2014-04-11 Tobias Burnus <burnus@net-b.de>
[official-gcc.git] / gcc / ChangeLog
blobfd681633384f581dfceb2d67268c62d1543ba736
1 2014-04-11  Tobias Burnus  <burnus@net-b.de>
3         PR c/60194
4         * doc/invoke.texi (-Wformat-signedness): Document it.
5         (Wformat=2): Mention that this enables -Wformat-signedness.
7 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
9         * common/config/epiphany/epiphany-common.c
10         (epiphany_option_optimization_table): Enable section anchors by
11         default at -O1 or higher.
12         * config/epiphany/epiphany.c (TARGET_MAX_ANCHOR_OFFSET): Define.
13         (TARGET_MIN_ANCHOR_OFFSET): Likewise.
14         (epiphany_rtx_costs) <SET>: For binary operators, the set as such
15         carries no extra cost.
16         (epiphany_legitimate_address_p): For BLKmode, apply SImode check.
17         * config/epiphany/epiphany.h (ASM_OUTPUT_DEF): Define.
18         * config/epiphany/predicates.md (memclob_operand): New predicate.
19         * config/epiphany/epiphany.md (stack_adjust_add, stack_adjust_str):
20         Use memclob_operand predicate and X constraint for operand 3.
22 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
24         * config/epiphany/epiphany.c (epiphany_rtx_cost): Compare
25         with CC_N_NE / CC_C_LTU / CC_C_GTU carries no extra cost for
26         its operands.
28 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
30         PR rtl-optimization/60651
31         * mode-switching.c (optimize_mode_switching): Make sure to emit
32         sets of a lower numbered entity before sets of a higher numbered
33         entity to a mode of the same or lower priority.
34         When creating a seginfo for a basic block that starts with a code
35         label, move the insertion point past the code label.
36         (new_seginfo): Document and enforce requirement that
37         NOTE_INSN_BASIC_BLOCK only appears for empty blocks.
38         * doc/tm.texi.in: Document ordering constraint for emitted mode sets.
39         * doc/tm.texi: Regenerate.
41 2014-01-11  Joern Rennecke  <joern.rennecke@embecosm.com>
43         PR target/60811
44         * config/arc/arc.c (arc_save_restore): Fix assert typo.
46 2013-04-11  Jakub Jelinek  <jakub@redhat.com>
48         * BASE-VER: Set to 4.10.0.
50 2014-04-11  Tobias Burnus  <burnus@net-b.de>
52         PR other/59055
53         * doc/bugreport.texi (Bugs): Remove nodes pointing to the
54         nirvana.
55         * doc/gcc.texi (Service): Update description in the @menu
56         * doc/invoke.texi (Option Summary): Remove misplaced and
57         duplicated @menu.
59 2014-04-11  Steve Ellcey  <sellcey@mips.com>
60             Jakub Jelinek  <jakub@redhat.com>
62         PR middle-end/60556
63         * expr.c (convert_move): Use emit_store_flag_force instead of
64         emit_store_flag.  Pass lowpart_mode instead of VOIDmode as 5th
65         argument to it.
67 2014-04-11  Richard Biener  <rguenther@suse.de>
69         PR middle-end/60797
70         * varasm.c (assemble_alias): Avoid endless error reporting
71         recursion by setting TREE_ASM_WRITTEN.
73 2014-04-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
75         * config/s390/s390.md: Add a splitter for NOT rtx.
77 2014-04-11  Jakub Jelinek  <jakub@redhat.com>
79         PR rtl-optimization/60663
80         * cse.c (cse_insn): Set src_volatile on ASM_OPERANDS in
81         PARALLEL.
83 2014-04-10  Jan Hubicka  <hubicka@ucw.cz>
84             Jakub Jelinek  <jakub@redhat.com>
86         PR lto/60567
87         * ipa.c (function_and_variable_visibility): Copy forced_by_abi flag from
88         decl_node to node.
90 2014-04-10  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
92         PR debug/60655
93         * config/arm/arm.c (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Define
94         (arm_const_not_ok_for_debug_p): Reject MINUS with SYM_REF's
95         ameliorating the cases where it can be.
97 2014-04-09  David Edelsohn  <dje.gcc@gmail.com>
99         Revert
100         2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
102         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
103         (loadsync_<mode>): Change mode.
104         (load_quadpti, store_quadpti): New.
105         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
106         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
107         * config/rs6000/predicates.md (quad_memory_operand): !TARGET_SYNC_TI.
109 2014-04-09  Cong Hou  <congh@google.com>
111         PR testsuite/60773
112         * doc/sourcebuild.texi (vect_widen_mult_si_to_di_pattern): Add
113         documentation.
115 2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
117         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
118         instead of vnor to exploit possible fusion opportunity in the
119         future.
120         (altivec_expand_vec_perm_const_le): Likewise.
122 2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
124         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
125         (loadsync_<mode>): Change mode.
126         (load_quadpti, store_quadpti): New.
127         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
128         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
130 2014-04-08  Richard Sandiford  <rdsandiford@googlemail.com>
132         PR target/60763
133         * config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
134         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
135         Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.
137 2014-04-08  Richard Biener  <rguenther@suse.de>
139         PR middle-end/60706
140         * tree-pretty-print.c (pp_double_int): For HWI32 hosts with
141         a 64bit widest int print double-int similar to on HWI64 hosts.
143 2014-04-08  Richard Biener  <rguenther@suse.de>
145         PR tree-optimization/60785
146         * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
147         default defs properly.
149 2014-04-08  Nathan Sidwell  <nathan@codesourcery.com>
151         * doc/invoke (Wnon-virtual-dtor): Update to match implementation.
152         (Weffc++): Likewise.
154 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
156         * ipa-devirt.c (maybe_record_node): When node is not recorded,
157         set completep to false rather than true.
159 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
161         PR target/60504
162         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
163         ARM_TARGET2_DWARF_FORMAT.
165 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
167         PR target/60609
168         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
169         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
170         ADDR_DIFF_VEC.
172 2014-04-07  Richard Biener  <rguenther@suse.de>
174         PR tree-optimization/60766
175         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
176         (may_eliminate_iv): Convert cand_value_at result to desired type.
178 2014-04-07  Jason Merrill  <jason@redhat.com>
180         PR c++/60731
181         * common.opt (-fno-gnu-unique): Add.
182         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
184 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
186         * haifa-sched.c: Fix outdated function reference and minor
187         grammar errors in introductory comment.
189 2014-04-07  Richard Biener  <rguenther@suse.de>
191         PR middle-end/60750
192         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
193         for noreturn calls.
194         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
196 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
198         PR debug/55794
199         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
200         size accounting for thunks.
201         (pa_asm_output_mi_thunk): Use final_start_function() and
202         final_end_function() to output function start and end directives.
204 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
206         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
207         specific ISA/ feature information. Remove short_sp and errata_skip ds.
208         Add avr_device_specific_features enum to have device specific info.
209         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
210         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
211         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
212         updated device specific info.
213         * config/avr/avr-mcus.def: Merge device specific details to
214         dev_attribute field.
215         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
216         errata_skip.
217         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
218         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
219         assembler if RMW isa supported by current device.
220         * config/avr/genmultilib.awk: Update as device info structure changed.
221         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
223 2014-04-04  Cong Hou  <congh@google.com>
225         PR tree-optimization/60656
226         * tree-vect-stmts.c (supportable_widening_operation):
227         Fix a bug that elements in a vector with vect_used_by_reduction
228         property are incorrectly reordered when the operation on it is not
229         consistant with the one in reduction operation.
231 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
233         PR rtl-optimization/60155
234         * gcse.c (record_set_data): New function.
235         (single_set_gcse): New function.
236         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
237         (hoist_code): Likewise.
238         (get_pressure_class_and_nregs): Likewise.
240 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
242         * explow.c (probe_stack_range): Emit a final optimization blockage.
244 2014-04-04  Anthony Green  <green@moxielogic.com>
246         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
247         typos.
249 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
251         PR ipa/59626
252         * lto-cgraph.c (input_overwrite_node): Check that partitioning
253         flags are set only during streaming.
254         * ipa.c (process_references, walk_polymorphic_call_targets,
255         symtab_remove_unreachable_nodes): Drop bodies of always inline
256         after early inlining.
257         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
259 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
260         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
262         PR debug/60655
263         * dwarf2out.c (const_ok_for_output_1): Reject expressions
264         containing a NOT.
266 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
268         PR bootstrap/60743
269         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
270         duration.
271         (cortex_a53_fdivd): Likewise.
273 2014-04-04  Martin Jambor  <mjambor@suse.cz>
275         PR ipa/60640
276         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
277         Adjust all callers.
278         * cgraph.c (clone_of_p): Also return true if thunks match.
279         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
280         cgraph_function_or_thunk_node and an obsolete comment.
281         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
282         file.
283         (build_function_decl_skip_args): Likewise.
284         (set_new_clone_decl_and_node_flags): New function.
285         (duplicate_thunk_for_node): Likewise.
286         (redirect_edge_duplicating_thunks): Likewise.
287         (cgraph_clone_node): New parameter args_to_skip, pass it to
288         redirect_edge_duplicating_thunks which is called instead of
289         cgraph_redirect_edge_callee.
290         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
291         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
293 2014-04-04  Jeff Law  <law@redhat.com>
295         PR target/60657
296         * config/arm/predicates.md (const_int_I_operand): New predicate.
297         (const_int_M_operand): Similarly.
298         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
299         const_int_operand.
300         (insv_t2, extv_reg, extzv_t2): Likewise.
301         (load_multiple_with_writeback): Similarly for const_int_I_operand.
302         (pop_multiple_with_writeback_and_return): Likewise.
303         (vfp_pop_multiple_with_writeback): Likewise
305 2014-04-04  Richard Biener  <rguenther@suse.de>
307         PR ipa/60746
308         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
309         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
310         non-GIMPLE_LABELs.
311         * gimplify.h (gimple_add_tmp_var_fn): Declare.
312         * gimplify.c (gimple_add_tmp_var_fn): New function.
313         * gimple-expr.h (create_tmp_reg_fn): Declare.
314         * gimple-expr.c (create_tmp_reg_fn): New function.
315         * gimple-low.c (record_vars_into): Don't change cfun.
316         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
317         code generation without cfun.
319 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
321         PR bootstrap/60719
322         * Makefile.in (install-driver): Fix shell scripting.
324 2014-04-03  Cong Hou  <congh@google.com>
326         PR tree-optimization/60505
327         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
328         threshold of number of iterations below which no vectorization will be
329         done.
330         * tree-vect-loop.c (new_loop_vec_info):
331         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
332         * tree-vect-loop.c (vect_analyze_loop_operations):
333         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
334         * tree-vect-loop.c (vect_transform_loop):
335         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
336         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
337         of iterations of the loop and see if we should build the epilogue.
339 2014-04-03  Richard Biener  <rguenther@suse.de>
341         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
342         member.
