re PR testsuite/60672 (FAIL: g++.dg/cpp1y/auto-fn25.C -std=gnu++1y (test for errors...
[official-gcc.git] / gcc / ChangeLog
blobf3b01bc362696a4242b36df4dabd035cb7447598
1 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
3         * config/rs6000/constraints.md (wD constraint): New constraint to
4         match the constant integer to get the top DImode/DFmode out of a
5         vector in a VSX register.
7         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
8         match the constant integer to get the top DImode/DFmode out of a
9         vector in a VSX register.
11         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
12         for ISA 2.07.
14         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15         vbpermq builtins.
17         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
18         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
20         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
21         Optimize vec_extract of 64-bit values, where the value being
22         extracted is in the top word, where we can use scalar
23         instructions.  Add direct move and store support.  Combine the big
24         endian/little endian vector select load support into a single
25         insn.
26         (vsx_extract_<mode>_internal1): Likewise.
27         (vsx_extract_<mode>_internal2): Likewise.
28         (vsx_extract_<mode>_load): Likewise.
29         (vsx_extract_<mode>_store): Likewise.
30         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
31         combined into vsx_extract_<mode>_load.
32         (vsx_extract_<mode>_one_le): Likewise.
34         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
35         define the top 64-bit vector element.
37         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
38         constraint.
40         PR target/60672
41         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
42         enable use of xxsldwi and xxpermdi builtin functions.
43         (vec_xxpermdi): Likewise.
45         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
46         Document use of vec_xxsldwi and vec_xxpermdi builtins.
48 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
50         PR rtl-optimization/60650
51         * lra-asign.c (find_hard_regno_for, spill_for): Add parameter
52         first_p.  Use it.
53         (find_spills_for): New.
54         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
55         Spill all pseudos on the second iteration.
57 2014-03-27  Marek Polacek  <polacek@redhat.com>
59         PR c/50347
60         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
61         types.
63 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
65         * config/s390/s390.c (s390_can_use_return_insn): Check for
66         call-saved FPRs on 31 bit.
68 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
70         PR middle-end/60682
71         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
72         if they need regimplification, just drop them instead of
73         calling gimple_regimplify_operands on them.
75 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
77         PR target/60580
78         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
79         (aarch64_frame_pointer_required): Adjust logic.
80         (aarch64_can_eliminate): Adjust logic.
81         (aarch64_override_options_after_change): Adjust logic.
83 2014-03-27  Dehao Chen  <dehao@google.com>
85         * ipa-inline.c (early_inliner): Update node's inline info.
87 2014-03-26  Dehao Chen  <dehao@google.com>
89         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
90         compiler inserted conditional jumps for NAN float check.
92 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
94         * ubsan.h (ubsan_create_data): Change second argument's type
95         to const location_t *.
96         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
97         _("<unknown>").
98         (ubsan_create_data): Change second argument to const location_t *PLOC.
99         Create Loc field whenever PLOC is non-NULL.
100         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
101         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
102         callers.
104         PR other/59545
105         * real.c (real_to_integer2): Change type of low to UHWI.
107 2014-03-26  Tobias Burnus  <burnus@net-b.de>
109         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
110         (CILK_SELF_SPECS): New define.
111         (driver_self_specs): Use it.
113 2014-03-26  Richard Biener  <rguenther@suse.de>
115         * tree-pretty-print.c (percent_K_format): Implement special
116         case for LTO and its stripped down BLOCK tree.
118 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
120         PR sanitizer/60636
121         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
123         * tree-vrp.c (simplify_internal_call_using_ranges): If only
124         one range is range_int_cst_p, but not both, at least optimize
125         addition/subtraction of 0 and multiplication by 0 or 1.
126         * gimple-fold.c (gimple_fold_call): Fold
127         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
128         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
129         INTEGER_CSTs, try to fold at least x * 0 and y - y.
131 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
133         PR rtl-optimization/60452
134         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
135         <case REG>: Return 1 for invalid offsets from the frame pointer.
137 2014-03-26  Marek Polacek  <polacek@redhat.com>
139         PR c/37428
140         * doc/extend.texi (C Extensions): Mention variable-length arrays in
141         a structure/union.
143 2014-03-26  Marek Polacek  <polacek@redhat.com>
145         PR c/39525
146         * doc/extend.texi (Designated Inits): Describe what happens to omitted
147         field members.
149 2014-03-26  Marek Polacek  <polacek@redhat.com>
151         PR other/59545
152         * ira-color.c (update_conflict_hard_regno_costs): Perform the
153         multiplication in unsigned type.
155 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
157         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
159 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
161         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
163 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
165         PR ipa/60315
166         * cif-code.def (UNREACHABLE) New code.
167         * ipa-inline.c (inline_small_functions): Skip edges to __builtlin_unreachable.
168         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
169         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
170         predicate to __bulitin_unreachable.
171         (set_cond_stmt_execution_predicate): Fix issue when invert_tree_comparison
172         returns ERROR_MARK.
173         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
174         propagate to inline clones.
175         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
176         to unreachable.
177         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
178         * cgraphclones.c (cgraph_clone_node): If call destination is already
179         ureachable, do not redirect it back.
180         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
181         unreachable.
183 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
185         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
186         Do not modify inline clones.
188 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
190         * config/i386/i386.md (general_sext_operand): New mode attr.
191         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
192         don't generate (sign_extend (const_int)).
193         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
194         operands[2].  Use We constraint instead of <i> and <general_sext_operand>
195         predicate instead of <general_operand>.
196         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
197         * config/i386/constraints.md (We): New constraint.
198         * config/i386/predicates.md (x86_64_sext_operand,
199         sext_operand): New predicates.
201 2014-03-25  Martin Jambor  <mjambor@suse.cz>
203         PR ipa/60600
204         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
205         inconsistent devirtualizations to __builtin_unreachable.
207 2014-03-25  Marek Polacek  <polacek@redhat.com>
209         PR c/35449
210         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
212 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
214         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
215         order of elements for big-endian.
217 2014-03-25  Richard Biener  <rguenther@suse.de>
219         PR middle-end/60635
220         * gimplify-me.c (gimple_regimplify_operands): Update the
221         re-gimplifed stmt.
223 2014-03-25  Martin Jambor  <mjambor@suse.cz>
225         PR ipa/59176
226         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
227         (lto_output_varpool_node): Likewise.
228         (input_overwrite_node): Likewise.
229         (input_varpool_node): Likewise.
231 2014-03-25  Richard Biener  <rguenther@suse.de>
233         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like
234         OPT_fpie.
235         (run_gcc): Likewise.
237 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
239         * combine.c (simplify_compare_const): Add MODE argument.
240         Handle mode_width 0 as very large mode_width.
241         (try_combine, simplify_comparison): Adjust callers.
243         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
244         type to avoid signed integer overflow.
245         * explow.c (plus_constant): Likewise.
247 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
249         * doc/generic.texi: Correct typos.
251 2014-03-24  Tobias Burnus  <burnus@net-b.de>
253         * doc/invoke.texi (-flto): Expand section about
254         using static libraries with LTO.
256 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
258         PR rtl-optimization/60501
259         * optabs.def (addptr3_optab): New optab.
260         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
261         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
262         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
264         * lra.c (emit_add3_insn): Use the addptr pattern if available.
266         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
268 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
270         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
271         _mm512_set1_pd.
273         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
274         (_mm256_undefined_ps): Define.
275         (_mm256_undefined_pd): Define.
276         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
277         (_mm_undefined_pd): Define.
278         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
279         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
280         (_mm512_undefined_ps): Define.
281         (_mm512_undefined_pd): Define.
282         Use _mm*_undefined_*.
283         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
285 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
287         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
288         (lshr_simd): DI mode added.
289         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
290         (aarch64_ushr_simddi): Likewise.
291         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
292         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
293         (vshrd_n_u64): Likewise.
295 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
297         * Makefile.in (s-macro_list): Depend on cc1.
299 2014-03-23  Teresa Johnson  <tejohnson@google.com>
301         * ipa-utils.c (ipa_print_order): Use specified dump file.
303 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
305         PR rtl-optimization/60601
306         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
308         * gcc.c (eval_spec_function): Initialize save_growing_value.
310 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
312         PR sanitizer/60613
313         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
314         code == MINUS_EXPR, never swap op0 with op1.
316         * toplev.c (init_local_tick): Avoid signed integer multiplication
317         overflow.
318         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
319         shift by first operand's bitsize.
321 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
323         PR target/60610
324         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
325         redefine to 1 or 0.
326         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
327         TARGET_ISA_64BIT_P(x).
329 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
331         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
332         pattern for vector nor instead of subtract from splat(-1).
333         (altivec_expand_vec_perm_const_le): Likewise.
335 2014-03-21  Richard Henderson  <rth@twiddle.net>
337         PR target/60598
338         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
339         related insns after epilogue_completed.
341 2014-03-21  Martin Jambor  <mjambor@suse.cz>
343         PR ipa/59176
344         * cgraph.h (symtab_node): New flag body_removed.
345         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
346         when removing bodies.
347         * symtab.c (dump_symtab_base): Dump body_removed flag.
348         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
349         had their bodies removed.
351 2014-03-21  Martin Jambor  <mjambor@suse.cz>
353         PR ipa/60419
354         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
355         in the border.
357 2014-03-21  Richard Biener  <rguenther@suse.de>
359         PR tree-optimization/60577
360         * tree-core.h (struct tree_base): Document nothrow_flag use
361         in VAR_DECL_NONALIASED.
362         * tree.h (VAR_DECL_NONALIASED): New.
363         (may_be_aliased): Adjust.
364         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
366 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
368         * expr.c (expand_expr_real_1): Remove outdated comment.
370 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
372         PR middle-end/60597
373         * ira.c (adjust_cleared_regs): Call copy_rtx on
374         *reg_equiv[REGNO (loc)].src_p before passing it to
375         simplify_replace_fn_rtx.
377         PR target/60568
378         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
379         into CONST, put pic register as first operand of PLUS.  Use
380         gen_const_mem for both 32-bit and 64-bit PIC got loads.
382 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
384         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
386 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
388         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
389         around for store forwarding issue in the FPU on the UT699.
390         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
391         loads and operations if -mfix-ut699 is specified.
392         (divtf3_hq): Tweak attribute.
393         (sqrttf2_hq): Likewise.
395 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
397         * calls.c (store_one_arg): Remove incorrect const qualification on the
398         type of the temporary.
399         * cfgexpand.c (expand_return): Likewise.
400         * expr.c (expand_constructor): Likewise.
401         (expand_expr_real_1): Likewise.
403 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
405         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
406         of parts.
408 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
410         PR target/60039
411         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
413 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
415         * config/arm/aarch-common-protos.h
416         (alu_cost_table): Fix spelling of "extend".
417         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
419 2014-03-19  Richard Biener  <rguenther@suse.de>
421         PR middle-end/60553
422         * tree-core.h (tree_type_common): Re-order pointer members
423         to reduce recursion depth during GC walks.
425 2014-03-19  Marek Polacek  <polacek@redhat.com>
427         PR sanitizer/60569
428         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
429         before accessing it.
431 2014-03-19  Richard Biener  <rguenther@suse.de>
433         PR lto/59543
434         * lto-streamer-in.c (input_function): In WPA stage do not drop
435         debug stmts.
437 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
439         PR tree-optimization/60559
440         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
441         with build_zero_cst assignment.
443 2014-03-18  Kai Tietz  <ktietz@redhat.com>
445         PR rtl-optimization/56356
446         * sdbout.c (sdbout_parms): Verify that parms'
447         incoming argument is valid.
448         (sdbout_reg_parms): Likewise.
450 2014-03-18  Richard Henderson  <rth@redhat.com>
452         PR target/60562
453         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
454         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
455         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
457 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
459         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
460         * doc/plugins.texi (Plugin callbacks): Mention
461         PLUGIN_INCLUDE_FILE.
462         Italicize plugin event names in description.  Explain that
463         PLUGIN_PRAGMAS has no sense for lto1. Explain
464         PLUGIN_INCLUDE_FILE.
465         Remind that no GCC functions should be called after
466         PLUGIN_FINISH.
467         Explain what pragmas with expansion are.
469 2014-03-18  Martin Liska  <mliska@suse.cz>
471         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
472         gimple call statement is update.
473         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
474         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
476 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
478         PR sanitizer/60557
479         * ubsan.c (ubsan_instrument_unreachable): Call
480         initialize_sanitizer_builtins.
481         (ubsan_pass): Likewise.
483         PR sanitizer/60535
484         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
485         varpool_finalize_decl instead of rest_of_decl_compilation.
487 2014-03-18  Richard Biener  <rguenther@suse.de>
489         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
490         by using bitmap_and_compl instead of bitmap_and_compl_into.
491         (df_rd_transfer_function): Likewise.
493 2014-03-18  Richard Biener  <rguenther@suse.de>
495         * doc/lto.texi (fresolution): Fix typo.
497 2014-03-18  Richard Biener  <rguenther@suse.de>
499         * doc/invoke.texi (flto): Update for changes in 4.9.
501 2014-03-18  Richard Biener  <rguenther@suse.de>
503         * doc/loop.texi: Remove section on the removed lambda framework.
504         Update loop docs with recent changes in preserving loop structure.
506 2014-03-18  Richard Biener  <rguenther@suse.de>
508         * doc/lto.texi (-fresolution): Document.
510 2014-03-18  Richard Biener  <rguenther@suse.de>
512         * doc/contrib.texi: Adjust my name.
514 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
516         PR ipa/58721
517         * internal-fn.c: Include diagnostic-core.h.
518         (expand_BUILTIN_EXPECT): New function.
519         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
520         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
521         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
522         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
523         IFN_BUILTIN_EXPECT.
524         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
525         Revert 3 argument __builtin_expect code.
526         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
527         * gimple-fold.c (gimple_fold_call): Likewise.
528         * tree.h (fold_builtin_expect): New prototype.
529         * builtins.c (build_builtin_expect_predicate): Add predictor
530         argument, if non-NULL, create 3 argument __builtin_expect.
531         (fold_builtin_expect): No longer static.  Add ARG2 argument,
532         pass it through to build_builtin_expect_predicate.
533         (fold_builtin_2): Adjust caller.
534         (fold_builtin_3): Handle BUILT_IN_EXPECT.
535         * internal-fn.def (BUILTIN_EXPECT): New.
537 2014-03-18  Tobias Burnus  <burnus@net-b.de>
539         PR ipa/58721
540         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
541         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
542         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
544 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
546         PR ipa/58721
547         * predict.c (combine_predictions_for_bb): Fix up formatting.
548         (expr_expected_value_1, expr_expected_value): Add predictor argument,
549         fill what it points to if non-NULL.
550         (tree_predict_by_opcode): Adjust caller, use the predictor.
551         * predict.def (PRED_COMPARE_AND_SWAP): Add.
553 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
555         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
556         proper constant for the store mode.
558 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
560         * symtab.c (change_decl_assembler_name): Fix transparent alias
561         chain construction.
563 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
565         * config/aarch64/aarch64.c: Correct the comments about the
566         aarch64 stack layout.
568 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
570         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
571         check for GF_OMP_FOR_KIND_FOR.
573 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
575         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
576         ymm and zmm register names.
578 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
580         PR target/60516
581         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
582         note creation for the 2010-08-31 changes.
584 2014-03-17  Marek Polacek  <polacek@redhat.com>
586         PR middle-end/60534
587         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
588         as -fno-tree-loop-vectorize.
589         (expand_omp_simd): Likewise.
591 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
593         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
594         (eligible_for_call_delay): New prototype.
595         * config/sparc/sparc.c (tls_call_delay): Rename into...
596         (eligible_for_call_delay): ...this.  Return false if the instruction
597         cannot be put in the delay slot of a branch.
598         (eligible_for_restore_insn): Simplify.
599         (eligible_for_return_delay): Return false if the instruction cannot be
600         put in the delay slot of a branch and simplify.
601         (eligible_for_sibcall_delay): Return false if the instruction cannot be
602         put in the delay slot of a branch.
603         * config/sparc/sparc.md (fix_ut699): New attribute.
604         (tls_call_delay): Delete.
605         (in_call_delay): Reimplement.
606         (eligible_for_sibcall_delay): Rename into...
607         (in_sibcall_delay): ...this.
608         (eligible_for_return_delay): Rename into...
609         (in_return_delay): ...this.
610         (in_branch_delay): Reimplement.
611         (in_uncond_branch_delay): Delete.
612         (in_annul_branch_delay): Delete.
614 2014-03-14  Richard Henderson  <rth@redhat.com>
616         PR target/60525
617         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
618         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
619         (*floathi<X87MODEF>2_i387_with_temp): Remove.
620         (floathi splitters): Remove.
621         (float<SWI48x>xf2): New pattern.
622         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
623         code that tried to handle DImode for 32-bit, but which was excluded
624         by the pattern's condition.  Drop allocation of stack temporary.
625         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
626         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
627         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
628         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
629         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
630         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
631         (*float<SWI48><MODEF>2_sse_interunit): Remove.
632         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
633         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
634         (*float<SWI48x><X87MODEF>2_i387): Remove.
635         (all float _with_temp splitters): Remove.
636         (*float<SWI48x><MODEF>2_i387): New pattern.
637         (*float<SWI48><MODEF>2_sse): New pattern.
638         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
639         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
641 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
642             Marek Polacek  <polacek@redhat.com>
644         PR middle-end/60484
645         * common.opt (dump_base_name_prefixed): New Variable.
646         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
647         if x_dump_base_name_prefixed is already set, set it at the end.
649 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
651         PR rtl-optimization/60508
652         * lra-constraints.c (get_reload_reg): Add new parameter
653         in_subreg_p.
654         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
655         Pass the new parameter values.
657 2014-03-14  Richard Biener  <rguenther@suse.de>
659         * common.opt: Revert unintented changes from r205065.
660         * opts.c: Likewise.
662 2014-03-14  Richard Biener  <rguenther@suse.de>
664         PR middle-end/60518
665         * cfghooks.c (split_block): Properly adjust all loops the
666         block was a latch of.
668 2014-03-14  Martin Jambor  <mjambor@suse.cz>
670         PR lto/60461
671         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
672         and simplify it.
674 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
676         PR target/59396
677         * config/avr/avr.c (avr_set_current_function): Pass function name
678         through default_strip_name_encoding before sanity checking instead
679         of skipping the first char of the assembler name.
681 2014-03-13  Richard Henderson  <rth@redhat.com>
683         PR debug/60438
684         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
685         (ix86_force_to_memory, ix86_free_from_memory): Remove.
686         * config/i386/i386-protos.h: Likewise.
687         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
688         in the expander instead of a splitter.
689         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
690         any possibility of requiring a memory.
691         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
692         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
693         (fp branch splitters): Update for ix86_split_fp_branch.
694         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
695         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
696         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
697         (*fop_<MODEF>_2_i387): Remove f/r alternative.
698         (*fop_<MODEF>_3_i387): Likewise.
699         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
700         (splitters for the fop_* register patterns): Remove.
701         (fscalexf4_i387): Rename from *fscalexf4_i387.
