shouldn't implement
1426         vec_widen_(s|u)mul_even/odd pair if it is less efficient
1427         than hi/lo pair.
1429 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1431         PR tree-optimization/59594
1432         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1433         a copy of the datarefs vector rather than the vector itself.
1435 2014-01-28  Jason Merrill  <jason@redhat.com>
1437         PR c++/53756
1438         * dwarf2out.c (auto_die): New static.
1439         (gen_type_die_with_usage): Handle C++1y 'auto'.
1440         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1441         on definition.
1443 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1445         PR target/59672
1446         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1447         (SPEC_X32): Likewise.
1448         (SPEC_64): Likewise.
1449         * config/i386/i386.c (ix86_option_override_internal): Turn off
1450         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1451         for TARGET_16BIT.
1452         (x86_file_start): Output .code16gcc for TARGET_16BIT.
1453         * config/i386/i386.h (TARGET_16BIT): New macro.
1454         (TARGET_16BIT_P): Likewise.
1455         * config/i386/i386.opt: Add m16.
1456         * doc/invoke.texi: Document -m16.
1458 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
1460         PR preprocessor/59935
1461         * input.c (location_get_source_line): Bail out on when line number
1462         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1464 2014-01-28  Richard Biener  <rguenther@suse.de>
1466         PR tree-optimization/58742
1467         * tree-ssa-forwprop.c (associate_plusminus): Handle
1468         pointer subtraction of the form (T)(P + A) - (T)P.
1470 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1472         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1473         at const_int_cost.
1475 2014-01-28  Richard Biener  <rguenther@suse.de>
1477         Revert
1478         2014-01-28  Richard Biener  <rguenther@suse.de>
1480         PR rtl-optimization/45364
1481         PR rtl-optimization/59890
1482         * var-tracking.c (local_get_addr_clear_given_value): Handle
1483         already cleared slot.
1484         (val_reset): Handle not allocated local_get_addr_cache.
1485         (vt_find_locations): Use post-order on the inverted CFG.
1487 2014-01-28  Richard Biener  <rguenther@suse.de>
1489         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1491 2014-01-28  Richard Biener  <rguenther@suse.de>
1493         PR rtl-optimization/45364
1494         PR rtl-optimization/59890
1495         * var-tracking.c (local_get_addr_clear_given_value): Handle
1496         already cleared slot.
1497         (val_reset): Handle not allocated local_get_addr_cache.
1498         (vt_find_locations): Use post-order on the inverted CFG.
1500 2014-01-28  Alan Modra  <amodra@gmail.com>
1502         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1503         * configure.ac <recursive call for build != host>: Define
1504         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1505         and LD_FOR_BUILD too.
1506         * configure: Regenerate.
1508 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
1510         * config/i386/i386.c (get_builtin_code_for_version): Separate
1511         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1512         Broadwell from Haswell.
1514 2014-01-27  Steve Ellcey  <sellcey@mips.com>
1516         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1517         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1518         * config/mips/mips.c (mips_option_override): Change setting
1519         of TARGET_DSP.
1520         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1521         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1522         Change from Mask to Var.
1524 2014-01-27  Jeff Law  <law@redhat.com>
1526         * ipa-inline.c (inline_small_functions): Fix typo.
1528 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
1530         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1531         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1532         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1533         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1534         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1535         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1536         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1537         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1538         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1539         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1540         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1541         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1542         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1543         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1544         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1545         (_mm512_storeu_epi64): Ditto.
1546         (_mm512_cmpge_epi32_mask): Ditto.
1547         (_mm512_cmpge_epu32_mask): Ditto.
1548         (_mm512_cmpge_epi64_mask): Ditto.
1549         (_mm512_cmpge_epu64_mask): Ditto.
1550         (_mm512_cmple_epi32_mask): Ditto.
1551         (_mm512_cmple_epu32_mask): Ditto.
1552         (_mm512_cmple_epi64_mask): Ditto.
1553         (_mm512_cmple_epu64_mask): Ditto.
1554         (_mm512_cmplt_epi32_mask): Ditto.
1555         (_mm512_cmplt_epu32_mask): Ditto.
1556         (_mm512_cmplt_epi64_mask): Ditto.
1557         (_mm512_cmplt_epu64_mask): Ditto.
1558         (_mm512_cmpneq_epi32_mask): Ditto.
1559         (_mm512_cmpneq_epu32_mask): Ditto.
1560         (_mm512_cmpneq_epi64_mask): Ditto.
1561         (_mm512_cmpneq_epu64_mask): Ditto.
1562         (_mm512_expand_pd): Ditto.
1563         (_mm512_expand_ps): Ditto.
1564         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1565         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1566         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1567         * config/i386/i386.c (ix86_builtins): Add
1568         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1569         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1570         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1571         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1572         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1573         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1574         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1575         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1576         IX86_BUILTIN_PMOVUSQW512_MEM.
1577         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1578         __builtin_ia32_pmovsqd512mem_mask,
1579         __builtin_ia32_pmovqd512mem_mask,
1580         __builtin_ia32_pmovusqw512mem_mask,
1581         __builtin_ia32_pmovsqw512mem_mask,
1582         __builtin_ia32_pmovqw512mem_mask,
1583         __builtin_ia32_pmovusdw512mem_mask,
1584         __builtin_ia32_pmovsdw512mem_mask,
1585         __builtin_ia32_pmovdw512mem_mask,
1586         __builtin_ia32_pmovqb512mem_mask,
1587         __builtin_ia32_pmovusqb512mem_mask,
1588         __builtin_ia32_pmovsqb512mem_mask,
1589         __builtin_ia32_pmovusdb512mem_mask,
1590         __builtin_ia32_pmovsdb512mem_mask,
1591         __builtin_ia32_pmovdb512mem_mask.
1592         (bdesc_args): Add __builtin_ia32_expanddf512,
1593         __builtin_ia32_expandsf512.
1594         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1595         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1596         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1597         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1598         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1599         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1600         (avx512f_<code>v8div16qi2_mask_store): This.
1601         (avx512f_expand<mode>): New.
1603 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
1605         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
1606         New.
1607         (_mm512_mask_prefetch_i64gather_pd): Ditto.
1608         (_mm512_prefetch_i32scatter_pd): Ditto.
1609         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
1610         (_mm512_prefetch_i64scatter_pd): Ditto.
1611         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
1612         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
1613         (_mm512_mask_prefetch_i64gather_ps): Ditto.
1614         (_mm512_prefetch_i32scatter_ps): Ditto.
1615         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
1616         (_mm512_prefetch_i64scatter_ps): Ditto.
1617         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
1618         * config/i386/i386-builtin-types.def: Define
1619         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
1620         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
1621         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
1622         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
1623         IX86_BUILTIN_SCATTERPFQPD.
1624         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
1625         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
1626         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
1627         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
1628         __builtin_ia32_scatterpfqps.
1629         (ix86_expand_builtin): Expand new built-ins.
1630         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
1631         fix memory access data type.
1632         (*avx512pf_gatherpf<mode>_mask): Ditto.
1633         (*avx512pf_gatherpf<mode>): Ditto.
1634         (avx512pf_scatterpf<mode>): Ditto.
1635         (*avx512pf_scatterpf<mode>_mask): Ditto.
1636         (*avx512pf_scatterpf<mode>): Ditto.
1637         (GATHER_SCATTER_SF_MEM_MODE): New.
1638         (avx512pf_gatherpf<mode>df): Ditto.
1639         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1640         (*avx512pf_scatterpf<mode>df): Ditto.
1642 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
1644         PR bootstrap/59934
1645         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
1646         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
1647         reached.