343         (streamer_tree_cache_create): Adjust.
344         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
345         to allow optional nodes array.
346         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
347         (streamer_tree_cache_append): Likewise.
348         (streamer_tree_cache_create): Create nodes array optionally
349         as specified by parameter.
350         * lto-streamer-out.c (create_output_block): Avoid maintaining
351         the node array in the writer cache.
352         (DFS_write_tree): Remove assertion.
353         (produce_asm_for_decls): Free the out decl state hash table
354         early.
355         * lto-streamer-in.c (lto_data_in_create): Adjust for
356         streamer_tree_cache_create prototype change.
358 2014-04-03  Richard Biener  <rguenther@suse.de>
360         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
361         set TREE_CHAIN to NULL_TREE.
363 2014-04-03  Richard Biener  <rguenther@suse.de>
365         PR tree-optimization/60740
366         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
367         over all GIMPLE_COND operands.
369 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
371         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
372         (Weffc++): Remove Scott's numbering, merge lists and reference
373         Wnon-virtual-dtor.
375         c-family/
377         cp/
378         * class.c (accessible_nvdtor_p): New.
379         (check_bases): Don't check base destructor here ...
380         (check_bases_and_members): ... check them here.  Trigger on
381         Wnon-virtual-dtor flag.
382         (finish_struct_1): Use accessible_nvdtor_p.
384         testsuite/
385         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
386         * g++.dg/warn/Wnvdtor-2.C: New.
387         * g++.dg/warn/Wnvdtor-3.C: New.
388         * g++.dg/warn/Wnvdtor-4.C: New.
389         * g++.dg/warn/Weff1.C: Delete.
390         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
391         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
393 2014-04-03  Nick Clifton  <nickc@redhat.com>
395         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
396         properly.
398 2014-04-03  Martin Jambor  <mjambor@suse.cz>
400         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
401         mention gcc_unreachable before failing.
402         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
403         removed symbols.
405 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
407         PR ipa/60659
408         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
409         code and instead mark the context inconsistent.
410         (possible_polymorphic_call_targets): For inconsistent contexts
411         return empty complete list.
413 2014-04-02  Anthony Green  <green@moxielogic.com>
415         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
416         (extendqisi2, extendhisi2): Define.
417         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
418         (WCHAR_TYPE): Change to unsigned int.
420 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
422         PR tree-optimization/60733
423         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
424         insertion point for PHI candidates to be the end of the feeding
425         block for the PHI argument.
427 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
429         PR rtl-optimization/60650
430         * lra-constraints.c (process_alt_operands): Decrease reject for
431         earlyclobber matching.
433 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
435         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
437 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
439         * config/spu/spu.c (pad_bb): Do not crash when the last
440         insn is CODE_FOR_blockage.
442 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
444         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
445         lies outside the target mode.
447 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
449         PR target/60735
450         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
451         software floating point or no floating point registers, do not
452         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
453         in GPRs that occurs after we tested for GPRs that would never be
454         true.
456         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
457         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
458         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
459         specifically allow DDmode, since that does not use the SPE SIMD
460         instructions.
462 2014-04-02  Richard Biener  <rguenther@suse.de>
464         PR middle-end/60729
465         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
466         MODE_INTs.  Properly use negv_optab.
467         (expand_abs): Likewise.
469 2014-04-02  Richard Biener  <rguenther@suse.de>
471         PR bootstrap/60719
472         * Makefile.in (install-driver): Guard extra installs with special
473         names properly.
475 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
477         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
478         Document vec_vgbbd.
480 2014-04-01  Richard Henderson  <rth@redhat.com>
482         PR target/60704
483         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
484         alternative enabled before register allocation.
486 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
488         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
489         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
490         typo.
491         (nios2_large_got_address): Remove unneeded 'sym' parameter.
492         (nios2_got_address): Update nios2_large_got_address call site.
493         (nios2_delegitimize_address): New function.
494         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
495         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
496         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
498 2014-04-01  Martin Husemann  <martin@duskware.de>
500         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
501         for -mabi=32.
503 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
505         PR rtl-optimization/60604
506         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
507         check from register_operand.
508         (register_operand): Redefine in terms of general_operand.
509         (nonmemory_operand): Use register_operand for the non-constant cases.
511 2014-04-01  Richard Biener  <rguenther@suse.de>
513         * gimple.h (struct gimple_statement_base): Align subcode to
514         16 bits.
516 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
518         * doc/invoke.texi (mapp-regs): Clarify.
520 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
522         * config/i386/avx512fintrin.h (__v32hi): Define type.
523         (__v64qi): Likewise.
524         (_mm512_set1_epi8): Define.
525         (_mm512_set1_epi16): Define.
526         (_mm512_set4_epi32): Define.
527         (_mm512_set4_epi64): Define.
528         (_mm512_set4_pd): Define.
529         (_mm512_set4_ps): Define.
530         (_mm512_setr4_epi64): Define.
531         (_mm512_setr4_epi32): Define.
532         (_mm512_setr4_pd): Define.
533         (_mm512_setr4_ps): Define.
534         (_mm512_setzero_epi32): Define.
536 2014-03-31  Martin Jambor  <mjambor@suse.cz>
538         PR middle-end/60647
539         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
540         callsite_arguments_match_p.  Updated all callers.  Also check types of
541         corresponding formal parameters and actual arguments.
542         (not_all_callers_have_enough_arguments_p) Renamed to
543         some_callers_have_mismatched_arguments_p.
545 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
547         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
549 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
551         PR target/60034
552         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
553         section anchor.
555 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
557         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
558         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
559         Split out
560         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
561         Use FMAMODE_NOVF512 mode iterator.
562         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
563         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
564         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
565         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
566         Split out
567         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
568         Use VF_128_256 mode iterator.
569         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
570         Ditto.
572 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
574         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
575         static chain if needed.
577 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
579         PR target/60697
580         * lra-constraints.c (index_part_to_reg): New.
581         (process_address): Use it.
583 2014-03-27  Jeff Law  <law@redhat.com>
584             Jakub Jelinek  <jakub@redhat.com>
586         PR target/60648
587         * expr.c (do_tablejump): Use simplify_gen_binary rather than
588         gen_rtx_{PLUS,MULT} to build up the address expression.
590         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
591         creating non-canonical RTL.
593 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
595         PR ipa/60243
596         * ipa-inline.c (want_inline_small_function_p): Short circuit large
597         functions; reorganize to make cheap checks first.
598         (inline_small_functions): Do not estimate growth when dumping;
599         it is expensive.
600         * ipa-inline.h (inline_summary): Add min_size.
601         (growth_likely_positive): New function.
602         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
603         (set_cond_stmt_execution_predicate): Cleanup.
604         (estimate_edge_size_and_time): Compute min_size.
605         (estimate_calls_size_and_time): Likewise.
606         (estimate_node_size_and_time): Likewise.
607         (inline_update_overall_summary): Update min_size.
608         (do_estimate_edge_time): Likewise.
609         (do_estimate_edge_size): Update.
610         (do_estimate_edge_hints): Update.
611         (growth_likely_positive): New function.
613 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
615         PR target/60693
616         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
617         also if addr has VOIDmode.
619 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
621         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
622         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
623         Declare extern.
624         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
625         instructions as well as AdvancedSIMD loads.
627 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
629         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
630         Use crypto_aese type.
631         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
632         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
633         crypto_aese, crypto_aesmc.  Move to types.md.
634         * config/arm/types.md (crypto_aes): Split into crypto_aese,
635         crypto_aesmc.
636         * config/arm/iterators.md (crypto_type): Likewise.
638 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
640         * cgraph.c: Include expr.h and tree-dfa.h.
641         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
642         remove LHS.
644 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
646         PR target/60675
647         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
648         regs from checking multi-reg pseudos.
650 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
652         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
654 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
656         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
657         if it would clobber the stack pointer, even temporarily.
659 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
661         * mode-switching.c: Make small adjustments to the top comment.
663 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
665         * config/rs6000/constraints.md (wD constraint): New constraint to
666         match the constant integer to get the top DImode/DFmode out of a
667         vector in a VSX register.
669         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
670         match the constant integer to get the top DImode/DFmode out of a
671         vector in a VSX register.
673         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
674         for ISA 2.07.
676         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
677         vbpermq builtins.
679         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
680         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
682         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
683         Optimize vec_extract of 64-bit values, where the value being
684         extracted is in the top word, where we can use scalar
685         instructions.  Add direct move and store support.  Combine the big
686         endian/little endian vector select load support into a single insn.
687         (vsx_extract_<mode>_internal1): Likewise.
688         (vsx_extract_<mode>_internal2): Likewise.
689         (vsx_extract_<mode>_load): Likewise.
690         (vsx_extract_<mode>_store): Likewise.
691         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
692         combined into vsx_extract_<mode>_load.
693         (vsx_extract_<mode>_one_le): Likewise.
695         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
696         define the top 64-bit vector element.
698         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
699         constraint.
701         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
702         Document vec_vbpermq builtin.
704         PR target/60672
705         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
706         enable use of xxsldwi and xxpermdi builtin functions.
707         (vec_xxpermdi): Likewise.
709         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
710         Document use of vec_xxsldwi and vec_xxpermdi builtins.
712 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
714         PR rtl-optimization/60650
715         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
716         first_p.  Use it.
717         (find_spills_for): New.
718         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
719         Spill all pseudos on the second iteration.
721 2014-03-27  Marek Polacek  <polacek@redhat.com>
723         PR c/50347
724         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
725         types.
727 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
729         * config/s390/s390.c (s390_can_use_return_insn): Check for
730         call-saved FPRs on 31 bit.
732 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
734         PR middle-end/60682
735         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
736         if they need regimplification, just drop them instead of
737         calling gimple_regimplify_operands on them.
739 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
741         PR target/60580
742         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
743         (aarch64_frame_pointer_required): Adjust logic.
744         (aarch64_can_eliminate): Adjust logic.
745         (aarch64_override_options_after_change): Adjust logic.
747 2014-03-27  Dehao Chen  <dehao@google.com>
749         * ipa-inline.c (early_inliner): Update node's inline info.
751 2014-03-26  Dehao Chen  <dehao@google.com>
753         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
754         compiler inserted conditional jumps for NAN float check.
756 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
758         * ubsan.h (ubsan_create_data): Change second argument's type
759         to const location_t *.
760         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
761         _("<unknown>").
762         (ubsan_create_data): Change second argument to const location_t *PLOC.
763         Create Loc field whenever PLOC is non-NULL.
764         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
765         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
766         callers.
768         PR other/59545
769         * real.c (real_to_integer2): Change type of low to UHWI.
771 2014-03-26  Tobias Burnus  <burnus@net-b.de>
773         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
774         (CILK_SELF_SPECS): New define.
775         (driver_self_specs): Use it.
777 2014-03-26  Richard Biener  <rguenther@suse.de>
779         * tree-pretty-print.c (percent_K_format): Implement special
780         case for LTO and its stripped down BLOCK tree.