702         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
703         
704 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
706         PR tree-optimization/59779
707         * tree-dfa.c (get_ref_base_and_extent): Use double_int
708         type for bitsize and maxsize instead of HOST_WIDE_INT.
710 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
712         PR rtl-optimization/57320
713         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
714         the CFG after thread_prologue_and_epilogue_insns.
716 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
718         PR rtl-optimization/57189
719         * lra-constraints.c (process_alt_operands): Disfavor spilling
720         vector pseudos.
722 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
724         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
725         messages.
727 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
729         PR tree-optimization/59025
730         PR middle-end/60418
731         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
732         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
734 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
736         PR target/60486
737         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
738         calls of avr_out_plus_1.
740 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
742         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
743         BB's single pred and update the father loop's latch info later.
745 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
747         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
748         (VEC_M): Likewise.
749         (VEC_N): Likewise.
750         (VEC_R): Likewise.
751         (VEC_base): Likewise.
752         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
753         registers, we need to swap double words in little endian mode.
755         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
756         to be a container mode for 128-bit integer operations added in ISA
757         2.07.  Unlike TImode and PTImode, the preferred register set is
758         the Altivec/VMX registers for the 128-bit operations.
760         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
761         declarations.
762         (rs6000_split_128bit_ok_p): Likewise.
764         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
765         macros for creating ISA 2.07 normal and overloaded builtin
766         functions with 3 arguments.
767         (BU_P8V_OVERLOAD_3): Likewise.
768         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
769         for use as overloaded functions.
770         (VPERM_1TI_UNS): Likewise.
771         (VSEL_1TI): Likewise.
772         (VSEL_1TI_UNS): Likewise.
773         (ST_INTERNAL_1ti): Likewise.
774         (LD_INTERNAL_1ti): Likewise.
775         (XXSEL_1TI): Likewise.
776         (XXSEL_1TI_UNS): Likewise.
777         (VPERM_1TI): Likewise.
778         (VPERM_1TI_UNS): Likewise.
779         (XXPERMDI_1TI): Likewise.
780         (SET_1TI): Likewise.
781         (LXVD2X_V1TI): Likewise.
782         (STXVD2X_V1TI): Likewise.
783         (VEC_INIT_V1TI): Likewise.
784         (VEC_SET_V1TI): Likewise.
785         (VEC_EXT_V1TI): Likewise.
786         (EQV_V1TI): Likewise.
787         (NAND_V1TI): Likewise.
788         (ORC_V1TI): Likewise.
789         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
790         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
791         overloaded builtin.
792         (VADDUQM): Likewise.
793         (VSUBCUQ): Likewise.
794         (VADDEUQM): Likewise.
795         (VADDECUQ): Likewise.
796         (VSUBEUQM): Likewise.
797         (VSUBECUQ): Likewise.
799         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
800         __int128_t and __uint128_t types.
801         (__uint128_type): Likewise.
802         (altivec_categorize_keyword): Add support for vector __int128_t,
803         vector __uint128_t, vector __int128, and vector unsigned __int128
804         as a container type for TImode operations that need to be done in
805         VSX/Altivec registers.
806         (rs6000_macro_to_expand): Likewise.
807         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
808         to support 128-bit integer instructions vaddcuq, vadduqm,
809         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
810         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
812         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
813         for V1TImode, and set up preferences to use VSX/Altivec registers.
814         Setup VSX reload handlers.
815         (rs6000_debug_reg_global): Likewise.
816         (rs6000_init_hard_regno_mode_ok): Likewise.
817         (rs6000_preferred_simd_mode): Likewise.
818         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
819         (easy_altivec_constant): Likewise.
820         (output_vec_const_move): Likewise.
821         (rs6000_expand_vector_set): Convert V1TImode set and extract to
822         simple move.
823         (rs6000_expand_vector_extract): Likewise.
824         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
825         addressing.
826         (rs6000_const_vec): Add support for V1TImode.
827         (rs6000_emit_le_vsx_load): Swap double words when loading or
828         storing TImode/V1TImode.
829         (rs6000_emit_le_vsx_store): Likewise.
830         (rs6000_emit_le_vsx_move): Likewise.
831         (rs6000_emit_move): Add support for V1TImode.
832         (altivec_expand_ld_builtin): Likewise.
833         (altivec_expand_st_builtin): Likewise.
834         (altivec_expand_vec_init_builtin): Likewise.
835         (altivec_expand_builtin): Likewise.
836         (rs6000_init_builtins): Add support for V1TImode type.  Add
837         support for ISA 2.07 128-bit integer builtins.  Define type names
838         for the VSX/Altivec vector types.
839         (altivec_init_builtins): Add support for overloaded vector
840         functions with V1TImode type.
841         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
842         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
843         external function.
844         (rs6000_split_128bit_ok_p): Likewise.
845         (rs6000_handle_altivec_attribute): Create V1TImode from vector
846         __int128_t and vector __uint128_t.
848         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
849         and mode attributes.
850         (VSX_M): Likewise.
851         (VSX_M2): Likewise.
852         (VSm): Likewise.
853         (VSs): Likewise.
854         (VSr): Likewise.
855         (VSv): Likewise.
856         (VS_scalar): Likewise.
857         (VS_double): Likewise.
858         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
860         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
861         we support the ISA 2.07 128-bit integer arithmetic instructions.
862         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
863         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
864         and TImode types for use with the builtin functions.
865         (V1TI_type_node): Likewise.
866         (unsigned_V1TI_type_node): Likewise.
867         (intTI_type_internal_node): Likewise.
868         (uintTI_type_internal_node): Likewise.
870         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
871         128-bit builtin functions.
872         (UNSPEC_VADDEUQM): Likewise.
873         (UNSPEC_VADDECUQ): Likewise.
874         (UNSPEC_VSUBCUQ): Likewise.
875         (UNSPEC_VSUBEUQM): Likewise.
876         (UNSPEC_VSUBECUQ): Likewise.
877         (VM): Add V1TImode to vector mode iterators.
878         (VM2): Likewise.
879         (VI_unit): Likewise.
880         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
881         (altivec_vaddcuq): Likewise.
882         (altivec_vsubuqm): Likewise.
883         (altivec_vsubcuq): Likewise.
884         (altivec_vaddeuqm): Likewise.
885         (altivec_vaddecuq): Likewise.
886         (altivec_vsubeuqm): Likewise.
887         (altivec_vsubecuq): Likewise.
889         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
890         mode iterators.
891         (BOOL_128): Likewise.
892         (BOOL_REGS_OUTPUT): Likewise.
893         (BOOL_REGS_OP1): Likewise.
894         (BOOL_REGS_OP2): Likewise.
895         (BOOL_REGS_UNARY): Likewise.
896         (BOOL_REGS_AND_CR0): Likewise.
898         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
899         128-bit integer builtin support.
900         (vec_vadduqm): Likewise.
901         (vec_vaddecuq): Likewise.
902         (vec_vaddeuqm): Likewise.
903         (vec_vsubecuq): Likewise.
904         (vec_vsubeuqm): Likewise.
905         (vec_vsubcuq): Likewise.
906         (vec_vsubuqm): Likewise.
908         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
909         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
910         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
911         128-bit integer add/subtract to ISA 2.07.
913 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
915         * config/arc/arc.c (arc_predicate_delay_insns):
916         Fix third argument passed to conditionalize_nonjump.
918 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
920         * config/aarch64/aarch64-builtins.c
921         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
922         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
923         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
924         instead of __builtin_lfloor.
925         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
927 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
929         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
930         (tree_ssa_ifcombine_bb_1): New function.
931         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
932         is an empty forwarder block to then_bb or vice versa and then_bb
933         and else_bb are effectively swapped.
935 2014-03-12  Christian Bruel  <christian.bruel@st.com>
937         PR target/60264
938         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
939         REG_CFA_DEF_CFA note.
940         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
941         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
943 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
945         PR tree-optimization/60454
946         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
948 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
950         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
951         Do not define target_cpu_default2 to generic.
952         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
953         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
954         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
956 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
957             Marc Glisse  <marc.glisse@inria.fr>
959         PR tree-optimization/60502
960         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
961         instead of build_low_bits_mask.
963 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
965         PR middle-end/60482
966         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
967         if there are multiple uses, but op doesn't live on E edge.
968         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
969         clobber stmts before __builtin_unreachable.
971 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
973         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
974         hard_frame_pointer_rtx.
975         * cse.c (cse_insn): Remove volatile check.
976         * cselib.c (cselib_process_insn): Likewise.
977         * dse.c (scan_insn): Likewise.
979 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
981         * config/arc/arc.c (conditionalize_nonjump): New function,
982         broken out of ...
983         (arc_ifcvt): ... this.
984         (arc_predicate_delay_insns): Use it.
986 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
988         * config/arc/predicates.md (extend_operand): During/after reload,
989         allow const_int_operand.
990         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
991         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
992         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
993         to "i".
994         (umulsi3_highpart_i): Likewise.
996 2014-03-11  Richard Biener  <rguenther@suse.de>
998         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
999         Add asserts to guard possible wrong-code bugs.
1001 2014-03-11  Richard Biener  <rguenther@suse.de>
1003         PR tree-optimization/60429
1004         PR tree-optimization/60485
1005         * tree-ssa-structalias.c (set_union_with_increment): Properly
1006         take into account all fields that overlap the shifted vars.
1007         (do_sd_constraint): Likewise.
1008         (do_ds_constraint): Likewise.
1009         (get_constraint_for_ptr_offset): Likewise.
1011 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1013         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1014         (nios2_compute_frame_layout):
1015         Add calculation of cfun->machine->fp_save_offset.
1016         (nios2_expand_prologue): Correct setting of frame pointer register
1017         in prologue.
1018         (nios2_expand_epilogue): Update recovery of stack pointer from
1019         frame pointer accordingly.
1020         (nios2_initial_elimination_offset): Update calculation of offset
1021         for eliminating to HARD_FRAME_POINTER_REGNUM.
1023 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1025         PR ipa/60457
1026         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1027         cgraph_get_create_node on VAR_DECLs.
1029 2014-03-10  Richard Biener  <rguenther@suse.de>
1031         PR middle-end/60474
1032         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1034 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1036         * config/vms/vms.opt (vms_float_format): New variable.
1038 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1040         * doc/invoke.texi (-fcilkplus): Update implementation status.
1042 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1043             Richard Biener  <rguenther@suse.de>
1045         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1046         consistently accross all TUs.
1047         (run_gcc): Enable -fshort-double automatically at link at link-time
1048         and disallow override.
1050 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1052         PR target/58271
1053         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1054         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1055         if they can't be used.
1057 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1059         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1060         for Solaris 11/x86 ld.
1061         * configure: Regenerate.
1063 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1065         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1066         (LIB_TLS_SPEC): Save as ld_tls_libs.
1067         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1068         (HAVE_AS_IX86_TLSLDM): New test.
1069         * configure, config.in: Regenerate.
1070         * config/i386/i386.c (legitimize_tls_address): Fall back to
1071         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1072         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1073         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1074         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1076 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1078         * common.opt (fira-loop-pressure): Mark as optimization.
1080 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1082         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1083         an OpenMP mappable type.
1085 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1087         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1088         MULTILIB_OSDIRNAMES is not defined.
1090 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1091             Meador Inge  <meadori@codesourcery.com>
1093         PR target/58595
1094         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1095         (arm_legitimize_address): Call legitimize_tls_address for any
1096         arm_tls_referenced_p expression, handle constant addend.  Call it
1097         before testing for !TARGET_ARM.
1098         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1100 2014-03-06  Richard Biener  <rguenther@suse.de>
1102         PR middle-end/60445
1103         PR lto/60424
1104         PR lto/60427
1105         Revert
1106         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1108         * tree-streamer.c (record_common_node): Assert we don't record
1109         nodes with type double.
1110         (preload_common_node): Skip type double, complex double and double
1111         pointer since it is now frontend dependent due to fshort-double option.
1113 2014-03-06  Richard Biener  <rguenther@suse.de>
1115         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1116         or -fno-lto is specified and the linker has full plugin support.
1117         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1118         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1119         * lto-wrapper.c (merge_and_complain): Merge compile-time
1120         optimization levels.
1121         (run_gcc): And pass it through to the link options.
1123 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
1125         PR debug/60381
1126         Revert:
1127         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1128         PR debug/59992
1129         * cselib.c (remove_useless_values): Skip to avoid quadratic
1130         behavior if the condition moved from...
1131         (cselib_process_insn): ... here holds.
1133 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1135         PR plugins/59335
1136         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1137         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1139         PR plugins/59335
1140         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1141         (TM_H): Add x86-tune.def.
1143 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1145         * config/aarch64/aarch64.c (generic_tunings):
1146         Use cortexa57_extra_costs.
1148 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1150         PR lto/60404
1151         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1152         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1153         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1154         cost for in_lto_p.
1156 2014-03-04  Heiher  <r@hev.cc>
1158         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1159         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1161 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1163         * config/i386/predicates.md (const2356_operand): Change to ...
1164         (const2367_operand): ... this.
1165         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1166         const2367_operand.
1167         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1168         (*avx512pf_scatterpf<mode>sf): Ditto.
1169         (avx512pf_scatterpf<mode>df): Ditto.
1170         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1171         (*avx512pf_scatterpf<mode>df): Ditto.
1172         * config/i386/i386.c (ix86_expand_builtin): Update
1173         incorrect hint operand error message.
1175 2014-03-04  Richard Biener  <rguenther@suse.de>
1177         * lto-section-in.c (lto_get_section_data): Fix const cast.
1179 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1181         * tree-streamer.c (record_common_node): Assert we don't record
1182         nodes with type double.
1183         (preload_common_node): Skip type double, complex double and double
1184         pointer since it is now frontend dependent due to fshort-double option.
1186 2014-03-04  Richard Biener  <rguenther@suse.de>
1188         PR lto/60405
1189         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1190         (lto_input_toplevel_asms): Likewise.
1191         * lto-section-in.c (lto_get_section_data): Instead do it here
1192         for every section.
1194 2014-03-04  Richard Biener  <rguenther@suse.de>
1196         PR tree-optimization/60382
1197         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1198         dead PHIs a reduction.
1200 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1202         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1203         hint value.
1204         (_mm_prefetch): Move out of GCC target("sse") pragma.
1205         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1206         GCC target("prfchw") pragma.
1207         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1208         for locality <= 2.
1209         * config/i386/i386.c (ix86_option_override_internal): Enable
1210         -mprfchw with -mprefetchwt1.
1212 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1214         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1215         Mark as varying.
1217 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1219         * opts.h (CL_PCH_IGNORE): Define.
1220         * targhooks.c (option_affects_pch_p):
1221         Return false for options that have CL_PCH_IGNORE set.
1222         * opt-functions.awk: Process PchIgnore.
1223         * doc/options.texi: Document PchIgnore.
1225         * config/arc/arc.opt (misize): Add PchIgnore property.
1227 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1229         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1230         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1231         constraint on constants to permit them being loaded into
1232         GENERAL_REGS or BASE_REGS.
1234 2014-03-03  Nick Clifton  <nickc@redhat.com>
1236         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1237         anti-cacnonical alternatives.
1238         (negandhi3_real): New pattern.
1239         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1241 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1243         * config/avr/avr-mcus.def: Remove atxmega16x1.
1244         * config/avr/avr-tables.opt: Regenerate.
1245         * config/avr/t-multilib: Regenerate.
1246         * doc/avr-mmcu.texi: Regenerate.
1248 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1249             Mircea Namolaru  <mircea.namolaru@inria.fr>
1251         PR tree-optimization/58028
1252         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1253         scalar dimensions.
1255 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1257         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1258         not handled by recognizers.
1260 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1262         PR middle-end/60175
1263         * function.c (expand_function_end): Don't emit
1264         clobber_return_register sequence if clobber_after is a BARRIER.
1265         * cfgexpand.c (construct_exit_block): Append instructions before
1266         return_label to prev_bb.
1268 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1270         * config/rs6000/constraints.md: Document reserved use of "wc".
1272 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1274         PR ipa/60150
1275         * ipa.c (function_and_variable_visibility): When dissolving comdat
1276         group, also set all symbols to local.
1278 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1280         PR ipa/60306
1282         Revert:
1283         2013-12-14   Jan Hubicka  <jh@suse.cz>
1284         PR middle-end/58477
1285         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1287 2014-03-02  Jon Beniston  <jon@beniston.com>
1289         PR bootstrap/48230
1290         PR bootstrap/50927
1291         PR bootstrap/52466
1292         PR target/46898
1293         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1294           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1295         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1296         (simple_return, *simple_return): New patterns
1297         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1298         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1300 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1302         * dwarf2out.c (gen_subprogram_die): Tidy.
1304 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1306         PR target/60071
1307         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1308         (*mov_t_msb_neg_negc): ... this new insn.
1310 2014-02-28  Jason Merrill  <jason@redhat.com>
1312         PR c++/58678
1313         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1314         function.
1316 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1318         PR c++/60314
1319         * dwarf2out.c (decltype_auto_die): New static.
1320         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1321         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1322         (is_cxx_auto): Likewise.
1324 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1326         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1327         we are not using general regs only.
1329 2014-02-28  Richard Biener  <rguenther@suse.de>
1331         PR target/60280
1332         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1333         previous fix and only allow to remove trivial pre-headers
1334         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1335         (remove_forwarder_block): Properly update the latch of a loop.
1337 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1339         PR debug/59992
1340         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1341         (cselib_preserved_hash_table): New.
1342         (preserve_constants_and_equivs): Move preserved vals to it.
1343         (cselib_find_slot): Look it up first.
1344         (cselib_init): Initialize it.
1345         (cselib_finish): Release it.
1346         (dump_cselib_table): Dump it.
1348 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1350         PR debug/59992
1351         * cselib.c (remove_useless_values): Skip to avoid quadratic
1352         behavior if the condition moved from...
1353         (cselib_process_insn): ... here holds.
1355 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1357         PR debug/57232
1358         * var-tracking.c (vt_initialize): Apply the same condition to
1359         preserve the CFA base value.
1361 2014-02-28  Joey Ye  <joey.ye@arm.com>
1363         PR target/PR60169
1364         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1365         if reload in progress or completed.
1367 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1369         PR middle-end/60147
1370         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1371         NAMELIST_DECL.
1373 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1375         * doc/tm.texi.in (Condition Code Status): Update documention for
1376         relative locations of cc0-setter and cc0-user.
1378 2014-02-27  Jeff Law  <law@redhat.com>
1380         PR rtl-optimization/52714
1381         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1382         into two independent simple sets, if I3 is a jump, ensure the
1383         pattern we place into I3 is a (set (pc) ...).
1385 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1386             Jeff Law  <law@redhat.com>
1388         PR rtl-optimization/49847
1389         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1390         are in different blocks.
1391         * doc/tm.texi (Condition Code Status): Update documention for
1392         relative locations of cc0-setter and cc0-user.
1394 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1396         PR target/59222
1397         * lra.c (lra_emit_add): Check SUBREG too.
1399 2014-02-27  Andreas Schwab  <schwab@suse.de>
1401         * config/m68k/m68k.c (m68k_option_override): Disable
1402         -flive-range-shrinkage for classic m68k.
