PR c/50347
[official-gcc.git] / gcc / ChangeLog
blob2c11c94014839dc3b94fef9c2c64dd1b798ea4bf
1 2014-03-27  Marek Polacek  <polacek@redhat.com>
3         PR c/50347
4         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
5         types.
7 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
9         * config/s390/s390.c (s390_can_use_return_insn): Check for
10         call-saved FPRs on 31 bit.
12 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
14         PR middle-end/60682
15         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
16         if they need regimplification, just drop them instead of
17         calling gimple_regimplify_operands on them.
19 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
21         PR target/60580
22         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
23         (aarch64_frame_pointer_required): Adjust logic.
24         (aarch64_can_eliminate): Adjust logic.
25         (aarch64_override_options_after_change): Adjust logic.
27 2014-03-27  Dehao Chen  <dehao@google.com>
29         * ipa-inline.c (early_inliner): Update node's inline info.
31 2014-03-26  Dehao Chen  <dehao@google.com>
33         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
34         compiler inserted conditional jumps for NAN float check.
36 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
38         * ubsan.h (ubsan_create_data): Change second argument's type
39         to const location_t *.
40         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
41         _("<unknown>").
42         (ubsan_create_data): Change second argument to const location_t *PLOC.
43         Create Loc field whenever PLOC is non-NULL.
44         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
45         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
46         callers.
48         PR other/59545
49         * real.c (real_to_integer2): Change type of low to UHWI.
51 2014-03-26  Tobias Burnus  <burnus@net-b.de>
53         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
54         (CILK_SELF_SPECS): New define.
55         (driver_self_specs): Use it.
57 2014-03-26  Richard Biener  <rguenther@suse.de>
59         * tree-pretty-print.c (percent_K_format): Implement special
60         case for LTO and its stripped down BLOCK tree.
62 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
64         PR sanitizer/60636
65         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
67         * tree-vrp.c (simplify_internal_call_using_ranges): If only
68         one range is range_int_cst_p, but not both, at least optimize
69         addition/subtraction of 0 and multiplication by 0 or 1.
70         * gimple-fold.c (gimple_fold_call): Fold
71         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
72         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
73         INTEGER_CSTs, try to fold at least x * 0 and y - y.
75 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
77         PR rtl-optimization/60452
78         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
79         <case REG>: Return 1 for invalid offsets from the frame pointer.
81 2014-03-26  Marek Polacek  <polacek@redhat.com>
83         PR c/37428
84         * doc/extend.texi (C Extensions): Mention variable-length arrays in
85         a structure/union.
87 2014-03-26  Marek Polacek  <polacek@redhat.com>
89         PR c/39525
90         * doc/extend.texi (Designated Inits): Describe what happens to omitted
91         field members.
93 2014-03-26  Marek Polacek  <polacek@redhat.com>
95         PR other/59545
96         * ira-color.c (update_conflict_hard_regno_costs): Perform the
97         multiplication in unsigned type.
99 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
101         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
103 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
105         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
107 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
109         PR ipa/60315
110         * cif-code.def (UNREACHABLE) New code.
111         * ipa-inline.c (inline_small_functions): Skip edges to __builtlin_unreachable.
112         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
113         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
114         predicate to __bulitin_unreachable.
115         (set_cond_stmt_execution_predicate): Fix issue when invert_tree_comparison
116         returns ERROR_MARK.
117         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
118         propagate to inline clones.
119         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
120         to unreachable.
121         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
122         * cgraphclones.c (cgraph_clone_node): If call destination is already
123         ureachable, do not redirect it back.
124         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
125         unreachable.
127 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
129         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
130         Do not modify inline clones.
132 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
134         * config/i386/i386.md (general_sext_operand): New mode attr.
135         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
136         don't generate (sign_extend (const_int)).
137         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
138         operands[2].  Use We constraint instead of <i> and <general_sext_operand>
139         predicate instead of <general_operand>.
140         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
141         * config/i386/constraints.md (We): New constraint.
142         * config/i386/predicates.md (x86_64_sext_operand,
143         sext_operand): New predicates.
145 2014-03-25  Martin Jambor  <mjambor@suse.cz>
147         PR ipa/60600
148         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
149         inconsistent devirtualizations to __builtin_unreachable.
151 2014-03-25  Marek Polacek  <polacek@redhat.com>
153         PR c/35449
154         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
156 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
158         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
159         order of elements for big-endian.
161 2014-03-25  Richard Biener  <rguenther@suse.de>
163         PR middle-end/60635
164         * gimplify-me.c (gimple_regimplify_operands): Update the
165         re-gimplifed stmt.
167 2014-03-25  Martin Jambor  <mjambor@suse.cz>
169         PR ipa/59176
170         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
171         (lto_output_varpool_node): Likewise.
172         (input_overwrite_node): Likewise.
173         (input_varpool_node): Likewise.
175 2014-03-25  Richard Biener  <rguenther@suse.de>
177         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like
178         OPT_fpie.
179         (run_gcc): Likewise.
181 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
183         * combine.c (simplify_compare_const): Add MODE argument.
184         Handle mode_width 0 as very large mode_width.
185         (try_combine, simplify_comparison): Adjust callers.
187         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
188         type to avoid signed integer overflow.
189         * explow.c (plus_constant): Likewise.
191 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
193         * doc/generic.texi: Correct typos.
195 2014-03-24  Tobias Burnus  <burnus@net-b.de>
197         * doc/invoke.texi (-flto): Expand section about
198         using static libraries with LTO.
200 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
202         PR rtl-optimization/60501
203         * optabs.def (addptr3_optab): New optab.
204         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
205         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
206         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
208         * lra.c (emit_add3_insn): Use the addptr pattern if available.
210         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
212 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
214         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
215         _mm512_set1_pd.
217         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
218         (_mm256_undefined_ps): Define.
219         (_mm256_undefined_pd): Define.
220         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
221         (_mm_undefined_pd): Define.
222         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
223         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
224         (_mm512_undefined_ps): Define.
225         (_mm512_undefined_pd): Define.
226         Use _mm*_undefined_*.
227         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
229 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
231         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
232         (lshr_simd): DI mode added.
233         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
234         (aarch64_ushr_simddi): Likewise.
235         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
236         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
237         (vshrd_n_u64): Likewise.
239 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
241         * Makefile.in (s-macro_list): Depend on cc1.
243 2014-03-23  Teresa Johnson  <tejohnson@google.com>
245         * ipa-utils.c (ipa_print_order): Use specified dump file.
247 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
249         PR rtl-optimization/60601
250         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
252         * gcc.c (eval_spec_function): Initialize save_growing_value.
254 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
256         PR sanitizer/60613
257         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
258         code == MINUS_EXPR, never swap op0 with op1.
260         * toplev.c (init_local_tick): Avoid signed integer multiplication
261         overflow.
262         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
263         shift by first operand's bitsize.
265 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
267         PR target/60610
268         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
269         redefine to 1 or 0.
270         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
271         TARGET_ISA_64BIT_P(x).
273 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
275         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
276         pattern for vector nor instead of subtract from splat(-1).
277         (altivec_expand_vec_perm_const_le): Likewise.
279 2014-03-21  Richard Henderson  <rth@twiddle.net>
281         PR target/60598
282         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
283         related insns after epilogue_completed.
285 2014-03-21  Martin Jambor  <mjambor@suse.cz>
287         PR ipa/59176
288         * cgraph.h (symtab_node): New flag body_removed.
289         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
290         when removing bodies.
291         * symtab.c (dump_symtab_base): Dump body_removed flag.
292         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
293         had their bodies removed.
295 2014-03-21  Martin Jambor  <mjambor@suse.cz>
297         PR ipa/60419
298         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
299         in the border.
301 2014-03-21  Richard Biener  <rguenther@suse.de>
303         PR tree-optimization/60577
304         * tree-core.h (struct tree_base): Document nothrow_flag use
305         in VAR_DECL_NONALIASED.
306         * tree.h (VAR_DECL_NONALIASED): New.
307         (may_be_aliased): Adjust.
308         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
310 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
312         * expr.c (expand_expr_real_1): Remove outdated comment.
314 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
316         PR middle-end/60597
317         * ira.c (adjust_cleared_regs): Call copy_rtx on
318         *reg_equiv[REGNO (loc)].src_p before passing it to
319         simplify_replace_fn_rtx.
321         PR target/60568
322         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
323         into CONST, put pic register as first operand of PLUS.  Use
324         gen_const_mem for both 32-bit and 64-bit PIC got loads.
326 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
328         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
330 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
332         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
333         around for store forwarding issue in the FPU on the UT699.
334         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
335         loads and operations if -mfix-ut699 is specified.
336         (divtf3_hq): Tweak attribute.
337         (sqrttf2_hq): Likewise.
339 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
341         * calls.c (store_one_arg): Remove incorrect const qualification on the
342         type of the temporary.
343         * cfgexpand.c (expand_return): Likewise.
344         * expr.c (expand_constructor): Likewise.
345         (expand_expr_real_1): Likewise.
347 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
349         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
350         of parts.
352 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
354         PR target/60039
355         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
357 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
359         * config/arm/aarch-common-protos.h
360         (alu_cost_table): Fix spelling of "extend".
361         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
363 2014-03-19  Richard Biener  <rguenther@suse.de>
365         PR middle-end/60553
366         * tree-core.h (tree_type_common): Re-order pointer members
367         to reduce recursion depth during GC walks.
369 2014-03-19  Marek Polacek  <polacek@redhat.com>
371         PR sanitizer/60569
372         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
373         before accessing it.
375 2014-03-19  Richard Biener  <rguenther@suse.de>
377         PR lto/59543
378         * lto-streamer-in.c (input_function): In WPA stage do not drop
379         debug stmts.
381 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
383         PR tree-optimization/60559
384         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
385         with build_zero_cst assignment.
387 2014-03-18  Kai Tietz  <ktietz@redhat.com>
389         PR rtl-optimization/56356
390         * sdbout.c (sdbout_parms): Verify that parms'
391         incoming argument is valid.
392         (sdbout_reg_parms): Likewise.
394 2014-03-18  Richard Henderson  <rth@redhat.com>
396         PR target/60562
397         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
398         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
399         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
401 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
403         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
404         * doc/plugins.texi (Plugin callbacks): Mention
405         PLUGIN_INCLUDE_FILE.
406         Italicize plugin event names in description.  Explain that
407         PLUGIN_PRAGMAS has no sense for lto1. Explain
408         PLUGIN_INCLUDE_FILE.
409         Remind that no GCC functions should be called after
410         PLUGIN_FINISH.
411         Explain what pragmas with expansion are.
413 2014-03-18  Martin Liska  <mliska@suse.cz>
415         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
416         gimple call statement is update.
417         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
418         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
420 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
422         PR sanitizer/60557
423         * ubsan.c (ubsan_instrument_unreachable): Call
424         initialize_sanitizer_builtins.
425         (ubsan_pass): Likewise.
427         PR sanitizer/60535
428         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
429         varpool_finalize_decl instead of rest_of_decl_compilation.
431 2014-03-18  Richard Biener  <rguenther@suse.de>
433         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
434         by using bitmap_and_compl instead of bitmap_and_compl_into.
435         (df_rd_transfer_function): Likewise.
437 2014-03-18  Richard Biener  <rguenther@suse.de>
439         * doc/lto.texi (fresolution): Fix typo.
441 2014-03-18  Richard Biener  <rguenther@suse.de>
443         * doc/invoke.texi (flto): Update for changes in 4.9.
445 2014-03-18  Richard Biener  <rguenther@suse.de>
447         * doc/loop.texi: Remove section on the removed lambda framework.
448         Update loop docs with recent changes in preserving loop structure.
450 2014-03-18  Richard Biener  <rguenther@suse.de>
452         * doc/lto.texi (-fresolution): Document.
454 2014-03-18  Richard Biener  <rguenther@suse.de>
456         * doc/contrib.texi: Adjust my name.
458 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
460         PR ipa/58721
461         * internal-fn.c: Include diagnostic-core.h.
462         (expand_BUILTIN_EXPECT): New function.
463         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
464         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
465         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
466         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
467         IFN_BUILTIN_EXPECT.
468         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
469         Revert 3 argument __builtin_expect code.
470         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
471         * gimple-fold.c (gimple_fold_call): Likewise.
472         * tree.h (fold_builtin_expect): New prototype.
473         * builtins.c (build_builtin_expect_predicate): Add predictor
474         argument, if non-NULL, create 3 argument __builtin_expect.
475         (fold_builtin_expect): No longer static.  Add ARG2 argument,
476         pass it through to build_builtin_expect_predicate.
477         (fold_builtin_2): Adjust caller.
478         (fold_builtin_3): Handle BUILT_IN_EXPECT.
479         * internal-fn.def (BUILTIN_EXPECT): New.
481 2014-03-18  Tobias Burnus  <burnus@net-b.de>
483         PR ipa/58721
484         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
485         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
486         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
488 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
490         PR ipa/58721
491         * predict.c (combine_predictions_for_bb): Fix up formatting.
492         (expr_expected_value_1, expr_expected_value): Add predictor argument,
493         fill what it points to if non-NULL.
494         (tree_predict_by_opcode): Adjust caller, use the predictor.
495         * predict.def (PRED_COMPARE_AND_SWAP): Add.
497 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
499         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
500         proper constant for the store mode.
502 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
504         * symtab.c (change_decl_assembler_name): Fix transparent alias
505         chain construction.
507 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
509         * config/aarch64/aarch64.c: Correct the comments about the
510         aarch64 stack layout.
512 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
514         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
515         check for GF_OMP_FOR_KIND_FOR.
517 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
519         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
520         ymm and zmm register names.
522 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
524         PR target/60516
525         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
526         note creation for the 2010-08-31 changes.
528 2014-03-17  Marek Polacek  <polacek@redhat.com>
530         PR middle-end/60534
531         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
532         as -fno-tree-loop-vectorize.
533         (expand_omp_simd): Likewise.
535 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
537         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
538         (eligible_for_call_delay): New prototype.
539         * config/sparc/sparc.c (tls_call_delay): Rename into...
540         (eligible_for_call_delay): ...this.  Return false if the instruction
541         cannot be put in the delay slot of a branch.
542         (eligible_for_restore_insn): Simplify.
543         (eligible_for_return_delay): Return false if the instruction cannot be
544         put in the delay slot of a branch and simplify.
545         (eligible_for_sibcall_delay): Return false if the instruction cannot be
546         put in the delay slot of a branch.
547         * config/sparc/sparc.md (fix_ut699): New attribute.
548         (tls_call_delay): Delete.
549         (in_call_delay): Reimplement.
550         (eligible_for_sibcall_delay): Rename into...
551         (in_sibcall_delay): ...this.
552         (eligible_for_return_delay): Rename into...
553         (in_return_delay): ...this.
554         (in_branch_delay): Reimplement.
555         (in_uncond_branch_delay): Delete.
556         (in_annul_branch_delay): Delete.
558 2014-03-14  Richard Henderson  <rth@redhat.com>
560         PR target/60525
561         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
562         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
563         (*floathi<X87MODEF>2_i387_with_temp): Remove.
564         (floathi splitters): Remove.
565         (float<SWI48x>xf2): New pattern.
566         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
567         code that tried to handle DImode for 32-bit, but which was excluded
568         by the pattern's condition.  Drop allocation of stack temporary.
569         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
570         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
571         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
572         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
573         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
574         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
575         (*float<SWI48><MODEF>2_sse_interunit): Remove.
576         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
577         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
578         (*float<SWI48x><X87MODEF>2_i387): Remove.
579         (all float _with_temp splitters): Remove.
580         (*float<SWI48x><MODEF>2_i387): New pattern.
581         (*float<SWI48><MODEF>2_sse): New pattern.
582         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
583         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
585 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
586             Marek Polacek  <polacek@redhat.com>
588         PR middle-end/60484
589         * common.opt (dump_base_name_prefixed): New Variable.
590         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
591         if x_dump_base_name_prefixed is already set, set it at the end.
593 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
595         PR rtl-optimization/60508
596         * lra-constraints.c (get_reload_reg): Add new parameter
597         in_subreg_p.
598         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
599         Pass the new parameter values.
601 2014-03-14  Richard Biener  <rguenther@suse.de>
603         * common.opt: Revert unintented changes from r205065.
604         * opts.c: Likewise.
606 2014-03-14  Richard Biener  <rguenther@suse.de>
608         PR middle-end/60518
609         * cfghooks.c (split_block): Properly adjust all loops the
610         block was a latch of.
612 2014-03-14  Martin Jambor  <mjambor@suse.cz>
614         PR lto/60461
615         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
616         and simplify it.
618 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
620         PR target/59396
621         * config/avr/avr.c (avr_set_current_function): Pass function name
622         through default_strip_name_encoding before sanity checking instead
623         of skipping the first char of the assembler name.
625 2014-03-13  Richard Henderson  <rth@redhat.com>
627         PR debug/60438
628         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
629         (ix86_force_to_memory, ix86_free_from_memory): Remove.
630         * config/i386/i386-protos.h: Likewise.
631         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
632         in the expander instead of a splitter.
633         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
634         any possibility of requiring a memory.
635         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
636         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
637         (fp branch splitters): Update for ix86_split_fp_branch.
638         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
639         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
640         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
641         (*fop_<MODEF>_2_i387): Remove f/r alternative.
642         (*fop_<MODEF>_3_i387): Likewise.
643         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
644         (splitters for the fop_* register patterns): Remove.
645         (fscalexf4_i387): Rename from *fscalexf4_i387.
646         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
647         
648 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
650         PR tree-optimization/59779
651         * tree-dfa.c (get_ref_base_and_extent): Use double_int
652         type for bitsize and maxsize instead of HOST_WIDE_INT.
654 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
656         PR rtl-optimization/57320
657         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
658         the CFG after thread_prologue_and_epilogue_insns.
660 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
662         PR rtl-optimization/57189
663         * lra-constraints.c (process_alt_operands): Disfavor spilling
664         vector pseudos.
666 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
668         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
669         messages.
671 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
673         PR tree-optimization/59025
674         PR middle-end/60418
675         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
676         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
678 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
680         PR target/60486
681         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
682         calls of avr_out_plus_1.
684 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
686         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
687         BB's single pred and update the father loop's latch info later.
689 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
691         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
692         (VEC_M): Likewise.
693         (VEC_N): Likewise.
694         (VEC_R): Likewise.
695         (VEC_base): Likewise.
696         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
697         registers, we need to swap double words in little endian mode.
699         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
700         to be a container mode for 128-bit integer operations added in ISA
701         2.07.  Unlike TImode and PTImode, the preferred register set is
702         the Altivec/VMX registers for the 128-bit operations.
704         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
705         declarations.
706         (rs6000_split_128bit_ok_p): Likewise.
708         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
709         macros for creating ISA 2.07 normal and overloaded builtin
710         functions with 3 arguments.
711         (BU_P8V_OVERLOAD_3): Likewise.