782 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
784         PR sanitizer/60636
785         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
787         * tree-vrp.c (simplify_internal_call_using_ranges): If only
788         one range is range_int_cst_p, but not both, at least optimize
789         addition/subtraction of 0 and multiplication by 0 or 1.
790         * gimple-fold.c (gimple_fold_call): Fold
791         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
792         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
793         INTEGER_CSTs, try to fold at least x * 0 and y - y.
795 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
797         PR rtl-optimization/60452
798         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
799         <case REG>: Return 1 for invalid offsets from the frame pointer.
801 2014-03-26  Marek Polacek  <polacek@redhat.com>
803         PR c/37428
804         * doc/extend.texi (C Extensions): Mention variable-length arrays in
805         a structure/union.
807 2014-03-26  Marek Polacek  <polacek@redhat.com>
809         PR c/39525
810         * doc/extend.texi (Designated Inits): Describe what happens to omitted
811         field members.
813 2014-03-26  Marek Polacek  <polacek@redhat.com>
815         PR other/59545
816         * ira-color.c (update_conflict_hard_regno_costs): Perform the
817         multiplication in unsigned type.
819 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
821         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
823 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
825         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
827 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
829         PR ipa/60315
830         * cif-code.def (UNREACHABLE) New code.
831         * ipa-inline.c (inline_small_functions): Skip edges to
832         __builtlin_unreachable.
833         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
834         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
835         predicate to __bulitin_unreachable.
836         (set_cond_stmt_execution_predicate): Fix issue when
837         invert_tree_comparison returns ERROR_MARK.
838         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
839         propagate to inline clones.
840         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
841         to unreachable.
842         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
843         * cgraphclones.c (cgraph_clone_node): If call destination is already
844         ureachable, do not redirect it back.
845         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
846         unreachable.
848 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
850         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
851         Do not modify inline clones.
853 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
855         * config/i386/i386.md (general_sext_operand): New mode attr.
856         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
857         don't generate (sign_extend (const_int)).
858         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
859         operands[2].  Use We constraint instead of <i> and
860         <general_sext_operand> predicate instead of <general_operand>.
861         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
862         * config/i386/constraints.md (We): New constraint.
863         * config/i386/predicates.md (x86_64_sext_operand,
864         sext_operand): New predicates.
866 2014-03-25  Martin Jambor  <mjambor@suse.cz>
868         PR ipa/60600
869         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
870         inconsistent devirtualizations to __builtin_unreachable.
872 2014-03-25  Marek Polacek  <polacek@redhat.com>
874         PR c/35449
875         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
877 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
879         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
880         order of elements for big-endian.
882 2014-03-25  Richard Biener  <rguenther@suse.de>
884         PR middle-end/60635
885         * gimplify-me.c (gimple_regimplify_operands): Update the
886         re-gimplifed stmt.
888 2014-03-25  Martin Jambor  <mjambor@suse.cz>
890         PR ipa/59176
891         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
892         (lto_output_varpool_node): Likewise.
893         (input_overwrite_node): Likewise.
894         (input_varpool_node): Likewise.
896 2014-03-25  Richard Biener  <rguenther@suse.de>
898         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
899         (run_gcc): Likewise.
901 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
903         * combine.c (simplify_compare_const): Add MODE argument.
904         Handle mode_width 0 as very large mode_width.
905         (try_combine, simplify_comparison): Adjust callers.
907         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
908         type to avoid signed integer overflow.
909         * explow.c (plus_constant): Likewise.
911 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
913         * doc/generic.texi: Correct typos.
915 2014-03-24  Tobias Burnus  <burnus@net-b.de>
917         * doc/invoke.texi (-flto): Expand section about
918         using static libraries with LTO.
920 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
922         PR rtl-optimization/60501
923         * optabs.def (addptr3_optab): New optab.
924         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
925         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
926         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
928         * lra.c (emit_add3_insn): Use the addptr pattern if available.
930         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
932 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
934         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
935         _mm512_set1_pd.
937         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
938         (_mm256_undefined_ps): Define.
939         (_mm256_undefined_pd): Define.
940         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
941         (_mm_undefined_pd): Define.
942         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
943         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
944         (_mm512_undefined_ps): Define.
945         (_mm512_undefined_pd): Define.
946         Use _mm*_undefined_*.
947         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
949 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
951         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
952         (lshr_simd): DI mode added.
953         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
954         (aarch64_ushr_simddi): Likewise.
955         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
956         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
957         (vshrd_n_u64): Likewise.
959 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
961         * Makefile.in (s-macro_list): Depend on cc1.
963 2014-03-23  Teresa Johnson  <tejohnson@google.com>
965         * ipa-utils.c (ipa_print_order): Use specified dump file.
967 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
969         PR rtl-optimization/60601
970         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
972         * gcc.c (eval_spec_function): Initialize save_growing_value.
974 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
976         PR sanitizer/60613
977         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
978         code == MINUS_EXPR, never swap op0 with op1.
980         * toplev.c (init_local_tick): Avoid signed integer multiplication
981         overflow.
982         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
983         shift by first operand's bitsize.
985 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
987         PR target/60610
988         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
989         redefine to 1 or 0.
990         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
991         TARGET_ISA_64BIT_P(x).
993 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
995         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
996         pattern for vector nor instead of subtract from splat(-1).
997         (altivec_expand_vec_perm_const_le): Likewise.
999 2014-03-21  Richard Henderson  <rth@twiddle.net>
1001         PR target/60598
1002         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
1003         related insns after epilogue_completed.
1005 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1007         PR ipa/59176
1008         * cgraph.h (symtab_node): New flag body_removed.
1009         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
1010         when removing bodies.
1011         * symtab.c (dump_symtab_base): Dump body_removed flag.
1012         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
1013         had their bodies removed.
1015 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1017         PR ipa/60419
1018         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
1019         in the border.
1021 2014-03-21  Richard Biener  <rguenther@suse.de>
1023         PR tree-optimization/60577
1024         * tree-core.h (struct tree_base): Document nothrow_flag use
1025         in VAR_DECL_NONALIASED.
1026         * tree.h (VAR_DECL_NONALIASED): New.
1027         (may_be_aliased): Adjust.
1028         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
1030 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1032         * expr.c (expand_expr_real_1): Remove outdated comment.
1034 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
1036         PR middle-end/60597
1037         * ira.c (adjust_cleared_regs): Call copy_rtx on
1038         *reg_equiv[REGNO (loc)].src_p before passing it to
1039         simplify_replace_fn_rtx.
1041         PR target/60568
1042         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
1043         into CONST, put pic register as first operand of PLUS.  Use
1044         gen_const_mem for both 32-bit and 64-bit PIC got loads.
1046 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1048         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
1050 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1052         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
1053         around for store forwarding issue in the FPU on the UT699.
1054         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
1055         loads and operations if -mfix-ut699 is specified.
1056         (divtf3_hq): Tweak attribute.
1057         (sqrttf2_hq): Likewise.
1059 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1061         * calls.c (store_one_arg): Remove incorrect const qualification on the
1062         type of the temporary.
1063         * cfgexpand.c (expand_return): Likewise.
1064         * expr.c (expand_constructor): Likewise.
1065         (expand_expr_real_1): Likewise.
1067 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1069         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
1070         of parts.
1072 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
1074         PR target/60039
1075         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
1077 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
1079         * config/arm/aarch-common-protos.h
1080         (alu_cost_table): Fix spelling of "extend".
1081         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
1083 2014-03-19  Richard Biener  <rguenther@suse.de>
1085         PR middle-end/60553
1086         * tree-core.h (tree_type_common): Re-order pointer members
1087         to reduce recursion depth during GC walks.
1089 2014-03-19  Marek Polacek  <polacek@redhat.com>
1091         PR sanitizer/60569
1092         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
1093         before accessing it.
1095 2014-03-19  Richard Biener  <rguenther@suse.de>
1097         PR lto/59543
1098         * lto-streamer-in.c (input_function): In WPA stage do not drop
1099         debug stmts.
1101 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
1103         PR tree-optimization/60559
1104         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
1105         with build_zero_cst assignment.
1107 2014-03-18  Kai Tietz  <ktietz@redhat.com>
1109         PR rtl-optimization/56356
1110         * sdbout.c (sdbout_parms): Verify that parms'
1111         incoming argument is valid.
1112         (sdbout_reg_parms): Likewise.
1114 2014-03-18  Richard Henderson  <rth@redhat.com>
1116         PR target/60562
1117         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
1118         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
1119         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
1121 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
1123         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
1124         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
1125         Italicize plugin event names in description.  Explain that
1126         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
1127         Remind that no GCC functions should be called after PLUGIN_FINISH.
1128         Explain what pragmas with expansion are.
1130 2014-03-18  Martin Liska  <mliska@suse.cz>
1132         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
1133         gimple call statement is update.
1134         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
1135         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
1137 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1139         PR sanitizer/60557
1140         * ubsan.c (ubsan_instrument_unreachable): Call
1141         initialize_sanitizer_builtins.
1142         (ubsan_pass): Likewise.
1144         PR sanitizer/60535
1145         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1146         varpool_finalize_decl instead of rest_of_decl_compilation.
1148 2014-03-18  Richard Biener  <rguenther@suse.de>
1150         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1151         by using bitmap_and_compl instead of bitmap_and_compl_into.
1152         (df_rd_transfer_function): Likewise.
1154 2014-03-18  Richard Biener  <rguenther@suse.de>
1156         * doc/lto.texi (fresolution): Fix typo.
1158 2014-03-18  Richard Biener  <rguenther@suse.de>
1160         * doc/invoke.texi (flto): Update for changes in 4.9.
1162 2014-03-18  Richard Biener  <rguenther@suse.de>
1164         * doc/loop.texi: Remove section on the removed lambda framework.
1165         Update loop docs with recent changes in preserving loop structure.
1167 2014-03-18  Richard Biener  <rguenther@suse.de>
1169         * doc/lto.texi (-fresolution): Document.
1171 2014-03-18  Richard Biener  <rguenther@suse.de>
1173         * doc/contrib.texi: Adjust my name.
1175 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1177         PR ipa/58721
1178         * internal-fn.c: Include diagnostic-core.h.
1179         (expand_BUILTIN_EXPECT): New function.
1180         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1181         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1182         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1183         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1184         IFN_BUILTIN_EXPECT.
1185         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1186         Revert 3 argument __builtin_expect code.
1187         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1188         * gimple-fold.c (gimple_fold_call): Likewise.
1189         * tree.h (fold_builtin_expect): New prototype.
1190         * builtins.c (build_builtin_expect_predicate): Add predictor
1191         argument, if non-NULL, create 3 argument __builtin_expect.
1192         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1193         pass it through to build_builtin_expect_predicate.
1194         (fold_builtin_2): Adjust caller.
1195         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1196         * internal-fn.def (BUILTIN_EXPECT): New.