1403         (m68k_override_options_after_change): Likewise.
1405 2014-02-27  Marek Polacek  <polacek@redhat.com>
1407         PR middle-end/59223
1408         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1409         -Wmaybe-uninitialized.
1411 2014-02-27  Alan Modra  <amodra@gmail.com>
1413         PR target/57936
1414         * reload1.c (emit_input_reload_insns): When reload_override_in,
1415         set old to rl->in_reg when rl->in_reg is a subreg.
1417 2014-02-26  Richard Biener  <rguenther@suse.de>
1419         PR bootstrap/60343
1420         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1422 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1424         * common/config/i386/predicates.md (const1256_operand): Remove.
1425         (const2356_operand): New.
1426         (const_1_to_2_operand): Remove.
1427         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1428         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1429         (*avx512pf_gatherpf<mode>sf): Ditto.
1430         (avx512pf_gatherpf<mode>df): Ditto.
1431         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1432         (*avx512pf_gatherpf<mode>df): Ditto.
1433         (avx512pf_scatterpf<mode>sf): Ditto.
1434         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1435         (*avx512pf_scatterpf<mode>sf): Ditto.
1436         (avx512pf_scatterpf<mode>df): Ditto.
1437         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1438         (*avx512pf_scatterpf<mode>df): Ditto.
1439         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1441 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1443         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1444         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1445         (_mm512_mask_testn_epi64_mask): Move to ...
1446         * config/i386/avx512cdintrin.h: Here.
1447         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1448         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1449         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1450         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1451         TARGET_AVX512F from TARGET_AVX512CD.
1453 2014-02-26  Richard Biener  <rguenther@suse.de>
1455         PR ipa/60327
1456         * ipa.c (walk_polymorphic_call_targets): Properly guard
1457         call to inline_update_overall_summary.
1459 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1461         PR target/60280
1462         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1463         and latches only if requested.  Fix latch if it is removed.
1464         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1465         LOOPS_HAVE_PREHEADERS.
1467 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1469         * builtins.c (expand_builtin_thread_pointer): Create a new target
1470         when the target is NULL.
1472 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1474         PR rtl-optimization/60317
1475         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1476         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1477         * lra-assigns.c: Include params.h.
1478         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1479         other reload pseudos considerations.
1481 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1483         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1484         to use canonical form for nor<mode>3.
1486 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1488         PR target/55426
1489         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1490         conversions.
1492 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1494         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1495         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1496         (ix86_handle_option): Handle OPT_mprefetchwt1.
1497         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1498         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1499         PREFETCHWT1 CPUID.
1500         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1501         OPTION_MASK_ISA_PREFETCHWT1.
1502         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1503         (PTA_PREFETCHWT1): New.
1504         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1505         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1506         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1507         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1508         (*prefetch_avx512pf_<mode>_: Change into ...
1509         (*prefetch_prefetchwt1_<mode>: This.
1510         * config/i386/i386.opt (mprefetchwt1): New.
1511         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1512         (_mm_prefetch): Handle intent to write.
1513         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1515 2014-02-25  Richard Biener  <rguenther@suse.de>
1517         PR middle-end/60291
1518         * emit-rtl.c (mem_attrs_htab): Remove.
1519         (mem_attrs_htab_hash): Likewise.
1520         (mem_attrs_htab_eq): Likewise.
1521         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1522         (init_emit_once): Do not allocate mem_attrs_htab.
1524 2014-02-25  Richard Biener  <rguenther@suse.de>
1526         PR lto/60319
1527         * lto-opts.c (lto_write_options): Output non-explicit conservative
1528         -fwrapv, -fno-trapv and -fno-strict-overflow.
1529         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1530         (run_gcc): And pass them through.
1532 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1534         * sel-sched.c (calculate_new_fences): New parameter ptime.
1535         Calculate it as a maximum over all fence cycles.
1536         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1537         Print the final schedule timing when sched_verbose.
1539 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1541         PR rtl-optimization/60292
1542         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1543         bit fot the fence instruction.
1545 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1547         * calls.h: Fix typo in comment.
1549 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1551         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1552         adjusting offsetable addresses.
1554 2014-02-24  Guozhi Wei  <carrot@google.com>
1556         * sparseset.h (sparseset_pop): Fix the wrong index.
1558 2014-02-24  Walter Lee  <walt@tilera.com>
1560         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1561         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1562         triplet.
1563         * common/config/tilegx/tilegx-common.c
1564         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1565         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1566         (LINK_SPEC): Ditto.
1567         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1568         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1569         (tilegx_gimplify_va_arg_expr): Handle big endian.
1570         (tilegx_expand_unaligned_load): Ditto.
1571         (tilegx_expand_unaligned_store): Ditto.
1572         (TARGET_RETURN_IN_MSB): New.
1573         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1574         (TARGET_ENDIAN_DEFAULT): New.
1575         (TARGET_BIG_ENDIAN): Handle big endian.
1576         (BYTES_BIG_ENDIAN): Ditto.
1577         (WORDS_BIG_ENDIAN): Ditto.
1578         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1579         (ENDIAN_SPEC): New.
1580         (EXTRA_SPECS): New.
1581         * config/tilegx/tilegx.md (extv): Handle big endian.
1582         (extzv): Ditto.
1583         (insn_st<n>): Ditto.
1584         (insn_st<n>_add<bitsuffix>): Ditto.
1585         (insn_stnt<n>): Ditto.
1586         (insn_stnt<n>_add<bitsuffix>):Ditto.
1587         (vec_interleave_highv8qi): Handle big endian.
1588         (vec_interleave_highv8qi_be): New.
1589         (vec_interleave_highv8qi_le): New.
1590         (insn_v1int_h): Handle big endian.
1591         (vec_interleave_lowv8qi): Handle big endian.
1592         (vec_interleave_lowv8qi_be): New.
1593         (vec_interleave_lowv8qi_le): New.
1594         (insn_v1int_l): Handle big endian.
1595         (vec_interleave_highv4hi): Handle big endian.
1596         (vec_interleave_highv4hi_be): New.
1597         (vec_interleave_highv4hi_le): New.
1598         (insn_v2int_h): Handle big endian.
1599         (vec_interleave_lowv4hi): Handle big endian.
1600         (vec_interleave_lowv4hi_be): New.
1601         (vec_interleave_lowv4hi_le): New.
1602         (insn_v2int_l): Handle big endian.
1603         (vec_interleave_highv2si): Handle big endian.
1604         (vec_interleave_highv2si_be): New.
1605         (vec_interleave_highv2si_le): New.
1606         (insn_v4int_h): Handle big endian.
1607         (vec_interleave_lowv2si): Handle big endian.
1608         (vec_interleave_lowv2si_be): New.
1609         (vec_interleave_lowv2si_le): New.
1610         (insn_v4int_l): Handle big endian.
1611         * config/tilegx/tilegx.opt (mbig-endian): New option.
1612         (mlittle-endian): New option.
1613         * doc/install.texi: Document tilegxbe-linux.
1614         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1616 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1618         PR ipa/60266
1619         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1620         there are no parameter descriptors.
1622 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1624         PR rtl-optimization/60268
1625         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1626         initialization to ...
1627         (sched_rgn_init): ... here.
1628         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1630 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1632         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1633         names.
1635 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1637         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1638         definition.
1640 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1642         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1643         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1645 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1647         * config/microblaze/predicates.md: Add cmp_op predicate.
1648         * config/microblaze/microblaze.md: Add branch_compare instruction
1649         which uses cmp_op predicate and emits cmp insn before branch.
1650         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1651         to microblaze_expand_conditional_branch and consolidate logic.
1652         (microblaze_expand_conditional_branch): emit branch_compare
1653         insn instead of handling cmp op separate from branch insn.
1655 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1657         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1658         to permit subregs.
1660 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1662         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1663         define_insn with define_expand and new define_insn
1664         *altivec_lve<VI_char>x_internal.
1665         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1666         and new define_insn *altivec_stve<VI_char>x_internal.
1667         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1668         prototype.
1669         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1670         lve*x built-ins.
1671         (altivec_expand_stvex_be): New function.
1673 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1675         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1676         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1677         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1678         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1680 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1682         PR target/60298
1683         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1684         instead of emit_move_insn.
1686 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1688         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1689         vspltw with vsldoi.
1690         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1691         gen_altivec_vsumsws.
1693 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1695         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1696         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1697         (altivec_lvxl_<mode>): New define_expand incorporating
1698         -maltivec=be semantics where needed.
1699         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1700         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1701         semantics where needed.
1702         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1703         (altivec_stvx_<mode>): New define_expand incorporating
1704         -maltivec=be semantics where needed.
1705         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1706         VM2 iterator instead of V4SI.
1707         (altivec_stvxl_<mode>): New define_expand incorporating
1708         -maltivec=be semantics where needed.
1709         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1710         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1711         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1712         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1713         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1714         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1715         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1716         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1717         ALTIVEC_BUILTIN_STVXL.
1718         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1719         (altivec_expand_stvx_be): Likewise.
1720         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1721         (altivec_expand_lvx_be): Likewise.
1722         (altivec_expand_stvx_be): Likewise.
1723         (altivec_expand_builtin): Add cases for
1724         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1725         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1726         (altivec_init_builtins): Add definitions for
1727         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1728         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1730 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1732         * doc/invoke.texi (mvirt, mno-virt): Document.
1733         * config/mips/mips.opt (mvirt): New option.
1734         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1736 2014-02-21  Richard Biener  <rguenther@suse.de>
1738         PR tree-optimization/60276
1739         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1740         (STMT_VINFO_MIN_NEG_DIST): New macro.
1741         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1742         STMT_VINFO_MIN_NEG_DIST.
1743         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1744         made for negative dependence distances still hold.
1746 2014-02-21  Richard Biener  <rguenther@suse.de>
1748         PR middle-end/60291
1749         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1750         DECL_INITIAL for globals not in the current function context.
1752 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1754         PR tree-optimization/56490
1755         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1756         * tree-ssa-uninit.c: Include params.h.
1757         (compute_control_dep_chain): Add num_calls argument, return false
1758         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1759         num_calls to recursive call.
1760         (find_predicates): Change dep_chain into normal array,
1761         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1762         variable and adjust compute_control_dep_chain caller.
1763         (find_def_preds): Likewise.
1765 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1767         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1768         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1770 2014-02-21  Nick Clifton  <nickc@redhat.com>
1772         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1773         (pushhi1): Likewise.
1774         (popqi1): Add mode to pre_dec.
1775         (pophi1): Likewise.
1777 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1779         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1780         mode for mask of V8SFmode permutation.
1782 2014-02-20  Richard Henderson  <rth@redhat.com>
1784         PR c++/60272
1785         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1786         a new pseudo for OLDVAL.
1788 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1790         PR target/57896
1791         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1792         gen_reg_rtx if d->testing_p.
1793         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1794         if d->testing_p and we will certainly return true.
1795         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1796         if d->testing_p.
1798 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1800         * emit-rtl.c (gen_reg_rtx): Assert that
1801         crtl->emit.regno_pointer_align_length is non-zero.
1803 2014-02-20  Richard Henderson  <rth@redhat.com>
1805         PR c++/60272
1806         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1807         on failure the store back into EXPECT.
1809 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1810             Sandra Loosemore  <sandra@codesourcery.com>
1812         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1813         * config/nios2/nios2.c (nios2_function_profiler): Add
1814         -fPIC (flag_pic == 2) support.
1815         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1816         (nios2_large_offset_p): New function.
1817         (nios2_unspec_reloc_p): Move up position, update to use
1818         nios2_large_offset_p.
1819         (nios2_unspec_address): Remove function.
1820         (nios2_unspec_offset): New function.
1821         (nios2_large_got_address): New function.
1822         (nios2_got_address): Add large offset support.
1823         (nios2_legitimize_tls_address): Update usage of removed and new
1824         functions.
1825         (nios2_symbol_binds_local_p): New function.
1826         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1827         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1828         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1829         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1830         processing for (const (unspec ...)).
1831         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1833 2014-02-20  Richard Biener  <rguenther@suse.de>
1835         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1836         doing the substitution.
1837         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1839 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1841         PR ipa/55260
1842         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1843         info when checking whether lattices are bottom.
1845 2014-02-20  Richard Biener  <rguenther@suse.de>
1847         PR middle-end/60221
1848         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1849         regions at -O0.
1851 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1853         PR ipa/58555
1854         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1855         parameter specifying the scaling.
1856         (inline_call): Update.
1857         (want_inline_recursively): Guard division by zero.
1858         (recursive_inlining): Update.
1859         * ipa-inline.h (clone_inlined_nodes): Update.
1861 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1863         PR target/60204
1864         * config/i386/i386.c (classify_argument): Pass structures of size
1865         64 bytes or less in register.
1867 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1868             Kirill Yukhin  <kirill.yukhin@intel.com>
1870         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1871         (_mm_rcp28_round_ss): Ditto.
1872         (_mm_rsqrt28_round_sd): Ditto.
1873         (_mm_rsqrt28_round_ss): Ditto.
1874         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1875         (_mm_rcp14_round_ss): Ditto.
1876         (_mm_rsqrt14_round_sd): Ditto.
1877         (_mm_rsqrt14_round_ss): Ditto.
1878         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1879         the first input operand, get rid of match_dup.
1880         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1881         attribute to sse.
1882         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1883         Ditto.
1884         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1885         operand as the first input operand, set type attribute.
1886         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1887         Set type attribute.
1888         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1889         operand as the first input operand, set type attribute.
1891 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1893         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1894         bit of zero.
1896 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1898         PR target/60207
1899         * config/i386/i386.c (construct_container): Remove TFmode check
1900         for X86_64_INTEGER_CLASS.
1902 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1904         PR target/59794
1905         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1906         only when -Wpsabi is enabled.
1908 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1910          PR target/59799
1911         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1912         passing arrays in registers are the same as for structs, so remove the
1913         special case for them.
1915 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1917         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1918         destination type, extract only the valid bits if the source type is not
1919         integral and has a different mode.
1921 2014-02-19  Richard Biener  <rguenther@suse.de>
1923         PR ipa/60243
1924         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1925         for all calls.
1927 2014-02-19  Richard Biener  <rguenther@suse.de>
1929         PR ipa/60243
1930         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1931         (ipa_modify_call_arguments): Emit an argument load explicitely and
1932         preserve virtual SSA form there and for the replacement call.
1933         Do not update SSA form nor free dominance info.
1935 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1937         * ipa.c (function_and_variable_visibility): Also clear WEAK
1938         flag when disolving COMDAT_GROUP.
1940 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1942         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1943         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1944         not devirtualizing.
1945         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1946         do more sanity checks.
1947         (detect_type_change): Return true when giving up early.
1948         (compute_complex_assign_jump_func): Fix type parameter of
1949         ipa_set_ancestor_jf.
1950         (compute_complex_ancestor_jump_func): Likewise.
1951         (update_jump_functions_after_inlining): Fix updating of
1952         ancestor function.
1953         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1955 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1957         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1958         inline clones when edge disappears.
1960 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1962         PR target/60203
1963         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1964         Split 64-bit moves into 2 patterns.  Do not allow the use of
1965         direct move for TDmode in little endian, since the decimal value
1966         has little endian bytes within a word, but the 64-bit pieces are
1967         ordered in a big endian fashion, and normal subreg's of TDmode are
1968         not allowed.
1969         (mov<mode>_64bit_dm): Likewise.
1970         (movtd_64bit_nodm): Likewise.
1972 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1974         PR tree-optimization/60174
1975         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1976         statement of an SSA_NAME that occurs in an abnormal PHI node.
1978 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1980         PR sanitizer/60142
1981         * final.c (SEEN_BB): Remove.
1982         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1983         (final_scan_insn): Don't force_source_line on second
1984         NOTE_INSN_BASIC_BLOCK.
1986 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1988         PR target/60205
1989         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1990         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1991         (type_natural_mode): Warn ABI change when %zmm register is not
1992         available for AVX512F vector value passing.
1994 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1996         PR target/60193
1997         * config/i386/i386.c (ix86_expand_prologue): Use value in
1998         rax register as displacement when restoring %r10 or %rax.
1999         Fix wrong offset when restoring both registers.
2001 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2003         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2004         assertion with conditional return.
2006 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2007             Uros Bizjak  <ubizjak@gmail.com>
2009         PR driver/60233
2010         * config/i386/driver-i386.c (host_detect_local_cpu): If
2011         YMM state is not saved by the OS, also clear has_f16c.  Move
2012         CPUID 0x80000001 handling before YMM state saving checking.
2014 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2016         PR rtl-optimization/58960
2017         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2018         factored out from ...
2019         (sched_init): ... here.
2020         (free_global_sched_pressure_data): New, factored out from ...
2021         (sched_finish): ... here.
2022         * sched-int.h (free_global_sched_pressure_data): Declare.
2023         * sched-rgn.c (nr_regions_initial): New static global.
2024         (haifa_find_rgns): Initialize it.
2025         (schedule_region): Disable sched-pressure for the newly
2026         generated regions.
2028 2014-02-17  Richard Biener  <rguenther@suse.de>
2030         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2031         release SSA defs of pattern stmts.
2033 2014-02-17  Richard Biener  <rguenther@suse.de>
2035         * tree-inline.c (expand_call_inline): Release the virtual
2036         operand defined by the call we are about to inline.
2038 2014-02-17  Richard Biener  <rguenther@suse.de>
2040         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2042 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2043             Ilya Tocar  <ilya.tocar@intel.com>
2045         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2046         arguments order in builtin.
2047         (_mm512_permutexvar_epi64): Ditto.
2048         (_mm512_mask_permutexvar_epi64): Ditto
2049         (_mm512_maskz_permutexvar_epi32): Ditto
2050         (_mm512_permutexvar_epi32): Ditto
2051         (_mm512_mask_permutexvar_epi32): Ditto
2053 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2055         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2056         (p8_vmrgow): Likewise.
2058 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2060         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2061         endian targets.
2063 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2065         PR target/60203
2066         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2067         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2068         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2069         using direct move instructions on ISA 2.07.  Also adjust
2070         instruction length for 64-bit.
2071         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2072         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2074 2014-02-15  Alan Modra  <amodra@gmail.com>
2076         PR target/58675
2077         PR target/57935
2078         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2079         find_replacement on parts of insn rtl that might be reloaded.
2081 2014-02-15  Richard Biener  <rguenther@suse.de>
2083         PR tree-optimization/60183
2084         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2085         (tree_ssa_phiprop): Calculate and free post-dominators.
2087 2014-02-14  Jeff Law  <law@redhat.com>
2089         PR rtl-optimization/60131
2090         * ree.c (get_extended_src_reg): New function.
2091         (combine_reaching_defs): Use it rather than assuming location of REG.
2092         (find_and_remove_re): Verify first operand of extension is
2093         a REG before adding the insns to the copy list.
2095 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2097         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2098         * configure: Regenerated.
2099         * config.in: Regenerated.
2100         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2101         instead of ASM_SHORT.