712         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
713         for use as overloaded functions.
714         (VPERM_1TI_UNS): Likewise.
715         (VSEL_1TI): Likewise.
716         (VSEL_1TI_UNS): Likewise.
717         (ST_INTERNAL_1ti): Likewise.
718         (LD_INTERNAL_1ti): Likewise.
719         (XXSEL_1TI): Likewise.
720         (XXSEL_1TI_UNS): Likewise.
721         (VPERM_1TI): Likewise.
722         (VPERM_1TI_UNS): Likewise.
723         (XXPERMDI_1TI): Likewise.
724         (SET_1TI): Likewise.
725         (LXVD2X_V1TI): Likewise.
726         (STXVD2X_V1TI): Likewise.
727         (VEC_INIT_V1TI): Likewise.
728         (VEC_SET_V1TI): Likewise.
729         (VEC_EXT_V1TI): Likewise.
730         (EQV_V1TI): Likewise.
731         (NAND_V1TI): Likewise.
732         (ORC_V1TI): Likewise.
733         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
734         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
735         overloaded builtin.
736         (VADDUQM): Likewise.
737         (VSUBCUQ): Likewise.
738         (VADDEUQM): Likewise.
739         (VADDECUQ): Likewise.
740         (VSUBEUQM): Likewise.
741         (VSUBECUQ): Likewise.
743         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
744         __int128_t and __uint128_t types.
745         (__uint128_type): Likewise.
746         (altivec_categorize_keyword): Add support for vector __int128_t,
747         vector __uint128_t, vector __int128, and vector unsigned __int128
748         as a container type for TImode operations that need to be done in
749         VSX/Altivec registers.
750         (rs6000_macro_to_expand): Likewise.
751         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
752         to support 128-bit integer instructions vaddcuq, vadduqm,
753         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
754         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
756         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
757         for V1TImode, and set up preferences to use VSX/Altivec registers.
758         Setup VSX reload handlers.
759         (rs6000_debug_reg_global): Likewise.
760         (rs6000_init_hard_regno_mode_ok): Likewise.
761         (rs6000_preferred_simd_mode): Likewise.
762         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
763         (easy_altivec_constant): Likewise.
764         (output_vec_const_move): Likewise.
765         (rs6000_expand_vector_set): Convert V1TImode set and extract to
766         simple move.
767         (rs6000_expand_vector_extract): Likewise.
768         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
769         addressing.
770         (rs6000_const_vec): Add support for V1TImode.
771         (rs6000_emit_le_vsx_load): Swap double words when loading or
772         storing TImode/V1TImode.
773         (rs6000_emit_le_vsx_store): Likewise.
774         (rs6000_emit_le_vsx_move): Likewise.
775         (rs6000_emit_move): Add support for V1TImode.
776         (altivec_expand_ld_builtin): Likewise.
777         (altivec_expand_st_builtin): Likewise.
778         (altivec_expand_vec_init_builtin): Likewise.
779         (altivec_expand_builtin): Likewise.
780         (rs6000_init_builtins): Add support for V1TImode type.  Add
781         support for ISA 2.07 128-bit integer builtins.  Define type names
782         for the VSX/Altivec vector types.
783         (altivec_init_builtins): Add support for overloaded vector
784         functions with V1TImode type.
785         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
786         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
787         external function.
788         (rs6000_split_128bit_ok_p): Likewise.
789         (rs6000_handle_altivec_attribute): Create V1TImode from vector
790         __int128_t and vector __uint128_t.
792         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
793         and mode attributes.
794         (VSX_M): Likewise.
795         (VSX_M2): Likewise.
796         (VSm): Likewise.
797         (VSs): Likewise.
798         (VSr): Likewise.
799         (VSv): Likewise.
800         (VS_scalar): Likewise.
801         (VS_double): Likewise.
802         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
804         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
805         we support the ISA 2.07 128-bit integer arithmetic instructions.
806         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
807         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
808         and TImode types for use with the builtin functions.
809         (V1TI_type_node): Likewise.
810         (unsigned_V1TI_type_node): Likewise.
811         (intTI_type_internal_node): Likewise.
812         (uintTI_type_internal_node): Likewise.
814         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
815         128-bit builtin functions.
816         (UNSPEC_VADDEUQM): Likewise.
817         (UNSPEC_VADDECUQ): Likewise.
818         (UNSPEC_VSUBCUQ): Likewise.
819         (UNSPEC_VSUBEUQM): Likewise.
820         (UNSPEC_VSUBECUQ): Likewise.
821         (VM): Add V1TImode to vector mode iterators.
822         (VM2): Likewise.
823         (VI_unit): Likewise.
824         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
825         (altivec_vaddcuq): Likewise.
826         (altivec_vsubuqm): Likewise.
827         (altivec_vsubcuq): Likewise.
828         (altivec_vaddeuqm): Likewise.
829         (altivec_vaddecuq): Likewise.
830         (altivec_vsubeuqm): Likewise.
831         (altivec_vsubecuq): Likewise.
833         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
834         mode iterators.
835         (BOOL_128): Likewise.
836         (BOOL_REGS_OUTPUT): Likewise.
837         (BOOL_REGS_OP1): Likewise.
838         (BOOL_REGS_OP2): Likewise.
839         (BOOL_REGS_UNARY): Likewise.
840         (BOOL_REGS_AND_CR0): Likewise.
842         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
843         128-bit integer builtin support.
844         (vec_vadduqm): Likewise.
845         (vec_vaddecuq): Likewise.
846         (vec_vaddeuqm): Likewise.
847         (vec_vsubecuq): Likewise.
848         (vec_vsubeuqm): Likewise.
849         (vec_vsubcuq): Likewise.
850         (vec_vsubuqm): Likewise.
852         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
853         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
854         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
855         128-bit integer add/subtract to ISA 2.07.
857 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
859         * config/arc/arc.c (arc_predicate_delay_insns):
860         Fix third argument passed to conditionalize_nonjump.
862 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
864         * config/aarch64/aarch64-builtins.c
865         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
866         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
867         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
868         instead of __builtin_lfloor.
869         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
871 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
873         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
874         (tree_ssa_ifcombine_bb_1): New function.
875         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
876         is an empty forwarder block to then_bb or vice versa and then_bb
877         and else_bb are effectively swapped.
879 2014-03-12  Christian Bruel  <christian.bruel@st.com>
881         PR target/60264
882         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
883         REG_CFA_DEF_CFA note.
884         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
885         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
887 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
889         PR tree-optimization/60454
890         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
892 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
894         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
895         Do not define target_cpu_default2 to generic.
896         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
897         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
898         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
900 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
901             Marc Glisse  <marc.glisse@inria.fr>
903         PR tree-optimization/60502
904         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
905         instead of build_low_bits_mask.
907 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
909         PR middle-end/60482
910         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
911         if there are multiple uses, but op doesn't live on E edge.
912         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
913         clobber stmts before __builtin_unreachable.
915 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
917         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
918         hard_frame_pointer_rtx.
919         * cse.c (cse_insn): Remove volatile check.
920         * cselib.c (cselib_process_insn): Likewise.
921         * dse.c (scan_insn): Likewise.
923 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
925         * config/arc/arc.c (conditionalize_nonjump): New function,
926         broken out of ...
927         (arc_ifcvt): ... this.
928         (arc_predicate_delay_insns): Use it.
930 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
932         * config/arc/predicates.md (extend_operand): During/after reload,
933         allow const_int_operand.
934         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
935         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
936         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
937         to "i".
938         (umulsi3_highpart_i): Likewise.
940 2014-03-11  Richard Biener  <rguenther@suse.de>
942         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
943         Add asserts to guard possible wrong-code bugs.
945 2014-03-11  Richard Biener  <rguenther@suse.de>
947         PR tree-optimization/60429
948         PR tree-optimization/60485
949         * tree-ssa-structalias.c (set_union_with_increment): Properly
950         take into account all fields that overlap the shifted vars.
951         (do_sd_constraint): Likewise.
952         (do_ds_constraint): Likewise.
953         (get_constraint_for_ptr_offset): Likewise.
955 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
957         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
958         (nios2_compute_frame_layout):
959         Add calculation of cfun->machine->fp_save_offset.
960         (nios2_expand_prologue): Correct setting of frame pointer register
961         in prologue.
962         (nios2_expand_epilogue): Update recovery of stack pointer from
963         frame pointer accordingly.
964         (nios2_initial_elimination_offset): Update calculation of offset
965         for eliminating to HARD_FRAME_POINTER_REGNUM.
967 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
969         PR ipa/60457
970         * ipa.c (symtab_remove_unreachable_nodes): Don't call
971         cgraph_get_create_node on VAR_DECLs.
973 2014-03-10  Richard Biener  <rguenther@suse.de>
975         PR middle-end/60474
976         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
978 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
980         * config/vms/vms.opt (vms_float_format): New variable.
982 2014-03-08  Tobias Burnus  <burnus@net-b.de>
984         * doc/invoke.texi (-fcilkplus): Update implementation status.
986 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
987             Richard Biener  <rguenther@suse.de>
989         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
990         consistently accross all TUs.
991         (run_gcc): Enable -fshort-double automatically at link at link-time
992         and disallow override.
994 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
996         PR target/58271
997         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
998         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
999         if they can't be used.
1001 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1003         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1004         for Solaris 11/x86 ld.
1005         * configure: Regenerate.
1007 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1009         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1010         (LIB_TLS_SPEC): Save as ld_tls_libs.
1011         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1012         (HAVE_AS_IX86_TLSLDM): New test.
1013         * configure, config.in: Regenerate.
1014         * config/i386/i386.c (legitimize_tls_address): Fall back to
1015         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1016         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1017         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1018         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1020 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1022         * common.opt (fira-loop-pressure): Mark as optimization.
1024 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1026         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1027         an OpenMP mappable type.
1029 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1031         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1032         MULTILIB_OSDIRNAMES is not defined.
1034 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1035             Meador Inge  <meadori@codesourcery.com>
1037         PR target/58595
1038         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1039         (arm_legitimize_address): Call legitimize_tls_address for any
1040         arm_tls_referenced_p expression, handle constant addend.  Call it
1041         before testing for !TARGET_ARM.
1042         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1044 2014-03-06  Richard Biener  <rguenther@suse.de>
1046         PR middle-end/60445
1047         PR lto/60424
1048         PR lto/60427
1049         Revert
1050         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1052         * tree-streamer.c (record_common_node): Assert we don't record
1053         nodes with type double.
1054         (preload_common_node): Skip type double, complex double and double
1055         pointer since it is now frontend dependent due to fshort-double option.
1057 2014-03-06  Richard Biener  <rguenther@suse.de>
1059         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1060         or -fno-lto is specified and the linker has full plugin support.
1061         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1062         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1063         * lto-wrapper.c (merge_and_complain): Merge compile-time
1064         optimization levels.
1065         (run_gcc): And pass it through to the link options.
1067 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
1069         PR debug/60381
1070         Revert:
1071         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1072         PR debug/59992
1073         * cselib.c (remove_useless_values): Skip to avoid quadratic
1074         behavior if the condition moved from...
1075         (cselib_process_insn): ... here holds.
1077 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1079         PR plugins/59335
1080         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1081         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1083         PR plugins/59335
1084         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1085         (TM_H): Add x86-tune.def.
1087 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1089         * config/aarch64/aarch64.c (generic_tunings):
1090         Use cortexa57_extra_costs.
1092 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1094         PR lto/60404
1095         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1096         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1097         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1098         cost for in_lto_p.
1100 2014-03-04  Heiher  <r@hev.cc>
1102         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1103         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1105 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1107         * config/i386/predicates.md (const2356_operand): Change to ...
1108         (const2367_operand): ... this.
1109         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1110         const2367_operand.
1111         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1112         (*avx512pf_scatterpf<mode>sf): Ditto.
1113         (avx512pf_scatterpf<mode>df): Ditto.
1114         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1115         (*avx512pf_scatterpf<mode>df): Ditto.
1116         * config/i386/i386.c (ix86_expand_builtin): Update
1117         incorrect hint operand error message.
1119 2014-03-04  Richard Biener  <rguenther@suse.de>
1121         * lto-section-in.c (lto_get_section_data): Fix const cast.
1123 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1125         * tree-streamer.c (record_common_node): Assert we don't record
1126         nodes with type double.
1127         (preload_common_node): Skip type double, complex double and double
1128         pointer since it is now frontend dependent due to fshort-double option.
1130 2014-03-04  Richard Biener  <rguenther@suse.de>
1132         PR lto/60405
1133         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1134         (lto_input_toplevel_asms): Likewise.
1135         * lto-section-in.c (lto_get_section_data): Instead do it here
1136         for every section.
1138 2014-03-04  Richard Biener  <rguenther@suse.de>
1140         PR tree-optimization/60382
1141         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1142         dead PHIs a reduction.
1144 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1146         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1147         hint value.
1148         (_mm_prefetch): Move out of GCC target("sse") pragma.
1149         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1150         GCC target("prfchw") pragma.
1151         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1152         for locality <= 2.
1153         * config/i386/i386.c (ix86_option_override_internal): Enable
1154         -mprfchw with -mprefetchwt1.
1156 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1158         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1159         Mark as varying.
1161 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1163         * opts.h (CL_PCH_IGNORE): Define.
1164         * targhooks.c (option_affects_pch_p):
1165         Return false for options that have CL_PCH_IGNORE set.
1166         * opt-functions.awk: Process PchIgnore.
1167         * doc/options.texi: Document PchIgnore.
1169         * config/arc/arc.opt (misize): Add PchIgnore property.
1171 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1173         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1174         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1175         constraint on constants to permit them being loaded into
1176         GENERAL_REGS or BASE_REGS.
1178 2014-03-03  Nick Clifton  <nickc@redhat.com>
1180         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1181         anti-cacnonical alternatives.
1182         (negandhi3_real): New pattern.
1183         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1185 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1187         * config/avr/avr-mcus.def: Remove atxmega16x1.
1188         * config/avr/avr-tables.opt: Regenerate.
1189         * config/avr/t-multilib: Regenerate.
1190         * doc/avr-mmcu.texi: Regenerate.
1192 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1193             Mircea Namolaru  <mircea.namolaru@inria.fr>
1195         PR tree-optimization/58028
1196         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1197         scalar dimensions.
1199 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1201         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1202         not handled by recognizers.
1204 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1206         PR middle-end/60175
1207         * function.c (expand_function_end): Don't emit
1208         clobber_return_register sequence if clobber_after is a BARRIER.
1209         * cfgexpand.c (construct_exit_block): Append instructions before
1210         return_label to prev_bb.
1212 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1214         * config/rs6000/constraints.md: Document reserved use of "wc".
1216 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1218         PR ipa/60150
1219         * ipa.c (function_and_variable_visibility): When dissolving comdat
1220         group, also set all symbols to local.
1222 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1224         PR ipa/60306
1226         Revert:
1227         2013-12-14   Jan Hubicka  <jh@suse.cz>
1228         PR middle-end/58477
1229         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1231 2014-03-02  Jon Beniston  <jon@beniston.com>
1233         PR bootstrap/48230
1234         PR bootstrap/50927
1235         PR bootstrap/52466
1236         PR target/46898
1237         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1238           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1239         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1240         (simple_return, *simple_return): New patterns
1241         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1242         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1244 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1246         * dwarf2out.c (gen_subprogram_die): Tidy.
1248 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1250         PR target/60071
1251         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1252         (*mov_t_msb_neg_negc): ... this new insn.
1254 2014-02-28  Jason Merrill  <jason@redhat.com>
1256         PR c++/58678
1257         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1258         function.
1260 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1262         PR c++/60314
1263         * dwarf2out.c (decltype_auto_die): New static.
1264         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1265         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1266         (is_cxx_auto): Likewise.
1268 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1270         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1271         we are not using general regs only.
1273 2014-02-28  Richard Biener  <rguenther@suse.de>
1275         PR target/60280
1276         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1277         previous fix and only allow to remove trivial pre-headers
1278         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1279         (remove_forwarder_block): Properly update the latch of a loop.
1281 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1283         PR debug/59992
1284         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1285         (cselib_preserved_hash_table): New.
1286         (preserve_constants_and_equivs): Move preserved vals to it.
1287         (cselib_find_slot): Look it up first.
1288         (cselib_init): Initialize it.
1289         (cselib_finish): Release it.
1290         (dump_cselib_table): Dump it.
1292 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1294         PR debug/59992
1295         * cselib.c (remove_useless_values): Skip to avoid quadratic
1296         behavior if the condition moved from...
1297         (cselib_process_insn): ... here holds.
1299 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1301         PR debug/57232
1302         * var-tracking.c (vt_initialize): Apply the same condition to
1303         preserve the CFA base value.
1305 2014-02-28  Joey Ye  <joey.ye@arm.com>
1307         PR target/PR60169
1308         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1309         if reload in progress or completed.
1311 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1313         PR middle-end/60147
1314         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1315         NAMELIST_DECL.
1317 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1319         * doc/tm.texi.in (Condition Code Status): Update documention for
1320         relative locations of cc0-setter and cc0-user.
1322 2014-02-27  Jeff Law  <law@redhat.com>
1324         PR rtl-optimization/52714
1325         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1326         into two independent simple sets, if I3 is a jump, ensure the
1327         pattern we place into I3 is a (set (pc) ...).
1329 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1330             Jeff Law  <law@redhat.com>
1332         PR rtl-optimization/49847
1333         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1334         are in different blocks.
1335         * doc/tm.texi (Condition Code Status): Update documention for
1336         relative locations of cc0-setter and cc0-user.
1338 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1340         PR target/59222
1341         * lra.c (lra_emit_add): Check SUBREG too.
1343 2014-02-27  Andreas Schwab  <schwab@suse.de>
1345         * config/m68k/m68k.c (m68k_option_override): Disable
1346         -flive-range-shrinkage for classic m68k.
1347         (m68k_override_options_after_change): Likewise.
1349 2014-02-27  Marek Polacek  <polacek@redhat.com>
1351         PR middle-end/59223
1352         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1353         -Wmaybe-uninitialized.
1355 2014-02-27  Alan Modra  <amodra@gmail.com>
1357         PR target/57936
1358         * reload1.c (emit_input_reload_insns): When reload_override_in,
1359         set old to rl->in_reg when rl->in_reg is a subreg.
1361 2014-02-26  Richard Biener  <rguenther@suse.de>
1363         PR bootstrap/60343
1364         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1366 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1368         * common/config/i386/predicates.md (const1256_operand): Remove.
1369         (const2356_operand): New.
1370         (const_1_to_2_operand): Remove.
1371         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1372         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1373         (*avx512pf_gatherpf<mode>sf): Ditto.
1374         (avx512pf_gatherpf<mode>df): Ditto.
1375         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1376         (*avx512pf_gatherpf<mode>df): Ditto.
1377         (avx512pf_scatterpf<mode>sf): Ditto.
1378         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1379         (*avx512pf_scatterpf<mode>sf): Ditto.