1649 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1651         * common/config/arm/arm-common.c
1652         (arm_rewrite_mcpu): Handle multiple names.
1653         * config/arm/arm.h
1654         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1656 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
1658         * gimple-builder.h (create_gimple_tmp): Delete.
1660 2014-01-27  Christian Bruel  <christian.bruel@st.com>
1662         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
1663         words comparisons.
1665 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
1667         * config/pa/pa.md (call): Generate indirect long calls to non-local
1668         functions when outputing 32-bit code.
1669         (call_value): Likewise except for special call to buggy powf function.
1671         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
1672         portable runtime and PIC indirect calls.
1673         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
1674         and PIC call sequences.  Use ldo instead of blr to set return register
1675         in PIC call sequence.
1677 2014-01-25  Walter Lee  <walt@tilera.com>
1679         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
1680         avoid clobbering a live register.
1682 2014-01-25  Walter Lee  <walt@tilera.com>
1684         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
1685         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
1686         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
1687         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
1689 2014-01-25  Walter Lee  <walt@tilera.com>
1691         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
1692         arguments on even registers.
1693         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
1694         STACK_BOUNDARY.
1695         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
1696         (BIGGEST_ALIGNMENT): Ditto.
1697         (BIGGEST_FIELD_ALIGNMENT): Ditto.
1699 2014-01-25  Walter Lee  <walt@tilera.com>
1701         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
1702         insns before bundling.
1703         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
1705 2014-01-25  Walter Lee  <walt@tilera.com>
1707         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
1708         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
1709         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
1711 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1713         * config/mips/constraints.md (kl): Delete.
1714         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
1715         define expands, using...
1716         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
1717         instructions for MIPS16.
1718         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
1719         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
1721 2014-01-25  Walter Lee  <walt@tilera.com>
1723         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
1724         (clzdi2): Ditto.
1725         (ffsdi2): Ditto.
1727 2014-01-25  Walter Lee  <walt@tilera.com>
1729         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
1730         (TARGET_EXPAND_TO_RTL_HOOK): Define.
1732 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
1734         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
1735         Handle XOR.
1737 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
1739         * print-rtl.c (in_call_function_usage): New var.
1740         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
1741         EXPR_LIST mode as mode and not as reg note name.
1743         PR middle-end/59561
1744         * cfgloopmanip.c (copy_loop_info): If
1745         loop->warned_aggressive_loop_optimizations, make sure
1746         the flag is set in target loop too.
1748 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
1750         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
1751         flag_cilkplus.
1752         * builtins.def: Likewise.
1753         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
1754         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
1755         * ira.c (ira_setup_eliminable_regset): Likewise.
1756         * omp-low.c (gate_expand_omp): Likewise.
1757         (execute_lower_omp): Likewise.
1758         (diagnose_sb_0): Likewise.
1759         (gate_diagnose_omp_blocks): Likewise.
1760         (simd_clone_clauses_extract): Likewise.
1761         (gate): Likewise.
1763 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1765         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
1766         correction for little endian...
1767         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
1768         here.
1770 2014-01-24  Jeff Law  <law@redhat.com>
1772         PR tree-optimization/59919
1773         * tree-vrp.c (find_assert_locations_1): Do not register asserts
1774         for non-returning calls.
1776 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
1778         * common/config/aarch64/aarch64-common.c
1779         (aarch64_rewrite_mcpu): Handle multiple names.
1780         * config/aarch64/aarch64.h
1781         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1783 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
1785         * input.c (add_file_to_cache_tab): Handle the case where fopen
1786         returns NULL.
1788 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
1790         PR target/59929
1791         * config/i386/i386.md (pushsf splitter): Get stack adjustment
1792         from push operand if code of push isn't PRE_DEC.
1794 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
1796         PR target/59909
1797         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
1798         -mquad-memory-atomic.  Update -mquad-memory documentation to say
1799         it is only used for non-atomic loads/stores.
1801         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
1802         -mquad-memory or -mquad-memory-atomic switches.
1804         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
1805         -mquad-memory-atomic to ISA 2.07 support.
1807         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
1808         to separate support of normal quad word memory operations (ldq, stq)
1809         from the atomic quad word memory operations.
1811         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
1812         support to separate non-atomic quad word operations from atomic
1813         quad word operations.  Disable non-atomic quad word operations in
1814         little endian mode so that we don't have to swap words after the
1815         load and before the store.
1816         (quad_load_store_p): Add comment about atomic quad word support.
1817         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
1818         options printed with -mdebug=reg.
1820         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
1821         -mquad-memory-atomic as the test for whether we have quad word
1822         atomic instructions.
1823         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
1824         or -mp8-vector are used, allow byte/half-word atomic operations.
1826         * config/rs6000/sync.md (load_lockedti): Insure that the address
1827         is a proper indexed or indirect address for the lqarx instruction.
1828         On little endian systems, swap the hi/lo registers after the lqarx
1829         instruction.
1830         (load_lockedpti): Use indexed_or_indirect_operand predicate to
1831         insure the address is valid for the lqarx instruction.
1832         (store_conditionalti): Insure that the address is a proper indexed
1833         or indirect address for the stqcrx. instruction.  On little endian
1834         systems, swap the hi/lo registers before doing the stqcrx.
1835         instruction.
1836         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
1837         insure the address is valid for the stqcrx. instruction.
1839         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
1840         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
1841         type of quad memory support is available.
1843 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
1845         PR regression/59915
1846         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
1847         there is a danger of looping.
1849 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
1851         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1852         force flag_ira_loop_pressure if set via command line.
1854 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1856         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
1857         (ashr_simd): New builtin handling DI mode.
1858         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
1859         (aarch64_sshr_simddi): New match pattern.
1860         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
1861         (vshrd_n_s64): Likewise.
1862         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
1864 2014-01-23  Nick Clifton  <nickc@redhat.com>
1866         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
1867         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
1868         favour of mcu specific scripts.
1869         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
1870         430x multilibs.
1872 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1873             Alex Velenko  <Alex.Velenko@arm.com>
1875         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
1876         (vaddv_s16): Likewise.
1877         (vaddv_s32): Likewise.
1878         (vaddv_u8): Likewise.
1879         (vaddv_u16): Likewise.
1880         (vaddv_u32): Likewise.
1881         (vaddvq_s8): Likewise.
1882         (vaddvq_s16): Likewise.
1883         (vaddvq_s32): Likewise.
1884         (vaddvq_s64): Likewise.
1885         (vaddvq_u8): Likewise.
1886         (vaddvq_u16): Likewise.
1887         (vaddvq_u32): Likewise.
1888         (vaddvq_u64): Likewise.
1889         (vaddv_f32): Likewise.
1890         (vaddvq_f32): Likewise.
1891         (vaddvq_f64): Likewise.
1892         (vmaxv_f32): Likewise.
1893         (vmaxv_s8): Likewise.
1894         (vmaxv_s16): Likewise.
1895         (vmaxv_s32): Likewise.
1896         (vmaxv_u8): Likewise.
1897         (vmaxv_u16): Likewise.
1898         (vmaxv_u32): Likewise.
1899         (vmaxvq_f32): Likewise.
1900         (vmaxvq_f64): Likewise.
1901         (vmaxvq_s8): Likewise.
1902         (vmaxvq_s16): Likewise.
1903         (vmaxvq_s32): Likewise.
1904         (vmaxvq_u8): Likewise.
1905         (vmaxvq_u16): Likewise.
1906         (vmaxvq_u32): Likewise.
1907         (vmaxnmv_f32): Likewise.