2103 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2104             Richard Earnshaw  <rearnsha@arm.com>
2106         PR rtl-optimization/59535
2107         * lra-constraints.c (process_alt_operands): Encourage alternative
2108         when unassigned pseudo class is superset of the alternative class.
2109         (inherit_reload_reg): Don't inherit when optimizing for code size.
2110         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2111         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2112         modes not less than 4 for Thumb1.
2114 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2116         PR pch/60010
2117         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2119 2014-02-14  Richard Biener  <rguenther@suse.de>
2121         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2122         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2123         Do not strip INDIRECT_REFs.
2125 2014-02-14  Richard Biener  <rguenther@suse.de>
2127         PR lto/60179
2128         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2129         DECL_FUNCTION_SPECIFIC_TARGET.
2130         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2131         * tree-streamer-out.c (pack_ts_target_option): Remove.
2132         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2133         (write_ts_function_decl_tree_pointers): Do not stream
2134         DECL_FUNCTION_SPECIFIC_TARGET.
2135         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2136         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2137         (lto_input_ts_function_decl_tree_pointers): Do not stream
2138         DECL_FUNCTION_SPECIFIC_TARGET.
2140 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2142         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2143         (get_initial_def_for_induction, vectorizable_induction): Ignore
2144         debug stmts when looking for exit_phi.
2145         (vectorizable_live_operation): Fix up condition.
2147 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2149         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2150         nreverse() because it changes the content of original tree list.
2152 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2154         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2155         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2157 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2159         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2160         GNU coding standards.
2162 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2164         PR debug/60152
2165         * dwarf2out.c (gen_subprogram_die): Don't call
2166         add_calling_convention_attribute if subr_die is old_die.
2168 2014-02-13  Sharad Singhai  <singhai@google.com>
2170         * doc/optinfo.texi: Fix order of nodes.
2172 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2174         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2175         operands[2], not operands[3].
2177 2014-02-13  Richard Biener  <rguenther@suse.de>
2179         PR bootstrap/59878
2180         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2181         mention the possibility of an in-tree build.
2182         (CLooG): Update recommended version to 0.18.1, mention the
2183         possibility of an in-tree build and clarify that the ISL
2184         bundled with CLooG does not work.
2186 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2188         PR target/43546
2189         * expr.c (compress_float_constant): If x is a hard register,
2190         extend into a pseudo and then move to x.
2192 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2194         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2195         caused by bad second argument to warning_at() with -mhotpatch and
2196         nested functions (e.g. with gfortran).
2198 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2200         * opts.c (option_name): Remove "enabled by default" rider.
2202 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2204         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2206 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2207             Uros Bizjak  <ubizjak@gmail.com>
2209         PR target/60151
2210         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2211         * configure: Regenerated.
2213 2014-02-12  Richard Biener  <rguenther@suse.de>
2215         * vec.c (vec_prefix::calculate_allocation): Move as
2216         inline variant to vec.h.
2217         (vec_prefix::calculate_allocation_1): New out-of-line version.
2218         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2219         (vec_prefix::m_has_auto_buf): Rename to ...
2220         (vec_prefix::m_using_auto_storage): ... this.
2221         (vec_prefix::calculate_allocation): Inline the easy cases
2222         and dispatch to calculate_allocation_1 which doesn't need the
2223         prefix address.
2224         (va_heap::reserve): Use gcc_checking_assert.
2225         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2226         m_using_auto_storage.
2227         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2228         member and adjust.
2229         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2230         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2231         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2233 2014-02-12  Richard Biener  <rguenther@suse.de>
2235         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2236         when we found a dependence.
2238 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2240         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2241         common code...
2242         (maybe_fold_stmt): ... into this new function.
2243         * omp-low.c (lower_omp): Update comment.
2245         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2246         last use.
2248         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2249         dereference.
2251 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2253         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2254         identifiers in comments.
2255         (cortexa53_extra_costs): Likewise.
2256         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2257         (cortexa7_extra_costs): Likewise.
2258         (cortexa12_extra_costs): Likewise.
2259         (cortexa15_extra_costs): Likewise.
2260         (v7m_extra_costs): Likewise.
2262 2014-02-12  Richard Biener  <rguenther@suse.de>
2264         PR middle-end/60092
2265         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2266         of posix_memalign being successful.
2267         (lower_stmt): Restrict lowering of posix_memalign to when
2268         -ftree-bit-ccp is enabled.
2270 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2272         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2273         arg_loc.
2274         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2276 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2278         PR rtl-optimization/60116
2279         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2280         other_insn once the combination has been validated.
2282 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2284         PR lto/59468
2285         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2286         and wrapper.
2287         * ipa-devirt.c: Include demangle.h
2288         (odr_violation_reported): New static variable.
2289         (add_type_duplicate): Update odr_violations.
2290         (maybe_record_node): Add completep parameter; update it.
2291         (record_target_from_binfo): Add COMPLETEP parameter;
2292         update it as needed.
2293         (possible_polymorphic_call_targets_1): Likewise.
2294         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2295         rename FINAL to COMPLETE.
2296         (record_targets_from_bases): Sanity check we found the binfo;
2297         fix COMPLETEP updating.
2298         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2299         parameter, fix computing of COMPLETEP.
2300         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2301         at LTO time do demangling.
2302         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2303         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2304         parameter.
2305         (gimple_get_virt_method_for_binfo): Likewise.
2306         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2307         gimple_get_virt_method_for_vtable): Update prototypes.
2309 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2311         PR target/49008
2312         * genautomata.c (add_presence_absence): Fix typo with
2313         {final_}presence_list.
2315 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2317         PR target/60137
2318         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2319         for VSX/Altivec vectors that land in GPR registers.
2321 2014-02-11  Richard Henderson  <rth@redhat.com>
2322             Jakub Jelinek  <jakub@redhat.com>
2324         PR debug/59776
2325         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2326         around drhs if type conversion to lacc->type is not useless.
2328 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2330         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2331         tuning struct.
2332         (cortex-a57.cortex-a53): Likewise.
2333         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2335 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2337         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2338         arm_restrict_it.
2340 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2342         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2343         add_options_for_arm_vfp3.
2345 2014-02-11  Jeff Law  <law@redhat.com>
2347         PR middle-end/54041
2348         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2349         object with an undesirable mode.
2351 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2353         PR libgomp/60107
2354         * config/i386/sol2-9.h: New file.
2355         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2356         *-*-solaris2.9*): Use it.
2358 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2360         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2361         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2363 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2365         * config/microblaze/microblaze.c: Extend mcpu version format
2367 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
2369         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2371 2014-02-10  Richard Henderson  <rth@redhat.com>
2373         PR target/59927
2374         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2375         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2376         ms-abi vs -mno-accumulate-outgoing-args.
2377         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2378         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2379         respect to ms-abi.
2381 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2383         PR middle-end/60080
2384         * cfgexpand.c (expand_asm_operands): Attach source location to
2385         ASM_INPUT rtx objects.
2386         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2388 2014-02-10  Nick Clifton  <nickc@redhat.com>
2390         * config/mn10300/mn10300.c (popcount): New function.
2391         (mn10300_expand_prologue): Include saved registers in stack usage
2392         count.
2394 2014-02-10  Jeff Law  <law@redhat.com>
2396         PR middle-end/52306
2397         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2398         when changing the SET_DEST of a prior insn to avoid an input reload.
2400 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2402         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2403         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2404         -mcall-openbsd, or -mcall-linux.
2405         (CC1_ENDIAN_BIG_SPEC): Remove.
2406         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2407         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2408         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2409         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2410         and %cc1_endian_default.
2411         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2413 2014-02-10  Richard Biener  <rguenther@suse.de>
2415         PR tree-optimization/60115
2416         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2417         MEM_REF handling.  Properly verify that the accesses are not
2418         out of the objects bound.
2420 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2422         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2423         coretex to cortex.
2425 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2427         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2428         proper constants and fix formatting.
2429         (possible_polymorphic_call_targets): Fix formatting.
2431 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2432             Ilya Tocar  <ilya.tocar@intel.com>
2434         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2435         (_mm512_loadu_epi32): Renamed into...
2436         (_mm512_loadu_si512): This.
2437         (_mm512_storeu_epi32): Renamed into...
2438         (_mm512_storeu_si512): This.
2439         (_mm512_maskz_ceil_ps): Removed.
2440         (_mm512_maskz_ceil_pd): Ditto.
2441         (_mm512_maskz_floor_ps): Ditto.
2442         (_mm512_maskz_floor_pd): Ditto.
2443         (_mm512_floor_round_ps): Ditto.
2444         (_mm512_floor_round_pd): Ditto.
2445         (_mm512_ceil_round_ps): Ditto.
2446         (_mm512_ceil_round_pd): Ditto.
2447         (_mm512_mask_floor_round_ps): Ditto.
2448         (_mm512_mask_floor_round_pd): Ditto.
2449         (_mm512_mask_ceil_round_ps): Ditto.
2450         (_mm512_mask_ceil_round_pd): Ditto.
2451         (_mm512_maskz_floor_round_ps): Ditto.
2452         (_mm512_maskz_floor_round_pd): Ditto.
2453         (_mm512_maskz_ceil_round_ps): Ditto.
2454         (_mm512_maskz_ceil_round_pd): Ditto.
2455         (_mm512_expand_pd): Ditto.
2456         (_mm512_expand_ps): Ditto.
2457         * config/i386/i386.c (ix86_builtins): Remove
2458         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2459         (bdesc_args): Ditto.
2460         * config/i386/predicates.md (const1256_operand): New.
2461         (const_1_to_2_operand): Ditto.
2462         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2463         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2464         (*avx512pf_gatherpf<mode>sf): Ditto.
2465         (avx512pf_gatherpf<mode>df): Ditto.
2466         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2467         (*avx512pf_gatherpf<mode>df): Ditto.
2468         (avx512pf_scatterpf<mode>sf): Ditto.
2469         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2470         (*avx512pf_scatterpf<mode>sf): Ditto.
2471         (avx512pf_scatterpf<mode>df): Ditto.
2472         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2473         (*avx512pf_scatterpf<mode>df): Ditto.
2474         (avx512f_expand<mode>): Removed.
2475         (<shift_insn><mode>3<mask_name>): Change predicate type.
2477 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2479         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2480         not at the end of datarefs vector use ordered_remove to avoid
2481         reordering datarefs vector.
2483         PR c/59984
2484         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2485         mark local addressable non-static vars as GOVD_PRIVATE
2486         instead of GOVD_LOCAL.
2487         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2488         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2489         than copying them.
2491         PR middle-end/60092
2492         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2493         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2494         assume_aligned or alloc_align attributes.
2495         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2496         arguments.  Handle also assume_aligned and alloc_align attributes.
2497         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2498         calls to functions with assume_aligned or alloc_align attributes.
2499         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2501 2014-02-08  Terry Guo  <terry.guo@arm.com>
2503         * doc/invoke.texi: Document ARM -march=armv7e-m.
2505 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2507         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2508         flag on __cilkrts_rethrow builtin.
2510         PR ipa/60026
2511         * ipa-cp.c (determine_versionability): Fail at -O0
2512         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2513         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2515         Revert:
2516         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2518         PR ipa/60026
2519         * tree-inline.c (copy_forbidden): Fail for
2520         __attribute__((optimize (0))) functions.
2522 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2524         * varpool.c: Include pointer-set.h.
2525         (varpool_remove_unreferenced_decls): Variables in other partitions
2526         will not be output; be however careful to not lose information
2527         about partitioning.
2529 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2531         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2532         lookup in the vtable constructor.
2534 2014-02-07  Jeff Law  <law@redhat.com>
2536         PR target/40977
2537         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2538         define_insn_and_split.
2540         * ipa-inline.c (inline_small_functions): Fix typos.
2542 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2544         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2545         (s390_can_use_return_insn): Declare.
2546         * config/s390/s390.h (EPILOGUE_USES): Define.
2547         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2548         instructions.
2549         (s390_chunkify_start): Handle return JUMP_LABELs.
2550         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2551         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2552         (s390_can_use_return_insn): New functions.
2553         (s390_fix_long_loop_prediction): Handle conditional returns.
2554         (TARGET_SET_UP_BY_PROLOGUE): Define.
2555         * config/s390/s390.md (ANY_RETURN): New code iterator.
2556         (*creturn, *csimple_return, return, simple_return): New patterns.
2558 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2560         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2561         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2562         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2563         REG_CFA_RESTORE list when deciding not to restore a register.
2565 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2567         * config/s390/s390.c: Include tree-pass.h and context.h.
2568         (s390_early_mach): New function, split out from...
2569         (s390_emit_prologue): ...here.
2570         (pass_data_s390_early_mach): New pass structure.
2571         (pass_s390_early_mach): New class.
2572         (s390_option_override): Create and register early_mach pass.
2573         Move to end of file.
2575 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2577         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2578         to match for the exit block.
2580 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2582         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2583         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2584         Reject misaligned operands.
2586 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2588         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2590 2014-02-07  Richard Biener  <rguenther@suse.de>
2592         PR middle-end/60092
2593         * gimple-low.c (lower_builtin_posix_memalign): New function.
2594         (lower_stmt): Call it to lower posix_memalign in a way
2595         to make alignment info accessible.
2597 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2599         PR c++/60082
2600         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2601         __builtin_setjmp_receiver.
2603 2014-02-07  Richard Biener  <rguenther@suse.de>
2605         PR middle-end/60092
2606         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2607         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2608         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2609         Handle BUILT_IN_POSIX_MEMALIGN.
2610         (find_func_clobbers): Likewise.
2611         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2612         (call_may_clobber_ref_p_1): Likewise.
2614 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2616         PR ipa/59918
2617         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2618         sanity check.
2620 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2622         PR ipa/59469
2623         * lto-cgraph.c (lto_output_node): Use
2624         symtab_get_symbol_partitioning_class.
2625         (lto_output_varpool_node): likewise.
2626         (symtab_get_symbol_partitioning_class): Move here from
2627         lto/lto-partition.c
2628         * cgraph.h (symbol_partitioning_class): Likewise.
2629         (symtab_get_symbol_partitioning_class): Declare.
2631 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2633         * ggc.h (ggc_internal_cleared_alloc): New macro.
2634         * vec.h (vec_safe_copy): Handle memory stats.
2635         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2636         * target-globals.c (save_target_globals): Likewise.
2638 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2640         PR target/60077
2641         * expr.c (emit_move_resolve_push): Export; be bit more selective
2642         on when to clear alias set.
2643         * expr.h (emit_move_resolve_push): Declare.
2644         * function.h (struct function): Add tail_call_marked.
2645         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2646         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2647         * config/i386/i386.md (TImode move expander): De not call
2648         ix86_expand_push.
2649         (FP push expanders): Preserve memory attributes.
2650         * config/i386/sse.md (push<mode>1): Remove.
2651         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2652         (ix86_expand_push): Remove.
2653         * config/i386/mmx.md (push<mode>1): Remove.
2655 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2657         PR rtl-optimization/60030
2658         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2659         lopart with paradoxical subreg before shifting it up by hprec.
2661 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2663         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2664         Remove extra newline at end of file.
2665         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2666         (arm_issue_rate): Handle cortexa57.
2667         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2668         (cortex-a57.cortex-a53): Likewise.
2670 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2672         PR target/59575
2673         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2674         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2675         bitmask.
2676         (arm_expand_prologue): Adjust all callers.
2677         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2678         info, registers also at the lowest numbered registers side.  Use
2679         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2680         XEXP.
2682         PR debug/59992
2683         * var-tracking.c (adjust_mems): Before adding a SET to
2684         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2686 2014-02-06  Alan Modra  <amodra@gmail.com>
2688         PR target/60032
2689         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2690         change SDmode to DDmode when lra_in_progress.
2692 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2694         PR middle-end/59150
2695         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2696         free_data_ref on the dr first, and before goto again also set dr
2697         to the next dr.  For simd_lane_access, free old datarefs[i] before
2698         overwriting it.  For get_vectype_for_scalar_type failure, don't
2699         free_data_ref if simd_lane_access.
2701         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2703         PR target/60062
2704         * tree.h (opts_for_fn): New inline function.
2705         (opt_for_fn): Define.
2706         * config/i386/i386.c (ix86_function_regparm): Use
2707         opt_for_fn (decl, optimize) instead of optimize.
2709 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2711         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2712         for SYMBOL_REF in large memory model.
2714 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2716         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2717         and crypto support.
2718         (cortex-a57): Likewise.
2719         (cortex-a57.cortex-a53): Likewise.
2721 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2722             Kugan Vivekanandarajah  <kuganv@linaro.org>
2724         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2725         unaligned_access.
2726         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2728 2014-02-06  Richard Biener  <rguenther@suse.de>
2730         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2731         set_loop_copy and initialize_original_copy_tables.
2733 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2735         * config/aarch64/aarch64-simd.md
2736         (aarch64_ashr_simddi): Change QI to SI.
2738 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2739             Jakub Jelinek  <jakub@redhat.com>
2741         PR middle-end/60013
2742         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2743         of the dataflow.
2745 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2747         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2748         CODE_FOR_altivec_vpku[hw]um to
2749         CODE_FOR_altivec_vpku[hw]um_direct.
2750         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2751         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2752         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2753         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2755 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2757         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2758         generation for -maltivec=be.
2759         (altivec_vsumsws): Simplify redundant test.
2761 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2763         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2764         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2765         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2766         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2767         gen_altivec_vpkuwum.
2768         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2769         BYTES_BIG_ENDIAN.
2770         (altivec_vpks<VI_char>ss): Likewise.
2771         (altivec_vpks<VI_char>us): Likewise.
2772         (altivec_vpku<VI_char>us): Likewise.
2773         (altivec_vpku<VI_char>um): Likewise.
2774         (altivec_vpku<VI_char>um_direct): New (copy of
2775         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2776         internal use).
2777         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2778         target is little endian and -maltivec=be is not specified.
2779         (*altivec_vupkhs<VU_char>_direct): New (copy of
2780         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2781         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2782         target is little endian and -maltivec=be is not specified.
2783         (*altivec_vupkls<VU_char>_direct): New (copy of
2784         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2785         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2786         little endian and -maltivec=be is not specified.
2787         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2788         little endian and -maltivec=be is not specified.
2790 2014-02-05  Richard Henderson  <rth@redhat.com>
2792         PR debug/52727
2793         * combine-stack-adj.c: Revert r206943.
2794         * sched-int.h (struct deps_desc): Add last_args_size.
2795         * sched-deps.c (init_deps): Initialize it.
2796         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2797         contain REG_ARGS_SIZE notes.
2799 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2801         * lto-cgraph.c (asm_nodes_output): Make global.
2802         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2803         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2804         (driver_handle_option): Handle OPT_fwpa.
2806 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2808         PR ipa/59947
2809         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2810         a comment typo and formatting issue.  If odr_hash hasn't been
2811         created, return vNULL and set *completep to false.
2813         PR middle-end/57499
2814         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2815         bb with no successors.
2817 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2819         PR target/59718
2820         * doc/invoke.texi (-march): Clarify documentation for ARM.
2821         (-mtune): Likewise.
2822         (-mcpu): Likewise.