1380         (avx512pf_scatterpf<mode>df): Ditto.
1381         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1382         (*avx512pf_scatterpf<mode>df): Ditto.
1383         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1385 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1387         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1388         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1389         (_mm512_mask_testn_epi64_mask): Move to ...
1390         * config/i386/avx512cdintrin.h: Here.
1391         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1392         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1393         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1394         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1395         TARGET_AVX512F from TARGET_AVX512CD.
1397 2014-02-26  Richard Biener  <rguenther@suse.de>
1399         PR ipa/60327
1400         * ipa.c (walk_polymorphic_call_targets): Properly guard
1401         call to inline_update_overall_summary.
1403 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1405         PR target/60280
1406         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1407         and latches only if requested.  Fix latch if it is removed.
1408         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1409         LOOPS_HAVE_PREHEADERS.
1411 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1413         * builtins.c (expand_builtin_thread_pointer): Create a new target
1414         when the target is NULL.
1416 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1418         PR rtl-optimization/60317
1419         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1420         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1421         * lra-assigns.c: Include params.h.
1422         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1423         other reload pseudos considerations.
1425 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1427         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1428         to use canonical form for nor<mode>3.
1430 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1432         PR target/55426
1433         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1434         conversions.
1436 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1438         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1439         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1440         (ix86_handle_option): Handle OPT_mprefetchwt1.
1441         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1442         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1443         PREFETCHWT1 CPUID.
1444         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1445         OPTION_MASK_ISA_PREFETCHWT1.
1446         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1447         (PTA_PREFETCHWT1): New.
1448         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1449         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1450         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1451         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1452         (*prefetch_avx512pf_<mode>_: Change into ...
1453         (*prefetch_prefetchwt1_<mode>: This.
1454         * config/i386/i386.opt (mprefetchwt1): New.
1455         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1456         (_mm_prefetch): Handle intent to write.
1457         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1459 2014-02-25  Richard Biener  <rguenther@suse.de>
1461         PR middle-end/60291
1462         * emit-rtl.c (mem_attrs_htab): Remove.
1463         (mem_attrs_htab_hash): Likewise.
1464         (mem_attrs_htab_eq): Likewise.
1465         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1466         (init_emit_once): Do not allocate mem_attrs_htab.
1468 2014-02-25  Richard Biener  <rguenther@suse.de>
1470         PR lto/60319
1471         * lto-opts.c (lto_write_options): Output non-explicit conservative
1472         -fwrapv, -fno-trapv and -fno-strict-overflow.
1473         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1474         (run_gcc): And pass them through.
1476 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1478         * sel-sched.c (calculate_new_fences): New parameter ptime.
1479         Calculate it as a maximum over all fence cycles.
1480         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1481         Print the final schedule timing when sched_verbose.
1483 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1485         PR rtl-optimization/60292
1486         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1487         bit fot the fence instruction.
1489 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1491         * calls.h: Fix typo in comment.
1493 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1495         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1496         adjusting offsetable addresses.
1498 2014-02-24  Guozhi Wei  <carrot@google.com>
1500         * sparseset.h (sparseset_pop): Fix the wrong index.
1502 2014-02-24  Walter Lee  <walt@tilera.com>
1504         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1505         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1506         triplet.
1507         * common/config/tilegx/tilegx-common.c
1508         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1509         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1510         (LINK_SPEC): Ditto.
1511         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1512         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1513         (tilegx_gimplify_va_arg_expr): Handle big endian.
1514         (tilegx_expand_unaligned_load): Ditto.
1515         (tilegx_expand_unaligned_store): Ditto.
1516         (TARGET_RETURN_IN_MSB): New.
1517         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1518         (TARGET_ENDIAN_DEFAULT): New.
1519         (TARGET_BIG_ENDIAN): Handle big endian.
1520         (BYTES_BIG_ENDIAN): Ditto.
1521         (WORDS_BIG_ENDIAN): Ditto.
1522         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1523         (ENDIAN_SPEC): New.
1524         (EXTRA_SPECS): New.
1525         * config/tilegx/tilegx.md (extv): Handle big endian.
1526         (extzv): Ditto.
1527         (insn_st<n>): Ditto.
1528         (insn_st<n>_add<bitsuffix>): Ditto.
1529         (insn_stnt<n>): Ditto.
1530         (insn_stnt<n>_add<bitsuffix>):Ditto.
1531         (vec_interleave_highv8qi): Handle big endian.
1532         (vec_interleave_highv8qi_be): New.
1533         (vec_interleave_highv8qi_le): New.
1534         (insn_v1int_h): Handle big endian.
1535         (vec_interleave_lowv8qi): Handle big endian.
1536         (vec_interleave_lowv8qi_be): New.
1537         (vec_interleave_lowv8qi_le): New.
1538         (insn_v1int_l): Handle big endian.
1539         (vec_interleave_highv4hi): Handle big endian.
1540         (vec_interleave_highv4hi_be): New.
1541         (vec_interleave_highv4hi_le): New.
1542         (insn_v2int_h): Handle big endian.
1543         (vec_interleave_lowv4hi): Handle big endian.
1544         (vec_interleave_lowv4hi_be): New.
1545         (vec_interleave_lowv4hi_le): New.
1546         (insn_v2int_l): Handle big endian.
1547         (vec_interleave_highv2si): Handle big endian.
1548         (vec_interleave_highv2si_be): New.
1549         (vec_interleave_highv2si_le): New.
1550         (insn_v4int_h): Handle big endian.
1551         (vec_interleave_lowv2si): Handle big endian.
1552         (vec_interleave_lowv2si_be): New.
1553         (vec_interleave_lowv2si_le): New.
1554         (insn_v4int_l): Handle big endian.
1555         * config/tilegx/tilegx.opt (mbig-endian): New option.
1556         (mlittle-endian): New option.
1557         * doc/install.texi: Document tilegxbe-linux.
1558         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1560 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1562         PR ipa/60266
1563         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1564         there are no parameter descriptors.
1566 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1568         PR rtl-optimization/60268
1569         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1570         initialization to ...
1571         (sched_rgn_init): ... here.
1572         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1574 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1576         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1577         names.
1579 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1581         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1582         definition.
1584 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1586         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1587         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1589 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1591         * config/microblaze/predicates.md: Add cmp_op predicate.
1592         * config/microblaze/microblaze.md: Add branch_compare instruction
1593         which uses cmp_op predicate and emits cmp insn before branch.
1594         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1595         to microblaze_expand_conditional_branch and consolidate logic.
1596         (microblaze_expand_conditional_branch): emit branch_compare
1597         insn instead of handling cmp op separate from branch insn.
1599 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1601         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1602         to permit subregs.
1604 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1606         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1607         define_insn with define_expand and new define_insn
1608         *altivec_lve<VI_char>x_internal.
1609         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1610         and new define_insn *altivec_stve<VI_char>x_internal.
1611         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1612         prototype.
1613         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1614         lve*x built-ins.
1615         (altivec_expand_stvex_be): New function.
1617 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1619         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1620         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1621         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1622         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1624 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1626         PR target/60298
1627         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1628         instead of emit_move_insn.
1630 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1632         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1633         vspltw with vsldoi.
1634         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1635         gen_altivec_vsumsws.
1637 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1639         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1640         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1641         (altivec_lvxl_<mode>): New define_expand incorporating
1642         -maltivec=be semantics where needed.
1643         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1644         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1645         semantics where needed.
1646         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1647         (altivec_stvx_<mode>): New define_expand incorporating
1648         -maltivec=be semantics where needed.
1649         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1650         VM2 iterator instead of V4SI.
1651         (altivec_stvxl_<mode>): New define_expand incorporating
1652         -maltivec=be semantics where needed.
1653         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1654         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1655         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1656         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1657         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1658         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1659         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1660         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1661         ALTIVEC_BUILTIN_STVXL.
1662         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1663         (altivec_expand_stvx_be): Likewise.
1664         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1665         (altivec_expand_lvx_be): Likewise.
1666         (altivec_expand_stvx_be): Likewise.
1667         (altivec_expand_builtin): Add cases for
1668         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1669         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1670         (altivec_init_builtins): Add definitions for
1671         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1672         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1674 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1676         * doc/invoke.texi (mvirt, mno-virt): Document.
1677         * config/mips/mips.opt (mvirt): New option.
1678         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1680 2014-02-21  Richard Biener  <rguenther@suse.de>
1682         PR tree-optimization/60276
1683         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1684         (STMT_VINFO_MIN_NEG_DIST): New macro.
1685         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1686         STMT_VINFO_MIN_NEG_DIST.
1687         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1688         made for negative dependence distances still hold.
1690 2014-02-21  Richard Biener  <rguenther@suse.de>
1692         PR middle-end/60291
1693         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1694         DECL_INITIAL for globals not in the current function context.
1696 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1698         PR tree-optimization/56490
1699         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1700         * tree-ssa-uninit.c: Include params.h.
1701         (compute_control_dep_chain): Add num_calls argument, return false
1702         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1703         num_calls to recursive call.
1704         (find_predicates): Change dep_chain into normal array,
1705         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1706         variable and adjust compute_control_dep_chain caller.
1707         (find_def_preds): Likewise.
1709 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1711         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1712         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1714 2014-02-21  Nick Clifton  <nickc@redhat.com>
1716         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1717         (pushhi1): Likewise.
1718         (popqi1): Add mode to pre_dec.
1719         (pophi1): Likewise.
1721 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1723         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1724         mode for mask of V8SFmode permutation.
1726 2014-02-20  Richard Henderson  <rth@redhat.com>
1728         PR c++/60272
1729         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1730         a new pseudo for OLDVAL.
1732 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1734         PR target/57896
1735         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1736         gen_reg_rtx if d->testing_p.
1737         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1738         if d->testing_p and we will certainly return true.
1739         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1740         if d->testing_p.
1742 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1744         * emit-rtl.c (gen_reg_rtx): Assert that
1745         crtl->emit.regno_pointer_align_length is non-zero.
1747 2014-02-20  Richard Henderson  <rth@redhat.com>
1749         PR c++/60272
1750         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1751         on failure the store back into EXPECT.
1753 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1754             Sandra Loosemore  <sandra@codesourcery.com>
1756         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1757         * config/nios2/nios2.c (nios2_function_profiler): Add
1758         -fPIC (flag_pic == 2) support.
1759         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1760         (nios2_large_offset_p): New function.
1761         (nios2_unspec_reloc_p): Move up position, update to use
1762         nios2_large_offset_p.
1763         (nios2_unspec_address): Remove function.
1764         (nios2_unspec_offset): New function.
1765         (nios2_large_got_address): New function.
1766         (nios2_got_address): Add large offset support.
1767         (nios2_legitimize_tls_address): Update usage of removed and new
1768         functions.
1769         (nios2_symbol_binds_local_p): New function.
1770         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1771         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1772         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1773         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1774         processing for (const (unspec ...)).
1775         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1777 2014-02-20  Richard Biener  <rguenther@suse.de>
1779         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1780         doing the substitution.
1781         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1783 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1785         PR ipa/55260
1786         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1787         info when checking whether lattices are bottom.
1789 2014-02-20  Richard Biener  <rguenther@suse.de>
1791         PR middle-end/60221
1792         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1793         regions at -O0.
1795 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1797         PR ipa/58555
1798         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1799         parameter specifying the scaling.
1800         (inline_call): Update.
1801         (want_inline_recursively): Guard division by zero.
1802         (recursive_inlining): Update.
1803         * ipa-inline.h (clone_inlined_nodes): Update.
1805 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1807         PR target/60204
1808         * config/i386/i386.c (classify_argument): Pass structures of size
1809         64 bytes or less in register.
1811 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1812             Kirill Yukhin  <kirill.yukhin@intel.com>
1814         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1815         (_mm_rcp28_round_ss): Ditto.
1816         (_mm_rsqrt28_round_sd): Ditto.
1817         (_mm_rsqrt28_round_ss): Ditto.
1818         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1819         (_mm_rcp14_round_ss): Ditto.
1820         (_mm_rsqrt14_round_sd): Ditto.
1821         (_mm_rsqrt14_round_ss): Ditto.
1822         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1823         the first input operand, get rid of match_dup.
1824         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1825         attribute to sse.
1826         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1827         Ditto.
1828         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1829         operand as the first input operand, set type attribute.
1830         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1831         Set type attribute.
1832         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1833         operand as the first input operand, set type attribute.
1835 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1837         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1838         bit of zero.
1840 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1842         PR target/60207
1843         * config/i386/i386.c (construct_container): Remove TFmode check
1844         for X86_64_INTEGER_CLASS.
1846 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1848         PR target/59794
1849         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1850         only when -Wpsabi is enabled.
1852 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1854          PR target/59799
1855         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1856         passing arrays in registers are the same as for structs, so remove the
1857         special case for them.
1859 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1861         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1862         destination type, extract only the valid bits if the source type is not
1863         integral and has a different mode.
1865 2014-02-19  Richard Biener  <rguenther@suse.de>
1867         PR ipa/60243
1868         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1869         for all calls.
1871 2014-02-19  Richard Biener  <rguenther@suse.de>
1873         PR ipa/60243
1874         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1875         (ipa_modify_call_arguments): Emit an argument load explicitely and
1876         preserve virtual SSA form there and for the replacement call.
1877         Do not update SSA form nor free dominance info.
1879 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1881         * ipa.c (function_and_variable_visibility): Also clear WEAK
1882         flag when disolving COMDAT_GROUP.
1884 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1886         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1887         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1888         not devirtualizing.
1889         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1890         do more sanity checks.
1891         (detect_type_change): Return true when giving up early.
1892         (compute_complex_assign_jump_func): Fix type parameter of
1893         ipa_set_ancestor_jf.
1894         (compute_complex_ancestor_jump_func): Likewise.
1895         (update_jump_functions_after_inlining): Fix updating of
1896         ancestor function.
1897         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1899 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1901         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1902         inline clones when edge disappears.
1904 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1906         PR target/60203
1907         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1908         Split 64-bit moves into 2 patterns.  Do not allow the use of
1909         direct move for TDmode in little endian, since the decimal value
1910         has little endian bytes within a word, but the 64-bit pieces are
1911         ordered in a big endian fashion, and normal subreg's of TDmode are
1912         not allowed.
1913         (mov<mode>_64bit_dm): Likewise.
1914         (movtd_64bit_nodm): Likewise.
1916 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1918         PR tree-optimization/60174
1919         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1920         statement of an SSA_NAME that occurs in an abnormal PHI node.
1922 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1924         PR sanitizer/60142
1925         * final.c (SEEN_BB): Remove.
1926         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1927         (final_scan_insn): Don't force_source_line on second
1928         NOTE_INSN_BASIC_BLOCK.
1930 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1932         PR target/60205
1933         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1934         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1935         (type_natural_mode): Warn ABI change when %zmm register is not
1936         available for AVX512F vector value passing.
1938 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1940         PR target/60193
1941         * config/i386/i386.c (ix86_expand_prologue): Use value in
1942         rax register as displacement when restoring %r10 or %rax.
1943         Fix wrong offset when restoring both registers.
1945 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1947         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1948         assertion with conditional return.
1950 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1951             Uros Bizjak  <ubizjak@gmail.com>
1953         PR driver/60233
1954         * config/i386/driver-i386.c (host_detect_local_cpu): If
1955         YMM state is not saved by the OS, also clear has_f16c.  Move
1956         CPUID 0x80000001 handling before YMM state saving checking.
1958 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1960         PR rtl-optimization/58960
1961         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1962         factored out from ...
1963         (sched_init): ... here.
1964         (free_global_sched_pressure_data): New, factored out from ...
1965         (sched_finish): ... here.
1966         * sched-int.h (free_global_sched_pressure_data): Declare.
1967         * sched-rgn.c (nr_regions_initial): New static global.
1968         (haifa_find_rgns): Initialize it.
1969         (schedule_region): Disable sched-pressure for the newly
1970         generated regions.
1972 2014-02-17  Richard Biener  <rguenther@suse.de>
1974         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1975         release SSA defs of pattern stmts.
1977 2014-02-17  Richard Biener  <rguenther@suse.de>
1979         * tree-inline.c (expand_call_inline): Release the virtual
1980         operand defined by the call we are about to inline.
1982 2014-02-17  Richard Biener  <rguenther@suse.de>
1984         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1986 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1987             Ilya Tocar  <ilya.tocar@intel.com>
1989         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1990         arguments order in builtin.
1991         (_mm512_permutexvar_epi64): Ditto.
1992         (_mm512_mask_permutexvar_epi64): Ditto
1993         (_mm512_maskz_permutexvar_epi32): Ditto
1994         (_mm512_permutexvar_epi32): Ditto
1995         (_mm512_mask_permutexvar_epi32): Ditto
1997 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1999         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2000         (p8_vmrgow): Likewise.
2002 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2004         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2005         endian targets.
2007 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2009         PR target/60203
2010         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2011         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2012         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2013         using direct move instructions on ISA 2.07.  Also adjust
2014         instruction length for 64-bit.
2015         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2016         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2018 2014-02-15  Alan Modra  <amodra@gmail.com>
2020         PR target/58675
2021         PR target/57935
2022         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2023         find_replacement on parts of insn rtl that might be reloaded.
2025 2014-02-15  Richard Biener  <rguenther@suse.de>
2027         PR tree-optimization/60183
2028         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2029         (tree_ssa_phiprop): Calculate and free post-dominators.
2031 2014-02-14  Jeff Law  <law@redhat.com>
2033         PR rtl-optimization/60131
2034         * ree.c (get_extended_src_reg): New function.
2035         (combine_reaching_defs): Use it rather than assuming location of REG.
2036         (find_and_remove_re): Verify first operand of extension is
2037         a REG before adding the insns to the copy list.
2039 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2041         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2042         * configure: Regenerated.
2043         * config.in: Regenerated.
2044         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2045         instead of ASM_SHORT.
2047 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2048             Richard Earnshaw  <rearnsha@arm.com>
2050         PR rtl-optimization/59535
2051         * lra-constraints.c (process_alt_operands): Encourage alternative
2052         when unassigned pseudo class is superset of the alternative class.
2053         (inherit_reload_reg): Don't inherit when optimizing for code size.
2054         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2055         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2056         modes not less than 4 for Thumb1.
2058 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2060         PR pch/60010
2061         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2063 2014-02-14  Richard Biener  <rguenther@suse.de>
2065         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2066         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2067         Do not strip INDIRECT_REFs.
2069 2014-02-14  Richard Biener  <rguenther@suse.de>
2071         PR lto/60179
2072         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2073         DECL_FUNCTION_SPECIFIC_TARGET.
2074         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2075         * tree-streamer-out.c (pack_ts_target_option): Remove.