1908         (vmaxnmvq_f32): Likewise.
1909         (vmaxnmvq_f64): Likewise.
1910         (vminv_f32): Likewise.
1911         (vminv_s8): Likewise.
1912         (vminv_s16): Likewise.
1913         (vminv_s32): Likewise.
1914         (vminv_u8): Likewise.
1915         (vminv_u16): Likewise.
1916         (vminv_u32): Likewise.
1917         (vminvq_f32): Likewise.
1918         (vminvq_f64): Likewise.
1919         (vminvq_s8): Likewise.
1920         (vminvq_s16): Likewise.
1921         (vminvq_s32): Likewise.
1922         (vminvq_u8): Likewise.
1923         (vminvq_u16): Likewise.
1924         (vminvq_u32): Likewise.
1925         (vminnmv_f32): Likewise.
1926         (vminnmvq_f32): Likewise.
1927         (vminnmvq_f64): Likewise.
1929 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
1931         * config/aarch64/aarch64-simd.md
1932         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
1933         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
1934         (*aarch64_mul3_elt<mode>): Likewise.
1935         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
1936         (*aarch64_mul3_elt_to_64v2df): Likewise.
1937         (*aarch64_mla_elt<mode>): Likewise.
1938         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
1939         (*aarch64_mls_elt<mode>): Likewise.
1940         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
1941         (*aarch64_fma4_elt<mode>): Likewise.
1942         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
1943         (*aarch64_fma4_elt_to_64v2df): Likewise.
1944         (*aarch64_fnma4_elt<mode>): Likewise.
1945         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
1946         (*aarch64_fnma4_elt_to_64v2df): Likewise.
1947         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
1948         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
1949         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
1950         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
1951         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
1952         (aarch64_sqdmull_lane<mode>_internal): Likewise.
1953         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
1955 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1957         * config/aarch64/aarch64-simd.md
1958         (aarch64_be_checked_get_lane<mode>): New define_expand.
1959         * config/aarch64/aarch64-simd-builtins.def
1960         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
1961         New builtin definition.
1962         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
1963         Use new safe be builtin.
1965 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
1967         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
1968         New define_insn.
1969         (aarch64_be_st1<mode>): Likewise.
1970         (aarch_ld1<VALL:mode>): Define_expand modified.
1971         (aarch_st1<VALL:mode>): Likewise.
1972         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
1973         (UNSPEC_ST1): Likewise.
1975 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
1977         * config/microblaze/microblaze.md: Add trap insn and attribute
1979 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
1981         PR preprocessor/58580
1982         * input.h (location_get_source_line): Take an additional line_size
1983         parameter.
1984         (void diagnostics_file_cache_fini): Declare new function.
1985         * input.c (struct fcache): New type.
1986         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
1987         New static constants.
1988         (diagnostic_file_cache_init, total_lines_num)
1989         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
1990         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
1991         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
1992         (get_next_line, read_next_line, goto_next_line, read_line_num):
1993         New static function definitions.
1994         (diagnostic_file_cache_fini): New function.
1995         (location_get_source_line): Take an additional output line_len
1996         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
1997         read_line_num.
1998         * diagnostic.c (diagnostic_finish): Call
1999         diagnostic_file_cache_fini.
2000         (adjust_line): Take an additional input parameter for the length
2001         of the line, rather than calculating it with strlen.
2002         (diagnostic_show_locus): Adjust the use of
2003         location_get_source_line and adjust_line with respect to their new
2004         signature.  While displaying a line now, do not stop at the first
2005         null byte.  Rather, display the zero byte as a space and keep
2006         going until we reach the size of the line.
2007         * Makefile.in: Add vec.o to OBJS-libcommon
2009 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2010             Ilya Tocar     <ilya.tocar@intel.com>
2012         * config/i386/avx512fintrin.h (_mm512_kmov): New.
2013         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2014         (__builtin_ia32_kmov16): Ditto.
2015         * config/i386/i386.md (UNSPEC_KMOV): New.
2016         (kmovw): Ditto.
2018 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2020         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2021         (_mm512_storeu_si512): Ditto.
2023 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
2025         PR target/52125
2026         * rtl.h (get_referenced_operands): Declare.
2027         * recog.c (get_referenced_operands): New function.
2028         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2029         operands have been referenced when recording LO_SUM references.
2031 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
2033         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2035 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2037         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2038         Enable for generic and recent AMD targets.
2040 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2042         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2043         ARG_SIZE note when adjustment was eliminated.
2045 2014-01-22  Jeff Law  <law@redhat.com>
2047         PR tree-optimization/59597
2048         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2049         in file.  Accept new argument REGISTERING and use it to modify
2050         dump output appropriately.
2051         (register_jump_thread): Corresponding changes.
2052         (mark_threaded_blocks): Reinstate code to cancel unprofitable
2053         thread paths involving joiner blocks.  Add code to dump cancelled
2054         jump threading paths.
2056 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
2058         PR rtl-optimization/59477
2059         * lra-constraints.c (inherit_in_ebb): Process call for living hard
2060         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
2062 2014-01-22  Tom Tromey  <tromey@redhat.com>
2064         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2065         PARAMS.
2066         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2068 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2070         PR rtl-optimization/59896
2071         * lra-constraints.c (process_alt_operands): Check unused note for
2072         matched operands of insn with no output reloads.
2074 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
2076         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2077         (mips_move_from_gpr_cost): Likewise.
2079 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2081         PR rtl-optimization/59858
2082         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2083         ira_class_hard_regs_num.
2084         (process_alt_operands): Increase reject for dying matched operand.
2086 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
2088         PR target/59003
2089         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2090         smaller than size, perform several stores or loads and stores
2091         at dst + count - size to store or copy all of size bytes, rather
2092         than just last modesize bytes.
2094 2014-01-20  DJ Delorie  <dj@redhat.com>
2096         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2097         that CLOBBERs are REGs before propogating their values.
2099 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2101         PR middle-end/59789
2102         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2103         (cgraph_inline_failed_type): New function.
2104         * cgraph.h (DEFCIFCODE): Add type.
2105         (cgraph_inline_failed_type_t): New enum.
2106         (cgraph_inline_failed_type): New prototype.
2107         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2108         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2109         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2110         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2111         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2112         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2113         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2114         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2115         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2116         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2117         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2118         OPTIMIZATION_MISMATCH.
2119         * tree-inline.c (expand_call_inline): Emit errors during
2120         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2122 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2124         PR target/59685
2125         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2126         mode attribute in insn output.
2128 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
2130         * output.h (output_constant): Delete.
2131         * varasm.c (output_constant): Make private.
2133 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
2135         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2137 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2139         PR middle-end/59860
2140         * tree.h (fold_builtin_strcat): New prototype.
2141         * builtins.c (fold_builtin_strcat): No longer static.  Add len
2142         argument, if non-NULL, don't call c_strlen.  Optimize
2143         directly into __builtin_memcpy instead of __builtin_strcpy.
2144         (fold_builtin_2): Adjust fold_builtin_strcat caller.
2145         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2147 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2149         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2150         for SImode_address_operand operands, having only a REG argument.
2152 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2154         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2155         loader name using mbig-endian.
2156         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2158 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2160         * doc/invoke.texi (-march): Clarify documentation for AArch64.
2161         (-mtune): Likewise.
2162         (-mcpu): Likewise.
2164 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
2166         * config/aarch64/aarch64-protos.h
2167         (aarch64_cannot_change_mode_class_ptr): Declare.
2168         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2169         aarch64_cannot_change_mode_class_ptr): New.
2170         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2171         backend hook aarch64_cannot_change_mode_class.