2824 2014-02-05  Richard Biener  <rguenther@suse.de>
2826         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2827         when not vectorizing because of too many alias checks.
2828         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2829         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2831 2014-02-05  Nick Clifton  <nickc@redhat.com>
2833         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2834         accept extended registers in any mode when compiling for the MN10300.
2836 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2838         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2839         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2840         sanitization attributes.
2841         (can_inline_edge_p): Likewise.
2842         (sanitize_attrs_match_for_inline_p): New function.
2844 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2846         * ipa-prop.c (detect_type_change): Shor circuit testing of
2847         type changes on THIS pointer.
2849 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2851         PR target/59777
2852         * config/pa/pa.c (legitimize_tls_address): Return original address
2853         if not passed a SYMBOL_REF rtx.
2854         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2855         addresses.
2856         (pa_emit_move_sequence): Simplify TLS source operands.
2857         (pa_legitimate_constant_p): Reject all TLS constants.
2858         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2859         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2861 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2863         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2864         groups when we know they are controlled by LTO.
2865         * varasm.c (default_binds_local_p_1): If object is in other partition,
2866         it will be resolved locally.
2868 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2870         * config/host-linux.c (linux_gt_pch_use_address): Don't
2871         use SSIZE_MAX because it is not always defined.
2873 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2875         PR bootstrap/59913
2876         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2877         threshold for pseudo splitting.
2878         (update_ebb_live_info): Process call argument hard registers and
2879         hard registers from insn definition too.
2880         (max_small_class_regs_num): New constant.
2881         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2882         reloads_num only for small register classes.  Don't split for
2883         outputs of jumps.
2885 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2887         PR ipa/60058
2888         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2889         is non-null.
2891 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2893         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2894         visibility is safe.
2896 2014-02-04  Marek Polacek  <polacek@redhat.com>
2898         * gdbinit.in (pel): Define.
2900 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2902         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2903         behavior.
2905 2014-02-04  Richard Biener  <rguenther@suse.de>
2907         PR lto/59723
2908         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2909         in function context local.
2910         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2911         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2912         similar to LTO_imported_decl_ref.
2914 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2916         PR tree-optimization/60002
2917         * cgraphclones.c (build_function_decl_skip_args): Clear
2918         DECL_LANG_SPECIFIC.
2920         PR tree-optimization/60023
2921         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2922         false to gsi_replace.
2923         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2924         has been in some EH region and vec_stmt could throw, add
2925         vec_stmt into the same EH region.
2926         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2927         has no lhs, ignore it.
2928         * internal-fn.c (expand_MASK_LOAD): Likewise.
2930         PR ipa/60026
2931         * tree-inline.c (copy_forbidden): Fail for
2932         __attribute__((optimize (0))) functions.
2934         PR other/58712
2935         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2936         is set, copy one less argument.
2937         (expand_simd_clones): Don't subtract clone_info->inbranch
2938         from simd_clone_struct_alloc argument.
2940         PR rtl-optimization/57915
2941         * recog.c (simplify_while_replacing): If all unary/binary/relational
2942         operation arguments are constant, attempt to simplify those.
2944         PR middle-end/59261
2945         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2946         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2948 2014-02-04  Richard Biener  <rguenther@suse.de>
2950         PR tree-optimization/60012
2951         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2952         TBAA disambiguation to all DDRs.
2954 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2956         PR target/59788
2957         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2958         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2960 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2962         PR ipa/59882
2963         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2965 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2967         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2968         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2970 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2972         PR ipa/59831
2973         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2974         to figure out targets of polymorphic calls with known decl.
2975         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2976         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2977         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2978         (get_polymorphic_call_info): ... here.
2979         (get_polymorphic_call_info_from_invariant): New function.
2981 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2983         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2984         lookup via vtable pointer; check for type consistency
2985         and turn inconsitent facts into UNREACHABLE.
2986         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2987         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2988         type inconsistent querries; return UNREACHABLE instead.
2990 2014-02-03  Richard Henderson  <rth@twiddle.net>
2992         PR tree-opt/59924
2993         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2994         already processed this node.
2995         (normalize_one_pred_1): Pass along mark_set.
2996         (normalize_one_pred): Create and destroy a pointer_set_t.
2997         (normalize_one_pred_chain): Likewise.
2999 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3001         PR gcov-profile/58602
3002         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3004 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3006         PR ipa/59831
3007         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3008         -fno-devirtualize; try to devirtualize by the knowledge of
3009         virtual table pointer given by aggregate propagation.
3010         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3011         (ipa_print_node_jump_functions): Dump also offset that
3012         is relevant for polymorphic calls.
3013         (determine_known_aggregate_parts): Add arg_type parameter; use it
3014         instead of determining the type from pointer type.
3015         (ipa_compute_jump_functions_for_edge): Update call of
3016         determine_known_aggregate_parts.
3017         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3018         (gimple_get_virt_method_for_binfo): ... here; simplify using
3019         vtable_pointer_value_to_vtable.
3020         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3021         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3022         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3023         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3024         POINTER_PLUS_EXPR.
3025         (vtable_pointer_value_to_binfo): ... here.
3026         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3028 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3030         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3031         redef of outer loop index variable.
3033 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3035         PR c++/53017
3036         PR c++/59211
3037         * doc/extend.texi (Function Attributes): Typo.
3039 2014-02-03  Cong Hou  <congh@google.com>
3041         PR tree-optimization/60000
3042         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3043         if the vectorized statement is a store.  A store statement can only
3044         appear at the end of pattern statements.
3046 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3048         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3049         (ix86_option_override_internal): Default long double to 64-bit for
3050         32-bit Bionic and to 128-bit for 64-bit Bionic.
3052         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3053         TARGET_LONG_DOUBLE_128 is true.
3054         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3056         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3057         (mlong-double-64): Negate -mlong-double-128.
3058         (mlong-double-128): New option.
3060         * config/i386/i386-c.c (ix86_target_macros): Define
3061         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3063         * doc/invoke.texi: Document -mlong-double-128.
3065 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3067         PR rtl-optimization/60024
3068         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3070 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3072         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3074 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3076         PR rtl-optimization/57662
3077         * sel-sched.c (code_motion_path_driver): Do not mark already not
3078         existing blocks in the visiting bitmap.
3080 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3082         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3083         on the insn being emitted.
3085 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3086             Will Deacon  <will.deacon@arm.com>
3088         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3090 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3092         * config/arm/arm-tables.opt: Regenerate.
3094 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3096         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3097         for vector types other than V16QImode.
3098         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3099         define_expand, and call altivec_expand_vec_perm_le when producing
3100         code with little endian element order.
3101         (*altivec_vperm_<mode>_internal): New insn having previous
3102         behavior of altivec_vperm_<mode>.
3103         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3104         altivec_expand_vec_perm_le when producing code with little endian
3105         element order.
3106         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3107         behavior of altivec_vperm_<mode>_uns.
3109 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3111         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3112         (altivec_vsumsws): Add handling for -maltivec=be with a little
3113         endian target.
3114         (altivec_vsumsws_direct): New.
3115         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3116         gen_altivec_vsumsws.
3118 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3120         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3121         vtable_pointer_value_to_binfo): New functions.
3122         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3123         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3125 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3127         * config/nios2/nios2.md (load_got_register): Initialize GOT
3128         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3129         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3131 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3133         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3134         preserverd by passthrough, do not propagate the type.
3136 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3138         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3139         (mips_atomic_assign_expand_fenv): New function.
3140         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3142 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3144         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3145         (__builtin_mips_set_fcsr): Likewise.
3146         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3147         MIPS_USI_FTYPE_VOID.
3148         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3149         (mips16_expand_set_fcsr): Likewise.
3150         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3151         (mips16_set_fcsr_stub): Likewise.
3152         (mips16_get_fcsr_one_only_stub): New class.
3153         (mips16_set_fcsr_one_only_stub): Likewise.
3154         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3155         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3156         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3157         (hard_float): New availability predicate.
3158         (mips_builtins): Add get_fcsr and set_fcsr.
3159         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3160         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3161         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3162         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3163         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3164         patterns.
3166 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3168         * config/mips/mips.c (mips_one_only_stub): New class.
3169         (mips_need_mips16_rdhwr_p): Replace with...
3170         (mips16_rdhwr_stub): ...this new variable.
3171         (mips16_stub_call_address): New function.
3172         (mips16_rdhwr_one_only_stub): New class.
3173         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3174         (mips_output_mips16_rdhwr): Delete.
3175         (mips_finish_stub): New function.
3176         (mips_code_end): Use it to handle rdhwr stubs.
3178 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3180         PR target/60017
3181         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3182         when calculating size of integer atomic types.
3184 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3186         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3188 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3190         PR tree-optimization/60003
3191         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3192         * profile.c (branch_prob): Use gimple_call_builtin_p
3193         to check for BUILT_IN_SETJMP_RECEIVER.
3194         * tree-inline.c (copy_bb): Call notice_special_calls.
3196 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3198         PR bootstrap/59985
3199         * lra-constraints.c (process_alt_operands): Update reload_sum only
3200         on the first pass.
3202 2014-01-31  Richard Henderson  <rth@redhat.com>
3204         PR middle-end/60004
3205         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3206         until after else_eh is processed.
3208 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3210         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3211         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3212         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3213         in smmintrin.h, remove them.
3214         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3215         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3216         * config/i386/i386.md (ROUND_SAE): Fix value.
3217         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3218         (const48_operand): New.
3219         * config/i386/subst.md (round), (round_expand): Use
3220         const_4_or_8_to_11_operand.
3221         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3223 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3225         * config/i386/constraints.md (Yk): Swap meaning with k.
3226         * config/i386/i386.md (movhi_internal): Change Yk to k.
3227         (movqi_internal): Ditto.
3228         (*k<logic><mode>): Ditto.
3229         (*andhi_1): Ditto.
3230         (*andqi_1): Ditto.
3231         (kandn<mode>): Ditto.
3232         (*<code>hi_1): Ditto.
3233         (*<code>qi_1): Ditto.
3234         (kxnor<mode>): Ditto.
3235         (kortestzhi): Ditto.
3236         (kortestchi): Ditto.
3237         (kunpckhi): Ditto.
3238         (*one_cmplhi2_1): Ditto.
3239         (*one_cmplqi2_1): Ditto.
3240         * config/i386/sse.md (): Change k to Yk.
3241         (avx512f_load<mode>_mask): Ditto.
3242         (avx512f_blendm<mode>): Ditto.
3243         (avx512f_store<mode>_mask): Ditto.
3244         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3245         (avx512f_storedqu<mode>_mask): Ditto.
3246         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3247         Ditto.
3248         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3249         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3250         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3251         (avx512f_maskcmp<mode>3): Ditto.
3252         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3253         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3254         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3255         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3256         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3257         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3258         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3259         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3260         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3261         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3262         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3263         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3264         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3265         (vec_extract_lo_<mode>_maskm): Ditto.
3266         (vec_extract_hi_<mode>_maskm): Ditto.
3267         (avx512f_vternlog<mode>_mask): Ditto.
3268         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3269         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3270         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3271         (avx512f_<code>v8div16qi2_mask): Ditto.
3272         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3273         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3274         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3275         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3276         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3277         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3278         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3279         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3280         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3281         (avx512cd_maskb_vec_dupv8di): Ditto.
3282         (avx512cd_maskw_vec_dupv16si): Ditto.
3283         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3284         (avx512f_vpermi2var<mode>3_mask): Ditto.
3285         (avx512f_vpermi2var<mode>3_mask): Ditto.
3286         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3287         (*avx512f_gathersi<mode>): Ditto.
3288         (*avx512f_gathersi<mode>_2): Ditto.
3289         (*avx512f_gatherdi<mode>): Ditto.
3290         (*avx512f_gatherdi<mode>_2): Ditto.
3291         (*avx512f_scattersi<mode>): Ditto.
3292         (*avx512f_scatterdi<mode>): Ditto.
3293         (avx512f_compress<mode>_mask): Ditto.
3294         (avx512f_compressstore<mode>_mask): Ditto.
3295         (avx512f_expand<mode>_mask): Ditto.
3296         * config/i386/subst.md (mask): Change k to Yk.
3297         (mask_scalar_merge): Ditto.
3298         (sd): Ditto.
3300 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3302         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3304 2014-01-31  Richard Biener  <rguenther@suse.de>
3306         PR middle-end/59990
3307         * builtins.c (fold_builtin_memory_op): Make sure to not
3308         use a floating-point mode or a boolean or enumeral type for
3309         the copy operation.
3311 2014-01-30  DJ Delorie  <dj@redhat.com>
3313         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3314         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3315         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3316         whenever main() has an epilogue.
3318 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3320         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3321         unused variable "field".
3322         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3323         (vsx_mergeh_<mode>): Likewise.
3324         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3325         (altivec_vmrghh): Likewise.
3326         (altivec_vmrghw): Likewise.
3327         (altivec_vmrglb): Likewise.
3328         (altivec_vmrglh): Likewise.
3329         (altivec_vmrglw): Likewise.
3330         (altivec_vspltb): Add missing uses.
3331         (altivec_vsplth): Likewise.
3332         (altivec_vspltw): Likewise.
3333         (altivec_vspltsf): Likewise.
3335 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3337         PR target/59923
3338         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3339         frame related instructions.
3341 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3343         PR rtl-optimization/59959
3344         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3345         any reload of register whose subreg is invalid.
3347 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3349         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3350         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3351         Add missing return type - void.
3353 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3355         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3356         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3357         remove element index adjustment for endian (now handled in vsx.md
3358         and altivec.md).
3359         (altivec_expand_vec_perm_const): Use
3360         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3361         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3362         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3363         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3364         define_expand and a new define_insn *altivec_vspltb_internal;
3365         adjust for -maltivec=be on a little endian target.
3366         (altivec_vspltb_direct): New.
3367         (altivec_vsplth): Divide into a define_expand and a new
3368         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3369         little endian target.
3370         (altivec_vsplth_direct): New.
3371         (altivec_vspltw): Divide into a define_expand and a new
3372         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3373         little endian target.
3374         (altivec_vspltw_direct): New.
3375         (altivec_vspltsf): Divide into a define_expand and a new
3376         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3377         a little endian target.
3379 2014-01-30  Richard Biener  <rguenther@suse.de>
3381         PR tree-optimization/59993
3382         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3383         can propagate form the earlier stmt and avoid the transform
3384         when the intermediate result is needed.
3386 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3388         * README.Portability: Fix typo.
3390 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
3392         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3393         comparison_operator with ordered_comparison_operator.
3395 2014-01-30  Nick Clifton  <nickc@redhat.com>
3397         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3398         Rename to mn10300_store_multiple_regs.
3399         * config/mn10300/mn10300.c: Likewise.
3400         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3401         store_multiple_regs.
3402         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3403         Call mn10300_store_multiple_regs.
3405 2014-01-30  Nick Clifton  <nickc@redhat.com>
3406             DJ Delorie  <dj@redhat.com>
3408         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3409         %fp 2 to keep registers after it properly word-aligned.
3410         (rl78_alloc_physical_registers_umul): Handle the case where both
3411         input operands are the same.
3413 2014-01-30  Richard Biener  <rguenther@suse.de>
3415         PR tree-optimization/59903
3416         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3417         check properly.
3419 2014-01-30  Jason Merrill  <jason@redhat.com>
3421         PR c++/59633
3422         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3424         PR c++/59645
3425         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3427 2014-01-30  Richard Biener  <rguenther@suse.de>
3429         PR tree-optimization/59951
3430         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3432 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3434         PR target/59784
3435         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3436         SFmode to DFmode case.
3438 2014-01-29  DJ Delorie  <dj@redhat.com>
3440         * config/msp430/msp430.opt (-minrt): New.
3441         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3442         if -minrt given.
3443         (ENDFILE_SPEC): Likewise.
3445 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3447         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3448         (estimate_function_body_sizes): Use it.
3450 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3452         PR c++/58561
3453         * dwarf2out.c (is_cxx_auto): New.
3454         (is_base_type): Use it.
3455         (gen_type_die_with_usage): Likewise.
3457 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3459         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3460         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3461         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3462         -maltivec=be with LE targets.
3463         (vsx_mergeh_<mode>): Likewise.
3464         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3465         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3466         (altivec_vmrghb): Replace with define_expand and new
3467         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3468         (altivec_vmrghb_direct): New define_insn.
3469         (altivec_vmrghh): Replace with define_expand and new
3470         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3471         (altivec_vmrghh_direct): New define_insn.
3472         (altivec_vmrghw): Replace with define_expand and new
3473         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3474         (altivec_vmrghw_direct): New define_insn.
3475         (*altivec_vmrghsf): Adjust for endianness.
3476         (altivec_vmrglb): Replace with define_expand and new
3477         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3478         (altivec_vmrglb_direct): New define_insn.
3479         (altivec_vmrglh): Replace with define_expand and new
3480         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3481         (altivec_vmrglh_direct): New define_insn.
3482         (altivec_vmrglw): Replace with define_expand and new
3483         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3484         (altivec_vmrglw_direct): New define_insn.
3485         (*altivec_vmrglsf): Adjust for endianness.
3486         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3487         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3488         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3489         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3490         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3491         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3492         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3493         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3495 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3497         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3498         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3499         whitespace.
3501 2014-01-29  Richard Biener  <rguenther@suse.de>
3503         PR tree-optimization/58742
3504         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3505         associate_pointerplus_align.
3506         (associate_pointerplus_diff): New function.
3507         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3508         and associate_pointerplus_diff.
3510 2014-01-29  Richard Biener  <rguenther@suse.de>
3512         * lto-streamer.h (LTO_major_version): Bump to 3.
3513         (LTO_minor_version): Reset to 0.
3515 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3517         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3518         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3519         (arm_file_start): Generate correct asm header for armv7ve.
3520         * config/arm/bpabi.h: Add multilib support for armv7ve.
3521         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3522         and cortex-a15 to armv7ve.
3523         * config/arm/t-aprofile: Add multilib support for armv7ve.
3524         * doc/invoke.texi: Document -march=armv7ve.
3526 2014-01-29  Richard Biener  <rguenther@suse.de>
3528         PR tree-optimization/58742
3529         * tree-ssa-forwprop.c (associate_plusminus): Return true
3530         if we changed sth, defer EH cleanup to ...
3531         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3532         (simplify_mult): New function.
3534 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3536         PR middle-end/59917
3537         PR tree-optimization/59920
3538         * tree.c (build_common_builtin_nodes): Remove
3539         __builtin_setjmp_dispatcher initialization.
3540         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3541         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3542         instead of gsi_after_labels + manually skipping debug stmts.
3543         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3544         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3545         * tree-inline.c (copy_edges_for_bb): Remove
3546         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3547         argument.  Ignore computed_goto_p stmts.  Don't call
3548         make_abnormal_goto_edges.  If a call might need abnormal edges
3549         for non-local gotos, see if it already has an edge to
3550         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3551         with true argument, don't do anything then, otherwise add
3552         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3553         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3554         caller.
3555         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3556         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3557         (lower_stmt): Don't set data->calls_builtin_setjmp.
3558         (lower_builtin_setjmp): Adjust comment.