1198 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1200         PR ipa/58721
1201         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1202         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1203         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1205 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1207         PR ipa/58721
1208         * predict.c (combine_predictions_for_bb): Fix up formatting.
1209         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1210         fill what it points to if non-NULL.
1211         (tree_predict_by_opcode): Adjust caller, use the predictor.
1212         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1214 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1216         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1217         proper constant for the store mode.
1219 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1221         * symtab.c (change_decl_assembler_name): Fix transparent alias
1222         chain construction.
1224 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1226         * config/aarch64/aarch64.c: Correct the comments about the
1227         aarch64 stack layout.
1229 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1231         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1232         check for GF_OMP_FOR_KIND_FOR.
1234 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1236         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1237         ymm and zmm register names.
1239 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1241         PR target/60516
1242         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1243         note creation for the 2010-08-31 changes.
1245 2014-03-17  Marek Polacek  <polacek@redhat.com>
1247         PR middle-end/60534
1248         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1249         as -fno-tree-loop-vectorize.
1250         (expand_omp_simd): Likewise.
1252 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1254         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1255         (eligible_for_call_delay): New prototype.
1256         * config/sparc/sparc.c (tls_call_delay): Rename into...
1257         (eligible_for_call_delay): ...this.  Return false if the instruction
1258         cannot be put in the delay slot of a branch.
1259         (eligible_for_restore_insn): Simplify.
1260         (eligible_for_return_delay): Return false if the instruction cannot be
1261         put in the delay slot of a branch and simplify.
1262         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1263         put in the delay slot of a branch.
1264         * config/sparc/sparc.md (fix_ut699): New attribute.
1265         (tls_call_delay): Delete.
1266         (in_call_delay): Reimplement.
1267         (eligible_for_sibcall_delay): Rename into...
1268         (in_sibcall_delay): ...this.
1269         (eligible_for_return_delay): Rename into...
1270         (in_return_delay): ...this.
1271         (in_branch_delay): Reimplement.
1272         (in_uncond_branch_delay): Delete.
1273         (in_annul_branch_delay): Delete.
1275 2014-03-14  Richard Henderson  <rth@redhat.com>
1277         PR target/60525
1278         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1279         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1280         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1281         (floathi splitters): Remove.
1282         (float<SWI48x>xf2): New pattern.
1283         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1284         code that tried to handle DImode for 32-bit, but which was excluded
1285         by the pattern's condition.  Drop allocation of stack temporary.
1286         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1287         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1288         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1289         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1290         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1291         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1292         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1293         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1294         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1295         (*float<SWI48x><X87MODEF>2_i387): Remove.
1296         (all float _with_temp splitters): Remove.
1297         (*float<SWI48x><MODEF>2_i387): New pattern.
1298         (*float<SWI48><MODEF>2_sse): New pattern.
1299         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1300         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1302 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1303             Marek Polacek  <polacek@redhat.com>
1305         PR middle-end/60484
1306         * common.opt (dump_base_name_prefixed): New Variable.
1307         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1308         if x_dump_base_name_prefixed is already set, set it at the end.
1310 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1312         PR rtl-optimization/60508
1313         * lra-constraints.c (get_reload_reg): Add new parameter
1314         in_subreg_p.
1315         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1316         Pass the new parameter values.
1318 2014-03-14  Richard Biener  <rguenther@suse.de>
1320         * common.opt: Revert unintented changes from r205065.
1321         * opts.c: Likewise.
1323 2014-03-14  Richard Biener  <rguenther@suse.de>
1325         PR middle-end/60518
1326         * cfghooks.c (split_block): Properly adjust all loops the
1327         block was a latch of.
1329 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1331         PR lto/60461
1332         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1333         and simplify it.
1335 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1337         PR target/59396
1338         * config/avr/avr.c (avr_set_current_function): Pass function name
1339         through default_strip_name_encoding before sanity checking instead
1340         of skipping the first char of the assembler name.
1342 2014-03-13  Richard Henderson  <rth@redhat.com>
1344         PR debug/60438
1345         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1346         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1347         * config/i386/i386-protos.h: Likewise.
1348         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1349         in the expander instead of a splitter.
1350         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1351         any possibility of requiring a memory.
1352         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1353         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1354         (fp branch splitters): Update for ix86_split_fp_branch.
1355         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1356         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1357         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1358         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1359         (*fop_<MODEF>_3_i387): Likewise.
1360         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1361         (splitters for the fop_* register patterns): Remove.
1362         (fscalexf4_i387): Rename from *fscalexf4_i387.
1363         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1365 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1367         PR tree-optimization/59779
1368         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1369         type for bitsize and maxsize instead of HOST_WIDE_INT.
1371 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1373         PR rtl-optimization/57320
1374         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1375         the CFG after thread_prologue_and_epilogue_insns.
1377 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1379         PR rtl-optimization/57189
1380         * lra-constraints.c (process_alt_operands): Disfavor spilling
1381         vector pseudos.
1383 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1385         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1387 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1389         PR tree-optimization/59025
1390         PR middle-end/60418
1391         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1392         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1394 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1396         PR target/60486
1397         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1398         calls of avr_out_plus_1.
1400 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1402         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1403         BB's single pred and update the father loop's latch info later.
1405 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1407         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1408         (VEC_M): Likewise.
1409         (VEC_N): Likewise.
1410         (VEC_R): Likewise.
1411         (VEC_base): Likewise.
1412         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1413         registers, we need to swap double words in little endian mode.
1415         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1416         to be a container mode for 128-bit integer operations added in ISA
1417         2.07.  Unlike TImode and PTImode, the preferred register set is
1418         the Altivec/VMX registers for the 128-bit operations.
1420         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1421         declarations.
1422         (rs6000_split_128bit_ok_p): Likewise.
1424         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1425         macros for creating ISA 2.07 normal and overloaded builtin
1426         functions with 3 arguments.
1427         (BU_P8V_OVERLOAD_3): Likewise.
1428         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1429         for use as overloaded functions.
1430         (VPERM_1TI_UNS): Likewise.
1431         (VSEL_1TI): Likewise.
1432         (VSEL_1TI_UNS): Likewise.
1433         (ST_INTERNAL_1ti): Likewise.
1434         (LD_INTERNAL_1ti): Likewise.
1435         (XXSEL_1TI): Likewise.
1436         (XXSEL_1TI_UNS): Likewise.
1437         (VPERM_1TI): Likewise.
1438         (VPERM_1TI_UNS): Likewise.
1439         (XXPERMDI_1TI): Likewise.
1440         (SET_1TI): Likewise.
1441         (LXVD2X_V1TI): Likewise.
1442         (STXVD2X_V1TI): Likewise.
1443         (VEC_INIT_V1TI): Likewise.
1444         (VEC_SET_V1TI): Likewise.
1445         (VEC_EXT_V1TI): Likewise.
1446         (EQV_V1TI): Likewise.
1447         (NAND_V1TI): Likewise.
1448         (ORC_V1TI): Likewise.
1449         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1450         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1451         overloaded builtin.
1452         (VADDUQM): Likewise.
1453         (VSUBCUQ): Likewise.
1454         (VADDEUQM): Likewise.
1455         (VADDECUQ): Likewise.
1456         (VSUBEUQM): Likewise.
1457         (VSUBECUQ): Likewise.
1459         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1460         __int128_t and __uint128_t types.
1461         (__uint128_type): Likewise.
1462         (altivec_categorize_keyword): Add support for vector __int128_t,
1463         vector __uint128_t, vector __int128, and vector unsigned __int128
1464         as a container type for TImode operations that need to be done in
1465         VSX/Altivec registers.
1466         (rs6000_macro_to_expand): Likewise.
1467         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1468         to support 128-bit integer instructions vaddcuq, vadduqm,
1469         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1470         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1472         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1473         for V1TImode, and set up preferences to use VSX/Altivec registers.
1474         Setup VSX reload handlers.
1475         (rs6000_debug_reg_global): Likewise.
1476         (rs6000_init_hard_regno_mode_ok): Likewise.
1477         (rs6000_preferred_simd_mode): Likewise.
1478         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1479         (easy_altivec_constant): Likewise.
1480         (output_vec_const_move): Likewise.
1481         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1482         simple move.
1483         (rs6000_expand_vector_extract): Likewise.
1484         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1485         addressing.
1486         (rs6000_const_vec): Add support for V1TImode.
1487         (rs6000_emit_le_vsx_load): Swap double words when loading or
1488         storing TImode/V1TImode.
1489         (rs6000_emit_le_vsx_store): Likewise.
1490         (rs6000_emit_le_vsx_move): Likewise.
1491         (rs6000_emit_move): Add support for V1TImode.
1492         (altivec_expand_ld_builtin): Likewise.
1493         (altivec_expand_st_builtin): Likewise.
1494         (altivec_expand_vec_init_builtin): Likewise.
1495         (altivec_expand_builtin): Likewise.
1496         (rs6000_init_builtins): Add support for V1TImode type.  Add
1497         support for ISA 2.07 128-bit integer builtins.  Define type names
1498         for the VSX/Altivec vector types.
1499         (altivec_init_builtins): Add support for overloaded vector
1500         functions with V1TImode type.
1501         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1502         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1503         external function.
1504         (rs6000_split_128bit_ok_p): Likewise.
1505         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1506         __int128_t and vector __uint128_t.
1508         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1509         and mode attributes.
1510         (VSX_M): Likewise.
1511         (VSX_M2): Likewise.
1512         (VSm): Likewise.
1513         (VSs): Likewise.
1514         (VSr): Likewise.
1515         (VSv): Likewise.
1516         (VS_scalar): Likewise.
1517         (VS_double): Likewise.
1518         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1520         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1521         we support the ISA 2.07 128-bit integer arithmetic instructions.
1522         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1523         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1524         and TImode types for use with the builtin functions.
1525         (V1TI_type_node): Likewise.
1526         (unsigned_V1TI_type_node): Likewise.
1527         (intTI_type_internal_node): Likewise.
1528         (uintTI_type_internal_node): Likewise.
1530         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1531         128-bit builtin functions.
1532         (UNSPEC_VADDEUQM): Likewise.
1533         (UNSPEC_VADDECUQ): Likewise.
1534         (UNSPEC_VSUBCUQ): Likewise.
1535         (UNSPEC_VSUBEUQM): Likewise.
1536         (UNSPEC_VSUBECUQ): Likewise.
1537         (VM): Add V1TImode to vector mode iterators.
1538         (VM2): Likewise.
1539         (VI_unit): Likewise.
1540         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1541         (altivec_vaddcuq): Likewise.
1542         (altivec_vsubuqm): Likewise.
1543         (altivec_vsubcuq): Likewise.
1544         (altivec_vaddeuqm): Likewise.
1545         (altivec_vaddecuq): Likewise.
1546         (altivec_vsubeuqm): Likewise.
1547         (altivec_vsubecuq): Likewise.
1549         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1550         mode iterators.
1551         (BOOL_128): Likewise.