2173 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2175         * common/config/aarch64/aarch64-common.c
2176         (aarch64_handle_option): Don't handle any option order logic here.
2177         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2178         selected_cpu, warn on architecture version mismatch.
2179         (aarch64_override_options): Fix parsing order for option strings.
2181 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2182             Iain Sandoe  <iain@codesourcery.com>
2184         PR bootstrap/59496
2185         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2186         warning.  Amend comment to reflect current functionality.
2188 2014-01-20  Richard Biener  <rguenther@suse.de>
2190         PR middle-end/59860
2191         * builtins.c (fold_builtin_strcat): Remove case better handled
2192         by tree-ssa-strlen.c.
2194 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2196         * config/aarch64/aarch64.opt
2197         (mcpu, march, mtune): Make case-insensitive.
2199 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2201         PR target/59880
2202         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2203         if operands[1] is a REG or ZERO_EXTEND of a REG.
2205 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2207         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2209 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2211         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2212         long non-pic millicode calls.
2214 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2216         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2218 2014-01-19  Kito Cheng  <kito@0xlab.org>
2220         * builtins.c (expand_movstr): Check movstr expand done or fail.
2222 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2223             H.J. Lu  <hongjiu.lu@intel.com>
2225         PR target/59379
2226         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2227         to DImode for zero-extended addresses.
2229 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2231         PR rtl-optimization/57763
2232         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2233         on the new indirect jump_insn and increment LABEL_NUSES (label).
2235 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2237         PR bootstrap/59580
2238         PR bootstrap/59583
2239         * config.gcc (x86_archs): New variable.
2240         (x86_64_archs): Likewise.
2241         (x86_cpus): Likewise.
2242         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2243         --with-arch/--with-cpu= options.
2244         Support --with-arch=/--with-cpu={nehalem,westmere,
2245         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2247 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2249         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2250         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2252 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2254         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2256 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2258         PR target/58944
2259         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2260         clear cpp_get_options (parse_in)->warn_unused_macros for
2261         ix86_target_macros_internal with cpp_define.
2263 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2265         * jump.c (delete_related_insns): Keep (use (insn))s.
2266         * reorg.c (redundant_insn): Check for barriers too.
2268 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2270         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2272 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2274         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2275         call to $$dyncall when TARGET_LONG_CALLS is true.
2277 2014-01-17  Jeff Law  <law@redhat.com>
2279         * ree.c (combine_set_extension): Temporarily disable test for
2280         changing number of hard registers.
2282 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2284         PR middle-end/58125
2285         * ipa-inline-analysis.c (inline_free_summary):
2286         Do not free summary of aliases.
2288 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2290         PR middle-end/59706
2291         * gimplify.c (gimplify_expr): Use create_tmp_var
2292         instead of create_tmp_var_raw.  If cond doesn't have
2293         integral type, don't add the IFN_ANNOTATE builtin at all.
2295 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2297         PR ipa/59736
2298         * ipa-cp.c (prev_edge_clone): New variable.
2299         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2300         Also resize prev_edge_clone vector.
2301         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2302         (ipcp_edge_removal_hook): New function.
2303         (ipcp_driver): Register ipcp_edge_removal_hook.
2305 2014-01-17  Andrew Pinski <apinski@cavium.com>
2306             Steve Ellcey  <sellcey@mips.com>
2308         PR target/59462
2309         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2310         of operator mode.
2312 2014-01-17  Jeff Law  <law@redhat.com>
2314         PR middle-end/57904
2315         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2316         so that pass_ccp runs first.
2318 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2320         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2321         (ix86_adjust_cost): Use !TARGET_XXX.
2322         (do_reorder_for_imul): Likewise.
2323         (swap_top_of_ready_list): Likewise.
2324         (ix86_sched_reorder): Likewise.
2326 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2328         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2329         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2330         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2331         (intel_memset): New.  Duplicate slm_memset.
2332         (intel_cost): New.  Duplicate slm_cost.
2333         (m_INTEL): New macro.
2334         (processor_target_table): Add "intel".
2335         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2336         with PROCESSOR_INTEL for "intel".
2337         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2338         PROCESSOR_SILVERMONT.
2339         (ix86_issue_rate): Likewise.
2340         (ix86_adjust_cost): Likewise.
2341         (ia32_multipass_dfa_lookahead): Likewise.
2342         (swap_top_of_ready_list): Likewise.
2343         (ix86_sched_reorder): Likewise.
2344         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2345         instead of TARGET_OPT_AGU.
2346         * config/i386/i386.h (TARGET_INTEL): New.
2347         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2348         (processor_type): Add PROCESSOR_INTEL.
2349         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2350         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2352 2014-01-17  Marek Polacek  <polacek@redhat.com>
2354         PR c/58346
2355         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2356         size is zero.
2358 2014-01-17  Richard Biener  <rguenther@suse.de>
2360         PR tree-optimization/46590
2361         * opts.c (default_options_table): Add entries for
2362         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2363         all enabled at -O1 but not for -Og.
2364         * common.opt (fbranch-count-reg): Remove Init(1).
2365         (fmove-loop-invariants): Likewise.
2366         (ftree-pta): Likewise.
2368 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2370         * config/i386/i386.c (ix86_data_alignment): For compatibility with
2371         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2372         decls to at least the GCC 4.8 used alignments.
2374         PR fortran/59440
2375         * tree-nested.c (convert_nonlocal_reference_stmt,
2376         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2377         of GIMPLE_BIND stmts, adjust associated decls.
2379 2014-01-17  Richard Biener  <rguenther@suse.de>
2381         PR tree-optimization/46590
2382         * vec.h (vec<>::bseach): New member function implementing
2383         binary search according to C89 bsearch.
2384         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2385         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2386         bitmap pointer again.  Make accesses_in_loop a flat array.
2387         (mem_ref_obstack): New global.
2388         (outermost_indep_loop): Adjust for mem_ref->stored changes.
2389         (mark_ref_stored): Likewise.
2390         (ref_indep_loop_p_2): Likewise.
2391         (set_ref_stored_in_loop): New helper function.
2392         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2393         (memref_free): Adjust.
2394         (record_mem_ref_loc): Simplify.
2395         (gather_mem_refs_stmt): Adjust.
2396         (sort_locs_in_loop_postorder_cmp): New function.
2397         (analyze_memory_references): Sort accesses_in_loop after
2398         loop postorder number.
2399         (find_ref_loc_in_loop_cmp): New function.
2400         (for_all_locs_in_loop): Find relevant cluster of locs in
2401         accesses_in_loop and iterate without recursion.
2402         (execute_sm): Avoid uninit warning.
2403         (struct ref_always_accessed): Simplify.
2404         (ref_always_accessed::operator ()): Likewise.
2405         (ref_always_accessed_p): Likewise.
2406         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2407         loop postorder numbers here.
2408         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2409         numbers.
2411 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2413         PR c++/57945
2414         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2415         on decls for which assemble_alias has been called.
2417 2014-01-17  Nick Clifton  <nickc@redhat.com>
2419         * config/msp430/msp430.opt: (mcpu): New option.
2420         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2421         (msp430_option_override): Parse target_cpu.  If the MCU name
2422         matches a generic string, clear target_mcu.
2423         (msp430_attr): Allow numeric interrupt values up to 63.
2424         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2425         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2426         option.
2427         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2428         Add mcpu matches.
2429         * config/msp430/msp430.md (popm): Use %J rather than %I.
2430         (addsi3): Use msp430_nonimmediate_operand for operand 2.
2431         (addhi_cy_i): Use immediate_operand for operand 2.