2076         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2077         (write_ts_function_decl_tree_pointers): Do not stream
2078         DECL_FUNCTION_SPECIFIC_TARGET.
2079         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2080         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2081         (lto_input_ts_function_decl_tree_pointers): Do not stream
2082         DECL_FUNCTION_SPECIFIC_TARGET.
2084 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2086         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2087         (get_initial_def_for_induction, vectorizable_induction): Ignore
2088         debug stmts when looking for exit_phi.
2089         (vectorizable_live_operation): Fix up condition.
2091 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2093         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2094         nreverse() because it changes the content of original tree list.
2096 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2098         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2099         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2101 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2103         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2104         GNU coding standards.
2106 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2108         PR debug/60152
2109         * dwarf2out.c (gen_subprogram_die): Don't call
2110         add_calling_convention_attribute if subr_die is old_die.
2112 2014-02-13  Sharad Singhai  <singhai@google.com>
2114         * doc/optinfo.texi: Fix order of nodes.
2116 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2118         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2119         operands[2], not operands[3].
2121 2014-02-13  Richard Biener  <rguenther@suse.de>
2123         PR bootstrap/59878
2124         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2125         mention the possibility of an in-tree build.
2126         (CLooG): Update recommended version to 0.18.1, mention the
2127         possibility of an in-tree build and clarify that the ISL
2128         bundled with CLooG does not work.
2130 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2132         PR target/43546
2133         * expr.c (compress_float_constant): If x is a hard register,
2134         extend into a pseudo and then move to x.
2136 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2138         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2139         caused by bad second argument to warning_at() with -mhotpatch and
2140         nested functions (e.g. with gfortran).
2142 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2144         * opts.c (option_name): Remove "enabled by default" rider.
2146 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2148         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2150 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2151             Uros Bizjak  <ubizjak@gmail.com>
2153         PR target/60151
2154         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2155         * configure: Regenerated.
2157 2014-02-12  Richard Biener  <rguenther@suse.de>
2159         * vec.c (vec_prefix::calculate_allocation): Move as
2160         inline variant to vec.h.
2161         (vec_prefix::calculate_allocation_1): New out-of-line version.
2162         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2163         (vec_prefix::m_has_auto_buf): Rename to ...
2164         (vec_prefix::m_using_auto_storage): ... this.
2165         (vec_prefix::calculate_allocation): Inline the easy cases
2166         and dispatch to calculate_allocation_1 which doesn't need the
2167         prefix address.
2168         (va_heap::reserve): Use gcc_checking_assert.
2169         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2170         m_using_auto_storage.
2171         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2172         member and adjust.
2173         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2174         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2175         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2177 2014-02-12  Richard Biener  <rguenther@suse.de>
2179         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2180         when we found a dependence.
2182 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2184         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2185         common code...
2186         (maybe_fold_stmt): ... into this new function.
2187         * omp-low.c (lower_omp): Update comment.
2189         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2190         last use.
2192         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2193         dereference.
2195 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2197         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2198         identifiers in comments.
2199         (cortexa53_extra_costs): Likewise.
2200         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2201         (cortexa7_extra_costs): Likewise.
2202         (cortexa12_extra_costs): Likewise.
2203         (cortexa15_extra_costs): Likewise.
2204         (v7m_extra_costs): Likewise.
2206 2014-02-12  Richard Biener  <rguenther@suse.de>
2208         PR middle-end/60092
2209         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2210         of posix_memalign being successful.
2211         (lower_stmt): Restrict lowering of posix_memalign to when
2212         -ftree-bit-ccp is enabled.
2214 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2216         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2217         arg_loc.
2218         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2220 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2222         PR rtl-optimization/60116
2223         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2224         other_insn once the combination has been validated.
2226 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2228         PR lto/59468
2229         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2230         and wrapper.
2231         * ipa-devirt.c: Include demangle.h
2232         (odr_violation_reported): New static variable.
2233         (add_type_duplicate): Update odr_violations.
2234         (maybe_record_node): Add completep parameter; update it.
2235         (record_target_from_binfo): Add COMPLETEP parameter;
2236         update it as needed.
2237         (possible_polymorphic_call_targets_1): Likewise.
2238         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2239         rename FINAL to COMPLETE.
2240         (record_targets_from_bases): Sanity check we found the binfo;
2241         fix COMPLETEP updating.
2242         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2243         parameter, fix computing of COMPLETEP.
2244         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2245         at LTO time do demangling.
2246         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2247         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2248         parameter.
2249         (gimple_get_virt_method_for_binfo): Likewise.
2250         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2251         gimple_get_virt_method_for_vtable): Update prototypes.
2253 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2255         PR target/49008
2256         * genautomata.c (add_presence_absence): Fix typo with
2257         {final_}presence_list.
2259 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2261         PR target/60137
2262         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2263         for VSX/Altivec vectors that land in GPR registers.
2265 2014-02-11  Richard Henderson  <rth@redhat.com>
2266             Jakub Jelinek  <jakub@redhat.com>
2268         PR debug/59776
2269         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2270         around drhs if type conversion to lacc->type is not useless.
2272 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2274         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2275         tuning struct.
2276         (cortex-a57.cortex-a53): Likewise.
2277         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2279 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2281         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2282         arm_restrict_it.
2284 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2286         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2287         add_options_for_arm_vfp3.
2289 2014-02-11  Jeff Law  <law@redhat.com>
2291         PR middle-end/54041
2292         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2293         object with an undesirable mode.
2295 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2297         PR libgomp/60107
2298         * config/i386/sol2-9.h: New file.
2299         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2300         *-*-solaris2.9*): Use it.
2302 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2304         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2305         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2307 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2309         * config/microblaze/microblaze.c: Extend mcpu version format
2311 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
2313         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2315 2014-02-10  Richard Henderson  <rth@redhat.com>
2317         PR target/59927
2318         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2319         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2320         ms-abi vs -mno-accumulate-outgoing-args.
2321         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2322         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2323         respect to ms-abi.
2325 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2327         PR middle-end/60080
2328         * cfgexpand.c (expand_asm_operands): Attach source location to
2329         ASM_INPUT rtx objects.
2330         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2332 2014-02-10  Nick Clifton  <nickc@redhat.com>
2334         * config/mn10300/mn10300.c (popcount): New function.
2335         (mn10300_expand_prologue): Include saved registers in stack usage
2336         count.
2338 2014-02-10  Jeff Law  <law@redhat.com>
2340         PR middle-end/52306
2341         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2342         when changing the SET_DEST of a prior insn to avoid an input reload.
2344 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2346         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2347         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2348         -mcall-openbsd, or -mcall-linux.
2349         (CC1_ENDIAN_BIG_SPEC): Remove.
2350         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2351         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2352         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2353         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2354         and %cc1_endian_default.
2355         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2357 2014-02-10  Richard Biener  <rguenther@suse.de>
2359         PR tree-optimization/60115
2360         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2361         MEM_REF handling.  Properly verify that the accesses are not
2362         out of the objects bound.
2364 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2366         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2367         coretex to cortex.
2369 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2371         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2372         proper constants and fix formatting.
2373         (possible_polymorphic_call_targets): Fix formatting.
2375 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2376             Ilya Tocar  <ilya.tocar@intel.com>
2378         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2379         (_mm512_loadu_epi32): Renamed into...
2380         (_mm512_loadu_si512): This.
2381         (_mm512_storeu_epi32): Renamed into...
2382         (_mm512_storeu_si512): This.
2383         (_mm512_maskz_ceil_ps): Removed.
2384         (_mm512_maskz_ceil_pd): Ditto.
2385         (_mm512_maskz_floor_ps): Ditto.
2386         (_mm512_maskz_floor_pd): Ditto.
2387         (_mm512_floor_round_ps): Ditto.
2388         (_mm512_floor_round_pd): Ditto.
2389         (_mm512_ceil_round_ps): Ditto.
2390         (_mm512_ceil_round_pd): Ditto.
2391         (_mm512_mask_floor_round_ps): Ditto.
2392         (_mm512_mask_floor_round_pd): Ditto.
2393         (_mm512_mask_ceil_round_ps): Ditto.
2394         (_mm512_mask_ceil_round_pd): Ditto.
2395         (_mm512_maskz_floor_round_ps): Ditto.
2396         (_mm512_maskz_floor_round_pd): Ditto.
2397         (_mm512_maskz_ceil_round_ps): Ditto.
2398         (_mm512_maskz_ceil_round_pd): Ditto.
2399         (_mm512_expand_pd): Ditto.
2400         (_mm512_expand_ps): Ditto.
2401         * config/i386/i386.c (ix86_builtins): Remove
2402         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2403         (bdesc_args): Ditto.
2404         * config/i386/predicates.md (const1256_operand): New.
2405         (const_1_to_2_operand): Ditto.
2406         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2407         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2408         (*avx512pf_gatherpf<mode>sf): Ditto.
2409         (avx512pf_gatherpf<mode>df): Ditto.
2410         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2411         (*avx512pf_gatherpf<mode>df): Ditto.
2412         (avx512pf_scatterpf<mode>sf): Ditto.
2413         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2414         (*avx512pf_scatterpf<mode>sf): Ditto.
2415         (avx512pf_scatterpf<mode>df): Ditto.
2416         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2417         (*avx512pf_scatterpf<mode>df): Ditto.
2418         (avx512f_expand<mode>): Removed.
2419         (<shift_insn><mode>3<mask_name>): Change predicate type.
2421 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2423         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2424         not at the end of datarefs vector use ordered_remove to avoid
2425         reordering datarefs vector.
2427         PR c/59984
2428         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2429         mark local addressable non-static vars as GOVD_PRIVATE
2430         instead of GOVD_LOCAL.
2431         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2432         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2433         than copying them.
2435         PR middle-end/60092
2436         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2437         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2438         assume_aligned or alloc_align attributes.
2439         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2440         arguments.  Handle also assume_aligned and alloc_align attributes.
2441         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2442         calls to functions with assume_aligned or alloc_align attributes.
2443         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2445 2014-02-08  Terry Guo  <terry.guo@arm.com>
2447         * doc/invoke.texi: Document ARM -march=armv7e-m.
2449 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2451         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2452         flag on __cilkrts_rethrow builtin.
2454         PR ipa/60026
2455         * ipa-cp.c (determine_versionability): Fail at -O0
2456         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2457         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2459         Revert:
2460         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2462         PR ipa/60026
2463         * tree-inline.c (copy_forbidden): Fail for
2464         __attribute__((optimize (0))) functions.
2466 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2468         * varpool.c: Include pointer-set.h.
2469         (varpool_remove_unreferenced_decls): Variables in other partitions
2470         will not be output; be however careful to not lose information
2471         about partitioning.
2473 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2475         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2476         lookup in the vtable constructor.
2478 2014-02-07  Jeff Law  <law@redhat.com>
2480         PR target/40977
2481         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2482         define_insn_and_split.
2484         * ipa-inline.c (inline_small_functions): Fix typos.
2486 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2488         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2489         (s390_can_use_return_insn): Declare.
2490         * config/s390/s390.h (EPILOGUE_USES): Define.
2491         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2492         instructions.
2493         (s390_chunkify_start): Handle return JUMP_LABELs.
2494         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2495         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2496         (s390_can_use_return_insn): New functions.
2497         (s390_fix_long_loop_prediction): Handle conditional returns.
2498         (TARGET_SET_UP_BY_PROLOGUE): Define.
2499         * config/s390/s390.md (ANY_RETURN): New code iterator.
2500         (*creturn, *csimple_return, return, simple_return): New patterns.
2502 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2504         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2505         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2506         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2507         REG_CFA_RESTORE list when deciding not to restore a register.
2509 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2511         * config/s390/s390.c: Include tree-pass.h and context.h.
2512         (s390_early_mach): New function, split out from...
2513         (s390_emit_prologue): ...here.
2514         (pass_data_s390_early_mach): New pass structure.
2515         (pass_s390_early_mach): New class.
2516         (s390_option_override): Create and register early_mach pass.
2517         Move to end of file.
2519 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2521         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2522         to match for the exit block.
2524 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2526         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2527         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2528         Reject misaligned operands.
2530 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2532         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2534 2014-02-07  Richard Biener  <rguenther@suse.de>
2536         PR middle-end/60092
2537         * gimple-low.c (lower_builtin_posix_memalign): New function.
2538         (lower_stmt): Call it to lower posix_memalign in a way
2539         to make alignment info accessible.
2541 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2543         PR c++/60082
2544         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2545         __builtin_setjmp_receiver.
2547 2014-02-07  Richard Biener  <rguenther@suse.de>
2549         PR middle-end/60092
2550         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2551         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2552         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2553         Handle BUILT_IN_POSIX_MEMALIGN.
2554         (find_func_clobbers): Likewise.
2555         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2556         (call_may_clobber_ref_p_1): Likewise.
2558 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2560         PR ipa/59918
2561         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2562         sanity check.
2564 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2566         PR ipa/59469
2567         * lto-cgraph.c (lto_output_node): Use
2568         symtab_get_symbol_partitioning_class.
2569         (lto_output_varpool_node): likewise.
2570         (symtab_get_symbol_partitioning_class): Move here from
2571         lto/lto-partition.c
2572         * cgraph.h (symbol_partitioning_class): Likewise.
2573         (symtab_get_symbol_partitioning_class): Declare.
2575 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2577         * ggc.h (ggc_internal_cleared_alloc): New macro.
2578         * vec.h (vec_safe_copy): Handle memory stats.
2579         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2580         * target-globals.c (save_target_globals): Likewise.
2582 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2584         PR target/60077
2585         * expr.c (emit_move_resolve_push): Export; be bit more selective
2586         on when to clear alias set.
2587         * expr.h (emit_move_resolve_push): Declare.
2588         * function.h (struct function): Add tail_call_marked.
2589         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2590         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2591         * config/i386/i386.md (TImode move expander): De not call
2592         ix86_expand_push.
2593         (FP push expanders): Preserve memory attributes.
2594         * config/i386/sse.md (push<mode>1): Remove.
2595         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2596         (ix86_expand_push): Remove.
2597         * config/i386/mmx.md (push<mode>1): Remove.
2599 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2601         PR rtl-optimization/60030
2602         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2603         lopart with paradoxical subreg before shifting it up by hprec.
2605 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2607         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2608         Remove extra newline at end of file.
2609         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2610         (arm_issue_rate): Handle cortexa57.
2611         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2612         (cortex-a57.cortex-a53): Likewise.
2614 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2616         PR target/59575
2617         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2618         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2619         bitmask.
2620         (arm_expand_prologue): Adjust all callers.
2621         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2622         info, registers also at the lowest numbered registers side.  Use
2623         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2624         XEXP.
2626         PR debug/59992
2627         * var-tracking.c (adjust_mems): Before adding a SET to
2628         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2630 2014-02-06  Alan Modra  <amodra@gmail.com>
2632         PR target/60032
2633         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2634         change SDmode to DDmode when lra_in_progress.
2636 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2638         PR middle-end/59150
2639         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2640         free_data_ref on the dr first, and before goto again also set dr
2641         to the next dr.  For simd_lane_access, free old datarefs[i] before
2642         overwriting it.  For get_vectype_for_scalar_type failure, don't
2643         free_data_ref if simd_lane_access.
2645         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2647         PR target/60062
2648         * tree.h (opts_for_fn): New inline function.
2649         (opt_for_fn): Define.
2650         * config/i386/i386.c (ix86_function_regparm): Use
2651         opt_for_fn (decl, optimize) instead of optimize.
2653 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2655         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2656         for SYMBOL_REF in large memory model.
2658 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2660         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2661         and crypto support.
2662         (cortex-a57): Likewise.
2663         (cortex-a57.cortex-a53): Likewise.
2665 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2666             Kugan Vivekanandarajah  <kuganv@linaro.org>
2668         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2669         unaligned_access.
2670         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2672 2014-02-06  Richard Biener  <rguenther@suse.de>
2674         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2675         set_loop_copy and initialize_original_copy_tables.
2677 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2679         * config/aarch64/aarch64-simd.md
2680         (aarch64_ashr_simddi): Change QI to SI.
2682 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2683             Jakub Jelinek  <jakub@redhat.com>
2685         PR middle-end/60013
2686         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2687         of the dataflow.
2689 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2691         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2692         CODE_FOR_altivec_vpku[hw]um to
2693         CODE_FOR_altivec_vpku[hw]um_direct.
2694         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2695         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2696         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2697         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2699 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2701         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2702         generation for -maltivec=be.
2703         (altivec_vsumsws): Simplify redundant test.
2705 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2707         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2708         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2709         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2710         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2711         gen_altivec_vpkuwum.
2712         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2713         BYTES_BIG_ENDIAN.
2714         (altivec_vpks<VI_char>ss): Likewise.
2715         (altivec_vpks<VI_char>us): Likewise.
2716         (altivec_vpku<VI_char>us): Likewise.
2717         (altivec_vpku<VI_char>um): Likewise.
2718         (altivec_vpku<VI_char>um_direct): New (copy of
2719         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2720         internal use).
2721         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2722         target is little endian and -maltivec=be is not specified.
2723         (*altivec_vupkhs<VU_char>_direct): New (copy of
2724         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2725         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2726         target is little endian and -maltivec=be is not specified.
2727         (*altivec_vupkls<VU_char>_direct): New (copy of
2728         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2729         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2730         little endian and -maltivec=be is not specified.
2731         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2732         little endian and -maltivec=be is not specified.
2734 2014-02-05  Richard Henderson  <rth@redhat.com>
2736         PR debug/52727
2737         * combine-stack-adj.c: Revert r206943.
2738         * sched-int.h (struct deps_desc): Add last_args_size.
2739         * sched-deps.c (init_deps): Initialize it.
2740         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2741         contain REG_ARGS_SIZE notes.
2743 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2745         * lto-cgraph.c (asm_nodes_output): Make global.
2746         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2747         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2748         (driver_handle_option): Handle OPT_fwpa.
2750 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2752         PR ipa/59947
2753         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2754         a comment typo and formatting issue.  If odr_hash hasn't been
2755         created, return vNULL and set *completep to false.
2757         PR middle-end/57499
2758         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2759         bb with no successors.
2761 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2763         PR target/59718
2764         * doc/invoke.texi (-march): Clarify documentation for ARM.
2765         (-mtune): Likewise.
2766         (-mcpu): Likewise.
2768 2014-02-05  Richard Biener  <rguenther@suse.de>
2770         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2771         when not vectorizing because of too many alias checks.
2772         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2773         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2775 2014-02-05  Nick Clifton  <nickc@redhat.com>
2777         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2778         accept extended registers in any mode when compiling for the MN10300.
2780 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2782         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2783         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2784         sanitization attributes.
2785         (can_inline_edge_p): Likewise.
2786         (sanitize_attrs_match_for_inline_p): New function.
2788 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2790         * ipa-prop.c (detect_type_change): Shor circuit testing of
2791         type changes on THIS pointer.
2793 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2795         PR target/59777
2796         * config/pa/pa.c (legitimize_tls_address): Return original address
2797         if not passed a SYMBOL_REF rtx.
2798         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2799         addresses.
2800         (pa_emit_move_sequence): Simplify TLS source operands.