1552         (BOOL_REGS_OUTPUT): Likewise.
1553         (BOOL_REGS_OP1): Likewise.
1554         (BOOL_REGS_OP2): Likewise.
1555         (BOOL_REGS_UNARY): Likewise.
1556         (BOOL_REGS_AND_CR0): Likewise.
1558         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1559         128-bit integer builtin support.
1560         (vec_vadduqm): Likewise.
1561         (vec_vaddecuq): Likewise.
1562         (vec_vaddeuqm): Likewise.
1563         (vec_vsubecuq): Likewise.
1564         (vec_vsubeuqm): Likewise.
1565         (vec_vsubcuq): Likewise.
1566         (vec_vsubuqm): Likewise.
1568         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1569         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1570         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1571         128-bit integer add/subtract to ISA 2.07.
1573 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1575         * config/arc/arc.c (arc_predicate_delay_insns):
1576         Fix third argument passed to conditionalize_nonjump.
1578 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1580         * config/aarch64/aarch64-builtins.c
1581         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1582         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1583         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1584         instead of __builtin_lfloor.
1585         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1587 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1589         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1590         (tree_ssa_ifcombine_bb_1): New function.
1591         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1592         is an empty forwarder block to then_bb or vice versa and then_bb
1593         and else_bb are effectively swapped.
1595 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1597         PR target/60264
1598         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1599         REG_CFA_DEF_CFA note.
1600         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1601         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1603 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1605         PR tree-optimization/60454
1606         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1608 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1610         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1611         Do not define target_cpu_default2 to generic.
1612         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1613         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1614         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1616 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1617             Marc Glisse  <marc.glisse@inria.fr>
1619         PR tree-optimization/60502
1620         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1621         instead of build_low_bits_mask.
1623 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1625         PR middle-end/60482
1626         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1627         if there are multiple uses, but op doesn't live on E edge.
1628         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1629         clobber stmts before __builtin_unreachable.
1631 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1633         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1634         hard_frame_pointer_rtx.
1635         * cse.c (cse_insn): Remove volatile check.
1636         * cselib.c (cselib_process_insn): Likewise.
1637         * dse.c (scan_insn): Likewise.
1639 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1641         * config/arc/arc.c (conditionalize_nonjump): New function,
1642         broken out of ...
1643         (arc_ifcvt): ... this.
1644         (arc_predicate_delay_insns): Use it.
1646 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1648         * config/arc/predicates.md (extend_operand): During/after reload,
1649         allow const_int_operand.
1650         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1651         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1652         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1653         to "i".
1654         (umulsi3_highpart_i): Likewise.
1656 2014-03-11  Richard Biener  <rguenther@suse.de>
1658         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1659         Add asserts to guard possible wrong-code bugs.
1661 2014-03-11  Richard Biener  <rguenther@suse.de>
1663         PR tree-optimization/60429
1664         PR tree-optimization/60485
1665         * tree-ssa-structalias.c (set_union_with_increment): Properly
1666         take into account all fields that overlap the shifted vars.
1667         (do_sd_constraint): Likewise.
1668         (do_ds_constraint): Likewise.
1669         (get_constraint_for_ptr_offset): Likewise.
1671 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1673         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1674         (nios2_compute_frame_layout):
1675         Add calculation of cfun->machine->fp_save_offset.
1676         (nios2_expand_prologue): Correct setting of frame pointer register
1677         in prologue.
1678         (nios2_expand_epilogue): Update recovery of stack pointer from
1679         frame pointer accordingly.
1680         (nios2_initial_elimination_offset): Update calculation of offset
1681         for eliminating to HARD_FRAME_POINTER_REGNUM.
1683 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1685         PR ipa/60457
1686         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1687         cgraph_get_create_node on VAR_DECLs.
1689 2014-03-10  Richard Biener  <rguenther@suse.de>
1691         PR middle-end/60474
1692         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1694 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1696         * config/vms/vms.opt (vms_float_format): New variable.
1698 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1700         * doc/invoke.texi (-fcilkplus): Update implementation status.
1702 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1703             Richard Biener  <rguenther@suse.de>
1705         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1706         consistently accross all TUs.
1707         (run_gcc): Enable -fshort-double automatically at link at link-time
1708         and disallow override.
1710 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1712         PR target/58271
1713         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1714         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1715         if they can't be used.
1717 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1719         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1720         for Solaris 11/x86 ld.
1721         * configure: Regenerate.
1723 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1725         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1726         (LIB_TLS_SPEC): Save as ld_tls_libs.
1727         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1728         (HAVE_AS_IX86_TLSLDM): New test.
1729         * configure, config.in: Regenerate.
1730         * config/i386/i386.c (legitimize_tls_address): Fall back to
1731         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1732         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1733         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1734         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1736 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1738         * common.opt (fira-loop-pressure): Mark as optimization.
1740 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1742         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1743         an OpenMP mappable type.
1745 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1747         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1748         MULTILIB_OSDIRNAMES is not defined.
1750 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1751             Meador Inge  <meadori@codesourcery.com>
1753         PR target/58595
1754         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1755         (arm_legitimize_address): Call legitimize_tls_address for any
1756         arm_tls_referenced_p expression, handle constant addend.  Call it
1757         before testing for !TARGET_ARM.
1758         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1760 2014-03-06  Richard Biener  <rguenther@suse.de>
1762         PR middle-end/60445
1763         PR lto/60424
1764         PR lto/60427
1765         Revert
1766         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1768         * tree-streamer.c (record_common_node): Assert we don't record
1769         nodes with type double.
1770         (preload_common_node): Skip type double, complex double and double
1771         pointer since it is now frontend dependent due to fshort-double option.
1773 2014-03-06  Richard Biener  <rguenther@suse.de>
1775         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1776         or -fno-lto is specified and the linker has full plugin support.
1777         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1778         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1779         * lto-wrapper.c (merge_and_complain): Merge compile-time
1780         optimization levels.
1781         (run_gcc): And pass it through to the link options.
1783 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1785         PR debug/60381
1786         Revert:
1787         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1788         PR debug/59992
1789         * cselib.c (remove_useless_values): Skip to avoid quadratic
1790         behavior if the condition moved from...
1791         (cselib_process_insn): ... here holds.
1793 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1795         PR plugins/59335
1796         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1797         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1799         PR plugins/59335
1800         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1801         (TM_H): Add x86-tune.def.
1803 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1805         * config/aarch64/aarch64.c (generic_tunings):
1806         Use cortexa57_extra_costs.
1808 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1810         PR lto/60404
1811         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1812         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1813         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1814         cost for in_lto_p.
1816 2014-03-04  Heiher  <r@hev.cc>
1818         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1819         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1821 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1823         * config/i386/predicates.md (const2356_operand): Change to ...
1824         (const2367_operand): ... this.
1825         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1826         const2367_operand.
1827         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1828         (*avx512pf_scatterpf<mode>sf): Ditto.
1829         (avx512pf_scatterpf<mode>df): Ditto.
1830         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1831         (*avx512pf_scatterpf<mode>df): Ditto.
1832         * config/i386/i386.c (ix86_expand_builtin): Update
1833         incorrect hint operand error message.
1835 2014-03-04  Richard Biener  <rguenther@suse.de>
1837         * lto-section-in.c (lto_get_section_data): Fix const cast.
1839 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1841         * tree-streamer.c (record_common_node): Assert we don't record
1842         nodes with type double.
1843         (preload_common_node): Skip type double, complex double and double
1844         pointer since it is now frontend dependent due to fshort-double option.
1846 2014-03-04  Richard Biener  <rguenther@suse.de>
1848         PR lto/60405
1849         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1850         (lto_input_toplevel_asms): Likewise.
1851         * lto-section-in.c (lto_get_section_data): Instead do it here
1852         for every section.
1854 2014-03-04  Richard Biener  <rguenther@suse.de>
1856         PR tree-optimization/60382
1857         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1858         dead PHIs a reduction.
1860 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1862         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1863         hint value.
1864         (_mm_prefetch): Move out of GCC target("sse") pragma.
1865         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1866         GCC target("prfchw") pragma.
1867         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1868         for locality <= 2.
1869         * config/i386/i386.c (ix86_option_override_internal): Enable
1870         -mprfchw with -mprefetchwt1.
1872 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1874         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1875         Mark as varying.
1877 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1879         * opts.h (CL_PCH_IGNORE): Define.
1880         * targhooks.c (option_affects_pch_p):
1881         Return false for options that have CL_PCH_IGNORE set.
1882         * opt-functions.awk: Process PchIgnore.
1883         * doc/options.texi: Document PchIgnore.
1885         * config/arc/arc.opt (misize): Add PchIgnore property.
1887 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1889         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1890         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1891         constraint on constants to permit them being loaded into
1892         GENERAL_REGS or BASE_REGS.
1894 2014-03-03  Nick Clifton  <nickc@redhat.com>
1896         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1897         anti-cacnonical alternatives.
1898         (negandhi3_real): New pattern.
1899         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1901 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1903         * config/avr/avr-mcus.def: Remove atxmega16x1.
1904         * config/avr/avr-tables.opt: Regenerate.
1905         * config/avr/t-multilib: Regenerate.
1906         * doc/avr-mmcu.texi: Regenerate.
1908 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1909             Mircea Namolaru  <mircea.namolaru@inria.fr>
1911         PR tree-optimization/58028
1912         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1913         scalar dimensions.
1915 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1917         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1918         not handled by recognizers.
1920 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1922         PR middle-end/60175
1923         * function.c (expand_function_end): Don't emit
1924         clobber_return_register sequence if clobber_after is a BARRIER.
1925         * cfgexpand.c (construct_exit_block): Append instructions before
1926         return_label to prev_bb.
1928 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1930         * config/rs6000/constraints.md: Document reserved use of "wc".
1932 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1934         PR ipa/60150
1935         * ipa.c (function_and_variable_visibility): When dissolving comdat
1936         group, also set all symbols to local.
1938 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1940         PR ipa/60306
1942         Revert:
1943         2013-12-14   Jan Hubicka  <jh@suse.cz>
1944         PR middle-end/58477
1945         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1947 2014-03-02  Jon Beniston  <jon@beniston.com>
1949         PR bootstrap/48230
1950         PR bootstrap/50927
1951         PR bootstrap/52466
1952         PR target/46898
1953         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1954           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1955         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1956         (simple_return, *simple_return): New patterns
1957         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1958         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1960 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1962         * dwarf2out.c (gen_subprogram_die): Tidy.
1964 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1966         PR target/60071
1967         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1968         (*mov_t_msb_neg_negc): ... this new insn.
1970 2014-02-28  Jason Merrill  <jason@redhat.com>
1972         PR c++/58678
1973         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1974         function.
1976 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1978         PR c++/60314
1979         * dwarf2out.c (decltype_auto_die): New static.
1980         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1981         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1982         (is_cxx_auto): Likewise.
1984 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1986         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1987         we are not using general regs only.