2432         * doc/invoke.texi: Document -mcpu option.
2434 2014-01-17  Richard Biener  <rguenther@suse.de>
2436         PR rtl-optimization/38518
2437         * df.h (df_analyze_loop): Declare.
2438         * df-core.c: Include cfgloop.h.
2439         (df_analyze_1): Split out main part of df_analyze.
2440         (df_analyze): Adjust.
2441         (loop_inverted_post_order_compute): New function.
2442         (loop_post_order_compute): Likewise.
2443         (df_analyze_loop): New function avoiding whole-function
2444         postorder computes.
2445         * loop-invariant.c (find_defs): Use df_analyze_loop.
2446         (find_invariants): Adjust.
2447         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2449 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
2451         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2452         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2454 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
2456         * ipa-ref.c (ipa_remove_stmt_references): Fix references
2457         traversal when removing references.
2459 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
2461         PR ipa/59775
2462         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2464 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
2466         PR middle-end/56791
2467         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2468         pushing a reload for an autoinc when we had previously reloaded an
2469         inner part of the address.
2471 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
2473         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2474         field.
2475         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2476         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2477         when not giving up or versioning for alias only because of
2478         loop->safelen.
2479         (vect_analyze_data_ref_dependences): Set to true.
2480         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2481         is a GIMPLE_PHI.
2482         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2483         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2484         to the condition.
2486         PR middle-end/58344
2487         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2489         PR target/59839
2490         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2491         operand 0 predicate for gathers, use a new pseudo as subtarget.
2493 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2495         PR middle-end/59609
2496         * lra-constraints.c (process_alt_operands): Add printing debug info.
2497         Check absence of input/output reloads for matched operands too.
2499 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2501         PR rtl-optimization/59835
2502         * ira.c (ira_init_register_move_cost): Increase cost for
2503         impossible modes.
2505 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
2507         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2509 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
2511         PR target/59780
2512         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2513         non-register objects.  Use gen_(high/low)part more consistently.
2514         Fix assertions.
2516 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
2518         PR target/59844
2519         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2520         endian support, remove tests for WORDS_BIG_ENDIAN.
2521         (p8_mfvsrd_3_<mode>): Likewise.
2522         (reload_gpr_from_vsx<mode>): Likewise.
2523         (reload_gpr_from_vsxsf): Likewise.
2524         (p8_mfvsrd_4_disf): Likewise.
2526 2014-01-16  Richard Biener  <rguenther@suse.de>
2528         PR rtl-optimization/46590
2529         * lcm.c (compute_antinout_edge): Use postorder iteration.
2530         (compute_laterin): Use inverted postorder iteration.
2532 2014-01-16  Nick Clifton  <nickc@redhat.com>
2534         PR middle-end/28865
2535         * varasm.c (output_constant): Return the number of bytes actually
2536         emitted.
2537         (output_constructor_array_range): Update the field size with the
2538         number of bytes emitted by output_constant.
2539         (output_constructor_regular_field): Likewise.  Also do not
2540         complain if the total number of bytes emitted is now greater
2541         than the expected fieldpos.
2542         * output.h (output_constant): Update prototype and descriptive comment.
2544 2014-01-16  Marek Polacek  <polacek@redhat.com>
2546         PR middle-end/59827
2547         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2548         it is error_mark_node.
2550 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
2552         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2553         VALID_AVX256_REG_OR_OI_MODE.
2555 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
2557         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2558         current procedure should be profiled.
2560 2014-01-15  Andrew Pinski  <apinski@cavium.com>
2562         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2563         of moving from/to the STACK_REG register class.
2565 2014-01-15  Richard Henderson <rth@redhat.com>
2567         PR debug/54694
2568         * reginfo.c (global_regs_decl): Globalize.
2569         * rtl.h (global_regs_decl): Declare.
2570         * ira.c (do_reload): Diagnose frame_pointer_needed and it
2571         reserved via global_regs.
2573 2014-01-15  Teresa Johnson  <tejohnson@google.com>
2575         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2577 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
2579         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2580         and vmulosh rather than call gen_vec_widen_smult_*.
2581         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2582         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2583         (vec_widen_smult_even_v16qi): Likewise.
2584         (vec_widen_umult_even_v8hi): Likewise.
2585         (vec_widen_smult_even_v8hi): Likewise.
2586         (vec_widen_umult_odd_v16qi): Likewise.
2587         (vec_widen_smult_odd_v16qi): Likewise.
2588         (vec_widen_umult_odd_v8hi): Likewise.
2589         (vec_widen_smult_odd_v8hi): Likewise.
2590         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2591         vmuloub rather than call gen_vec_widen_umult_*.
2592         (vec_widen_umult_lo_v16qi): Likewise.
2593         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2594         vmulosb rather than call gen_vec_widen_smult_*.
2595         (vec_widen_smult_lo_v16qi): Likewise.
2596         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2597         rather than call gen_vec_widen_umult_*.
2598         (vec_widen_umult_lo_v8hi): Likewise.
2599         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2600         rather than call gen_vec_widen_smult_*.
2601         (vec_widen_smult_lo_v8hi): Likewise.
2603 2014-01-15  Jeff Law  <law@redhat.com>
2605         PR tree-optimization/59747
2606         * ree.c (find_and_remove_re): Properly handle case where a second
2607         eliminated extension requires widening a copy created for elimination
2608         of a prior extension.
2609         (combine_set_extension): Ensure that the number of hard regs needed
2610         for a destination register does not change when we widen it.
2612 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2614         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
2615         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
2616         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
2617         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
2618         (avr-*-rtems*): Likewise.
2619         (bfin*-rtems*): Likewise.
2620         (moxie-*-rtems*): Likewise.
2621         (h8300-*-rtems*): Likewise.
2622         (i[34567]86-*-rtems*): Likewise.
2623         (lm32-*-rtems*): Likewise.
2624         (m32r-*-rtems*): Likewise.
2625         (m68k-*-rtems*): Likewise.
2626         (microblaze*-*-rtems*): Likewise.
2627         (mips*-*-rtems*): Likewise.
2628         (powerpc-*-rtems*): Likewise.
2629         (sh-*-rtems*): Likewise.
2630         (sparc-*-rtems*): Likewise.
2631         (sparc64-*-rtems*): Likewise.
2632         (v850-*-rtems*): Likewise.
2633         (m32c-*-rtems*): Likewise.
2635 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
2637         PR rtl-optimization/59511
2638         * ira.c (ira_init_register_move_cost): Use memory costs for some
2639         cases of register move cost calculations.
2640         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
2641         instead of BB frequency.
2642         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
2643         * lra-assigns.c (find_hard_regno_for): Ditto.
2645 2014-01-15  Richard Biener  <rguenther@suse.de>
2647         PR tree-optimization/59822
2648         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
2649         (vectorizable_load): Use it to hoist defs of uses of invariant
2650         loads out of the loop.
2652 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
2653             Kugan Vivekanandarajah  <kuganv@linaro.org>
2655         PR target/59695
2656         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
2657         truncation.
2659 2014-01-15  Richard Biener  <rguenther@suse.de>
2661         PR rtl-optimization/59802
2662         * lcm.c (compute_available): Use inverted postorder to seed
2663         the initial worklist.
2665 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2667         PR target/59803
2668         * config/s390/s390.c (s390_preferred_reload_class): Don't return
2669         ADDR_REGS for invalid symrefs in non-PIC code.
2671 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
2673         PR other/58712
2674         * builtins.c (determine_block_size): Initialize *probable_max_size
2675         even if len_rtx is CONST_INT.