2801         (pa_legitimate_constant_p): Reject all TLS constants.
2802         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2803         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2805 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2807         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2808         groups when we know they are controlled by LTO.
2809         * varasm.c (default_binds_local_p_1): If object is in other partition,
2810         it will be resolved locally.
2812 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2814         * config/host-linux.c (linux_gt_pch_use_address): Don't
2815         use SSIZE_MAX because it is not always defined.
2817 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2819         PR bootstrap/59913
2820         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2821         threshold for pseudo splitting.
2822         (update_ebb_live_info): Process call argument hard registers and
2823         hard registers from insn definition too.
2824         (max_small_class_regs_num): New constant.
2825         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2826         reloads_num only for small register classes.  Don't split for
2827         outputs of jumps.
2829 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2831         PR ipa/60058
2832         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2833         is non-null.
2835 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2837         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2838         visibility is safe.
2840 2014-02-04  Marek Polacek  <polacek@redhat.com>
2842         * gdbinit.in (pel): Define.
2844 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2846         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2847         behavior.
2849 2014-02-04  Richard Biener  <rguenther@suse.de>
2851         PR lto/59723
2852         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2853         in function context local.
2854         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2855         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2856         similar to LTO_imported_decl_ref.
2858 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2860         PR tree-optimization/60002
2861         * cgraphclones.c (build_function_decl_skip_args): Clear
2862         DECL_LANG_SPECIFIC.
2864         PR tree-optimization/60023
2865         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2866         false to gsi_replace.
2867         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2868         has been in some EH region and vec_stmt could throw, add
2869         vec_stmt into the same EH region.
2870         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2871         has no lhs, ignore it.
2872         * internal-fn.c (expand_MASK_LOAD): Likewise.
2874         PR ipa/60026
2875         * tree-inline.c (copy_forbidden): Fail for
2876         __attribute__((optimize (0))) functions.
2878         PR other/58712
2879         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2880         is set, copy one less argument.
2881         (expand_simd_clones): Don't subtract clone_info->inbranch
2882         from simd_clone_struct_alloc argument.
2884         PR rtl-optimization/57915
2885         * recog.c (simplify_while_replacing): If all unary/binary/relational
2886         operation arguments are constant, attempt to simplify those.
2888         PR middle-end/59261
2889         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2890         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2892 2014-02-04  Richard Biener  <rguenther@suse.de>
2894         PR tree-optimization/60012
2895         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2896         TBAA disambiguation to all DDRs.
2898 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2900         PR target/59788
2901         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2902         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2904 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2906         PR ipa/59882
2907         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2909 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2911         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2912         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2914 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2916         PR ipa/59831
2917         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2918         to figure out targets of polymorphic calls with known decl.
2919         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2920         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2921         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2922         (get_polymorphic_call_info): ... here.
2923         (get_polymorphic_call_info_from_invariant): New function.
2925 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2927         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2928         lookup via vtable pointer; check for type consistency
2929         and turn inconsitent facts into UNREACHABLE.
2930         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2931         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2932         type inconsistent querries; return UNREACHABLE instead.
2934 2014-02-03  Richard Henderson  <rth@twiddle.net>
2936         PR tree-opt/59924
2937         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2938         already processed this node.
2939         (normalize_one_pred_1): Pass along mark_set.
2940         (normalize_one_pred): Create and destroy a pointer_set_t.
2941         (normalize_one_pred_chain): Likewise.
2943 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2945         PR gcov-profile/58602
2946         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2948 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2950         PR ipa/59831
2951         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2952         -fno-devirtualize; try to devirtualize by the knowledge of
2953         virtual table pointer given by aggregate propagation.
2954         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2955         (ipa_print_node_jump_functions): Dump also offset that
2956         is relevant for polymorphic calls.
2957         (determine_known_aggregate_parts): Add arg_type parameter; use it
2958         instead of determining the type from pointer type.
2959         (ipa_compute_jump_functions_for_edge): Update call of
2960         determine_known_aggregate_parts.
2961         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2962         (gimple_get_virt_method_for_binfo): ... here; simplify using
2963         vtable_pointer_value_to_vtable.
2964         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2965         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2966         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2967         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2968         POINTER_PLUS_EXPR.
2969         (vtable_pointer_value_to_binfo): ... here.
2970         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2972 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2974         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2975         redef of outer loop index variable.
2977 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2979         PR c++/53017
2980         PR c++/59211
2981         * doc/extend.texi (Function Attributes): Typo.
2983 2014-02-03  Cong Hou  <congh@google.com>
2985         PR tree-optimization/60000
2986         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2987         if the vectorized statement is a store.  A store statement can only
2988         appear at the end of pattern statements.
2990 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2992         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2993         (ix86_option_override_internal): Default long double to 64-bit for
2994         32-bit Bionic and to 128-bit for 64-bit Bionic.
2996         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2997         TARGET_LONG_DOUBLE_128 is true.
2998         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3000         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3001         (mlong-double-64): Negate -mlong-double-128.
3002         (mlong-double-128): New option.
3004         * config/i386/i386-c.c (ix86_target_macros): Define
3005         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3007         * doc/invoke.texi: Document -mlong-double-128.
3009 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3011         PR rtl-optimization/60024
3012         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3014 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3016         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3018 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3020         PR rtl-optimization/57662
3021         * sel-sched.c (code_motion_path_driver): Do not mark already not
3022         existing blocks in the visiting bitmap.
3024 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3026         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3027         on the insn being emitted.
3029 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3030             Will Deacon  <will.deacon@arm.com>
3032         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3034 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3036         * config/arm/arm-tables.opt: Regenerate.
3038 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3040         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3041         for vector types other than V16QImode.
3042         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3043         define_expand, and call altivec_expand_vec_perm_le when producing
3044         code with little endian element order.
3045         (*altivec_vperm_<mode>_internal): New insn having previous
3046         behavior of altivec_vperm_<mode>.
3047         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3048         altivec_expand_vec_perm_le when producing code with little endian
3049         element order.
3050         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3051         behavior of altivec_vperm_<mode>_uns.
3053 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3055         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3056         (altivec_vsumsws): Add handling for -maltivec=be with a little
3057         endian target.
3058         (altivec_vsumsws_direct): New.
3059         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3060         gen_altivec_vsumsws.
3062 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3064         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3065         vtable_pointer_value_to_binfo): New functions.
3066         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3067         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3069 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3071         * config/nios2/nios2.md (load_got_register): Initialize GOT
3072         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3073         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3075 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3077         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3078         preserverd by passthrough, do not propagate the type.
3080 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3082         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3083         (mips_atomic_assign_expand_fenv): New function.
3084         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3086 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3088         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3089         (__builtin_mips_set_fcsr): Likewise.
3090         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3091         MIPS_USI_FTYPE_VOID.
3092         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3093         (mips16_expand_set_fcsr): Likewise.
3094         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3095         (mips16_set_fcsr_stub): Likewise.
3096         (mips16_get_fcsr_one_only_stub): New class.
3097         (mips16_set_fcsr_one_only_stub): Likewise.
3098         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3099         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3100         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3101         (hard_float): New availability predicate.
3102         (mips_builtins): Add get_fcsr and set_fcsr.
3103         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3104         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3105         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3106         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3107         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3108         patterns.
3110 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3112         * config/mips/mips.c (mips_one_only_stub): New class.
3113         (mips_need_mips16_rdhwr_p): Replace with...
3114         (mips16_rdhwr_stub): ...this new variable.
3115         (mips16_stub_call_address): New function.
3116         (mips16_rdhwr_one_only_stub): New class.
3117         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3118         (mips_output_mips16_rdhwr): Delete.
3119         (mips_finish_stub): New function.
3120         (mips_code_end): Use it to handle rdhwr stubs.
3122 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3124         PR target/60017
3125         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3126         when calculating size of integer atomic types.
3128 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3130         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3132 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3134         PR tree-optimization/60003
3135         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3136         * profile.c (branch_prob): Use gimple_call_builtin_p
3137         to check for BUILT_IN_SETJMP_RECEIVER.
3138         * tree-inline.c (copy_bb): Call notice_special_calls.
3140 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3142         PR bootstrap/59985
3143         * lra-constraints.c (process_alt_operands): Update reload_sum only
3144         on the first pass.
3146 2014-01-31  Richard Henderson  <rth@redhat.com>
3148         PR middle-end/60004
3149         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3150         until after else_eh is processed.
3152 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3154         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3155         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3156         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3157         in smmintrin.h, remove them.
3158         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3159         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3160         * config/i386/i386.md (ROUND_SAE): Fix value.
3161         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3162         (const48_operand): New.
3163         * config/i386/subst.md (round), (round_expand): Use
3164         const_4_or_8_to_11_operand.
3165         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3167 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3169         * config/i386/constraints.md (Yk): Swap meaning with k.
3170         * config/i386/i386.md (movhi_internal): Change Yk to k.
3171         (movqi_internal): Ditto.
3172         (*k<logic><mode>): Ditto.
3173         (*andhi_1): Ditto.
3174         (*andqi_1): Ditto.
3175         (kandn<mode>): Ditto.
3176         (*<code>hi_1): Ditto.
3177         (*<code>qi_1): Ditto.
3178         (kxnor<mode>): Ditto.
3179         (kortestzhi): Ditto.
3180         (kortestchi): Ditto.
3181         (kunpckhi): Ditto.
3182         (*one_cmplhi2_1): Ditto.
3183         (*one_cmplqi2_1): Ditto.
3184         * config/i386/sse.md (): Change k to Yk.
3185         (avx512f_load<mode>_mask): Ditto.
3186         (avx512f_blendm<mode>): Ditto.
3187         (avx512f_store<mode>_mask): Ditto.
3188         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3189         (avx512f_storedqu<mode>_mask): Ditto.
3190         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3191         Ditto.
3192         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3193         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3194         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3195         (avx512f_maskcmp<mode>3): Ditto.
3196         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3197         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3198         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3199         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3200         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3201         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3202         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3203         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3204         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3205         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3206         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3207         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3208         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3209         (vec_extract_lo_<mode>_maskm): Ditto.
3210         (vec_extract_hi_<mode>_maskm): Ditto.
3211         (avx512f_vternlog<mode>_mask): Ditto.
3212         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3213         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3214         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3215         (avx512f_<code>v8div16qi2_mask): Ditto.
3216         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3217         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3218         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3219         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3220         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3221         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3222         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3223         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3224         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3225         (avx512cd_maskb_vec_dupv8di): Ditto.
3226         (avx512cd_maskw_vec_dupv16si): Ditto.
3227         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3228         (avx512f_vpermi2var<mode>3_mask): Ditto.
3229         (avx512f_vpermi2var<mode>3_mask): Ditto.
3230         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3231         (*avx512f_gathersi<mode>): Ditto.
3232         (*avx512f_gathersi<mode>_2): Ditto.
3233         (*avx512f_gatherdi<mode>): Ditto.
3234         (*avx512f_gatherdi<mode>_2): Ditto.
3235         (*avx512f_scattersi<mode>): Ditto.
3236         (*avx512f_scatterdi<mode>): Ditto.
3237         (avx512f_compress<mode>_mask): Ditto.
3238         (avx512f_compressstore<mode>_mask): Ditto.
3239         (avx512f_expand<mode>_mask): Ditto.
3240         * config/i386/subst.md (mask): Change k to Yk.
3241         (mask_scalar_merge): Ditto.
3242         (sd): Ditto.
3244 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3246         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3248 2014-01-31  Richard Biener  <rguenther@suse.de>
3250         PR middle-end/59990
3251         * builtins.c (fold_builtin_memory_op): Make sure to not
3252         use a floating-point mode or a boolean or enumeral type for
3253         the copy operation.
3255 2014-01-30  DJ Delorie  <dj@redhat.com>
3257         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3258         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3259         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3260         whenever main() has an epilogue.
3262 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3264         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3265         unused variable "field".
3266         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3267         (vsx_mergeh_<mode>): Likewise.
3268         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3269         (altivec_vmrghh): Likewise.
3270         (altivec_vmrghw): Likewise.
3271         (altivec_vmrglb): Likewise.
3272         (altivec_vmrglh): Likewise.
3273         (altivec_vmrglw): Likewise.
3274         (altivec_vspltb): Add missing uses.
3275         (altivec_vsplth): Likewise.
3276         (altivec_vspltw): Likewise.
3277         (altivec_vspltsf): Likewise.
3279 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3281         PR target/59923
3282         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3283         frame related instructions.
3285 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3287         PR rtl-optimization/59959
3288         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3289         any reload of register whose subreg is invalid.
3291 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3293         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3294         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3295         Add missing return type - void.
3297 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3299         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3300         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3301         remove element index adjustment for endian (now handled in vsx.md
3302         and altivec.md).
3303         (altivec_expand_vec_perm_const): Use
3304         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3305         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3306         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3307         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3308         define_expand and a new define_insn *altivec_vspltb_internal;
3309         adjust for -maltivec=be on a little endian target.
3310         (altivec_vspltb_direct): New.
3311         (altivec_vsplth): Divide into a define_expand and a new
3312         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3313         little endian target.
3314         (altivec_vsplth_direct): New.
3315         (altivec_vspltw): Divide into a define_expand and a new
3316         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3317         little endian target.
3318         (altivec_vspltw_direct): New.
3319         (altivec_vspltsf): Divide into a define_expand and a new
3320         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3321         a little endian target.
3323 2014-01-30  Richard Biener  <rguenther@suse.de>
3325         PR tree-optimization/59993
3326         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3327         can propagate form the earlier stmt and avoid the transform
3328         when the intermediate result is needed.
3330 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3332         * README.Portability: Fix typo.
3334 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
3336         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3337         comparison_operator with ordered_comparison_operator.
3339 2014-01-30  Nick Clifton  <nickc@redhat.com>
3341         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3342         Rename to mn10300_store_multiple_regs.
3343         * config/mn10300/mn10300.c: Likewise.
3344         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3345         store_multiple_regs.
3346         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3347         Call mn10300_store_multiple_regs.
3349 2014-01-30  Nick Clifton  <nickc@redhat.com>
3350             DJ Delorie  <dj@redhat.com>
3352         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3353         %fp 2 to keep registers after it properly word-aligned.
3354         (rl78_alloc_physical_registers_umul): Handle the case where both
3355         input operands are the same.
3357 2014-01-30  Richard Biener  <rguenther@suse.de>
3359         PR tree-optimization/59903
3360         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3361         check properly.
3363 2014-01-30  Jason Merrill  <jason@redhat.com>
3365         PR c++/59633
3366         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3368         PR c++/59645
3369         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3371 2014-01-30  Richard Biener  <rguenther@suse.de>
3373         PR tree-optimization/59951
3374         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3376 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3378         PR target/59784
3379         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3380         SFmode to DFmode case.
3382 2014-01-29  DJ Delorie  <dj@redhat.com>
3384         * config/msp430/msp430.opt (-minrt): New.
3385         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3386         if -minrt given.
3387         (ENDFILE_SPEC): Likewise.
3389 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3391         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3392         (estimate_function_body_sizes): Use it.
3394 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3396         PR c++/58561
3397         * dwarf2out.c (is_cxx_auto): New.
3398         (is_base_type): Use it.
3399         (gen_type_die_with_usage): Likewise.
3401 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3403         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3404         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3405         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3406         -maltivec=be with LE targets.
3407         (vsx_mergeh_<mode>): Likewise.
3408         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3409         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3410         (altivec_vmrghb): Replace with define_expand and new
3411         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3412         (altivec_vmrghb_direct): New define_insn.
3413         (altivec_vmrghh): Replace with define_expand and new
3414         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3415         (altivec_vmrghh_direct): New define_insn.
3416         (altivec_vmrghw): Replace with define_expand and new
3417         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3418         (altivec_vmrghw_direct): New define_insn.
3419         (*altivec_vmrghsf): Adjust for endianness.
3420         (altivec_vmrglb): Replace with define_expand and new
3421         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3422         (altivec_vmrglb_direct): New define_insn.
3423         (altivec_vmrglh): Replace with define_expand and new
3424         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3425         (altivec_vmrglh_direct): New define_insn.
3426         (altivec_vmrglw): Replace with define_expand and new
3427         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3428         (altivec_vmrglw_direct): New define_insn.
3429         (*altivec_vmrglsf): Adjust for endianness.
3430         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3431         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3432         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3433         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3434         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3435         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3436         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3437         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3439 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3441         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3442         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3443         whitespace.
3445 2014-01-29  Richard Biener  <rguenther@suse.de>
3447         PR tree-optimization/58742
3448         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3449         associate_pointerplus_align.
3450         (associate_pointerplus_diff): New function.
3451         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3452         and associate_pointerplus_diff.
3454 2014-01-29  Richard Biener  <rguenther@suse.de>
3456         * lto-streamer.h (LTO_major_version): Bump to 3.
3457         (LTO_minor_version): Reset to 0.
3459 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3461         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3462         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3463         (arm_file_start): Generate correct asm header for armv7ve.
3464         * config/arm/bpabi.h: Add multilib support for armv7ve.
3465         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3466         and cortex-a15 to armv7ve.
3467         * config/arm/t-aprofile: Add multilib support for armv7ve.
3468         * doc/invoke.texi: Document -march=armv7ve.
3470 2014-01-29  Richard Biener  <rguenther@suse.de>
3472         PR tree-optimization/58742
3473         * tree-ssa-forwprop.c (associate_plusminus): Return true
3474         if we changed sth, defer EH cleanup to ...
3475         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3476         (simplify_mult): New function.
3478 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3480         PR middle-end/59917
3481         PR tree-optimization/59920
3482         * tree.c (build_common_builtin_nodes): Remove
3483         __builtin_setjmp_dispatcher initialization.
3484         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3485         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3486         instead of gsi_after_labels + manually skipping debug stmts.
3487         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3488         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3489         * tree-inline.c (copy_edges_for_bb): Remove
3490         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3491         argument.  Ignore computed_goto_p stmts.  Don't call
3492         make_abnormal_goto_edges.  If a call might need abnormal edges
3493         for non-local gotos, see if it already has an edge to
3494         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3495         with true argument, don't do anything then, otherwise add
3496         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3497         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3498         caller.
3499         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3500         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3501         (lower_stmt): Don't set data->calls_builtin_setjmp.
3502         (lower_builtin_setjmp): Adjust comment.
3503         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3504         * tree-cfg.c (found_computed_goto): Remove.
3505         (factor_computed_gotos): Remove.
3506         (make_goto_expr_edges): Return bool, true for computed gotos.
3507         Don't call make_abnormal_goto_edges.
3508         (build_gimple_cfg): Don't set found_computed_goto, don't call
3509         factor_computed_gotos.
3510         (computed_goto_p): No longer static.
3511         (make_blocks): Don't set found_computed_goto.
3512         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3513         (make_edges): If make_goto_expr_edges returns true, push bb
3514         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3515         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3516         vector.  Record mapping between bbs and OpenMP regions if there
3517         are any, adjust make_gimple_omp_edges caller.  Call
3518         handle_abnormal_edges.