1989 2014-02-28  Richard Biener  <rguenther@suse.de>
1991         PR target/60280
1992         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1993         previous fix and only allow to remove trivial pre-headers
1994         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1995         (remove_forwarder_block): Properly update the latch of a loop.
1997 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1999         PR debug/59992
2000         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
2001         (cselib_preserved_hash_table): New.
2002         (preserve_constants_and_equivs): Move preserved vals to it.
2003         (cselib_find_slot): Look it up first.
2004         (cselib_init): Initialize it.
2005         (cselib_finish): Release it.
2006         (dump_cselib_table): Dump it.
2008 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2010         PR debug/59992
2011         * cselib.c (remove_useless_values): Skip to avoid quadratic
2012         behavior if the condition moved from...
2013         (cselib_process_insn): ... here holds.
2015 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2017         PR debug/57232
2018         * var-tracking.c (vt_initialize): Apply the same condition to
2019         preserve the CFA base value.
2021 2014-02-28  Joey Ye  <joey.ye@arm.com>
2023         PR target/PR60169
2024         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
2025         if reload in progress or completed.
2027 2014-02-28  Tobias Burnus  <burnus@net-b.de>
2029         PR middle-end/60147
2030         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
2031         NAMELIST_DECL.
2033 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
2035         * doc/tm.texi.in (Condition Code Status): Update documention for
2036         relative locations of cc0-setter and cc0-user.
2038 2014-02-27  Jeff Law  <law@redhat.com>
2040         PR rtl-optimization/52714
2041         * combine.c (try_combine): When splitting an unrecognized PARALLEL
2042         into two independent simple sets, if I3 is a jump, ensure the
2043         pattern we place into I3 is a (set (pc) ...).
2045 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
2046             Jeff Law  <law@redhat.com>
2048         PR rtl-optimization/49847
2049         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
2050         are in different blocks.
2051         * doc/tm.texi (Condition Code Status): Update documention for
2052         relative locations of cc0-setter and cc0-user.
2054 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
2056         PR target/59222
2057         * lra.c (lra_emit_add): Check SUBREG too.
2059 2014-02-27  Andreas Schwab  <schwab@suse.de>
2061         * config/m68k/m68k.c (m68k_option_override): Disable
2062         -flive-range-shrinkage for classic m68k.
2063         (m68k_override_options_after_change): Likewise.
2065 2014-02-27  Marek Polacek  <polacek@redhat.com>
2067         PR middle-end/59223
2068         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
2069         -Wmaybe-uninitialized.
2071 2014-02-27  Alan Modra  <amodra@gmail.com>
2073         PR target/57936
2074         * reload1.c (emit_input_reload_insns): When reload_override_in,
2075         set old to rl->in_reg when rl->in_reg is a subreg.
2077 2014-02-26  Richard Biener  <rguenther@suse.de>
2079         PR bootstrap/60343
2080         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
2082 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2084         * common/config/i386/predicates.md (const1256_operand): Remove.
2085         (const2356_operand): New.
2086         (const_1_to_2_operand): Remove.
2087         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2088         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2089         (*avx512pf_gatherpf<mode>sf): Ditto.
2090         (avx512pf_gatherpf<mode>df): Ditto.
2091         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2092         (*avx512pf_gatherpf<mode>df): Ditto.
2093         (avx512pf_scatterpf<mode>sf): Ditto.
2094         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2095         (*avx512pf_scatterpf<mode>sf): Ditto.
2096         (avx512pf_scatterpf<mode>df): Ditto.
2097         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2098         (*avx512pf_scatterpf<mode>df): Ditto.
2099         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
2101 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
2103         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
2104         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
2105         (_mm512_mask_testn_epi64_mask): Move to ...
2106         * config/i386/avx512cdintrin.h: Here.
2107         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
2108         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
2109         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
2110         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
2111         TARGET_AVX512F from TARGET_AVX512CD.
2113 2014-02-26  Richard Biener  <rguenther@suse.de>
2115         PR ipa/60327
2116         * ipa.c (walk_polymorphic_call_targets): Properly guard
2117         call to inline_update_overall_summary.
2119 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
2121         PR target/60280
2122         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
2123         and latches only if requested.  Fix latch if it is removed.
2124         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
2125         LOOPS_HAVE_PREHEADERS.
2127 2014-02-25  Andrew Pinski  <apinski@cavium.com>
2129         * builtins.c (expand_builtin_thread_pointer): Create a new target
2130         when the target is NULL.
2132 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
2134         PR rtl-optimization/60317
2135         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2136         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2137         * lra-assigns.c: Include params.h.
2138         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
2139         other reload pseudos considerations.
2141 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2143         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
2144         to use canonical form for nor<mode>3.
2146 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2148         PR target/55426
2149         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2150         conversions.
2152 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2154         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2155         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2156         (ix86_handle_option): Handle OPT_mprefetchwt1.
2157         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2158         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2159         PREFETCHWT1 CPUID.
2160         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2161         OPTION_MASK_ISA_PREFETCHWT1.
2162         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2163         (PTA_PREFETCHWT1): New.
2164         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2165         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2166         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2167         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2168         (*prefetch_avx512pf_<mode>_: Change into ...
2169         (*prefetch_prefetchwt1_<mode>: This.
2170         * config/i386/i386.opt (mprefetchwt1): New.
2171         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2172         (_mm_prefetch): Handle intent to write.
2173         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2175 2014-02-25  Richard Biener  <rguenther@suse.de>
2177         PR middle-end/60291
2178         * emit-rtl.c (mem_attrs_htab): Remove.
2179         (mem_attrs_htab_hash): Likewise.
2180         (mem_attrs_htab_eq): Likewise.
2181         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2182         (init_emit_once): Do not allocate mem_attrs_htab.
2184 2014-02-25  Richard Biener  <rguenther@suse.de>
2186         PR lto/60319
2187         * lto-opts.c (lto_write_options): Output non-explicit conservative
2188         -fwrapv, -fno-trapv and -fno-strict-overflow.
2189         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2190         (run_gcc): And pass them through.
2192 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2194         * sel-sched.c (calculate_new_fences): New parameter ptime.
2195         Calculate it as a maximum over all fence cycles.
2196         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2197         Print the final schedule timing when sched_verbose.
2199 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2201         PR rtl-optimization/60292
2202         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2203         bit fot the fence instruction.
2205 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2207         * calls.h: Fix typo in comment.
2209 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2211         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2212         adjusting offsetable addresses.
2214 2014-02-24  Guozhi Wei  <carrot@google.com>
2216         * sparseset.h (sparseset_pop): Fix the wrong index.
2218 2014-02-24  Walter Lee  <walt@tilera.com>
2220         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2221         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2222         triplet.
2223         * common/config/tilegx/tilegx-common.c
2224         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2225         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2226         (LINK_SPEC): Ditto.
2227         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2228         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2229         (tilegx_gimplify_va_arg_expr): Handle big endian.
2230         (tilegx_expand_unaligned_load): Ditto.
2231         (tilegx_expand_unaligned_store): Ditto.
2232         (TARGET_RETURN_IN_MSB): New.
2233         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2234         (TARGET_ENDIAN_DEFAULT): New.
2235         (TARGET_BIG_ENDIAN): Handle big endian.
2236         (BYTES_BIG_ENDIAN): Ditto.
2237         (WORDS_BIG_ENDIAN): Ditto.
2238         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2239         (ENDIAN_SPEC): New.
2240         (EXTRA_SPECS): New.
2241         * config/tilegx/tilegx.md (extv): Handle big endian.
2242         (extzv): Ditto.
2243         (insn_st<n>): Ditto.
2244         (insn_st<n>_add<bitsuffix>): Ditto.
2245         (insn_stnt<n>): Ditto.
2246         (insn_stnt<n>_add<bitsuffix>):Ditto.
2247         (vec_interleave_highv8qi): Handle big endian.
2248         (vec_interleave_highv8qi_be): New.
2249         (vec_interleave_highv8qi_le): New.
2250         (insn_v1int_h): Handle big endian.
2251         (vec_interleave_lowv8qi): Handle big endian.
2252         (vec_interleave_lowv8qi_be): New.
2253         (vec_interleave_lowv8qi_le): New.
2254         (insn_v1int_l): Handle big endian.
2255         (vec_interleave_highv4hi): Handle big endian.
2256         (vec_interleave_highv4hi_be): New.
2257         (vec_interleave_highv4hi_le): New.
2258         (insn_v2int_h): Handle big endian.
2259         (vec_interleave_lowv4hi): Handle big endian.
2260         (vec_interleave_lowv4hi_be): New.
2261         (vec_interleave_lowv4hi_le): New.
2262         (insn_v2int_l): Handle big endian.
2263         (vec_interleave_highv2si): Handle big endian.
2264         (vec_interleave_highv2si_be): New.
2265         (vec_interleave_highv2si_le): New.
2266         (insn_v4int_h): Handle big endian.
2267         (vec_interleave_lowv2si): Handle big endian.
2268         (vec_interleave_lowv2si_be): New.
2269         (vec_interleave_lowv2si_le): New.
2270         (insn_v4int_l): Handle big endian.
2271         * config/tilegx/tilegx.opt (mbig-endian): New option.
2272         (mlittle-endian): New option.
2273         * doc/install.texi: Document tilegxbe-linux.
2274         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2276 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2278         PR ipa/60266
2279         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2280         there are no parameter descriptors.
2282 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2284         PR rtl-optimization/60268
2285         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2286         initialization to ...
2287         (sched_rgn_init): ... here.
2288         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2290 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2292         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2293         names.
2295 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2297         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2298         definition.
2300 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2302         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2303         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2305 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2307         * config/microblaze/predicates.md: Add cmp_op predicate.
2308         * config/microblaze/microblaze.md: Add branch_compare instruction
2309         which uses cmp_op predicate and emits cmp insn before branch.
2310         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2311         to microblaze_expand_conditional_branch and consolidate logic.
2312         (microblaze_expand_conditional_branch): emit branch_compare
2313         insn instead of handling cmp op separate from branch insn.
2315 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2317         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2318         to permit subregs.
2320 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2322         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2323         define_insn with define_expand and new define_insn
2324         *altivec_lve<VI_char>x_internal.
2325         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2326         and new define_insn *altivec_stve<VI_char>x_internal.
2327         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2328         prototype.
2329         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2330         lve*x built-ins.
2331         (altivec_expand_stvex_be): New function.
2333 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2335         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2336         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2337         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2338         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2340 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2342         PR target/60298
2343         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2344         instead of emit_move_insn.
2346 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2348         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2349         vspltw with vsldoi.
2350         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2351         gen_altivec_vsumsws.
2353 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2355         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2356         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2357         (altivec_lvxl_<mode>): New define_expand incorporating
2358         -maltivec=be semantics where needed.
2359         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2360         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2361         semantics where needed.
2362         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2363         (altivec_stvx_<mode>): New define_expand incorporating
2364         -maltivec=be semantics where needed.