2677 2014-01-14  Andrew Pinski  <apinski@cavium.com>
2679         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
2680         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
2681         (cortexa53_tunings): Likewise.
2682         (aarch64_sched_issue_rate): New function.
2683         (TARGET_SCHED_ISSUE_RATE): Define.
2685 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2687         * ira-costs.c (find_costs_and_classes): Add missed
2688         ira_init_register_move_cost_if_necessary.
2690 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
2692         PR target/59787
2693         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
2695 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
2697         PR target/59794
2698         * config/i386/i386.c (type_natural_mode): Add a bool parameter
2699         to indicate if type is used for function return value.  Warn ABI
2700         change if the vector mode isn't available for function return value.
2701         (ix86_function_arg_advance): Pass false to type_natural_mode.
2702         (ix86_function_arg): Likewise.
2703         (ix86_gimplify_va_arg): Likewise.
2704         (function_arg_32): Don't warn ABI change.
2705         (ix86_function_value): Pass true to type_natural_mode.
2706         (ix86_return_in_memory): Likewise.
2707         (ix86_struct_value_rtx): Removed.
2708         (TARGET_STRUCT_VALUE_RTX): Likewise.
2710 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2712         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
2713         converting a conditional jump into a conditional return.
2715 2014-01-14  Richard Biener  <rguenther@suse.de>
2717         PR tree-optimization/58921
2718         PR tree-optimization/59006
2719         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
2720         hoisting invariant stmts.
2721         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
2722         invariant loads on the preheader edge if possible.
2724 2014-01-14  Joey Ye  <joey.ye@arm.com>
2726         * doc/plugin.texi (Building GCC plugins): Update to C++.
2728 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
2730         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
2731         (_mm_rcp28_round_ss): Ditto.
2732         (_mm_rsqrt28_round_sd): Ditto.
2733         (_mm_rsqrt28_round_ss): Ditto.
2734         (_mm_rcp28_sd): Ditto.
2735         (_mm_rcp28_ss): Ditto.
2736         (_mm_rsqrt28_sd): Ditto.
2737         (_mm_rsqrt28_ss): Ditto.
2738         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
2739         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
2740         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
2741         (IX86_BUILTIN_RCP28SD): Ditto.
2742         (IX86_BUILTIN_RCP28SS): Ditto.
2743         (IX86_BUILTIN_RSQRT28SD): Ditto.
2744         (IX86_BUILTIN_RSQRT28SS): Ditto.
2745         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
2746         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
2747         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
2748         (ix86_expand_special_args_builtin): Expand new FTYPE.
2749         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
2750         (srcp14<mode>): Make insn unary.
2751         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
2752         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
2753         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2754         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
2755         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
2756         Fix rounding: make it SAE only.
2757         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2758         Ditto.
2759         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2760         Ditto.
2761         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
2762         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
2763         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2764         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
2765         (round_saeonly_mask_scalar_operand4): Ditto.
2766         (round_saeonly_mask_scalar_op3): Ditto.
2767         (round_saeonly_mask_scalar_op4): Ditto.
2769 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2771         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2772         Implement -maltivec=be for vec_insert and vec_extract.
2774 2014-01-10  DJ Delorie  <dj@redhat.com>
2776         * config/msp430/msp430.md (call_internal): Don't allow memory
2777         references with SP as the base register.
2778         (call_value_internal): Likewise.
2779         * config/msp430/constraints.md (Yc): New.  For memory references
2780         that don't use SP as a base register.
2782         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
2783         "an integer without a # prefix"
2784         * config/msp430/msp430.md (epilogue_helper): Use it.
2786 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2788         PR target/59617
2789         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
2790         AVX512F gather builtins.
2791         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
2792         on gather decls with INTEGER_TYPE masktype.
2793         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
2794         directly into the builtin rather than hoisting it before loop.
2796         PR tree-optimization/59387
2797         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
2798         (scev_const_prop): If folded_casts and type has undefined overflow,
2799         use force_gimple_operand instead of force_gimple_operand_gsi and
2800         for each added stmt if it is assign with
2801         arith_code_with_undefined_signed_overflow, call
2802         rewrite_to_defined_overflow.
2803         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
2804         gimple-fold.h instead.
2805         (arith_code_with_undefined_signed_overflow,
2806         rewrite_to_defined_overflow): Moved to ...
2807         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
2808         rewrite_to_defined_overflow): ... here.  No longer static.
2809         Include gimplify-me.h.
2810         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
2811         rewrite_to_defined_overflow): New prototypes.
2813 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2815         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
2817 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
2819         * builtins.c (get_object_alignment_2): Minor tweak.
2820         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
2822 2014-01-13  Christian Bruel  <christian.bruel@st.com>
2824         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
2825         optimized non constant lengths.
2827 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
2829         PR libgomp/59194
2830         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
2831         load as __atomic_load_N if possible.
2833 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2835         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
2836         target parameter.
2837         (rs6000_expand_builtin): Adjust call.
2839 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
2841         PR target/58115
2842         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
2843         * config/rs6000/rs6000.c: Include target-globals.h.
2844         (rs6000_set_current_function): Instead of doing target_reinit
2845         unconditionally, use save_target_globals_default_opts and
2846         restore_target_globals.
2848         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
2849         FPSCR.
2850         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
2851         (rs6000_expand_builtin): Handle mffs and mtfsf.
2852         (rs6000_init_builtins): Define mffs and mtfsf.
2853         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
2854         (rs6000_mffs): New pattern.
2855         (rs6000_mtfsf): New pattern.
2857 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
2859         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
2860         Start narrowing with START.  Apply candidate-use pair
2861         and check overall cost in narrowing.
2862         (iv_ca_prune): Pass new argument.
2864 2014-01-10  Jeff Law  <law@redhat.com>
2866         PR middle-end/59743
2867         * ree.c (combine_reaching_defs): Ensure the defining statement
2868         occurs before the extension when optimizing extensions with
2869         different source and destination hard registers.
2871 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
2873         PR ipa/58585
2874         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
2875         vtables into the type inheritance graph.
2877 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2879         PR rtl-optimization/59754
2880         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
2881         modes in the REGNO != REGNO case.
2883 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2885         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
2887 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2889         PR tree-optimization/59745
2890         * tree-predcom.c (tree_predictive_commoning_loop): Call
2891         free_affine_expand_cache if giving up because components is NULL.
2893         * target-globals.c (save_target_globals): Allocate < 4KB structs using
2894         GC in payload of target_globals struct instead of allocating them on
2895         the heap and the larger structs separately using GC.
2896         * target-globals.h (struct target_globals): Make regs, hard_regs,
2897         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
2898         of GTY((skip)) and change type to void *.
2899         (reset_target_globals): Cast loads from those fields to corresponding
2900         types.
2902 2014-01-10  Steve Ellcey  <sellcey@mips.com>
2904         PR plugins/59335
2905         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
2906         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
2907         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
2909 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
2911         PR target/59744
2912         * aarch64-modes.def (CC_Zmode): New flags mode.
2913         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
2914         represents an equality.
2915         (aarch64_get_condition_code): Handle CC_Zmode.
2916         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
2918 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2920         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
2921         extraction in good case.
2923 2014-01-10  Richard Biener  <rguenther@suse.de>
2925         PR tree-optimization/59374
2926         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
2927         checking after SLP discovery.  Mark stmts not participating
2928         in any SLP instance properly.
2930 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2932         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
2933         when handling a SET rtx.
2935 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2937         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
2938         (cortex-a57): Likewise.
2939         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
2941 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2943         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
2944         non-iwmmxt builtins.