3519         (make_abnormal_goto_edges): Remove.
3520         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3521         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3522         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3523         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3524         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3525         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3526         filling *region also set *region_idx to (*region)->entry->index.
3528         PR other/58712
3529         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3530         For REGs set ORIGINAL_REGNO.
3532 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3534         * doc/md.texi: Mention that a target shouldn't implement
3535         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3536         than hi/lo pair.
3538 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3540         PR tree-optimization/59594
3541         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3542         a copy of the datarefs vector rather than the vector itself.
3544 2014-01-28  Jason Merrill  <jason@redhat.com>
3546         PR c++/53756
3547         * dwarf2out.c (auto_die): New static.
3548         (gen_type_die_with_usage): Handle C++1y 'auto'.
3549         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3550         on definition.
3552 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3554         PR target/59672
3555         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3556         (SPEC_X32): Likewise.
3557         (SPEC_64): Likewise.
3558         * config/i386/i386.c (ix86_option_override_internal): Turn off
3559         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3560         for TARGET_16BIT.
3561         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3562         * config/i386/i386.h (TARGET_16BIT): New macro.
3563         (TARGET_16BIT_P): Likewise.
3564         * config/i386/i386.opt: Add m16.
3565         * doc/invoke.texi: Document -m16.
3567 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3569         PR preprocessor/59935
3570         * input.c (location_get_source_line): Bail out on when line number
3571         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3573 2014-01-28  Richard Biener  <rguenther@suse.de>
3575         PR tree-optimization/58742
3576         * tree-ssa-forwprop.c (associate_plusminus): Handle
3577         pointer subtraction of the form (T)(P + A) - (T)P.
3579 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3581         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3582         at const_int_cost.
3584 2014-01-28  Richard Biener  <rguenther@suse.de>
3586         Revert
3587         2014-01-28  Richard Biener  <rguenther@suse.de>
3589         PR rtl-optimization/45364
3590         PR rtl-optimization/59890
3591         * var-tracking.c (local_get_addr_clear_given_value): Handle
3592         already cleared slot.
3593         (val_reset): Handle not allocated local_get_addr_cache.
3594         (vt_find_locations): Use post-order on the inverted CFG.
3596 2014-01-28  Richard Biener  <rguenther@suse.de>
3598         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3600 2014-01-28  Richard Biener  <rguenther@suse.de>
3602         PR rtl-optimization/45364
3603         PR rtl-optimization/59890
3604         * var-tracking.c (local_get_addr_clear_given_value): Handle
3605         already cleared slot.
3606         (val_reset): Handle not allocated local_get_addr_cache.
3607         (vt_find_locations): Use post-order on the inverted CFG.
3609 2014-01-28  Alan Modra  <amodra@gmail.com>
3611         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3612         * configure.ac <recursive call for build != host>: Define
3613         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3614         and LD_FOR_BUILD too.
3615         * configure: Regenerate.
3617 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3619         * config/i386/i386.c (get_builtin_code_for_version): Separate
3620         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3621         Broadwell from Haswell.
3623 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3625         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3626         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3627         * config/mips/mips.c (mips_option_override): Change setting
3628         of TARGET_DSP.
3629         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3630         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3631         Change from Mask to Var.
3633 2014-01-27  Jeff Law  <law@redhat.com>
3635         * ipa-inline.c (inline_small_functions): Fix typo.
3637 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3639         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3640         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3641         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3642         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3643         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3644         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3645         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3646         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3647         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3648         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3649         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3650         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3651         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3652         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3653         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3654         (_mm512_storeu_epi64): Ditto.
3655         (_mm512_cmpge_epi32_mask): Ditto.
3656         (_mm512_cmpge_epu32_mask): Ditto.
3657         (_mm512_cmpge_epi64_mask): Ditto.
3658         (_mm512_cmpge_epu64_mask): Ditto.
3659         (_mm512_cmple_epi32_mask): Ditto.
3660         (_mm512_cmple_epu32_mask): Ditto.
3661         (_mm512_cmple_epi64_mask): Ditto.
3662         (_mm512_cmple_epu64_mask): Ditto.
3663         (_mm512_cmplt_epi32_mask): Ditto.
3664         (_mm512_cmplt_epu32_mask): Ditto.
3665         (_mm512_cmplt_epi64_mask): Ditto.
3666         (_mm512_cmplt_epu64_mask): Ditto.
3667         (_mm512_cmpneq_epi32_mask): Ditto.
3668         (_mm512_cmpneq_epu32_mask): Ditto.
3669         (_mm512_cmpneq_epi64_mask): Ditto.
3670         (_mm512_cmpneq_epu64_mask): Ditto.
3671         (_mm512_expand_pd): Ditto.
3672         (_mm512_expand_ps): Ditto.
3673         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3674         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3675         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3676         * config/i386/i386.c (ix86_builtins): Add
3677         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3678         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3679         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3680         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3681         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3682         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3683         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3684         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3685         IX86_BUILTIN_PMOVUSQW512_MEM.
3686         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3687         __builtin_ia32_pmovsqd512mem_mask,
3688         __builtin_ia32_pmovqd512mem_mask,
3689         __builtin_ia32_pmovusqw512mem_mask,
3690         __builtin_ia32_pmovsqw512mem_mask,
3691         __builtin_ia32_pmovqw512mem_mask,
3692         __builtin_ia32_pmovusdw512mem_mask,
3693         __builtin_ia32_pmovsdw512mem_mask,
3694         __builtin_ia32_pmovdw512mem_mask,
3695         __builtin_ia32_pmovqb512mem_mask,
3696         __builtin_ia32_pmovusqb512mem_mask,
3697         __builtin_ia32_pmovsqb512mem_mask,
3698         __builtin_ia32_pmovusdb512mem_mask,
3699         __builtin_ia32_pmovsdb512mem_mask,
3700         __builtin_ia32_pmovdb512mem_mask.
3701         (bdesc_args): Add __builtin_ia32_expanddf512,
3702         __builtin_ia32_expandsf512.
3703         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3704         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3705         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3706         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3707         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3708         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3709         (avx512f_<code>v8div16qi2_mask_store): This.
3710         (avx512f_expand<mode>): New.
3712 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3714         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3715         New.
3716         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3717         (_mm512_prefetch_i32scatter_pd): Ditto.
3718         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3719         (_mm512_prefetch_i64scatter_pd): Ditto.
3720         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3721         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3722         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3723         (_mm512_prefetch_i32scatter_ps): Ditto.
3724         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3725         (_mm512_prefetch_i64scatter_ps): Ditto.
3726         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3727         * config/i386/i386-builtin-types.def: Define
3728         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3729         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3730         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3731         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3732         IX86_BUILTIN_SCATTERPFQPD.
3733         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3734         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3735         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3736         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3737         __builtin_ia32_scatterpfqps.
3738         (ix86_expand_builtin): Expand new built-ins.
3739         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3740         fix memory access data type.
3741         (*avx512pf_gatherpf<mode>_mask): Ditto.
3742         (*avx512pf_gatherpf<mode>): Ditto.
3743         (avx512pf_scatterpf<mode>): Ditto.
3744         (*avx512pf_scatterpf<mode>_mask): Ditto.
3745         (*avx512pf_scatterpf<mode>): Ditto.
3746         (GATHER_SCATTER_SF_MEM_MODE): New.
3747         (avx512pf_gatherpf<mode>df): Ditto.
3748         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3749         (*avx512pf_scatterpf<mode>df): Ditto.
3751 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3753         PR bootstrap/59934
3754         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3755         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3756         reached.
3758 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3760         * common/config/arm/arm-common.c
3761         (arm_rewrite_mcpu): Handle multiple names.
3762         * config/arm/arm.h
3763         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3765 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3767         * gimple-builder.h (create_gimple_tmp): Delete.
3769 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3771         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3772         words comparisons.
3774 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3776         * config/pa/pa.md (call): Generate indirect long calls to non-local
3777         functions when outputing 32-bit code.
3778         (call_value): Likewise except for special call to buggy powf function.
3780         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3781         portable runtime and PIC indirect calls.
3782         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3783         and PIC call sequences.  Use ldo instead of blr to set return register
3784         in PIC call sequence.
3786 2014-01-25  Walter Lee  <walt@tilera.com>
3788         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3789         avoid clobbering a live register.
3791 2014-01-25  Walter Lee  <walt@tilera.com>
3793         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3794         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3795         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3796         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3798 2014-01-25  Walter Lee  <walt@tilera.com>
3800         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3801         arguments on even registers.
3802         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3803         STACK_BOUNDARY.
3804         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3805         (BIGGEST_ALIGNMENT): Ditto.
3806         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3808 2014-01-25  Walter Lee  <walt@tilera.com>
3810         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3811         insns before bundling.
3812         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3814 2014-01-25  Walter Lee  <walt@tilera.com>
3816         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3817         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3818         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3820 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3822         * config/mips/constraints.md (kl): Delete.
3823         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3824         define expands, using...
3825         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3826         instructions for MIPS16.
3827         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3828         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3830 2014-01-25  Walter Lee  <walt@tilera.com>
3832         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3833         (clzdi2): Ditto.
3834         (ffsdi2): Ditto.
3836 2014-01-25  Walter Lee  <walt@tilera.com>
3838         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3839         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3841 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3843         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3844         Handle XOR.
3846 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3848         * print-rtl.c (in_call_function_usage): New var.
3849         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3850         EXPR_LIST mode as mode and not as reg note name.
3852         PR middle-end/59561
3853         * cfgloopmanip.c (copy_loop_info): If
3854         loop->warned_aggressive_loop_optimizations, make sure
3855         the flag is set in target loop too.
3857 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3859         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3860         flag_cilkplus.
3861         * builtins.def: Likewise.
3862         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3863         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3864         * ira.c (ira_setup_eliminable_regset): Likewise.
3865         * omp-low.c (gate_expand_omp): Likewise.
3866         (execute_lower_omp): Likewise.
3867         (diagnose_sb_0): Likewise.
3868         (gate_diagnose_omp_blocks): Likewise.
3869         (simd_clone_clauses_extract): Likewise.
3870         (gate): Likewise.
3872 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3874         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3875         correction for little endian...
3876         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3877         here.
3879 2014-01-24  Jeff Law  <law@redhat.com>
3881         PR tree-optimization/59919
3882         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3883         for non-returning calls.
3885 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3887         * common/config/aarch64/aarch64-common.c
3888         (aarch64_rewrite_mcpu): Handle multiple names.
3889         * config/aarch64/aarch64.h
3890         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3892 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3894         * input.c (add_file_to_cache_tab): Handle the case where fopen
3895         returns NULL.
3897 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3899         PR target/59929
3900         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3901         from push operand if code of push isn't PRE_DEC.
3903 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3905         PR target/59909
3906         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3907         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3908         it is only used for non-atomic loads/stores.
3910         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3911         -mquad-memory or -mquad-memory-atomic switches.
3913         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3914         -mquad-memory-atomic to ISA 2.07 support.
3916         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3917         to separate support of normal quad word memory operations (ldq, stq)
3918         from the atomic quad word memory operations.
3920         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3921         support to separate non-atomic quad word operations from atomic
3922         quad word operations.  Disable non-atomic quad word operations in
3923         little endian mode so that we don't have to swap words after the
3924         load and before the store.
3925         (quad_load_store_p): Add comment about atomic quad word support.
3926         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3927         options printed with -mdebug=reg.
3929         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3930         -mquad-memory-atomic as the test for whether we have quad word
3931         atomic instructions.
3932         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3933         or -mp8-vector are used, allow byte/half-word atomic operations.
3935         * config/rs6000/sync.md (load_lockedti): Insure that the address
3936         is a proper indexed or indirect address for the lqarx instruction.
3937         On little endian systems, swap the hi/lo registers after the lqarx
3938         instruction.
3939         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3940         insure the address is valid for the lqarx instruction.
3941         (store_conditionalti): Insure that the address is a proper indexed
3942         or indirect address for the stqcrx. instruction.  On little endian
3943         systems, swap the hi/lo registers before doing the stqcrx.
3944         instruction.
3945         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3946         insure the address is valid for the stqcrx. instruction.
3948         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3949         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3950         type of quad memory support is available.
3952 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3954         PR regression/59915
3955         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3956         there is a danger of looping.
3958 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3960         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3961         force flag_ira_loop_pressure if set via command line.
3963 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3965         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3966         (ashr_simd): New builtin handling DI mode.
3967         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3968         (aarch64_sshr_simddi): New match pattern.
3969         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3970         (vshrd_n_s64): Likewise.
3971         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3973 2014-01-23  Nick Clifton  <nickc@redhat.com>
3975         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3976         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3977         favour of mcu specific scripts.
3978         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3979         430x multilibs.
3981 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3982             Alex Velenko  <Alex.Velenko@arm.com>
3984         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3985         (vaddv_s16): Likewise.
3986         (vaddv_s32): Likewise.
3987         (vaddv_u8): Likewise.
3988         (vaddv_u16): Likewise.
3989         (vaddv_u32): Likewise.
3990         (vaddvq_s8): Likewise.
3991         (vaddvq_s16): Likewise.
3992         (vaddvq_s32): Likewise.
3993         (vaddvq_s64): Likewise.
3994         (vaddvq_u8): Likewise.
3995         (vaddvq_u16): Likewise.
3996         (vaddvq_u32): Likewise.
3997         (vaddvq_u64): Likewise.
3998         (vaddv_f32): Likewise.
3999         (vaddvq_f32): Likewise.
4000         (vaddvq_f64): Likewise.
4001         (vmaxv_f32): Likewise.
4002         (vmaxv_s8): Likewise.
4003         (vmaxv_s16): Likewise.
4004         (vmaxv_s32): Likewise.
4005         (vmaxv_u8): Likewise.
4006         (vmaxv_u16): Likewise.
4007         (vmaxv_u32): Likewise.
4008         (vmaxvq_f32): Likewise.
4009         (vmaxvq_f64): Likewise.
4010         (vmaxvq_s8): Likewise.
4011         (vmaxvq_s16): Likewise.
4012         (vmaxvq_s32): Likewise.
4013         (vmaxvq_u8): Likewise.
4014         (vmaxvq_u16): Likewise.
4015         (vmaxvq_u32): Likewise.
4016         (vmaxnmv_f32): Likewise.
4017         (vmaxnmvq_f32): Likewise.
4018         (vmaxnmvq_f64): Likewise.
4019         (vminv_f32): Likewise.
4020         (vminv_s8): Likewise.
4021         (vminv_s16): Likewise.
4022         (vminv_s32): Likewise.
4023         (vminv_u8): Likewise.
4024         (vminv_u16): Likewise.
4025         (vminv_u32): Likewise.
4026         (vminvq_f32): Likewise.
4027         (vminvq_f64): Likewise.
4028         (vminvq_s8): Likewise.
4029         (vminvq_s16): Likewise.
4030         (vminvq_s32): Likewise.
4031         (vminvq_u8): Likewise.
4032         (vminvq_u16): Likewise.
4033         (vminvq_u32): Likewise.
4034         (vminnmv_f32): Likewise.
4035         (vminnmvq_f32): Likewise.
4036         (vminnmvq_f64): Likewise.
4038 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4040         * config/aarch64/aarch64-simd.md
4041         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4042         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4043         (*aarch64_mul3_elt<mode>): Likewise.
4044         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4045         (*aarch64_mul3_elt_to_64v2df): Likewise.
4046         (*aarch64_mla_elt<mode>): Likewise.
4047         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4048         (*aarch64_mls_elt<mode>): Likewise.
4049         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4050         (*aarch64_fma4_elt<mode>): Likewise.
4051         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4052         (*aarch64_fma4_elt_to_64v2df): Likewise.
4053         (*aarch64_fnma4_elt<mode>): Likewise.
4054         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4055         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4056         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4057         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4058         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4059         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4060         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4061         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4062         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4064 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4066         * config/aarch64/aarch64-simd.md
4067         (aarch64_be_checked_get_lane<mode>): New define_expand.
4068         * config/aarch64/aarch64-simd-builtins.def
4069         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4070         New builtin definition.
4071         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4072         Use new safe be builtin.
4074 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4076         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4077         New define_insn.
4078         (aarch64_be_st1<mode>): Likewise.
4079         (aarch_ld1<VALL:mode>): Define_expand modified.
4080         (aarch_st1<VALL:mode>): Likewise.
4081         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4082         (UNSPEC_ST1): Likewise.
4084 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
4086         * config/microblaze/microblaze.md: Add trap insn and attribute
4088 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4090         PR preprocessor/58580
4091         * input.h (location_get_source_line): Take an additional line_size
4092         parameter.
4093         (void diagnostics_file_cache_fini): Declare new function.
4094         * input.c (struct fcache): New type.
4095         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4096         New static constants.
4097         (diagnostic_file_cache_init, total_lines_num)
4098         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4099         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4100         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4101         (get_next_line, read_next_line, goto_next_line, read_line_num):
4102         New static function definitions.
4103         (diagnostic_file_cache_fini): New function.
4104         (location_get_source_line): Take an additional output line_len
4105         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4106         read_line_num.
4107         * diagnostic.c (diagnostic_finish): Call
4108         diagnostic_file_cache_fini.
4109         (adjust_line): Take an additional input parameter for the length
4110         of the line, rather than calculating it with strlen.
4111         (diagnostic_show_locus): Adjust the use of
4112         location_get_source_line and adjust_line with respect to their new
4113         signature.  While displaying a line now, do not stop at the first
4114         null byte.  Rather, display the zero byte as a space and keep
4115         going until we reach the size of the line.
4116         * Makefile.in: Add vec.o to OBJS-libcommon
4118 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4119             Ilya Tocar     <ilya.tocar@intel.com>
4121         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4122         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4123         (__builtin_ia32_kmov16): Ditto.
4124         * config/i386/i386.md (UNSPEC_KMOV): New.
4125         (kmovw): Ditto.
4127 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4129         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4130         (_mm512_storeu_si512): Ditto.
4132 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4134         PR target/52125
4135         * rtl.h (get_referenced_operands): Declare.
4136         * recog.c (get_referenced_operands): New function.
4137         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4138         operands have been referenced when recording LO_SUM references.
4140 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
4142         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4144 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4146         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4147         Enable for generic and recent AMD targets.
4149 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4151         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4152         ARG_SIZE note when adjustment was eliminated.
4154 2014-01-22  Jeff Law  <law@redhat.com>
4156         PR tree-optimization/59597
4157         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4158         in file.  Accept new argument REGISTERING and use it to modify
4159         dump output appropriately.
4160         (register_jump_thread): Corresponding changes.
4161         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4162         thread paths involving joiner blocks.  Add code to dump cancelled
4163         jump threading paths.
4165 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4167         PR rtl-optimization/59477
4168         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4169         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4171 2014-01-22  Tom Tromey  <tromey@redhat.com>
4173         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4174         PARAMS.