2365         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2366         VM2 iterator instead of V4SI.
2367         (altivec_stvxl_<mode>): New define_expand incorporating
2368         -maltivec=be semantics where needed.
2369         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2370         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2371         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2372         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2373         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2374         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2375         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2376         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2377         ALTIVEC_BUILTIN_STVXL.
2378         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2379         (altivec_expand_stvx_be): Likewise.
2380         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2381         (altivec_expand_lvx_be): Likewise.
2382         (altivec_expand_stvx_be): Likewise.
2383         (altivec_expand_builtin): Add cases for
2384         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2385         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2386         (altivec_init_builtins): Add definitions for
2387         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2388         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2390 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2392         * doc/invoke.texi (mvirt, mno-virt): Document.
2393         * config/mips/mips.opt (mvirt): New option.
2394         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2396 2014-02-21  Richard Biener  <rguenther@suse.de>
2398         PR tree-optimization/60276
2399         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2400         (STMT_VINFO_MIN_NEG_DIST): New macro.
2401         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2402         STMT_VINFO_MIN_NEG_DIST.
2403         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2404         made for negative dependence distances still hold.
2406 2014-02-21  Richard Biener  <rguenther@suse.de>
2408         PR middle-end/60291
2409         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2410         DECL_INITIAL for globals not in the current function context.
2412 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2414         PR tree-optimization/56490
2415         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2416         * tree-ssa-uninit.c: Include params.h.
2417         (compute_control_dep_chain): Add num_calls argument, return false
2418         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2419         num_calls to recursive call.
2420         (find_predicates): Change dep_chain into normal array,
2421         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2422         variable and adjust compute_control_dep_chain caller.
2423         (find_def_preds): Likewise.
2425 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2427         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2428         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2430 2014-02-21  Nick Clifton  <nickc@redhat.com>
2432         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2433         (pushhi1): Likewise.
2434         (popqi1): Add mode to pre_dec.
2435         (pophi1): Likewise.
2437 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2439         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2440         mode for mask of V8SFmode permutation.
2442 2014-02-20  Richard Henderson  <rth@redhat.com>
2444         PR c++/60272
2445         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2446         a new pseudo for OLDVAL.
2448 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2450         PR target/57896
2451         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2452         gen_reg_rtx if d->testing_p.
2453         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2454         if d->testing_p and we will certainly return true.
2455         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2456         if d->testing_p.
2458 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2460         * emit-rtl.c (gen_reg_rtx): Assert that
2461         crtl->emit.regno_pointer_align_length is non-zero.
2463 2014-02-20  Richard Henderson  <rth@redhat.com>
2465         PR c++/60272
2466         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2467         on failure the store back into EXPECT.
2469 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2470             Sandra Loosemore  <sandra@codesourcery.com>
2472         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2473         * config/nios2/nios2.c (nios2_function_profiler): Add
2474         -fPIC (flag_pic == 2) support.
2475         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2476         (nios2_large_offset_p): New function.
2477         (nios2_unspec_reloc_p): Move up position, update to use
2478         nios2_large_offset_p.
2479         (nios2_unspec_address): Remove function.
2480         (nios2_unspec_offset): New function.
2481         (nios2_large_got_address): New function.
2482         (nios2_got_address): Add large offset support.
2483         (nios2_legitimize_tls_address): Update usage of removed and new
2484         functions.
2485         (nios2_symbol_binds_local_p): New function.
2486         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2487         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2488         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2489         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2490         processing for (const (unspec ...)).
2491         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2493 2014-02-20  Richard Biener  <rguenther@suse.de>
2495         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2496         doing the substitution.
2497         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2499 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2501         PR ipa/55260
2502         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2503         info when checking whether lattices are bottom.
2505 2014-02-20  Richard Biener  <rguenther@suse.de>
2507         PR middle-end/60221
2508         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2509         regions at -O0.
2511 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2513         PR ipa/58555
2514         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2515         parameter specifying the scaling.
2516         (inline_call): Update.
2517         (want_inline_recursively): Guard division by zero.
2518         (recursive_inlining): Update.
2519         * ipa-inline.h (clone_inlined_nodes): Update.
2521 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2523         PR target/60204
2524         * config/i386/i386.c (classify_argument): Pass structures of size
2525         64 bytes or less in register.
2527 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2528             Kirill Yukhin  <kirill.yukhin@intel.com>
2530         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2531         (_mm_rcp28_round_ss): Ditto.
2532         (_mm_rsqrt28_round_sd): Ditto.
2533         (_mm_rsqrt28_round_ss): Ditto.
2534         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2535         (_mm_rcp14_round_ss): Ditto.
2536         (_mm_rsqrt14_round_sd): Ditto.
2537         (_mm_rsqrt14_round_ss): Ditto.
2538         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2539         the first input operand, get rid of match_dup.
2540         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2541         attribute to sse.
2542         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2543         Ditto.
2544         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2545         operand as the first input operand, set type attribute.
2546         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2547         Set type attribute.
2548         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2549         operand as the first input operand, set type attribute.
2551 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2553         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2554         bit of zero.
2556 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2558         PR target/60207
2559         * config/i386/i386.c (construct_container): Remove TFmode check
2560         for X86_64_INTEGER_CLASS.
2562 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2564         PR target/59794
2565         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2566         only when -Wpsabi is enabled.
2568 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2570          PR target/59799
2571         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2572         passing arrays in registers are the same as for structs, so remove the
2573         special case for them.
2575 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2577         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2578         destination type, extract only the valid bits if the source type is not
2579         integral and has a different mode.
2581 2014-02-19  Richard Biener  <rguenther@suse.de>
2583         PR ipa/60243
2584         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2585         for all calls.
2587 2014-02-19  Richard Biener  <rguenther@suse.de>
2589         PR ipa/60243
2590         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2591         (ipa_modify_call_arguments): Emit an argument load explicitely and
2592         preserve virtual SSA form there and for the replacement call.
2593         Do not update SSA form nor free dominance info.
2595 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2597         * ipa.c (function_and_variable_visibility): Also clear WEAK
2598         flag when disolving COMDAT_GROUP.
2600 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2602         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2603         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2604         not devirtualizing.
2605         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2606         do more sanity checks.
2607         (detect_type_change): Return true when giving up early.
2608         (compute_complex_assign_jump_func): Fix type parameter of
2609         ipa_set_ancestor_jf.
2610         (compute_complex_ancestor_jump_func): Likewise.
2611         (update_jump_functions_after_inlining): Fix updating of
2612         ancestor function.
2613         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2615 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2617         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2618         inline clones when edge disappears.
2620 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2622         PR target/60203
2623         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2624         Split 64-bit moves into 2 patterns.  Do not allow the use of
2625         direct move for TDmode in little endian, since the decimal value
2626         has little endian bytes within a word, but the 64-bit pieces are
2627         ordered in a big endian fashion, and normal subreg's of TDmode are
2628         not allowed.
2629         (mov<mode>_64bit_dm): Likewise.
2630         (movtd_64bit_nodm): Likewise.
2632 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2634         PR tree-optimization/60174
2635         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2636         statement of an SSA_NAME that occurs in an abnormal PHI node.
2638 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2640         PR sanitizer/60142
2641         * final.c (SEEN_BB): Remove.
2642         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2643         (final_scan_insn): Don't force_source_line on second
2644         NOTE_INSN_BASIC_BLOCK.
2646 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2648         PR target/60205
2649         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2650         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2651         (type_natural_mode): Warn ABI change when %zmm register is not
2652         available for AVX512F vector value passing.
2654 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2656         PR target/60193
2657         * config/i386/i386.c (ix86_expand_prologue): Use value in
2658         rax register as displacement when restoring %r10 or %rax.
2659         Fix wrong offset when restoring both registers.
2661 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2663         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2664         assertion with conditional return.
2666 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2667             Uros Bizjak  <ubizjak@gmail.com>
2669         PR driver/60233
2670         * config/i386/driver-i386.c (host_detect_local_cpu): If
2671         YMM state is not saved by the OS, also clear has_f16c.  Move
2672         CPUID 0x80000001 handling before YMM state saving checking.
2674 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2676         PR rtl-optimization/58960
2677         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2678         factored out from ...
2679         (sched_init): ... here.
2680         (free_global_sched_pressure_data): New, factored out from ...
2681         (sched_finish): ... here.
2682         * sched-int.h (free_global_sched_pressure_data): Declare.
2683         * sched-rgn.c (nr_regions_initial): New static global.
2684         (haifa_find_rgns): Initialize it.
2685         (schedule_region): Disable sched-pressure for the newly
2686         generated regions.
2688 2014-02-17  Richard Biener  <rguenther@suse.de>
2690         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2691         release SSA defs of pattern stmts.
2693 2014-02-17  Richard Biener  <rguenther@suse.de>
2695         * tree-inline.c (expand_call_inline): Release the virtual
2696         operand defined by the call we are about to inline.
2698 2014-02-17  Richard Biener  <rguenther@suse.de>
2700         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2702 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2703             Ilya Tocar  <ilya.tocar@intel.com>
2705         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2706         arguments order in builtin.
2707         (_mm512_permutexvar_epi64): Ditto.
2708         (_mm512_mask_permutexvar_epi64): Ditto
2709         (_mm512_maskz_permutexvar_epi32): Ditto
2710         (_mm512_permutexvar_epi32): Ditto
2711         (_mm512_mask_permutexvar_epi32): Ditto
2713 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2715         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2716         (p8_vmrgow): Likewise.
2718 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2720         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2721         endian targets.
2723 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2725         PR target/60203
2726         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2727         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2728         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2729         using direct move instructions on ISA 2.07.  Also adjust
2730         instruction length for 64-bit.
2731         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2732         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2734 2014-02-15  Alan Modra  <amodra@gmail.com>
2736         PR target/58675
2737         PR target/57935
2738         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2739         find_replacement on parts of insn rtl that might be reloaded.
2741 2014-02-15  Richard Biener  <rguenther@suse.de>
2743         PR tree-optimization/60183
2744         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2745         (tree_ssa_phiprop): Calculate and free post-dominators.
2747 2014-02-14  Jeff Law  <law@redhat.com>
2749         PR rtl-optimization/60131
2750         * ree.c (get_extended_src_reg): New function.
2751         (combine_reaching_defs): Use it rather than assuming location of REG.
2752         (find_and_remove_re): Verify first operand of extension is
2753         a REG before adding the insns to the copy list.
2755 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2757         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2758         * configure: Regenerated.
2759         * config.in: Regenerated.
2760         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2761         instead of ASM_SHORT.
2763 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2764             Richard Earnshaw  <rearnsha@arm.com>
2766         PR rtl-optimization/59535
2767         * lra-constraints.c (process_alt_operands): Encourage alternative
2768         when unassigned pseudo class is superset of the alternative class.
2769         (inherit_reload_reg): Don't inherit when optimizing for code size.
2770         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2771         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2772         modes not less than 4 for Thumb1.