3559         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3560         * tree-cfg.c (found_computed_goto): Remove.
3561         (factor_computed_gotos): Remove.
3562         (make_goto_expr_edges): Return bool, true for computed gotos.
3563         Don't call make_abnormal_goto_edges.
3564         (build_gimple_cfg): Don't set found_computed_goto, don't call
3565         factor_computed_gotos.
3566         (computed_goto_p): No longer static.
3567         (make_blocks): Don't set found_computed_goto.
3568         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3569         (make_edges): If make_goto_expr_edges returns true, push bb
3570         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3571         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3572         vector.  Record mapping between bbs and OpenMP regions if there
3573         are any, adjust make_gimple_omp_edges caller.  Call
3574         handle_abnormal_edges.
3575         (make_abnormal_goto_edges): Remove.
3576         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3577         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3578         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3579         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3580         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3581         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3582         filling *region also set *region_idx to (*region)->entry->index.
3584         PR other/58712
3585         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3586         For REGs set ORIGINAL_REGNO.
3588 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3590         * doc/md.texi: Mention that a target shouldn't implement
3591         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3592         than hi/lo pair.
3594 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3596         PR tree-optimization/59594
3597         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3598         a copy of the datarefs vector rather than the vector itself.
3600 2014-01-28  Jason Merrill  <jason@redhat.com>
3602         PR c++/53756
3603         * dwarf2out.c (auto_die): New static.
3604         (gen_type_die_with_usage): Handle C++1y 'auto'.
3605         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3606         on definition.
3608 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3610         PR target/59672
3611         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3612         (SPEC_X32): Likewise.
3613         (SPEC_64): Likewise.
3614         * config/i386/i386.c (ix86_option_override_internal): Turn off
3615         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3616         for TARGET_16BIT.
3617         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3618         * config/i386/i386.h (TARGET_16BIT): New macro.
3619         (TARGET_16BIT_P): Likewise.
3620         * config/i386/i386.opt: Add m16.
3621         * doc/invoke.texi: Document -m16.
3623 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3625         PR preprocessor/59935
3626         * input.c (location_get_source_line): Bail out on when line number
3627         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3629 2014-01-28  Richard Biener  <rguenther@suse.de>
3631         PR tree-optimization/58742
3632         * tree-ssa-forwprop.c (associate_plusminus): Handle
3633         pointer subtraction of the form (T)(P + A) - (T)P.
3635 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3637         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3638         at const_int_cost.
3640 2014-01-28  Richard Biener  <rguenther@suse.de>
3642         Revert
3643         2014-01-28  Richard Biener  <rguenther@suse.de>
3645         PR rtl-optimization/45364
3646         PR rtl-optimization/59890
3647         * var-tracking.c (local_get_addr_clear_given_value): Handle
3648         already cleared slot.
3649         (val_reset): Handle not allocated local_get_addr_cache.
3650         (vt_find_locations): Use post-order on the inverted CFG.
3652 2014-01-28  Richard Biener  <rguenther@suse.de>
3654         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3656 2014-01-28  Richard Biener  <rguenther@suse.de>
3658         PR rtl-optimization/45364
3659         PR rtl-optimization/59890
3660         * var-tracking.c (local_get_addr_clear_given_value): Handle
3661         already cleared slot.
3662         (val_reset): Handle not allocated local_get_addr_cache.
3663         (vt_find_locations): Use post-order on the inverted CFG.
3665 2014-01-28  Alan Modra  <amodra@gmail.com>
3667         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3668         * configure.ac <recursive call for build != host>: Define
3669         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3670         and LD_FOR_BUILD too.
3671         * configure: Regenerate.
3673 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3675         * config/i386/i386.c (get_builtin_code_for_version): Separate
3676         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3677         Broadwell from Haswell.
3679 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3681         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3682         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3683         * config/mips/mips.c (mips_option_override): Change setting
3684         of TARGET_DSP.
3685         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3686         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3687         Change from Mask to Var.
3689 2014-01-27  Jeff Law  <law@redhat.com>
3691         * ipa-inline.c (inline_small_functions): Fix typo.
3693 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3695         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3696         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3697         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3698         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3699         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3700         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3701         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3702         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3703         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3704         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3705         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3706         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3707         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3708         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3709         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3710         (_mm512_storeu_epi64): Ditto.
3711         (_mm512_cmpge_epi32_mask): Ditto.
3712         (_mm512_cmpge_epu32_mask): Ditto.
3713         (_mm512_cmpge_epi64_mask): Ditto.
3714         (_mm512_cmpge_epu64_mask): Ditto.
3715         (_mm512_cmple_epi32_mask): Ditto.
3716         (_mm512_cmple_epu32_mask): Ditto.
3717         (_mm512_cmple_epi64_mask): Ditto.
3718         (_mm512_cmple_epu64_mask): Ditto.
3719         (_mm512_cmplt_epi32_mask): Ditto.
3720         (_mm512_cmplt_epu32_mask): Ditto.
3721         (_mm512_cmplt_epi64_mask): Ditto.
3722         (_mm512_cmplt_epu64_mask): Ditto.
3723         (_mm512_cmpneq_epi32_mask): Ditto.
3724         (_mm512_cmpneq_epu32_mask): Ditto.
3725         (_mm512_cmpneq_epi64_mask): Ditto.
3726         (_mm512_cmpneq_epu64_mask): Ditto.
3727         (_mm512_expand_pd): Ditto.
3728         (_mm512_expand_ps): Ditto.
3729         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3730         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3731         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3732         * config/i386/i386.c (ix86_builtins): Add
3733         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3734         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3735         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3736         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3737         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3738         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3739         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3740         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3741         IX86_BUILTIN_PMOVUSQW512_MEM.
3742         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3743         __builtin_ia32_pmovsqd512mem_mask,
3744         __builtin_ia32_pmovqd512mem_mask,
3745         __builtin_ia32_pmovusqw512mem_mask,
3746         __builtin_ia32_pmovsqw512mem_mask,
3747         __builtin_ia32_pmovqw512mem_mask,
3748         __builtin_ia32_pmovusdw512mem_mask,
3749         __builtin_ia32_pmovsdw512mem_mask,
3750         __builtin_ia32_pmovdw512mem_mask,
3751         __builtin_ia32_pmovqb512mem_mask,
3752         __builtin_ia32_pmovusqb512mem_mask,
3753         __builtin_ia32_pmovsqb512mem_mask,
3754         __builtin_ia32_pmovusdb512mem_mask,
3755         __builtin_ia32_pmovsdb512mem_mask,
3756         __builtin_ia32_pmovdb512mem_mask.
3757         (bdesc_args): Add __builtin_ia32_expanddf512,
3758         __builtin_ia32_expandsf512.
3759         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3760         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3761         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3762         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3763         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3764         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3765         (avx512f_<code>v8div16qi2_mask_store): This.
3766         (avx512f_expand<mode>): New.
3768 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3770         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3771         New.
3772         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3773         (_mm512_prefetch_i32scatter_pd): Ditto.
3774         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3775         (_mm512_prefetch_i64scatter_pd): Ditto.
3776         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3777         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3778         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3779         (_mm512_prefetch_i32scatter_ps): Ditto.
3780         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3781         (_mm512_prefetch_i64scatter_ps): Ditto.
3782         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3783         * config/i386/i386-builtin-types.def: Define
3784         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3785         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3786         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3787         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3788         IX86_BUILTIN_SCATTERPFQPD.
3789         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3790         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3791         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3792         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3793         __builtin_ia32_scatterpfqps.
3794         (ix86_expand_builtin): Expand new built-ins.
3795         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3796         fix memory access data type.
3797         (*avx512pf_gatherpf<mode>_mask): Ditto.
3798         (*avx512pf_gatherpf<mode>): Ditto.
3799         (avx512pf_scatterpf<mode>): Ditto.
3800         (*avx512pf_scatterpf<mode>_mask): Ditto.
3801         (*avx512pf_scatterpf<mode>): Ditto.
3802         (GATHER_SCATTER_SF_MEM_MODE): New.
3803         (avx512pf_gatherpf<mode>df): Ditto.
3804         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3805         (*avx512pf_scatterpf<mode>df): Ditto.
3807 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3809         PR bootstrap/59934
3810         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3811         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3812         reached.
3814 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3816         * common/config/arm/arm-common.c
3817         (arm_rewrite_mcpu): Handle multiple names.
3818         * config/arm/arm.h
3819         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3821 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3823         * gimple-builder.h (create_gimple_tmp): Delete.
3825 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3827         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3828         words comparisons.
3830 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3832         * config/pa/pa.md (call): Generate indirect long calls to non-local
3833         functions when outputing 32-bit code.
3834         (call_value): Likewise except for special call to buggy powf function.
3836         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3837         portable runtime and PIC indirect calls.
3838         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3839         and PIC call sequences.  Use ldo instead of blr to set return register
3840         in PIC call sequence.
3842 2014-01-25  Walter Lee  <walt@tilera.com>
3844         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3845         avoid clobbering a live register.
3847 2014-01-25  Walter Lee  <walt@tilera.com>
3849         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3850         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3851         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3852         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3854 2014-01-25  Walter Lee  <walt@tilera.com>
3856         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3857         arguments on even registers.
3858         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3859         STACK_BOUNDARY.
3860         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3861         (BIGGEST_ALIGNMENT): Ditto.
3862         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3864 2014-01-25  Walter Lee  <walt@tilera.com>
3866         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3867         insns before bundling.
3868         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3870 2014-01-25  Walter Lee  <walt@tilera.com>
3872         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3873         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3874         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3876 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3878         * config/mips/constraints.md (kl): Delete.
3879         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3880         define expands, using...
3881         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3882         instructions for MIPS16.
3883         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3884         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3886 2014-01-25  Walter Lee  <walt@tilera.com>
3888         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3889         (clzdi2): Ditto.
3890         (ffsdi2): Ditto.
3892 2014-01-25  Walter Lee  <walt@tilera.com>
3894         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3895         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3897 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3899         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3900         Handle XOR.
3902 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3904         * print-rtl.c (in_call_function_usage): New var.
3905         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3906         EXPR_LIST mode as mode and not as reg note name.
3908         PR middle-end/59561
3909         * cfgloopmanip.c (copy_loop_info): If
3910         loop->warned_aggressive_loop_optimizations, make sure
3911         the flag is set in target loop too.
3913 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3915         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3916         flag_cilkplus.
3917         * builtins.def: Likewise.
3918         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3919         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3920         * ira.c (ira_setup_eliminable_regset): Likewise.
3921         * omp-low.c (gate_expand_omp): Likewise.
3922         (execute_lower_omp): Likewise.
3923         (diagnose_sb_0): Likewise.
3924         (gate_diagnose_omp_blocks): Likewise.
3925         (simd_clone_clauses_extract): Likewise.
3926         (gate): Likewise.
3928 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3930         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3931         correction for little endian...
3932         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3933         here.
3935 2014-01-24  Jeff Law  <law@redhat.com>
3937         PR tree-optimization/59919
3938         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3939         for non-returning calls.
3941 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3943         * common/config/aarch64/aarch64-common.c
3944         (aarch64_rewrite_mcpu): Handle multiple names.
3945         * config/aarch64/aarch64.h
3946         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3948 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3950         * input.c (add_file_to_cache_tab): Handle the case where fopen
3951         returns NULL.
3953 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3955         PR target/59929
3956         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3957         from push operand if code of push isn't PRE_DEC.
3959 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3961         PR target/59909
3962         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3963         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3964         it is only used for non-atomic loads/stores.
3966         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3967         -mquad-memory or -mquad-memory-atomic switches.
3969         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3970         -mquad-memory-atomic to ISA 2.07 support.
3972         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3973         to separate support of normal quad word memory operations (ldq, stq)
3974         from the atomic quad word memory operations.
3976         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3977         support to separate non-atomic quad word operations from atomic
3978         quad word operations.  Disable non-atomic quad word operations in
3979         little endian mode so that we don't have to swap words after the
3980         load and before the store.
3981         (quad_load_store_p): Add comment about atomic quad word support.
3982         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3983         options printed with -mdebug=reg.
3985         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3986         -mquad-memory-atomic as the test for whether we have quad word
3987         atomic instructions.
3988         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3989         or -mp8-vector are used, allow byte/half-word atomic operations.
3991         * config/rs6000/sync.md (load_lockedti): Insure that the address
3992         is a proper indexed or indirect address for the lqarx instruction.
3993         On little endian systems, swap the hi/lo registers after the lqarx
3994         instruction.
3995         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3996         insure the address is valid for the lqarx instruction.
3997         (store_conditionalti): Insure that the address is a proper indexed
3998         or indirect address for the stqcrx. instruction.  On little endian
3999         systems, swap the hi/lo registers before doing the stqcrx.
4000         instruction.
4001         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4002         insure the address is valid for the stqcrx. instruction.
4004         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4005         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4006         type of quad memory support is available.
4008 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4010         PR regression/59915
4011         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4012         there is a danger of looping.
4014 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4016         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4017         force flag_ira_loop_pressure if set via command line.
4019 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4021         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4022         (ashr_simd): New builtin handling DI mode.
4023         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4024         (aarch64_sshr_simddi): New match pattern.
4025         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4026         (vshrd_n_s64): Likewise.
4027         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4029 2014-01-23  Nick Clifton  <nickc@redhat.com>
4031         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4032         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4033         favour of mcu specific scripts.
4034         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4035         430x multilibs.
4037 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4038             Alex Velenko  <Alex.Velenko@arm.com>
4040         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4041         (vaddv_s16): Likewise.
4042         (vaddv_s32): Likewise.
4043         (vaddv_u8): Likewise.
4044         (vaddv_u16): Likewise.
4045         (vaddv_u32): Likewise.
4046         (vaddvq_s8): Likewise.
4047         (vaddvq_s16): Likewise.
4048         (vaddvq_s32): Likewise.
4049         (vaddvq_s64): Likewise.
4050         (vaddvq_u8): Likewise.
4051         (vaddvq_u16): Likewise.
4052         (vaddvq_u32): Likewise.
4053         (vaddvq_u64): Likewise.
4054         (vaddv_f32): Likewise.
4055         (vaddvq_f32): Likewise.
4056         (vaddvq_f64): Likewise.
4057         (vmaxv_f32): Likewise.
4058         (vmaxv_s8): Likewise.
4059         (vmaxv_s16): Likewise.
4060         (vmaxv_s32): Likewise.
4061         (vmaxv_u8): Likewise.
4062         (vmaxv_u16): Likewise.
4063         (vmaxv_u32): Likewise.
4064         (vmaxvq_f32): Likewise.
4065         (vmaxvq_f64): Likewise.
4066         (vmaxvq_s8): Likewise.
4067         (vmaxvq_s16): Likewise.
4068         (vmaxvq_s32): Likewise.
4069         (vmaxvq_u8): Likewise.
4070         (vmaxvq_u16): Likewise.
4071         (vmaxvq_u32): Likewise.
4072         (vmaxnmv_f32): Likewise.
4073         (vmaxnmvq_f32): Likewise.
4074         (vmaxnmvq_f64): Likewise.
4075         (vminv_f32): Likewise.
4076         (vminv_s8): Likewise.
4077         (vminv_s16): Likewise.
4078         (vminv_s32): Likewise.
4079         (vminv_u8): Likewise.
4080         (vminv_u16): Likewise.
4081         (vminv_u32): Likewise.
4082         (vminvq_f32): Likewise.
4083         (vminvq_f64): Likewise.
4084         (vminvq_s8): Likewise.
4085         (vminvq_s16): Likewise.
4086         (vminvq_s32): Likewise.
4087         (vminvq_u8): Likewise.
4088         (vminvq_u16): Likewise.
4089         (vminvq_u32): Likewise.
4090         (vminnmv_f32): Likewise.
4091         (vminnmvq_f32): Likewise.
4092         (vminnmvq_f64): Likewise.
4094 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4096         * config/aarch64/aarch64-simd.md
4097         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4098         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4099         (*aarch64_mul3_elt<mode>): Likewise.
4100         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4101         (*aarch64_mul3_elt_to_64v2df): Likewise.
4102         (*aarch64_mla_elt<mode>): Likewise.
4103         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4104         (*aarch64_mls_elt<mode>): Likewise.
4105         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4106         (*aarch64_fma4_elt<mode>): Likewise.
4107         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4108         (*aarch64_fma4_elt_to_64v2df): Likewise.
4109         (*aarch64_fnma4_elt<mode>): Likewise.
4110         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4111         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4112         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4113         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4114         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4115         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4116         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4117         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4118         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4120 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4122         * config/aarch64/aarch64-simd.md
4123         (aarch64_be_checked_get_lane<mode>): New define_expand.
4124         * config/aarch64/aarch64-simd-builtins.def
4125         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4126         New builtin definition.
4127         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4128         Use new safe be builtin.
4130 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4132         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4133         New define_insn.
4134         (aarch64_be_st1<mode>): Likewise.
4135         (aarch_ld1<VALL:mode>): Define_expand modified.
4136         (aarch_st1<VALL:mode>): Likewise.
4137         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4138         (UNSPEC_ST1): Likewise.
4140 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
4142         * config/microblaze/microblaze.md: Add trap insn and attribute
4144 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4146         PR preprocessor/58580
4147         * input.h (location_get_source_line): Take an additional line_size
4148         parameter.
4149         (void diagnostics_file_cache_fini): Declare new function.
4150         * input.c (struct fcache): New type.
4151         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4152         New static constants.
4153         (diagnostic_file_cache_init, total_lines_num)
4154         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4155         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4156         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4157         (get_next_line, read_next_line, goto_next_line, read_line_num):
4158         New static function definitions.
4159         (diagnostic_file_cache_fini): New function.
4160         (location_get_source_line): Take an additional output line_len
4161         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4162         read_line_num.
4163         * diagnostic.c (diagnostic_finish): Call
4164         diagnostic_file_cache_fini.
4165         (adjust_line): Take an additional input parameter for the length
4166         of the line, rather than calculating it with strlen.
4167         (diagnostic_show_locus): Adjust the use of
4168         location_get_source_line and adjust_line with respect to their new
4169         signature.  While displaying a line now, do not stop at the first
4170         null byte.  Rather, display the zero byte as a space and keep
4171         going until we reach the size of the line.
4172         * Makefile.in: Add vec.o to OBJS-libcommon
4174 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4175             Ilya Tocar     <ilya.tocar@intel.com>
4177         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4178         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4179         (__builtin_ia32_kmov16): Ditto.
4180         * config/i386/i386.md (UNSPEC_KMOV): New.
4181         (kmovw): Ditto.
4183 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4185         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4186         (_mm512_storeu_si512): Ditto.
4188 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4190         PR target/52125
4191         * rtl.h (get_referenced_operands): Declare.
4192         * recog.c (get_referenced_operands): New function.
4193         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4194         operands have been referenced when recording LO_SUM references.
4196 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
4198         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4200 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4202         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4203         Enable for generic and recent AMD targets.
4205 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4207         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4208         ARG_SIZE note when adjustment was eliminated.
4210 2014-01-22  Jeff Law  <law@redhat.com>
4212         PR tree-optimization/59597
4213         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4214         in file.  Accept new argument REGISTERING and use it to modify
4215         dump output appropriately.