4175         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4177 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4179         PR rtl-optimization/59896
4180         * lra-constraints.c (process_alt_operands): Check unused note for
4181         matched operands of insn with no output reloads.
4183 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4185         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4186         (mips_move_from_gpr_cost): Likewise.
4188 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4190         PR rtl-optimization/59858
4191         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4192         ira_class_hard_regs_num.
4193         (process_alt_operands): Increase reject for dying matched operand.
4195 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4197         PR target/59003
4198         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4199         smaller than size, perform several stores or loads and stores
4200         at dst + count - size to store or copy all of size bytes, rather
4201         than just last modesize bytes.
4203 2014-01-20  DJ Delorie  <dj@redhat.com>
4205         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4206         that CLOBBERs are REGs before propogating their values.
4208 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4210         PR middle-end/59789
4211         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4212         (cgraph_inline_failed_type): New function.
4213         * cgraph.h (DEFCIFCODE): Add type.
4214         (cgraph_inline_failed_type_t): New enum.
4215         (cgraph_inline_failed_type): New prototype.
4216         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4217         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4218         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4219         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4220         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4221         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4222         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4223         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4224         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4225         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4226         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4227         OPTIMIZATION_MISMATCH.
4228         * tree-inline.c (expand_call_inline): Emit errors during
4229         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4231 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4233         PR target/59685
4234         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4235         mode attribute in insn output.
4237 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4239         * output.h (output_constant): Delete.
4240         * varasm.c (output_constant): Make private.
4242 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4244         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4246 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4248         PR middle-end/59860
4249         * tree.h (fold_builtin_strcat): New prototype.
4250         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4251         argument, if non-NULL, don't call c_strlen.  Optimize
4252         directly into __builtin_memcpy instead of __builtin_strcpy.
4253         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4254         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4256 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4258         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4259         for SImode_address_operand operands, having only a REG argument.
4261 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4263         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4264         loader name using mbig-endian.
4265         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4267 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4269         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4270         (-mtune): Likewise.
4271         (-mcpu): Likewise.
4273 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4275         * config/aarch64/aarch64-protos.h
4276         (aarch64_cannot_change_mode_class_ptr): Declare.
4277         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4278         aarch64_cannot_change_mode_class_ptr): New.
4279         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4280         backend hook aarch64_cannot_change_mode_class.
4282 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4284         * common/config/aarch64/aarch64-common.c
4285         (aarch64_handle_option): Don't handle any option order logic here.
4286         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4287         selected_cpu, warn on architecture version mismatch.
4288         (aarch64_override_options): Fix parsing order for option strings.
4290 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4291             Iain Sandoe  <iain@codesourcery.com>
4293         PR bootstrap/59496
4294         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4295         warning.  Amend comment to reflect current functionality.
4297 2014-01-20  Richard Biener  <rguenther@suse.de>
4299         PR middle-end/59860
4300         * builtins.c (fold_builtin_strcat): Remove case better handled
4301         by tree-ssa-strlen.c.
4303 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4305         * config/aarch64/aarch64.opt
4306         (mcpu, march, mtune): Make case-insensitive.
4308 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4310         PR target/59880
4311         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4312         if operands[1] is a REG or ZERO_EXTEND of a REG.
4314 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4316         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4318 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4320         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4321         long non-pic millicode calls.
4323 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4325         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4327 2014-01-19  Kito Cheng  <kito@0xlab.org>
4329         * builtins.c (expand_movstr): Check movstr expand done or fail.
4331 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4332             H.J. Lu  <hongjiu.lu@intel.com>
4334         PR target/59379
4335         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4336         to DImode for zero-extended addresses.
4338 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4340         PR rtl-optimization/57763
4341         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4342         on the new indirect jump_insn and increment LABEL_NUSES (label).
4344 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4346         PR bootstrap/59580
4347         PR bootstrap/59583
4348         * config.gcc (x86_archs): New variable.
4349         (x86_64_archs): Likewise.
4350         (x86_cpus): Likewise.
4351         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4352         --with-arch/--with-cpu= options.
4353         Support --with-arch=/--with-cpu={nehalem,westmere,
4354         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4356 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4358         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4359         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4361 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4363         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4365 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4367         PR target/58944
4368         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4369         clear cpp_get_options (parse_in)->warn_unused_macros for
4370         ix86_target_macros_internal with cpp_define.
4372 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4374         * jump.c (delete_related_insns): Keep (use (insn))s.
4375         * reorg.c (redundant_insn): Check for barriers too.
4377 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4379         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4381 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4383         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4384         call to $$dyncall when TARGET_LONG_CALLS is true.
4386 2014-01-17  Jeff Law  <law@redhat.com>
4388         * ree.c (combine_set_extension): Temporarily disable test for
4389         changing number of hard registers.
4391 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4393         PR middle-end/58125
4394         * ipa-inline-analysis.c (inline_free_summary):
4395         Do not free summary of aliases.
4397 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4399         PR middle-end/59706
4400         * gimplify.c (gimplify_expr): Use create_tmp_var
4401         instead of create_tmp_var_raw.  If cond doesn't have
4402         integral type, don't add the IFN_ANNOTATE builtin at all.
4404 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4406         PR ipa/59736
4407         * ipa-cp.c (prev_edge_clone): New variable.
4408         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4409         Also resize prev_edge_clone vector.
4410         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4411         (ipcp_edge_removal_hook): New function.
4412         (ipcp_driver): Register ipcp_edge_removal_hook.
4414 2014-01-17  Andrew Pinski <apinski@cavium.com>
4415             Steve Ellcey  <sellcey@mips.com>
4417         PR target/59462
4418         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4419         of operator mode.
4421 2014-01-17  Jeff Law  <law@redhat.com>
4423         PR middle-end/57904
4424         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4425         so that pass_ccp runs first.
4427 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4429         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4430         (ix86_adjust_cost): Use !TARGET_XXX.
4431         (do_reorder_for_imul): Likewise.
4432         (swap_top_of_ready_list): Likewise.
4433         (ix86_sched_reorder): Likewise.
4435 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4437         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4438         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4439         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4440         (intel_memset): New.  Duplicate slm_memset.
4441         (intel_cost): New.  Duplicate slm_cost.
4442         (m_INTEL): New macro.
4443         (processor_target_table): Add "intel".
4444         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4445         with PROCESSOR_INTEL for "intel".
4446         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4447         PROCESSOR_SILVERMONT.
4448         (ix86_issue_rate): Likewise.
4449         (ix86_adjust_cost): Likewise.
4450         (ia32_multipass_dfa_lookahead): Likewise.
4451         (swap_top_of_ready_list): Likewise.
4452         (ix86_sched_reorder): Likewise.
4453         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4454         instead of TARGET_OPT_AGU.
4455         * config/i386/i386.h (TARGET_INTEL): New.
4456         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4457         (processor_type): Add PROCESSOR_INTEL.
4458         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4459         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4461 2014-01-17  Marek Polacek  <polacek@redhat.com>
4463         PR c/58346
4464         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4465         size is zero.
4467 2014-01-17  Richard Biener  <rguenther@suse.de>
4469         PR tree-optimization/46590
4470         * opts.c (default_options_table): Add entries for
4471         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4472         all enabled at -O1 but not for -Og.
4473         * common.opt (fbranch-count-reg): Remove Init(1).
4474         (fmove-loop-invariants): Likewise.
4475         (ftree-pta): Likewise.
4477 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4479         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4480         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4481         decls to at least the GCC 4.8 used alignments.
4483         PR fortran/59440
4484         * tree-nested.c (convert_nonlocal_reference_stmt,
4485         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4486         of GIMPLE_BIND stmts, adjust associated decls.
4488 2014-01-17  Richard Biener  <rguenther@suse.de>
4490         PR tree-optimization/46590
4491         * vec.h (vec<>::bseach): New member function implementing
4492         binary search according to C89 bsearch.
4493         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4494         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4495         bitmap pointer again.  Make accesses_in_loop a flat array.
4496         (mem_ref_obstack): New global.
4497         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4498         (mark_ref_stored): Likewise.
4499         (ref_indep_loop_p_2): Likewise.
4500         (set_ref_stored_in_loop): New helper function.
4501         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4502         (memref_free): Adjust.
4503         (record_mem_ref_loc): Simplify.
4504         (gather_mem_refs_stmt): Adjust.
4505         (sort_locs_in_loop_postorder_cmp): New function.
4506         (analyze_memory_references): Sort accesses_in_loop after
4507         loop postorder number.
4508         (find_ref_loc_in_loop_cmp): New function.
4509         (for_all_locs_in_loop): Find relevant cluster of locs in
4510         accesses_in_loop and iterate without recursion.
4511         (execute_sm): Avoid uninit warning.
4512         (struct ref_always_accessed): Simplify.
4513         (ref_always_accessed::operator ()): Likewise.
4514         (ref_always_accessed_p): Likewise.
4515         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4516         loop postorder numbers here.
4517         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4518         numbers.
4520 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4522         PR c++/57945
4523         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4524         on decls for which assemble_alias has been called.
4526 2014-01-17  Nick Clifton  <nickc@redhat.com>
4528         * config/msp430/msp430.opt: (mcpu): New option.
4529         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4530         (msp430_option_override): Parse target_cpu.  If the MCU name
4531         matches a generic string, clear target_mcu.
4532         (msp430_attr): Allow numeric interrupt values up to 63.
4533         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4534         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4535         option.
4536         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4537         Add mcpu matches.
4538         * config/msp430/msp430.md (popm): Use %J rather than %I.
4539         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4540         (addhi_cy_i): Use immediate_operand for operand 2.
4541         * doc/invoke.texi: Document -mcpu option.
4543 2014-01-17  Richard Biener  <rguenther@suse.de>
4545         PR rtl-optimization/38518
4546         * df.h (df_analyze_loop): Declare.
4547         * df-core.c: Include cfgloop.h.
4548         (df_analyze_1): Split out main part of df_analyze.
4549         (df_analyze): Adjust.
4550         (loop_inverted_post_order_compute): New function.
4551         (loop_post_order_compute): Likewise.
4552         (df_analyze_loop): New function avoiding whole-function
4553         postorder computes.
4554         * loop-invariant.c (find_defs): Use df_analyze_loop.
4555         (find_invariants): Adjust.
4556         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4558 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4560         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4561         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4563 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4565         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4566         traversal when removing references.
4568 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4570         PR ipa/59775
4571         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4573 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4575         PR middle-end/56791
4576         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4577         pushing a reload for an autoinc when we had previously reloaded an
4578         inner part of the address.
4580 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4582         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4583         field.
4584         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4585         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4586         when not giving up or versioning for alias only because of
4587         loop->safelen.
4588         (vect_analyze_data_ref_dependences): Set to true.
4589         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4590         is a GIMPLE_PHI.
4591         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4592         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4593         to the condition.
4595         PR middle-end/58344
4596         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4598         PR target/59839
4599         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4600         operand 0 predicate for gathers, use a new pseudo as subtarget.
4602 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4604         PR middle-end/59609
4605         * lra-constraints.c (process_alt_operands): Add printing debug info.
4606         Check absence of input/output reloads for matched operands too.
4608 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4610         PR rtl-optimization/59835
4611         * ira.c (ira_init_register_move_cost): Increase cost for
4612         impossible modes.
4614 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4616         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4618 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4620         PR target/59780
4621         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4622         non-register objects.  Use gen_(high/low)part more consistently.
4623         Fix assertions.
4625 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4627         PR target/59844
4628         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4629         endian support, remove tests for WORDS_BIG_ENDIAN.
4630         (p8_mfvsrd_3_<mode>): Likewise.
4631         (reload_gpr_from_vsx<mode>): Likewise.
4632         (reload_gpr_from_vsxsf): Likewise.
4633         (p8_mfvsrd_4_disf): Likewise.
4635 2014-01-16  Richard Biener  <rguenther@suse.de>
4637         PR rtl-optimization/46590
4638         * lcm.c (compute_antinout_edge): Use postorder iteration.
4639         (compute_laterin): Use inverted postorder iteration.
4641 2014-01-16  Nick Clifton  <nickc@redhat.com>
4643         PR middle-end/28865
4644         * varasm.c (output_constant): Return the number of bytes actually
4645         emitted.
4646         (output_constructor_array_range): Update the field size with the
4647         number of bytes emitted by output_constant.
4648         (output_constructor_regular_field): Likewise.  Also do not
4649         complain if the total number of bytes emitted is now greater
4650         than the expected fieldpos.
4651         * output.h (output_constant): Update prototype and descriptive comment.
4653 2014-01-16  Marek Polacek  <polacek@redhat.com>
4655         PR middle-end/59827
4656         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4657         it is error_mark_node.
4659 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4661         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4662         VALID_AVX256_REG_OR_OI_MODE.
4664 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4666         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4667         current procedure should be profiled.
4669 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4671         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4672         of moving from/to the STACK_REG register class.
4674 2014-01-15  Richard Henderson  <rth@redhat.com>
4676         PR debug/54694
4677         * reginfo.c (global_regs_decl): Globalize.
4678         * rtl.h (global_regs_decl): Declare.
4679         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4680         reserved via global_regs.
4682 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4684         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4686 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4688         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4689         and vmulosh rather than call gen_vec_widen_smult_*.
4690         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4691         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4692         (vec_widen_smult_even_v16qi): Likewise.
4693         (vec_widen_umult_even_v8hi): Likewise.
4694         (vec_widen_smult_even_v8hi): Likewise.
4695         (vec_widen_umult_odd_v16qi): Likewise.
4696         (vec_widen_smult_odd_v16qi): Likewise.
4697         (vec_widen_umult_odd_v8hi): Likewise.
4698         (vec_widen_smult_odd_v8hi): Likewise.
4699         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4700         vmuloub rather than call gen_vec_widen_umult_*.
4701         (vec_widen_umult_lo_v16qi): Likewise.
4702         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4703         vmulosb rather than call gen_vec_widen_smult_*.
4704         (vec_widen_smult_lo_v16qi): Likewise.
4705         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4706         rather than call gen_vec_widen_umult_*.
4707         (vec_widen_umult_lo_v8hi): Likewise.
4708         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4709         rather than call gen_vec_widen_smult_*.
4710         (vec_widen_smult_lo_v8hi): Likewise.
4712 2014-01-15  Jeff Law  <law@redhat.com>
4714         PR tree-optimization/59747
4715         * ree.c (find_and_remove_re): Properly handle case where a second
4716         eliminated extension requires widening a copy created for elimination
4717         of a prior extension.
4718         (combine_set_extension): Ensure that the number of hard regs needed
4719         for a destination register does not change when we widen it.
4721 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4723         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4724         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4725         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4726         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4727         (avr-*-rtems*): Likewise.
4728         (bfin*-rtems*): Likewise.
4729         (moxie-*-rtems*): Likewise.
4730         (h8300-*-rtems*): Likewise.
4731         (i[34567]86-*-rtems*): Likewise.
4732         (lm32-*-rtems*): Likewise.
4733         (m32r-*-rtems*): Likewise.
4734         (m68k-*-rtems*): Likewise.
4735         (microblaze*-*-rtems*): Likewise.
4736         (mips*-*-rtems*): Likewise.
4737         (powerpc-*-rtems*): Likewise.
4738         (sh-*-rtems*): Likewise.
4739         (sparc-*-rtems*): Likewise.
4740         (sparc64-*-rtems*): Likewise.
4741         (v850-*-rtems*): Likewise.
4742         (m32c-*-rtems*): Likewise.
4744 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4746         PR rtl-optimization/59511
4747         * ira.c (ira_init_register_move_cost): Use memory costs for some
4748         cases of register move cost calculations.
4749         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4750         instead of BB frequency.
4751         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4752         * lra-assigns.c (find_hard_regno_for): Ditto.
4754 2014-01-15  Richard Biener  <rguenther@suse.de>
4756         PR tree-optimization/59822
4757         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4758         (vectorizable_load): Use it to hoist defs of uses of invariant
4759         loads out of the loop.
4761 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4762             Kugan Vivekanandarajah  <kuganv@linaro.org>
4764         PR target/59695
4765         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4766         truncation.
4768 2014-01-15  Richard Biener  <rguenther@suse.de>
4770         PR rtl-optimization/59802
4771         * lcm.c (compute_available): Use inverted postorder to seed
4772         the initial worklist.
4774 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4776         PR target/59803
4777         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4778         ADDR_REGS for invalid symrefs in non-PIC code.
4780 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4782         PR other/58712
4783         * builtins.c (determine_block_size): Initialize *probable_max_size
4784         even if len_rtx is CONST_INT.
4786 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4788         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4789         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4790         (cortexa53_tunings): Likewise.
4791         (aarch64_sched_issue_rate): New function.
4792         (TARGET_SCHED_ISSUE_RATE): Define.
4794 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4796         * ira-costs.c (find_costs_and_classes): Add missed
4797         ira_init_register_move_cost_if_necessary.
4799 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4801         PR target/59787
4802         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4804 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4806         PR target/59794
4807         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4808         to indicate if type is used for function return value.  Warn ABI
4809         change if the vector mode isn't available for function return value.
4810         (ix86_function_arg_advance): Pass false to type_natural_mode.
4811         (ix86_function_arg): Likewise.
4812         (ix86_gimplify_va_arg): Likewise.
4813         (function_arg_32): Don't warn ABI change.
4814         (ix86_function_value): Pass true to type_natural_mode.
4815         (ix86_return_in_memory): Likewise.
4816         (ix86_struct_value_rtx): Removed.
4817         (TARGET_STRUCT_VALUE_RTX): Likewise.
4819 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4821         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4822         converting a conditional jump into a conditional return.
4824 2014-01-14  Richard Biener  <rguenther@suse.de>
4826         PR tree-optimization/58921
4827         PR tree-optimization/59006
4828         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4829         hoisting invariant stmts.
4830         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4831         invariant loads on the preheader edge if possible.
4833 2014-01-14  Joey Ye  <joey.ye@arm.com>
4835         * doc/plugin.texi (Building GCC plugins): Update to C++.
4837 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4839         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4840         (_mm_rcp28_round_ss): Ditto.
4841         (_mm_rsqrt28_round_sd): Ditto.
4842         (_mm_rsqrt28_round_ss): Ditto.
4843         (_mm_rcp28_sd): Ditto.
4844         (_mm_rcp28_ss): Ditto.
4845         (_mm_rsqrt28_sd): Ditto.
4846         (_mm_rsqrt28_ss): Ditto.
4847         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4848         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4849         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4850         (IX86_BUILTIN_RCP28SD): Ditto.
4851         (IX86_BUILTIN_RCP28SS): Ditto.
4852         (IX86_BUILTIN_RSQRT28SD): Ditto.
4853         (IX86_BUILTIN_RSQRT28SS): Ditto.
4854         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4855         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4856         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4857         (ix86_expand_special_args_builtin): Expand new FTYPE.
4858         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4859         (srcp14<mode>): Make insn unary.