2946 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
2948         PR ipa/58252
2949         PR ipa/59226
2950         * ipa-devirt.c record_target_from_binfo): Take as argument
2951         stack of binfos and lookup matching one for virtual inheritance.
2952         (possible_polymorphic_call_targets_1): Update.
2954 2014-01-10  Huacai Chen  <chenhc@lemote.com>
2956         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
2957         kernel strings for Loongson-2E/2F/3A.
2959 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
2961         PR middle-end/59670
2962         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
2963         is_gimple_call before calling gimple_call_internal_p.
2965 2014-01-09  Steve Ellcey  <sellcey@mips.com>
2967         * Makefile.in (TREE_FLOW_H): Remove.
2968         (TREE_SSA_H): Add file names from tree-flow.h.
2969         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
2970         * tree.h: Remove tree-flow.h reference.
2971         * hash-table.h: Remove tree-flow.h reference.
2972         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
2973         reference with tree-ssa-loop.h.
2975 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2977         * doc/invoke.texi: Add -maltivec={be,le} options, and document
2978         default element-order behavior for -maltivec.
2979         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
2980         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
2981         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
2982         when targeting big endian, at least for now.
2983         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
2985 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
2987         PR middle-end/47735
2988         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
2989         var satisfies use_register_for_decl, just take into account type
2990         alignment, rather than decl alignment.
2992         PR tree-optimization/59622
2993         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
2994         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
2995         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
2996         Don't devirtualize for inplace at all.  For targets.length () == 1,
2997         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
2999 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3001         * config/i386/i386.md (cpu): Remove the unused btver1.
3003 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3005         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3007 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3009         PR target/58115
3010         * tree-core.h (struct target_globals): New forward declaration.
3011         (struct tree_target_option): Add globals field.
3012         * tree.h (TREE_TARGET_GLOBALS): Define.
3013         (prepare_target_option_nodes_for_pch): New prototype.
3014         * target-globals.h (struct target_globals): Define even if
3015         !SWITCHABLE_TARGET.
3016         * tree.c (prepare_target_option_node_for_pch,
3017         prepare_target_option_nodes_for_pch): New functions.
3018         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3019         * config/i386/i386.c: Include target-globals.h.
3020         (ix86_set_current_function): Instead of doing target_reinit
3021         unconditionally, use save_target_globals_default_opts and
3022         restore_target_globals.
3024 2014-01-09  Richard Biener  <rguenther@suse.de>
3026         PR tree-optimization/59715
3027         * tree-cfg.h (split_critical_edges): Declare.
3028         * tree-cfg.c (split_critical_edges): Export.
3029         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3031 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
3033         * cfgexpand.c (expand_stack_vars): Optionally disable
3034         asan stack protection.
3035         (expand_used_vars): Likewise.
3036         (partition_stack_vars): Likewise.
3037         * asan.c (asan_emit_stack_protection): Optionally disable
3038         after return stack usage.
3039         (instrument_derefs): Optionally disable memory access instrumentation.
3040         (instrument_builtin_call): Likewise.
3041         (instrument_strlen_call): Likewise.
3042         (asan_protect_global): Optionally disable global variables protection.
3043         * doc/invoke.texi: Added doc for new options.
3044         * params.def: Added new options.
3045         * params.h: Likewise.
3047 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3049         PR rtl-optimization/59724
3050         * ifcvt.c (cond_exec_process_if_block): Don't call
3051         flow_find_head_matching_sequence with 0 longest_match.
3052         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3053         non-active insns if !stop_after.
3054         (try_head_merge_bb): Revert 2014-01-07 changes.
3056 2014-01-08  Jeff Law  <law@redhat.com>
3058         * ree.c (get_sub_rtx): New function, extracted from...
3059         (merge_def_and_ext): Here.
3060         (combine_reaching_defs): Use get_sub_rtx.
3062 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
3064         * cgraph.h (varpool_variable_node): Do not choke on null node.
3066 2014-01-08  Catherine Moore  <clm@codesourcery.com>
3068         * config/mips/mips.md (simple_return): Attempt to use JRC
3069         for microMIPS.
3070         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3072 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3074         PR rtl-optimization/59137
3075         * reorg.c (steal_delay_list_from_target): Call update_block for
3076         elided insns.
3077         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3079 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3081         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3082         two duplicate entries.
3084 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3086         Revert:
3087         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
3089         * config/mips/mips.c (mips_truncated_op_cost): New function.
3090         (mips_rtx_costs): Adjust test for BADDU.
3091         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3093         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
3095         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3096         (*baddu_si): ...this new pattern.
3098 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3100         PR ipa/59722
3101         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3103 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3105         PR middle-end/57748
3106         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3107         inner_reference_p.
3108         (expand_expr, expand_normal): Adjust.
3109         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3110         inner_reference_p. Use inner_reference_p to expand inner references.
3111         (store_expr): Adjust.
3112         * cfgexpand.c (expand_call_stmt): Adjust.
3114 2014-01-08  Rong Xu  <xur@google.com>
3116         * gcov-io.c (gcov_var): Move from gcov-io.h.
3117         (gcov_position): Ditto.
3118         (gcov_is_error): Ditto.
3119         (gcov_rewrite): Ditto.
3120         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3121         only part to libgcc/libgcov.h.
3123 2014-01-08  Marek Polacek  <polacek@redhat.com>
3125         PR middle-end/59669
3126         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3128 2014-01-08  Marek Polacek  <polacek@redhat.com>
3130         PR sanitizer/59667
3131         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3133 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3135         PR rtl-optimization/59649
3136         * stor-layout.c (get_mode_bounds): For BImode return
3137         0 and STORE_FLAG_VALUE.
3139 2014-01-08  Richard Biener  <rguenther@suse.de>
3141         PR middle-end/59630
3142         * gimple.h (is_gimple_builtin_call): Remove.
3143         (gimple_builtin_call_types_compatible_p): New.
3144         (gimple_call_builtin_p): New overload.
3145         * gimple.c (is_gimple_builtin_call): Remove.
3146         (validate_call): Rename to ...
3147         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
3148         check return types.
3149         (validate_type): New static function.
3150         (gimple_call_builtin_p): New overload and adjust.
3151         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3152         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
3153         (gimple_fold_stmt_to_constant_1): Likewise.
3154         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3156 2014-01-08  Richard Biener  <rguenther@suse.de>
3158         PR middle-end/59471
3159         * gimplify.c (gimplify_expr): Gimplify register-register type
3160         VIEW_CONVERT_EXPRs to separate stmts.
3162 2014-01-07  Jeff Law  <law@redhat.com>
3164         PR middle-end/53623
3165         * ree.c (combine_set_extension): Handle case where source
3166         and destination registers in an extension insn are different.
3167         (combine_reaching_defs): Allow source and destination registers
3168         in extension to be different under limited circumstances.
3169         (add_removable_extension): Remove restriction that the
3170         source and destination registers in the extension are the same.
3171         (find_and_remove_re): Emit a copy from the extension's
3172         destination to its source after the defining insn if
3173         the source and destination registers are different.
3175         PR middle-end/59285
3176         * ifcvt.c (merge_if_block): If we are merging a block with more than
3177         one successor with a block with no successors, remove any BARRIER
3178         after the second block.
3180 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3182         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3184 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3186         PR target/59652
3187         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3188         for 14-bit register offsets when INT14_OK_STRICT is false.
3190 2014-01-07  Roland Stigge  <stigge@antcom.de>
3191             Michael Meissner  <meissner@linux.vnet.ibm.com>
3193         PR 57386/target
3194         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3195         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3197 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3199         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3200         -mcpu.