4216         (register_jump_thread): Corresponding changes.
4217         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4218         thread paths involving joiner blocks.  Add code to dump cancelled
4219         jump threading paths.
4221 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4223         PR rtl-optimization/59477
4224         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4225         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4227 2014-01-22  Tom Tromey  <tromey@redhat.com>
4229         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4230         PARAMS.
4231         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4233 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4235         PR rtl-optimization/59896
4236         * lra-constraints.c (process_alt_operands): Check unused note for
4237         matched operands of insn with no output reloads.
4239 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4241         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4242         (mips_move_from_gpr_cost): Likewise.
4244 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4246         PR rtl-optimization/59858
4247         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4248         ira_class_hard_regs_num.
4249         (process_alt_operands): Increase reject for dying matched operand.
4251 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4253         PR target/59003
4254         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4255         smaller than size, perform several stores or loads and stores
4256         at dst + count - size to store or copy all of size bytes, rather
4257         than just last modesize bytes.
4259 2014-01-20  DJ Delorie  <dj@redhat.com>
4261         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4262         that CLOBBERs are REGs before propogating their values.
4264 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4266         PR middle-end/59789
4267         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4268         (cgraph_inline_failed_type): New function.
4269         * cgraph.h (DEFCIFCODE): Add type.
4270         (cgraph_inline_failed_type_t): New enum.
4271         (cgraph_inline_failed_type): New prototype.
4272         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4273         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4274         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4275         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4276         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4277         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4278         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4279         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4280         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4281         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4282         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4283         OPTIMIZATION_MISMATCH.
4284         * tree-inline.c (expand_call_inline): Emit errors during
4285         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4287 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4289         PR target/59685
4290         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4291         mode attribute in insn output.
4293 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4295         * output.h (output_constant): Delete.
4296         * varasm.c (output_constant): Make private.
4298 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4300         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4302 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4304         PR middle-end/59860
4305         * tree.h (fold_builtin_strcat): New prototype.
4306         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4307         argument, if non-NULL, don't call c_strlen.  Optimize
4308         directly into __builtin_memcpy instead of __builtin_strcpy.
4309         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4310         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4312 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4314         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4315         for SImode_address_operand operands, having only a REG argument.
4317 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4319         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4320         loader name using mbig-endian.
4321         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4323 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4325         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4326         (-mtune): Likewise.
4327         (-mcpu): Likewise.
4329 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4331         * config/aarch64/aarch64-protos.h
4332         (aarch64_cannot_change_mode_class_ptr): Declare.
4333         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4334         aarch64_cannot_change_mode_class_ptr): New.
4335         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4336         backend hook aarch64_cannot_change_mode_class.
4338 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4340         * common/config/aarch64/aarch64-common.c
4341         (aarch64_handle_option): Don't handle any option order logic here.
4342         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4343         selected_cpu, warn on architecture version mismatch.
4344         (aarch64_override_options): Fix parsing order for option strings.
4346 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4347             Iain Sandoe  <iain@codesourcery.com>
4349         PR bootstrap/59496
4350         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4351         warning.  Amend comment to reflect current functionality.
4353 2014-01-20  Richard Biener  <rguenther@suse.de>
4355         PR middle-end/59860
4356         * builtins.c (fold_builtin_strcat): Remove case better handled
4357         by tree-ssa-strlen.c.
4359 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4361         * config/aarch64/aarch64.opt
4362         (mcpu, march, mtune): Make case-insensitive.
4364 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4366         PR target/59880
4367         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4368         if operands[1] is a REG or ZERO_EXTEND of a REG.
4370 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4372         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4374 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4376         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4377         long non-pic millicode calls.
4379 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4381         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4383 2014-01-19  Kito Cheng  <kito@0xlab.org>
4385         * builtins.c (expand_movstr): Check movstr expand done or fail.
4387 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4388             H.J. Lu  <hongjiu.lu@intel.com>
4390         PR target/59379
4391         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4392         to DImode for zero-extended addresses.
4394 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4396         PR rtl-optimization/57763
4397         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4398         on the new indirect jump_insn and increment LABEL_NUSES (label).
4400 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4402         PR bootstrap/59580
4403         PR bootstrap/59583
4404         * config.gcc (x86_archs): New variable.
4405         (x86_64_archs): Likewise.
4406         (x86_cpus): Likewise.
4407         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4408         --with-arch/--with-cpu= options.
4409         Support --with-arch=/--with-cpu={nehalem,westmere,
4410         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4412 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4414         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4415         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4417 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4419         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4421 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4423         PR target/58944
4424         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4425         clear cpp_get_options (parse_in)->warn_unused_macros for
4426         ix86_target_macros_internal with cpp_define.
4428 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4430         * jump.c (delete_related_insns): Keep (use (insn))s.
4431         * reorg.c (redundant_insn): Check for barriers too.
4433 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4435         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4437 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4439         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4440         call to $$dyncall when TARGET_LONG_CALLS is true.
4442 2014-01-17  Jeff Law  <law@redhat.com>
4444         * ree.c (combine_set_extension): Temporarily disable test for
4445         changing number of hard registers.
4447 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4449         PR middle-end/58125
4450         * ipa-inline-analysis.c (inline_free_summary):
4451         Do not free summary of aliases.
4453 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4455         PR middle-end/59706
4456         * gimplify.c (gimplify_expr): Use create_tmp_var
4457         instead of create_tmp_var_raw.  If cond doesn't have
4458         integral type, don't add the IFN_ANNOTATE builtin at all.
4460 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4462         PR ipa/59736
4463         * ipa-cp.c (prev_edge_clone): New variable.
4464         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4465         Also resize prev_edge_clone vector.
4466         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4467         (ipcp_edge_removal_hook): New function.
4468         (ipcp_driver): Register ipcp_edge_removal_hook.
4470 2014-01-17  Andrew Pinski <apinski@cavium.com>
4471             Steve Ellcey  <sellcey@mips.com>
4473         PR target/59462
4474         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4475         of operator mode.
4477 2014-01-17  Jeff Law  <law@redhat.com>
4479         PR middle-end/57904
4480         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4481         so that pass_ccp runs first.
4483 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4485         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4486         (ix86_adjust_cost): Use !TARGET_XXX.
4487         (do_reorder_for_imul): Likewise.
4488         (swap_top_of_ready_list): Likewise.
4489         (ix86_sched_reorder): Likewise.
4491 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4493         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4494         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4495         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4496         (intel_memset): New.  Duplicate slm_memset.
4497         (intel_cost): New.  Duplicate slm_cost.
4498         (m_INTEL): New macro.
4499         (processor_target_table): Add "intel".
4500         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4501         with PROCESSOR_INTEL for "intel".
4502         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4503         PROCESSOR_SILVERMONT.
4504         (ix86_issue_rate): Likewise.
4505         (ix86_adjust_cost): Likewise.
4506         (ia32_multipass_dfa_lookahead): Likewise.
4507         (swap_top_of_ready_list): Likewise.
4508         (ix86_sched_reorder): Likewise.
4509         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4510         instead of TARGET_OPT_AGU.
4511         * config/i386/i386.h (TARGET_INTEL): New.
4512         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4513         (processor_type): Add PROCESSOR_INTEL.
4514         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4515         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4517 2014-01-17  Marek Polacek  <polacek@redhat.com>
4519         PR c/58346
4520         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4521         size is zero.
4523 2014-01-17  Richard Biener  <rguenther@suse.de>
4525         PR tree-optimization/46590
4526         * opts.c (default_options_table): Add entries for
4527         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4528         all enabled at -O1 but not for -Og.
4529         * common.opt (fbranch-count-reg): Remove Init(1).
4530         (fmove-loop-invariants): Likewise.
4531         (ftree-pta): Likewise.
4533 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4535         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4536         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4537         decls to at least the GCC 4.8 used alignments.
4539         PR fortran/59440
4540         * tree-nested.c (convert_nonlocal_reference_stmt,
4541         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4542         of GIMPLE_BIND stmts, adjust associated decls.
4544 2014-01-17  Richard Biener  <rguenther@suse.de>
4546         PR tree-optimization/46590
4547         * vec.h (vec<>::bseach): New member function implementing
4548         binary search according to C89 bsearch.
4549         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4550         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4551         bitmap pointer again.  Make accesses_in_loop a flat array.
4552         (mem_ref_obstack): New global.
4553         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4554         (mark_ref_stored): Likewise.
4555         (ref_indep_loop_p_2): Likewise.
4556         (set_ref_stored_in_loop): New helper function.
4557         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4558         (memref_free): Adjust.
4559         (record_mem_ref_loc): Simplify.
4560         (gather_mem_refs_stmt): Adjust.
4561         (sort_locs_in_loop_postorder_cmp): New function.
4562         (analyze_memory_references): Sort accesses_in_loop after
4563         loop postorder number.
4564         (find_ref_loc_in_loop_cmp): New function.
4565         (for_all_locs_in_loop): Find relevant cluster of locs in
4566         accesses_in_loop and iterate without recursion.
4567         (execute_sm): Avoid uninit warning.
4568         (struct ref_always_accessed): Simplify.
4569         (ref_always_accessed::operator ()): Likewise.
4570         (ref_always_accessed_p): Likewise.
4571         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4572         loop postorder numbers here.
4573         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4574         numbers.
4576 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4578         PR c++/57945
4579         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4580         on decls for which assemble_alias has been called.
4582 2014-01-17  Nick Clifton  <nickc@redhat.com>
4584         * config/msp430/msp430.opt: (mcpu): New option.
4585         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4586         (msp430_option_override): Parse target_cpu.  If the MCU name
4587         matches a generic string, clear target_mcu.
4588         (msp430_attr): Allow numeric interrupt values up to 63.
4589         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4590         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4591         option.
4592         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4593         Add mcpu matches.
4594         * config/msp430/msp430.md (popm): Use %J rather than %I.
4595         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4596         (addhi_cy_i): Use immediate_operand for operand 2.
4597         * doc/invoke.texi: Document -mcpu option.
4599 2014-01-17  Richard Biener  <rguenther@suse.de>
4601         PR rtl-optimization/38518
4602         * df.h (df_analyze_loop): Declare.
4603         * df-core.c: Include cfgloop.h.
4604         (df_analyze_1): Split out main part of df_analyze.
4605         (df_analyze): Adjust.
4606         (loop_inverted_post_order_compute): New function.
4607         (loop_post_order_compute): Likewise.
4608         (df_analyze_loop): New function avoiding whole-function
4609         postorder computes.
4610         * loop-invariant.c (find_defs): Use df_analyze_loop.
4611         (find_invariants): Adjust.
4612         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4614 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4616         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4617         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4619 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4621         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4622         traversal when removing references.
4624 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4626         PR ipa/59775
4627         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4629 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4631         PR middle-end/56791
4632         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4633         pushing a reload for an autoinc when we had previously reloaded an
4634         inner part of the address.
4636 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4638         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4639         field.
4640         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4641         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4642         when not giving up or versioning for alias only because of
4643         loop->safelen.
4644         (vect_analyze_data_ref_dependences): Set to true.
4645         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4646         is a GIMPLE_PHI.
4647         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4648         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4649         to the condition.
4651         PR middle-end/58344
4652         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4654         PR target/59839
4655         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4656         operand 0 predicate for gathers, use a new pseudo as subtarget.
4658 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4660         PR middle-end/59609
4661         * lra-constraints.c (process_alt_operands): Add printing debug info.
4662         Check absence of input/output reloads for matched operands too.
4664 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4666         PR rtl-optimization/59835
4667         * ira.c (ira_init_register_move_cost): Increase cost for
4668         impossible modes.
4670 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4672         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4674 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4676         PR target/59780
4677         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4678         non-register objects.  Use gen_(high/low)part more consistently.
4679         Fix assertions.
4681 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4683         PR target/59844
4684         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4685         endian support, remove tests for WORDS_BIG_ENDIAN.
4686         (p8_mfvsrd_3_<mode>): Likewise.
4687         (reload_gpr_from_vsx<mode>): Likewise.
4688         (reload_gpr_from_vsxsf): Likewise.
4689         (p8_mfvsrd_4_disf): Likewise.
4691 2014-01-16  Richard Biener  <rguenther@suse.de>
4693         PR rtl-optimization/46590
4694         * lcm.c (compute_antinout_edge): Use postorder iteration.
4695         (compute_laterin): Use inverted postorder iteration.
4697 2014-01-16  Nick Clifton  <nickc@redhat.com>
4699         PR middle-end/28865
4700         * varasm.c (output_constant): Return the number of bytes actually
4701         emitted.
4702         (output_constructor_array_range): Update the field size with the
4703         number of bytes emitted by output_constant.
4704         (output_constructor_regular_field): Likewise.  Also do not
4705         complain if the total number of bytes emitted is now greater
4706         than the expected fieldpos.
4707         * output.h (output_constant): Update prototype and descriptive comment.
4709 2014-01-16  Marek Polacek  <polacek@redhat.com>
4711         PR middle-end/59827
4712         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4713         it is error_mark_node.
4715 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4717         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4718         VALID_AVX256_REG_OR_OI_MODE.
4720 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4722         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4723         current procedure should be profiled.
4725 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4727         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4728         of moving from/to the STACK_REG register class.
4730 2014-01-15  Richard Henderson  <rth@redhat.com>
4732         PR debug/54694
4733         * reginfo.c (global_regs_decl): Globalize.
4734         * rtl.h (global_regs_decl): Declare.
4735         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4736         reserved via global_regs.
4738 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4740         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4742 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4744         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4745         and vmulosh rather than call gen_vec_widen_smult_*.
4746         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4747         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4748         (vec_widen_smult_even_v16qi): Likewise.
4749         (vec_widen_umult_even_v8hi): Likewise.
4750         (vec_widen_smult_even_v8hi): Likewise.
4751         (vec_widen_umult_odd_v16qi): Likewise.
4752         (vec_widen_smult_odd_v16qi): Likewise.
4753         (vec_widen_umult_odd_v8hi): Likewise.
4754         (vec_widen_smult_odd_v8hi): Likewise.
4755         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4756         vmuloub rather than call gen_vec_widen_umult_*.
4757         (vec_widen_umult_lo_v16qi): Likewise.
4758         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4759         vmulosb rather than call gen_vec_widen_smult_*.
4760         (vec_widen_smult_lo_v16qi): Likewise.
4761         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4762         rather than call gen_vec_widen_umult_*.
4763         (vec_widen_umult_lo_v8hi): Likewise.
4764         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4765         rather than call gen_vec_widen_smult_*.
4766         (vec_widen_smult_lo_v8hi): Likewise.
4768 2014-01-15  Jeff Law  <law@redhat.com>
4770         PR tree-optimization/59747
4771         * ree.c (find_and_remove_re): Properly handle case where a second
4772         eliminated extension requires widening a copy created for elimination
4773         of a prior extension.
4774         (combine_set_extension): Ensure that the number of hard regs needed
4775         for a destination register does not change when we widen it.
4777 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4779         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4780         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4781         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4782         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4783         (avr-*-rtems*): Likewise.
4784         (bfin*-rtems*): Likewise.
4785         (moxie-*-rtems*): Likewise.
4786         (h8300-*-rtems*): Likewise.
4787         (i[34567]86-*-rtems*): Likewise.
4788         (lm32-*-rtems*): Likewise.
4789         (m32r-*-rtems*): Likewise.
4790         (m68k-*-rtems*): Likewise.
4791         (microblaze*-*-rtems*): Likewise.
4792         (mips*-*-rtems*): Likewise.
4793         (powerpc-*-rtems*): Likewise.
4794         (sh-*-rtems*): Likewise.
4795         (sparc-*-rtems*): Likewise.
4796         (sparc64-*-rtems*): Likewise.
4797         (v850-*-rtems*): Likewise.
4798         (m32c-*-rtems*): Likewise.
4800 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4802         PR rtl-optimization/59511
4803         * ira.c (ira_init_register_move_cost): Use memory costs for some
4804         cases of register move cost calculations.
4805         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4806         instead of BB frequency.
4807         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4808         * lra-assigns.c (find_hard_regno_for): Ditto.
4810 2014-01-15  Richard Biener  <rguenther@suse.de>
4812         PR tree-optimization/59822
4813         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4814         (vectorizable_load): Use it to hoist defs of uses of invariant
4815         loads out of the loop.
4817 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4818             Kugan Vivekanandarajah  <kuganv@linaro.org>
4820         PR target/59695
4821         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4822         truncation.
4824 2014-01-15  Richard Biener  <rguenther@suse.de>
4826         PR rtl-optimization/59802
4827         * lcm.c (compute_available): Use inverted postorder to seed
4828         the initial worklist.
4830 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4832         PR target/59803
4833         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4834         ADDR_REGS for invalid symrefs in non-PIC code.
4836 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4838         PR other/58712
4839         * builtins.c (determine_block_size): Initialize *probable_max_size
4840         even if len_rtx is CONST_INT.
4842 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4844         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4845         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4846         (cortexa53_tunings): Likewise.
4847         (aarch64_sched_issue_rate): New function.
4848         (TARGET_SCHED_ISSUE_RATE): Define.
4850 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4852         * ira-costs.c (find_costs_and_classes): Add missed
4853         ira_init_register_move_cost_if_necessary.
4855 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4857         PR target/59787
4858         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4860 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4862         PR target/59794
4863         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4864         to indicate if type is used for function return value.  Warn ABI
4865         change if the vector mode isn't available for function return value.
4866         (ix86_function_arg_advance): Pass false to type_natural_mode.
4867         (ix86_function_arg): Likewise.
4868         (ix86_gimplify_va_arg): Likewise.
4869         (function_arg_32): Don't warn ABI change.
4870         (ix86_function_value): Pass true to type_natural_mode.
4871         (ix86_return_in_memory): Likewise.
4872         (ix86_struct_value_rtx): Removed.
4873         (TARGET_STRUCT_VALUE_RTX): Likewise.
4875 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4877         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4878         converting a conditional jump into a conditional return.
4880 2014-01-14  Richard Biener  <rguenther@suse.de>
4882         PR tree-optimization/58921
4883         PR tree-optimization/59006
4884         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4885         hoisting invariant stmts.
4886         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4887         invariant loads on the preheader edge if possible.
4889 2014-01-14  Joey Ye  <joey.ye@arm.com>
4891         * doc/plugin.texi (Building GCC plugins): Update to C++.
4893 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4895         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4896         (_mm_rcp28_round_ss): Ditto.
4897         (_mm_rsqrt28_round_sd): Ditto.
4898         (_mm_rsqrt28_round_ss): Ditto.
4899         (_mm_rcp28_sd): Ditto.
4900         (_mm_rcp28_ss): Ditto.
4901         (_mm_rsqrt28_sd): Ditto.
4902         (_mm_rsqrt28_ss): Ditto.
4903         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4904         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4905         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4906         (IX86_BUILTIN_RCP28SD): Ditto.
4907         (IX86_BUILTIN_RCP28SS): Ditto.
4908         (IX86_BUILTIN_RSQRT28SD): Ditto.