4860         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4861         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4862         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4863         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4864         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4865         Fix rounding: make it SAE only.
4866         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4867         Ditto.
4868         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4869         Ditto.
4870         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4871         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4872         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4873         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4874         (round_saeonly_mask_scalar_operand4): Ditto.
4875         (round_saeonly_mask_scalar_op3): Ditto.
4876         (round_saeonly_mask_scalar_op4): Ditto.
4878 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4880         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4881         Implement -maltivec=be for vec_insert and vec_extract.
4883 2014-01-10  DJ Delorie  <dj@redhat.com>
4885         * config/msp430/msp430.md (call_internal): Don't allow memory
4886         references with SP as the base register.
4887         (call_value_internal): Likewise.
4888         * config/msp430/constraints.md (Yc): New.  For memory references
4889         that don't use SP as a base register.
4891         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4892         "an integer without a # prefix"
4893         * config/msp430/msp430.md (epilogue_helper): Use it.
4895 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4897         PR target/59617
4898         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4899         AVX512F gather builtins.
4900         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4901         on gather decls with INTEGER_TYPE masktype.
4902         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4903         directly into the builtin rather than hoisting it before loop.
4905         PR tree-optimization/59387
4906         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4907         (scev_const_prop): If folded_casts and type has undefined overflow,
4908         use force_gimple_operand instead of force_gimple_operand_gsi and
4909         for each added stmt if it is assign with
4910         arith_code_with_undefined_signed_overflow, call
4911         rewrite_to_defined_overflow.
4912         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4913         gimple-fold.h instead.
4914         (arith_code_with_undefined_signed_overflow,
4915         rewrite_to_defined_overflow): Moved to ...
4916         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4917         rewrite_to_defined_overflow): ... here.  No longer static.
4918         Include gimplify-me.h.
4919         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4920         rewrite_to_defined_overflow): New prototypes.
4922 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4924         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4926 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4928         * builtins.c (get_object_alignment_2): Minor tweak.
4929         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4931 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4933         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4934         optimized non constant lengths.
4936 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4938         PR libgomp/59194
4939         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4940         load as __atomic_load_N if possible.
4942 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4944         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4945         target parameter.
4946         (rs6000_expand_builtin): Adjust call.
4948 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4950         PR target/58115
4951         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4952         * config/rs6000/rs6000.c: Include target-globals.h.
4953         (rs6000_set_current_function): Instead of doing target_reinit
4954         unconditionally, use save_target_globals_default_opts and
4955         restore_target_globals.
4957         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4958         FPSCR.
4959         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4960         (rs6000_expand_builtin): Handle mffs and mtfsf.
4961         (rs6000_init_builtins): Define mffs and mtfsf.
4962         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4963         (rs6000_mffs): New pattern.
4964         (rs6000_mtfsf): New pattern.
4966 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4968         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4969         Start narrowing with START.  Apply candidate-use pair
4970         and check overall cost in narrowing.
4971         (iv_ca_prune): Pass new argument.
4973 2014-01-10  Jeff Law  <law@redhat.com>
4975         PR middle-end/59743
4976         * ree.c (combine_reaching_defs): Ensure the defining statement
4977         occurs before the extension when optimizing extensions with
4978         different source and destination hard registers.
4980 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4982         PR ipa/58585
4983         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4984         vtables into the type inheritance graph.
4986 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4988         PR rtl-optimization/59754
4989         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4990         modes in the REGNO != REGNO case.
4992 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4994         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4996 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4998         PR tree-optimization/59745
4999         * tree-predcom.c (tree_predictive_commoning_loop): Call
5000         free_affine_expand_cache if giving up because components is NULL.
5002         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5003         GC in payload of target_globals struct instead of allocating them on
5004         the heap and the larger structs separately using GC.
5005         * target-globals.h (struct target_globals): Make regs, hard_regs,
5006         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5007         of GTY((skip)) and change type to void *.
5008         (reset_target_globals): Cast loads from those fields to corresponding
5009         types.
5011 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5013         PR plugins/59335
5014         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5015         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5016         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5018 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5020         PR target/59744
5021         * aarch64-modes.def (CC_Zmode): New flags mode.
5022         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5023         represents an equality.
5024         (aarch64_get_condition_code): Handle CC_Zmode.
5025         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5027 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5029         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5030         extraction in good case.
5032 2014-01-10  Richard Biener  <rguenther@suse.de>
5034         PR tree-optimization/59374
5035         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5036         checking after SLP discovery.  Mark stmts not participating
5037         in any SLP instance properly.
5039 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5041         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5042         when handling a SET rtx.
5044 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5046         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5047         (cortex-a57): Likewise.
5048         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5050 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5052         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5053         non-iwmmxt builtins.
5055 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5057         PR ipa/58252
5058         PR ipa/59226
5059         * ipa-devirt.c record_target_from_binfo): Take as argument
5060         stack of binfos and lookup matching one for virtual inheritance.
5061         (possible_polymorphic_call_targets_1): Update.
5063 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5065         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5066         kernel strings for Loongson-2E/2F/3A.
5068 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5070         PR middle-end/59670
5071         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5072         is_gimple_call before calling gimple_call_internal_p.
5074 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5076         * Makefile.in (TREE_FLOW_H): Remove.
5077         (TREE_SSA_H): Add file names from tree-flow.h.
5078         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5079         * tree.h: Remove tree-flow.h reference.
5080         * hash-table.h: Remove tree-flow.h reference.
5081         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5082         reference with tree-ssa-loop.h.
5084 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5086         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5087         default element-order behavior for -maltivec.
5088         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5089         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5090         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5091         when targeting big endian, at least for now.
5092         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5094 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5096         PR middle-end/47735
5097         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5098         var satisfies use_register_for_decl, just take into account type
5099         alignment, rather than decl alignment.
5101         PR tree-optimization/59622
5102         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5103         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5104         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5105         Don't devirtualize for inplace at all.  For targets.length () == 1,
5106         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5108 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5110         * config/i386/i386.md (cpu): Remove the unused btver1.
5112 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5114         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5116 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5118         PR target/58115
5119         * tree-core.h (struct target_globals): New forward declaration.
5120         (struct tree_target_option): Add globals field.
5121         * tree.h (TREE_TARGET_GLOBALS): Define.
5122         (prepare_target_option_nodes_for_pch): New prototype.
5123         * target-globals.h (struct target_globals): Define even if
5124         !SWITCHABLE_TARGET.
5125         * tree.c (prepare_target_option_node_for_pch,
5126         prepare_target_option_nodes_for_pch): New functions.
5127         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5128         * config/i386/i386.c: Include target-globals.h.
5129         (ix86_set_current_function): Instead of doing target_reinit
5130         unconditionally, use save_target_globals_default_opts and
5131         restore_target_globals.
5133 2014-01-09  Richard Biener  <rguenther@suse.de>
5135         PR tree-optimization/59715
5136         * tree-cfg.h (split_critical_edges): Declare.
5137         * tree-cfg.c (split_critical_edges): Export.
5138         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5140 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5142         * cfgexpand.c (expand_stack_vars): Optionally disable
5143         asan stack protection.
5144         (expand_used_vars): Likewise.
5145         (partition_stack_vars): Likewise.
5146         * asan.c (asan_emit_stack_protection): Optionally disable
5147         after return stack usage.
5148         (instrument_derefs): Optionally disable memory access instrumentation.
5149         (instrument_builtin_call): Likewise.
5150         (instrument_strlen_call): Likewise.
5151         (asan_protect_global): Optionally disable global variables protection.
5152         * doc/invoke.texi: Added doc for new options.
5153         * params.def: Added new options.
5154         * params.h: Likewise.
5156 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5158         PR rtl-optimization/59724
5159         * ifcvt.c (cond_exec_process_if_block): Don't call
5160         flow_find_head_matching_sequence with 0 longest_match.
5161         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5162         non-active insns if !stop_after.
5163         (try_head_merge_bb): Revert 2014-01-07 changes.
5165 2014-01-08  Jeff Law  <law@redhat.com>
5167         * ree.c (get_sub_rtx): New function, extracted from...
5168         (merge_def_and_ext): Here.
5169         (combine_reaching_defs): Use get_sub_rtx.
5171 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5173         * cgraph.h (varpool_variable_node): Do not choke on null node.
5175 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5177         * config/mips/mips.md (simple_return): Attempt to use JRC
5178         for microMIPS.
5179         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5181 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5183         PR rtl-optimization/59137
5184         * reorg.c (steal_delay_list_from_target): Call update_block for
5185         elided insns.
5186         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5188 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5190         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5191         two duplicate entries.
5193 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5195         Revert:
5196         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5198         * config/mips/mips.c (mips_truncated_op_cost): New function.
5199         (mips_rtx_costs): Adjust test for BADDU.
5200         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5202         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5204         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5205         (*baddu_si): ...this new pattern.
5207 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5209         PR ipa/59722
5210         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5212 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5214         PR middle-end/57748
5215         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5216         inner_reference_p.
5217         (expand_expr, expand_normal): Adjust.
5218         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5219         inner_reference_p. Use inner_reference_p to expand inner references.
5220         (store_expr): Adjust.
5221         * cfgexpand.c (expand_call_stmt): Adjust.
5223 2014-01-08  Rong Xu  <xur@google.com>
5225         * gcov-io.c (gcov_var): Move from gcov-io.h.
5226         (gcov_position): Ditto.
5227         (gcov_is_error): Ditto.
5228         (gcov_rewrite): Ditto.
5229         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5230         only part to libgcc/libgcov.h.
5232 2014-01-08  Marek Polacek  <polacek@redhat.com>
5234         PR middle-end/59669
5235         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5237 2014-01-08  Marek Polacek  <polacek@redhat.com>
5239         PR sanitizer/59667
5240         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5242 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5244         PR rtl-optimization/59649
5245         * stor-layout.c (get_mode_bounds): For BImode return
5246         0 and STORE_FLAG_VALUE.
5248 2014-01-08  Richard Biener  <rguenther@suse.de>
5250         PR middle-end/59630
5251         * gimple.h (is_gimple_builtin_call): Remove.
5252         (gimple_builtin_call_types_compatible_p): New.
5253         (gimple_call_builtin_p): New overload.
5254         * gimple.c (is_gimple_builtin_call): Remove.
5255         (validate_call): Rename to ...
5256         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5257         check return types.
5258         (validate_type): New static function.
5259         (gimple_call_builtin_p): New overload and adjust.
5260         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5261         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5262         (gimple_fold_stmt_to_constant_1): Likewise.
5263         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5265 2014-01-08  Richard Biener  <rguenther@suse.de>
5267         PR middle-end/59471
5268         * gimplify.c (gimplify_expr): Gimplify register-register type
5269         VIEW_CONVERT_EXPRs to separate stmts.
5271 2014-01-07  Jeff Law  <law@redhat.com>
5273         PR middle-end/53623
5274         * ree.c (combine_set_extension): Handle case where source
5275         and destination registers in an extension insn are different.
5276         (combine_reaching_defs): Allow source and destination registers
5277         in extension to be different under limited circumstances.
5278         (add_removable_extension): Remove restriction that the
5279         source and destination registers in the extension are the same.
5280         (find_and_remove_re): Emit a copy from the extension's
5281         destination to its source after the defining insn if
5282         the source and destination registers are different.
5284         PR middle-end/59285
5285         * ifcvt.c (merge_if_block): If we are merging a block with more than
5286         one successor with a block with no successors, remove any BARRIER
5287         after the second block.
5289 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
5291         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5293 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5295         PR target/59652
5296         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5297         for 14-bit register offsets when INT14_OK_STRICT is false.
5299 2014-01-07  Roland Stigge  <stigge@antcom.de>
5300             Michael Meissner  <meissner@linux.vnet.ibm.com>
5302         PR 57386/target
5303         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5304         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5306 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5308         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5309         -mcpu.
5311 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5313         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5314         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5315         rtx is const0_rtx or not.
5317 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5319         PR target/58115
5320         * target-globals.c (save_target_globals): Remove this_fn_optab
5321         handling.
5322         * toplev.c: Include optabs.h.
5323         (target_reinit): Temporarily restore the global options if another
5324         set of options are in force.
5326 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5328         PR rtl-optimization/58668
5329         * cfgcleanup.c (flow_find_cross_jump): Don't count
5330         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5331         to determine what is counted.
5332         (flow_find_head_matching_sequence): Use active_insn_p to determine
5333         what is counted.
5334         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5335         counting change.
5336         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5337         determine what is counted.
5339         PR tree-optimization/59643
5340         * tree-predcom.c (split_data_refs_to_components): If one dr is
5341         read and one write, determine_offset fails and the write isn't
5342         in the bad component, just put the read into the bad component.
5344 2014-01-07  Mike Stump  <mikestump@comcast.net>
5345             Jakub Jelinek  <jakub@redhat.com>
5347         PR pch/59436
5348         * tree-core.h (struct tree_optimization_option): Change optabs
5349         type from unsigned char * to void *.
5350         * optabs.c (init_tree_optimization_optabs): Adjust
5351         TREE_OPTIMIZATION_OPTABS initialization.
5353 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5355         PR target/59644
5356         * config/i386/i386.h (struct machine_function): Add
5357         no_drap_save_restore field.
5358         * config/i386/i386.c (ix86_save_reg): Use
5359         !cfun->machine->no_drap_save_restore instead of
5360         crtl->stack_realign_needed.
5361         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5362         this function clears frame_pointer_needed.  Set
5363         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5364         and DRAP reg is needed.
5366 2014-01-06  Marek Polacek  <polacek@redhat.com>
5368         PR c/57773
5369         * doc/implement-c.texi: Mention that other integer types are
5370         permitted as bit-field types in strictly conforming mode.
5372 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5374         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5375         is newly allocated.
5377 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5379         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5381 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5383         PR ipa/59008
5384         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5385         to int.
5386         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5388 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5390         PR debug/59350
5391         PR debug/59510
5392         * var-tracking.c (add_stores): Preserve the value of the source even if
5393         we don't record the store.
5395 2014-01-06  Terry Guo  <terry.guo@arm.com>
5397         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5399 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5401         PR bootstrap/59541
5402         * config/darwin.c (darwin_function_section): Adjust return values to
5403         correspond to optimisation changes made in r206070.
5405 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5407         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5408         from prefetch_block tune setting.
5409         (nocona_cost): Correct size of prefetch block to 64.
5411 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5413         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5414         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5415         used to save the static chain register in the computation of the offset
5416         from which the FP registers need to be restored.
5418 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5420         PR tree-optimization/59519
5421         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5422         ICE if get_current_def (current_new_name) is already non-NULL, as long
5423         as it is a phi result of some other phi in *new_exit_bb that has
5424         the same argument.
5426         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5427         or vmovdqu* for misaligned_operand.
5428         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5429         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5430         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5431         aligned_mem for AVX512F masked aligned load and store builtins and for
5432         non-temporal moves.
5434 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5436         PR tree-optimization/59651
5437         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5438         Address range for negative step should be added by TYPE_SIZE_UNIT.
5440 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5442         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5443         overlapping registers also for registers other than the stack pointer.
5445 2014-01-03  Marek Polacek  <polacek@redhat.com>
5447         PR other/59661
5448         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5449         __builtin_FILE.
5451 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5453         PR target/59625
5454         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5455         asm goto as jump.
5457         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5458         (push splitter): Use <P:MODE_SIZE> instead of
5459         GET_MODE_SIZE (<P:MODE>mode).
5460         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5461         (mov -1, reg peephole2): Likewise.
5462         * config/i386/sse.md (*mov<mode>_internal,
5463         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5464         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5465         *<code><mode>3, *andnot<mode>3<mask_name>,
5466         <mask_codefor><code><mode>3<mask_name>): Likewise.
5467         * config/i386/subst.md (mask_mode512bit_condition,
5468         sd_mask_mode512bit_condition): Likewise.
5470 2014-01-02  Xinliang David Li  <davidxl@google.com>
5472         PR tree-optimization/59303
5473         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5474         (dump_predicates): Better output format.
5475         (pred_equal_p): New function.
5476         (is_neq_relop_p): Ditto.
5477         (is_neq_zero_form_p): Ditto.
5478         (pred_expr_equal_p): Ditto.
5479         (pred_neg_p): Ditto.
5480         (simplify_pred): Ditto.
5481         (simplify_preds_2): Ditto.
5482         (simplify_preds_3): Ditto.
5483         (simplify_preds_4): Ditto.
5484         (simplify_preds): Ditto.
5485         (push_pred): Ditto.
5486         (push_to_worklist): Ditto.
5487         (get_pred_info_from_cmp): Ditto.
5488         (is_degenerated_phi): Ditto.
5489         (normalize_one_pred_1): Ditto.
5490         (normalize_one_pred): Ditto.
5491         (normalize_one_pred_chain): Ditto.
5492         (normalize_preds): Ditto.
5493         (normalize_cond_1): Remove function.
5494         (normalize_cond): Ditto.
5495         (is_gcond_subset_of): Ditto.
5496         (is_subset_of_any): Ditto.
5497         (is_or_set_subset_of): Ditto.
5498         (is_and_set_subset_of): Ditto.
5499         (is_norm_cond_subset_of): Ditto.
5500         (pred_chain_length_cmp): Ditto.
5501         (convert_control_dep_chain_into_preds): Type change.
5502         (find_predicates): Ditto.
5503         (find_def_preds): Ditto.
5504         (destroy_predicates_vecs): Ditto.
5505         (find_matching_predicates_in_rest_chains): Ditto.
5506         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5507         (is_pred_expr_subset): Ditto.
5508         (is_pred_chain_subset_of): Ditto.
5509         (is_included_in): Ditto.
5510         (is_superset_of): Ditto.
5512 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5514         Update copyright years.
5516 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5518         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5519         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5520         config/arc/arc.md, config/arc/arc.opt,
5521         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5522         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5523         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5524         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5525         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5526         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5527         vtable-verify.c, vtable-verify.h: Use the standard form for the
5528         copyright notice.
5530 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5532         * gcc.c (process_command): Update copyright notice dates.
5533         * gcov-dump.c: Ditto.
5534         * gcov.c: Ditto.
5535         * doc/cpp.texi: Bump @copying's copyright year.
5536         * doc/cppinternals.texi: Ditto.
5537         * doc/gcc.texi: Ditto.
5538         * doc/gccint.texi: Ditto.
5539         * doc/gcov.texi: Ditto.
5540         * doc/install.texi: Ditto.
5541         * doc/invoke.texi: Ditto.
5543 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5545         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5547 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5549         * config/i386/sse.md (*mov<mode>_internal): Guard
5550         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5552         PR rtl-optimization/59647
5553         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5554         new_rtx into UNSIGNED_FLOAT rtxes.
5556 Copyright (C) 2014 Free Software Foundation, Inc.
5558 Copying and distribution of this file, with or without modification,
5559 are permitted in any medium without royalty provided the copyright
5560 notice and this notice are preserved.