3202 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3204         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3205         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3206         rtx is const0_rtx or not.
3208 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3210         PR target/58115
3211         * target-globals.c (save_target_globals): Remove this_fn_optab
3212         handling.
3213         * toplev.c: Include optabs.h.
3214         (target_reinit): Temporarily restore the global options if another
3215         set of options are in force.
3217 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3219         PR rtl-optimization/58668
3220         * cfgcleanup.c (flow_find_cross_jump): Don't count
3221         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3222         to determine what is counted.
3223         (flow_find_head_matching_sequence): Use active_insn_p to determine
3224         what is counted.
3225         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3226         counting change.
3227         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3228         determine what is counted.
3230         PR tree-optimization/59643
3231         * tree-predcom.c (split_data_refs_to_components): If one dr is
3232         read and one write, determine_offset fails and the write isn't
3233         in the bad component, just put the read into the bad component.
3235 2014-01-07  Mike Stump  <mikestump@comcast.net>
3236             Jakub Jelinek  <jakub@redhat.com>
3238         PR pch/59436
3239         * tree-core.h (struct tree_optimization_option): Change optabs
3240         type from unsigned char * to void *.
3241         * optabs.c (init_tree_optimization_optabs): Adjust
3242         TREE_OPTIMIZATION_OPTABS initialization.
3244 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3246         PR target/59644
3247         * config/i386/i386.h (struct machine_function): Add
3248         no_drap_save_restore field.
3249         * config/i386/i386.c (ix86_save_reg): Use
3250         !cfun->machine->no_drap_save_restore instead of
3251         crtl->stack_realign_needed.
3252         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3253         this function clears frame_pointer_needed.  Set
3254         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3255         and DRAP reg is needed.
3257 2014-01-06  Marek Polacek  <polacek@redhat.com>
3259         PR c/57773
3260         * doc/implement-c.texi: Mention that other integer types are
3261         permitted as bit-field types in strictly conforming mode.
3263 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3265         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3266         is newly allocated.
3268 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3270         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3272 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3274         PR ipa/59008
3275         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3276         to int.
3277         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3279 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3281         PR debug/59350
3282         PR debug/59510
3283         * var-tracking.c (add_stores): Preserve the value of the source even if
3284         we don't record the store.
3286 2014-01-06  Terry Guo  <terry.guo@arm.com>
3288         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3290 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3292         PR bootstrap/59541
3293         * config/darwin.c (darwin_function_section): Adjust return values to
3294         correspond to optimisation changes made in r206070.
3296 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3298         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3299         from prefetch_block tune setting.
3300         (nocona_cost): Correct size of prefetch block to 64.
3302 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3304         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3305         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3306         used to save the static chain register in the computation of the offset
3307         from which the FP registers need to be restored.
3309 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3311         PR tree-optimization/59519
3312         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3313         ICE if get_current_def (current_new_name) is already non-NULL, as long
3314         as it is a phi result of some other phi in *new_exit_bb that has
3315         the same argument.
3317         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3318         or vmovdqu* for misaligned_operand.
3319         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3320         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3321         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3322         aligned_mem for AVX512F masked aligned load and store builtins and for
3323         non-temporal moves.
3325 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3327         PR tree-optimization/59651
3328         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3329         Address range for negative step should be added by TYPE_SIZE_UNIT.
3331 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3333         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3334         overlapping registers also for registers other than the stack pointer.
3336 2014-01-03  Marek Polacek  <polacek@redhat.com>
3338         PR other/59661
3339         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3340         __builtin_FILE.
3342 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
3344         PR target/59625
3345         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3346         asm goto as jump.
3348         * config/i386/i386.md (MODE_SIZE): New mode attribute.
3349         (push splitter): Use <P:MODE_SIZE> instead of
3350         GET_MODE_SIZE (<P:MODE>mode).
3351         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3352         (mov -1, reg peephole2): Likewise.
3353         * config/i386/sse.md (*mov<mode>_internal,
3354         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3355         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3356         *<code><mode>3, *andnot<mode>3<mask_name>,
3357         <mask_codefor><code><mode>3<mask_name>): Likewise.
3358         * config/i386/subst.md (mask_mode512bit_condition,
3359         sd_mask_mode512bit_condition): Likewise.
3361 2014-01-02  Xinliang David Li  <davidxl@google.com>
3363         PR tree-optimization/59303
3364         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3365         (dump_predicates): Better output format.
3366         (pred_equal_p): New function.
3367         (is_neq_relop_p): Ditto.
3368         (is_neq_zero_form_p): Ditto.
3369         (pred_expr_equal_p): Ditto.
3370         (pred_neg_p): Ditto.
3371         (simplify_pred): Ditto.
3372         (simplify_preds_2): Ditto.
3373         (simplify_preds_3): Ditto.
3374         (simplify_preds_4): Ditto.
3375         (simplify_preds): Ditto.
3376         (push_pred): Ditto.
3377         (push_to_worklist): Ditto.
3378         (get_pred_info_from_cmp): Ditto.
3379         (is_degenerated_phi): Ditto.
3380         (normalize_one_pred_1): Ditto.
3381         (normalize_one_pred): Ditto.
3382         (normalize_one_pred_chain): Ditto.
3383         (normalize_preds): Ditto.
3384         (normalize_cond_1): Remove function.
3385         (normalize_cond): Ditto.
3386         (is_gcond_subset_of): Ditto.
3387         (is_subset_of_any): Ditto.
3388         (is_or_set_subset_of): Ditto.
3389         (is_and_set_subset_of): Ditto.
3390         (is_norm_cond_subset_of): Ditto.
3391         (pred_chain_length_cmp): Ditto.
3392         (convert_control_dep_chain_into_preds): Type change.
3393         (find_predicates): Ditto.
3394         (find_def_preds): Ditto.
3395         (destroy_predicates_vecs): Ditto.
3396         (find_matching_predicates_in_rest_chains): Ditto.
3397         (use_pred_not_overlap_with_undef_path_pred): Ditto.
3398         (is_pred_expr_subset): Ditto.
3399         (is_pred_chain_subset_of): Ditto.
3400         (is_included_in): Ditto.
3401         (is_superset_of): Ditto.
3403 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3405         Update copyright years.
3407 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3409         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3410         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3411         config/arc/arc.md, config/arc/arc.opt,
3412         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3413         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3414         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3415         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3416         config/linux-protos.h, config/linux.c, config/winnt-c.c,
3417         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3418         vtable-verify.c, vtable-verify.h: Use the standard form for the
3419         copyright notice.
3421 2014-01-02  Tobias Burnus  <burnus@net-b.de>
3423         * gcc.c (process_command): Update copyright notice dates.
3424         * gcov-dump.c: Ditto.
3425         * gcov.c: Ditto.
3426         * doc/cpp.texi: Bump @copying's copyright year.
3427         * doc/cppinternals.texi: Ditto.
3428         * doc/gcc.texi: Ditto.
3429         * doc/gccint.texi: Ditto.
3430         * doc/gcov.texi: Ditto.
3431         * doc/install.texi: Ditto.
3432         * doc/invoke.texi: Ditto.
3434 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3436         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3438 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
3440         * config/i386/sse.md (*mov<mode>_internal): Guard
3441         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3443         PR rtl-optimization/59647
3444         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3445         new_rtx into UNSIGNED_FLOAT rtxes.
3447 Copyright (C) 2014 Free Software Foundation, Inc.
3449 Copying and distribution of this file, with or without modification,
3450 are permitted in any medium without royalty provided the copyright
3451 notice and this notice are preserved.