4909         (IX86_BUILTIN_RSQRT28SS): Ditto.
4910         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4911         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4912         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4913         (ix86_expand_special_args_builtin): Expand new FTYPE.
4914         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4915         (srcp14<mode>): Make insn unary.
4916         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4917         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4918         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4919         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4920         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4921         Fix rounding: make it SAE only.
4922         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4923         Ditto.
4924         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4925         Ditto.
4926         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4927         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4928         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4929         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4930         (round_saeonly_mask_scalar_operand4): Ditto.
4931         (round_saeonly_mask_scalar_op3): Ditto.
4932         (round_saeonly_mask_scalar_op4): Ditto.
4934 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4936         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4937         Implement -maltivec=be for vec_insert and vec_extract.
4939 2014-01-10  DJ Delorie  <dj@redhat.com>
4941         * config/msp430/msp430.md (call_internal): Don't allow memory
4942         references with SP as the base register.
4943         (call_value_internal): Likewise.
4944         * config/msp430/constraints.md (Yc): New.  For memory references
4945         that don't use SP as a base register.
4947         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4948         "an integer without a # prefix"
4949         * config/msp430/msp430.md (epilogue_helper): Use it.
4951 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4953         PR target/59617
4954         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4955         AVX512F gather builtins.
4956         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4957         on gather decls with INTEGER_TYPE masktype.
4958         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4959         directly into the builtin rather than hoisting it before loop.
4961         PR tree-optimization/59387
4962         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4963         (scev_const_prop): If folded_casts and type has undefined overflow,
4964         use force_gimple_operand instead of force_gimple_operand_gsi and
4965         for each added stmt if it is assign with
4966         arith_code_with_undefined_signed_overflow, call
4967         rewrite_to_defined_overflow.
4968         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4969         gimple-fold.h instead.
4970         (arith_code_with_undefined_signed_overflow,
4971         rewrite_to_defined_overflow): Moved to ...
4972         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4973         rewrite_to_defined_overflow): ... here.  No longer static.
4974         Include gimplify-me.h.
4975         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4976         rewrite_to_defined_overflow): New prototypes.
4978 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4980         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4982 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4984         * builtins.c (get_object_alignment_2): Minor tweak.
4985         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4987 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4989         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4990         optimized non constant lengths.
4992 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4994         PR libgomp/59194
4995         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4996         load as __atomic_load_N if possible.
4998 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5000         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5001         target parameter.
5002         (rs6000_expand_builtin): Adjust call.
5004 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5006         PR target/58115
5007         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5008         * config/rs6000/rs6000.c: Include target-globals.h.
5009         (rs6000_set_current_function): Instead of doing target_reinit
5010         unconditionally, use save_target_globals_default_opts and
5011         restore_target_globals.
5013         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5014         FPSCR.
5015         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5016         (rs6000_expand_builtin): Handle mffs and mtfsf.
5017         (rs6000_init_builtins): Define mffs and mtfsf.
5018         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5019         (rs6000_mffs): New pattern.
5020         (rs6000_mtfsf): New pattern.
5022 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5024         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5025         Start narrowing with START.  Apply candidate-use pair
5026         and check overall cost in narrowing.
5027         (iv_ca_prune): Pass new argument.
5029 2014-01-10  Jeff Law  <law@redhat.com>
5031         PR middle-end/59743
5032         * ree.c (combine_reaching_defs): Ensure the defining statement
5033         occurs before the extension when optimizing extensions with
5034         different source and destination hard registers.
5036 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5038         PR ipa/58585
5039         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5040         vtables into the type inheritance graph.
5042 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5044         PR rtl-optimization/59754
5045         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5046         modes in the REGNO != REGNO case.
5048 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5050         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5052 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5054         PR tree-optimization/59745
5055         * tree-predcom.c (tree_predictive_commoning_loop): Call
5056         free_affine_expand_cache if giving up because components is NULL.
5058         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5059         GC in payload of target_globals struct instead of allocating them on
5060         the heap and the larger structs separately using GC.
5061         * target-globals.h (struct target_globals): Make regs, hard_regs,
5062         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5063         of GTY((skip)) and change type to void *.
5064         (reset_target_globals): Cast loads from those fields to corresponding
5065         types.
5067 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5069         PR plugins/59335
5070         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5071         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5072         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5074 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5076         PR target/59744
5077         * aarch64-modes.def (CC_Zmode): New flags mode.
5078         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5079         represents an equality.
5080         (aarch64_get_condition_code): Handle CC_Zmode.
5081         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5083 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5085         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5086         extraction in good case.
5088 2014-01-10  Richard Biener  <rguenther@suse.de>
5090         PR tree-optimization/59374
5091         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5092         checking after SLP discovery.  Mark stmts not participating
5093         in any SLP instance properly.
5095 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5097         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5098         when handling a SET rtx.
5100 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5102         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5103         (cortex-a57): Likewise.
5104         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5106 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5108         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5109         non-iwmmxt builtins.
5111 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5113         PR ipa/58252
5114         PR ipa/59226
5115         * ipa-devirt.c record_target_from_binfo): Take as argument
5116         stack of binfos and lookup matching one for virtual inheritance.
5117         (possible_polymorphic_call_targets_1): Update.
5119 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5121         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5122         kernel strings for Loongson-2E/2F/3A.
5124 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5126         PR middle-end/59670
5127         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5128         is_gimple_call before calling gimple_call_internal_p.
5130 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5132         * Makefile.in (TREE_FLOW_H): Remove.
5133         (TREE_SSA_H): Add file names from tree-flow.h.
5134         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5135         * tree.h: Remove tree-flow.h reference.
5136         * hash-table.h: Remove tree-flow.h reference.
5137         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5138         reference with tree-ssa-loop.h.
5140 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5142         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5143         default element-order behavior for -maltivec.
5144         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5145         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5146         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5147         when targeting big endian, at least for now.
5148         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5150 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5152         PR middle-end/47735
5153         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5154         var satisfies use_register_for_decl, just take into account type
5155         alignment, rather than decl alignment.
5157         PR tree-optimization/59622
5158         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5159         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5160         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5161         Don't devirtualize for inplace at all.  For targets.length () == 1,
5162         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5164 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5166         * config/i386/i386.md (cpu): Remove the unused btver1.
5168 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5170         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5172 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5174         PR target/58115
5175         * tree-core.h (struct target_globals): New forward declaration.
5176         (struct tree_target_option): Add globals field.
5177         * tree.h (TREE_TARGET_GLOBALS): Define.
5178         (prepare_target_option_nodes_for_pch): New prototype.
5179         * target-globals.h (struct target_globals): Define even if
5180         !SWITCHABLE_TARGET.
5181         * tree.c (prepare_target_option_node_for_pch,
5182         prepare_target_option_nodes_for_pch): New functions.
5183         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5184         * config/i386/i386.c: Include target-globals.h.
5185         (ix86_set_current_function): Instead of doing target_reinit
5186         unconditionally, use save_target_globals_default_opts and
5187         restore_target_globals.
5189 2014-01-09  Richard Biener  <rguenther@suse.de>
5191         PR tree-optimization/59715
5192         * tree-cfg.h (split_critical_edges): Declare.
5193         * tree-cfg.c (split_critical_edges): Export.
5194         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5196 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5198         * cfgexpand.c (expand_stack_vars): Optionally disable
5199         asan stack protection.
5200         (expand_used_vars): Likewise.
5201         (partition_stack_vars): Likewise.
5202         * asan.c (asan_emit_stack_protection): Optionally disable
5203         after return stack usage.
5204         (instrument_derefs): Optionally disable memory access instrumentation.
5205         (instrument_builtin_call): Likewise.
5206         (instrument_strlen_call): Likewise.
5207         (asan_protect_global): Optionally disable global variables protection.
5208         * doc/invoke.texi: Added doc for new options.
5209         * params.def: Added new options.
5210         * params.h: Likewise.
5212 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5214         PR rtl-optimization/59724
5215         * ifcvt.c (cond_exec_process_if_block): Don't call
5216         flow_find_head_matching_sequence with 0 longest_match.
5217         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5218         non-active insns if !stop_after.
5219         (try_head_merge_bb): Revert 2014-01-07 changes.
5221 2014-01-08  Jeff Law  <law@redhat.com>
5223         * ree.c (get_sub_rtx): New function, extracted from...
5224         (merge_def_and_ext): Here.
5225         (combine_reaching_defs): Use get_sub_rtx.
5227 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5229         * cgraph.h (varpool_variable_node): Do not choke on null node.
5231 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5233         * config/mips/mips.md (simple_return): Attempt to use JRC
5234         for microMIPS.
5235         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5237 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5239         PR rtl-optimization/59137
5240         * reorg.c (steal_delay_list_from_target): Call update_block for
5241         elided insns.
5242         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5244 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5246         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5247         two duplicate entries.
5249 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5251         Revert:
5252         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5254         * config/mips/mips.c (mips_truncated_op_cost): New function.
5255         (mips_rtx_costs): Adjust test for BADDU.
5256         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5258         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5260         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5261         (*baddu_si): ...this new pattern.
5263 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5265         PR ipa/59722
5266         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5268 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5270         PR middle-end/57748
5271         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5272         inner_reference_p.
5273         (expand_expr, expand_normal): Adjust.
5274         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5275         inner_reference_p. Use inner_reference_p to expand inner references.
5276         (store_expr): Adjust.
5277         * cfgexpand.c (expand_call_stmt): Adjust.
5279 2014-01-08  Rong Xu  <xur@google.com>
5281         * gcov-io.c (gcov_var): Move from gcov-io.h.
5282         (gcov_position): Ditto.
5283         (gcov_is_error): Ditto.
5284         (gcov_rewrite): Ditto.
5285         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5286         only part to libgcc/libgcov.h.
5288 2014-01-08  Marek Polacek  <polacek@redhat.com>
5290         PR middle-end/59669
5291         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5293 2014-01-08  Marek Polacek  <polacek@redhat.com>
5295         PR sanitizer/59667
5296         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5298 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5300         PR rtl-optimization/59649
5301         * stor-layout.c (get_mode_bounds): For BImode return
5302         0 and STORE_FLAG_VALUE.
5304 2014-01-08  Richard Biener  <rguenther@suse.de>
5306         PR middle-end/59630
5307         * gimple.h (is_gimple_builtin_call): Remove.
5308         (gimple_builtin_call_types_compatible_p): New.
5309         (gimple_call_builtin_p): New overload.
5310         * gimple.c (is_gimple_builtin_call): Remove.
5311         (validate_call): Rename to ...
5312         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5313         check return types.
5314         (validate_type): New static function.
5315         (gimple_call_builtin_p): New overload and adjust.
5316         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5317         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5318         (gimple_fold_stmt_to_constant_1): Likewise.
5319         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5321 2014-01-08  Richard Biener  <rguenther@suse.de>
5323         PR middle-end/59471
5324         * gimplify.c (gimplify_expr): Gimplify register-register type
5325         VIEW_CONVERT_EXPRs to separate stmts.
5327 2014-01-07  Jeff Law  <law@redhat.com>
5329         PR middle-end/53623
5330         * ree.c (combine_set_extension): Handle case where source
5331         and destination registers in an extension insn are different.
5332         (combine_reaching_defs): Allow source and destination registers
5333         in extension to be different under limited circumstances.
5334         (add_removable_extension): Remove restriction that the
5335         source and destination registers in the extension are the same.
5336         (find_and_remove_re): Emit a copy from the extension's
5337         destination to its source after the defining insn if
5338         the source and destination registers are different.
5340         PR middle-end/59285
5341         * ifcvt.c (merge_if_block): If we are merging a block with more than
5342         one successor with a block with no successors, remove any BARRIER
5343         after the second block.
5345 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
5347         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5349 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5351         PR target/59652
5352         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5353         for 14-bit register offsets when INT14_OK_STRICT is false.
5355 2014-01-07  Roland Stigge  <stigge@antcom.de>
5356             Michael Meissner  <meissner@linux.vnet.ibm.com>
5358         PR 57386/target
5359         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5360         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5362 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5364         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5365         -mcpu.
5367 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5369         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5370         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5371         rtx is const0_rtx or not.
5373 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5375         PR target/58115
5376         * target-globals.c (save_target_globals): Remove this_fn_optab
5377         handling.
5378         * toplev.c: Include optabs.h.
5379         (target_reinit): Temporarily restore the global options if another
5380         set of options are in force.
5382 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5384         PR rtl-optimization/58668
5385         * cfgcleanup.c (flow_find_cross_jump): Don't count
5386         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5387         to determine what is counted.
5388         (flow_find_head_matching_sequence): Use active_insn_p to determine
5389         what is counted.
5390         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5391         counting change.
5392         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5393         determine what is counted.
5395         PR tree-optimization/59643
5396         * tree-predcom.c (split_data_refs_to_components): If one dr is
5397         read and one write, determine_offset fails and the write isn't
5398         in the bad component, just put the read into the bad component.
5400 2014-01-07  Mike Stump  <mikestump@comcast.net>
5401             Jakub Jelinek  <jakub@redhat.com>
5403         PR pch/59436
5404         * tree-core.h (struct tree_optimization_option): Change optabs
5405         type from unsigned char * to void *.
5406         * optabs.c (init_tree_optimization_optabs): Adjust
5407         TREE_OPTIMIZATION_OPTABS initialization.
5409 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5411         PR target/59644
5412         * config/i386/i386.h (struct machine_function): Add
5413         no_drap_save_restore field.
5414         * config/i386/i386.c (ix86_save_reg): Use
5415         !cfun->machine->no_drap_save_restore instead of
5416         crtl->stack_realign_needed.
5417         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5418         this function clears frame_pointer_needed.  Set
5419         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5420         and DRAP reg is needed.
5422 2014-01-06  Marek Polacek  <polacek@redhat.com>
5424         PR c/57773
5425         * doc/implement-c.texi: Mention that other integer types are
5426         permitted as bit-field types in strictly conforming mode.
5428 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5430         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5431         is newly allocated.
5433 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5435         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5437 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5439         PR ipa/59008
5440         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5441         to int.
5442         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5444 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5446         PR debug/59350
5447         PR debug/59510
5448         * var-tracking.c (add_stores): Preserve the value of the source even if
5449         we don't record the store.
5451 2014-01-06  Terry Guo  <terry.guo@arm.com>
5453         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5455 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5457         PR bootstrap/59541
5458         * config/darwin.c (darwin_function_section): Adjust return values to
5459         correspond to optimisation changes made in r206070.
5461 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5463         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5464         from prefetch_block tune setting.
5465         (nocona_cost): Correct size of prefetch block to 64.
5467 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5469         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5470         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5471         used to save the static chain register in the computation of the offset
5472         from which the FP registers need to be restored.
5474 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5476         PR tree-optimization/59519
5477         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5478         ICE if get_current_def (current_new_name) is already non-NULL, as long
5479         as it is a phi result of some other phi in *new_exit_bb that has
5480         the same argument.
5482         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5483         or vmovdqu* for misaligned_operand.
5484         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5485         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5486         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5487         aligned_mem for AVX512F masked aligned load and store builtins and for
5488         non-temporal moves.
5490 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5492         PR tree-optimization/59651
5493         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5494         Address range for negative step should be added by TYPE_SIZE_UNIT.
5496 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5498         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5499         overlapping registers also for registers other than the stack pointer.
5501 2014-01-03  Marek Polacek  <polacek@redhat.com>
5503         PR other/59661
5504         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5505         __builtin_FILE.
5507 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5509         PR target/59625
5510         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5511         asm goto as jump.
5513         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5514         (push splitter): Use <P:MODE_SIZE> instead of
5515         GET_MODE_SIZE (<P:MODE>mode).
5516         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5517         (mov -1, reg peephole2): Likewise.
5518         * config/i386/sse.md (*mov<mode>_internal,
5519         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5520         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5521         *<code><mode>3, *andnot<mode>3<mask_name>,
5522         <mask_codefor><code><mode>3<mask_name>): Likewise.
5523         * config/i386/subst.md (mask_mode512bit_condition,
5524         sd_mask_mode512bit_condition): Likewise.
5526 2014-01-02  Xinliang David Li  <davidxl@google.com>
5528         PR tree-optimization/59303
5529         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5530         (dump_predicates): Better output format.
5531         (pred_equal_p): New function.
5532         (is_neq_relop_p): Ditto.
5533         (is_neq_zero_form_p): Ditto.
5534         (pred_expr_equal_p): Ditto.
5535         (pred_neg_p): Ditto.
5536         (simplify_pred): Ditto.
5537         (simplify_preds_2): Ditto.
5538         (simplify_preds_3): Ditto.
5539         (simplify_preds_4): Ditto.
5540         (simplify_preds): Ditto.
5541         (push_pred): Ditto.
5542         (push_to_worklist): Ditto.
5543         (get_pred_info_from_cmp): Ditto.
5544         (is_degenerated_phi): Ditto.
5545         (normalize_one_pred_1): Ditto.
5546         (normalize_one_pred): Ditto.
5547         (normalize_one_pred_chain): Ditto.
5548         (normalize_preds): Ditto.
5549         (normalize_cond_1): Remove function.
5550         (normalize_cond): Ditto.
5551         (is_gcond_subset_of): Ditto.
5552         (is_subset_of_any): Ditto.
5553         (is_or_set_subset_of): Ditto.
5554         (is_and_set_subset_of): Ditto.
5555         (is_norm_cond_subset_of): Ditto.
5556         (pred_chain_length_cmp): Ditto.
5557         (convert_control_dep_chain_into_preds): Type change.
5558         (find_predicates): Ditto.
5559         (find_def_preds): Ditto.
5560         (destroy_predicates_vecs): Ditto.
5561         (find_matching_predicates_in_rest_chains): Ditto.
5562         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5563         (is_pred_expr_subset): Ditto.
5564         (is_pred_chain_subset_of): Ditto.
5565         (is_included_in): Ditto.
5566         (is_superset_of): Ditto.
5568 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5570         Update copyright years.
5572 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5574         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5575         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5576         config/arc/arc.md, config/arc/arc.opt,
5577         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5578         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5579         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5580         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5581         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5582         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5583         vtable-verify.c, vtable-verify.h: Use the standard form for the
5584         copyright notice.
5586 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5588         * gcc.c (process_command): Update copyright notice dates.
5589         * gcov-dump.c: Ditto.
5590         * gcov.c: Ditto.
5591         * doc/cpp.texi: Bump @copying's copyright year.
5592         * doc/cppinternals.texi: Ditto.
5593         * doc/gcc.texi: Ditto.
5594         * doc/gccint.texi: Ditto.
5595         * doc/gcov.texi: Ditto.
5596         * doc/install.texi: Ditto.
5597         * doc/invoke.texi: Ditto.
5599 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5601         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5603 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5605         * config/i386/sse.md (*mov<mode>_internal): Guard
5606         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5608         PR rtl-optimization/59647
5609         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5610         new_rtx into UNSIGNED_FLOAT rtxes.
5612 Copyright (C) 2014 Free Software Foundation, Inc.
5614 Copying and distribution of this file, with or without modification,
5615 are permitted in any medium without royalty provided the copyright
5616 notice and this notice are preserved.