2774 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2776         PR pch/60010
2777         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2779 2014-02-14  Richard Biener  <rguenther@suse.de>
2781         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2782         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2783         Do not strip INDIRECT_REFs.
2785 2014-02-14  Richard Biener  <rguenther@suse.de>
2787         PR lto/60179
2788         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2789         DECL_FUNCTION_SPECIFIC_TARGET.
2790         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2791         * tree-streamer-out.c (pack_ts_target_option): Remove.
2792         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2793         (write_ts_function_decl_tree_pointers): Do not stream
2794         DECL_FUNCTION_SPECIFIC_TARGET.
2795         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2796         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2797         (lto_input_ts_function_decl_tree_pointers): Do not stream
2798         DECL_FUNCTION_SPECIFIC_TARGET.
2800 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2802         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2803         (get_initial_def_for_induction, vectorizable_induction): Ignore
2804         debug stmts when looking for exit_phi.
2805         (vectorizable_live_operation): Fix up condition.
2807 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2809         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2810         nreverse() because it changes the content of original tree list.
2812 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2814         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2815         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2817 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2819         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2820         GNU coding standards.
2822 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2824         PR debug/60152
2825         * dwarf2out.c (gen_subprogram_die): Don't call
2826         add_calling_convention_attribute if subr_die is old_die.
2828 2014-02-13  Sharad Singhai  <singhai@google.com>
2830         * doc/optinfo.texi: Fix order of nodes.
2832 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2834         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2835         operands[2], not operands[3].
2837 2014-02-13  Richard Biener  <rguenther@suse.de>
2839         PR bootstrap/59878
2840         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2841         mention the possibility of an in-tree build.
2842         (CLooG): Update recommended version to 0.18.1, mention the
2843         possibility of an in-tree build and clarify that the ISL
2844         bundled with CLooG does not work.
2846 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2848         PR target/43546
2849         * expr.c (compress_float_constant): If x is a hard register,
2850         extend into a pseudo and then move to x.
2852 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2854         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2855         caused by bad second argument to warning_at() with -mhotpatch and
2856         nested functions (e.g. with gfortran).
2858 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2860         * opts.c (option_name): Remove "enabled by default" rider.
2862 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2864         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2866 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2867             Uros Bizjak  <ubizjak@gmail.com>
2869         PR target/60151
2870         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2871         * configure: Regenerated.
2873 2014-02-12  Richard Biener  <rguenther@suse.de>
2875         * vec.c (vec_prefix::calculate_allocation): Move as
2876         inline variant to vec.h.
2877         (vec_prefix::calculate_allocation_1): New out-of-line version.
2878         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2879         (vec_prefix::m_has_auto_buf): Rename to ...
2880         (vec_prefix::m_using_auto_storage): ... this.
2881         (vec_prefix::calculate_allocation): Inline the easy cases
2882         and dispatch to calculate_allocation_1 which doesn't need the
2883         prefix address.
2884         (va_heap::reserve): Use gcc_checking_assert.
2885         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2886         m_using_auto_storage.
2887         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2888         member and adjust.
2889         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2890         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2891         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2893 2014-02-12  Richard Biener  <rguenther@suse.de>
2895         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2896         when we found a dependence.
2898 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2900         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2901         common code...
2902         (maybe_fold_stmt): ... into this new function.
2903         * omp-low.c (lower_omp): Update comment.
2905         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2906         last use.
2908         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2909         dereference.
2911 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2913         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2914         identifiers in comments.
2915         (cortexa53_extra_costs): Likewise.
2916         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2917         (cortexa7_extra_costs): Likewise.
2918         (cortexa12_extra_costs): Likewise.
2919         (cortexa15_extra_costs): Likewise.
2920         (v7m_extra_costs): Likewise.
2922 2014-02-12  Richard Biener  <rguenther@suse.de>
2924         PR middle-end/60092
2925         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2926         of posix_memalign being successful.
2927         (lower_stmt): Restrict lowering of posix_memalign to when
2928         -ftree-bit-ccp is enabled.
2930 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2932         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2933         arg_loc.
2934         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2936 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2938         PR rtl-optimization/60116
2939         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2940         other_insn once the combination has been validated.
2942 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2944         PR lto/59468
2945         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2946         and wrapper.
2947         * ipa-devirt.c: Include demangle.h
2948         (odr_violation_reported): New static variable.
2949         (add_type_duplicate): Update odr_violations.
2950         (maybe_record_node): Add completep parameter; update it.
2951         (record_target_from_binfo): Add COMPLETEP parameter;
2952         update it as needed.
2953         (possible_polymorphic_call_targets_1): Likewise.
2954         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2955         rename FINAL to COMPLETE.
2956         (record_targets_from_bases): Sanity check we found the binfo;
2957         fix COMPLETEP updating.
2958         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2959         parameter, fix computing of COMPLETEP.
2960         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2961         at LTO time do demangling.
2962         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2963         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2964         parameter.
2965         (gimple_get_virt_method_for_binfo): Likewise.
2966         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2967         gimple_get_virt_method_for_vtable): Update prototypes.
2969 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2971         PR target/49008
2972         * genautomata.c (add_presence_absence): Fix typo with
2973         {final_}presence_list.
2975 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2977         PR target/60137
2978         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2979         for VSX/Altivec vectors that land in GPR registers.
2981 2014-02-11  Richard Henderson  <rth@redhat.com>
2982             Jakub Jelinek  <jakub@redhat.com>
2984         PR debug/59776
2985         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2986         around drhs if type conversion to lacc->type is not useless.
2988 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2990         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2991         tuning struct.
2992         (cortex-a57.cortex-a53): Likewise.
2993         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2995 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2997         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2998         arm_restrict_it.
3000 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
3002         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
3003         add_options_for_arm_vfp3.
3005 2014-02-11  Jeff Law  <law@redhat.com>
3007         PR middle-end/54041
3008         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
3009         object with an undesirable mode.
3011 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3013         PR libgomp/60107
3014         * config/i386/sol2-9.h: New file.
3015         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
3016         *-*-solaris2.9*): Use it.
3018 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3020         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
3021         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
3023 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3025         * config/microblaze/microblaze.c: Extend mcpu version format
3027 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
3029         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
3031 2014-02-10  Richard Henderson  <rth@redhat.com>
3033         PR target/59927
3034         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
3035         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
3036         ms-abi vs -mno-accumulate-outgoing-args.
3037         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
3038         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
3039         respect to ms-abi.
3041 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3043         PR middle-end/60080
3044         * cfgexpand.c (expand_asm_operands): Attach source location to
3045         ASM_INPUT rtx objects.
3046         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
3048 2014-02-10  Nick Clifton  <nickc@redhat.com>
3050         * config/mn10300/mn10300.c (popcount): New function.
3051         (mn10300_expand_prologue): Include saved registers in stack usage
3052         count.
3054 2014-02-10  Jeff Law  <law@redhat.com>
3056         PR middle-end/52306
3057         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
3058         when changing the SET_DEST of a prior insn to avoid an input reload.
3060 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
3062         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
3063         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
3064         -mcall-openbsd, or -mcall-linux.
3065         (CC1_ENDIAN_BIG_SPEC): Remove.
3066         (CC1_ENDIAN_LITTLE_SPEC): Remove.
3067         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3068         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
3069         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
3070         and %cc1_endian_default.
3071         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3073 2014-02-10  Richard Biener  <rguenther@suse.de>
3075         PR tree-optimization/60115
3076         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3077         MEM_REF handling.  Properly verify that the accesses are not
3078         out of the objects bound.
3080 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3082         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
3083         coretex to cortex.
3085 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
3087         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
3088         proper constants and fix formatting.
3089         (possible_polymorphic_call_targets): Fix formatting.
3091 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
3092             Ilya Tocar  <ilya.tocar@intel.com>
3094         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
3095         (_mm512_loadu_epi32): Renamed into...
3096         (_mm512_loadu_si512): This.
3097         (_mm512_storeu_epi32): Renamed into...
3098         (_mm512_storeu_si512): This.
3099         (_mm512_maskz_ceil_ps): Removed.
3100         (_mm512_maskz_ceil_pd): Ditto.
3101         (_mm512_maskz_floor_ps): Ditto.
3102         (_mm512_maskz_floor_pd): Ditto.
3103         (_mm512_floor_round_ps): Ditto.
3104         (_mm512_floor_round_pd): Ditto.
3105         (_mm512_ceil_round_ps): Ditto.
3106         (_mm512_ceil_round_pd): Ditto.
3107         (_mm512_mask_floor_round_ps): Ditto.
3108         (_mm512_mask_floor_round_pd): Ditto.
3109         (_mm512_mask_ceil_round_ps): Ditto.
3110         (_mm512_mask_ceil_round_pd): Ditto.
3111         (_mm512_maskz_floor_round_ps): Ditto.
3112         (_mm512_maskz_floor_round_pd): Ditto.
3113         (_mm512_maskz_ceil_round_ps): Ditto.
3114         (_mm512_maskz_ceil_round_pd): Ditto.
3115         (_mm512_expand_pd): Ditto.
3116         (_mm512_expand_ps): Ditto.
3117         * config/i386/i386.c (ix86_builtins): Remove
3118         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
3119         (bdesc_args): Ditto.
3120         * config/i386/predicates.md (const1256_operand): New.
3121         (const_1_to_2_operand): Ditto.
3122         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
3123         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3124         (*avx512pf_gatherpf<mode>sf): Ditto.
3125         (avx512pf_gatherpf<mode>df): Ditto.
3126         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3127         (*avx512pf_gatherpf<mode>df): Ditto.
3128         (avx512pf_scatterpf<mode>sf): Ditto.
3129         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3130         (*avx512pf_scatterpf<mode>sf): Ditto.
3131         (avx512pf_scatterpf<mode>df): Ditto.
3132         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3133         (*avx512pf_scatterpf<mode>df): Ditto.
3134         (avx512f_expand<mode>): Removed.
3135         (<shift_insn><mode>3<mask_name>): Change predicate type.
3137 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3139         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
3140         not at the end of datarefs vector use ordered_remove to avoid
3141         reordering datarefs vector.
3143         PR c/59984
3144         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
3145         mark local addressable non-static vars as GOVD_PRIVATE
3146         instead of GOVD_LOCAL.
3147         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3148         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3149         than copying them.
3151         PR middle-end/60092
3152         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3153         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3154         assume_aligned or alloc_align attributes.
3155         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3156         arguments.  Handle also assume_aligned and alloc_align attributes.
3157         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3158         calls to functions with assume_aligned or alloc_align attributes.
3159         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3161 2014-02-08  Terry Guo  <terry.guo@arm.com>
3163         * doc/invoke.texi: Document ARM -march=armv7e-m.
3165 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3167         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3168         flag on __cilkrts_rethrow builtin.
3170         PR ipa/60026
3171         * ipa-cp.c (determine_versionability): Fail at -O0
3172         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3173         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3175         Revert:
3176         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3178         PR ipa/60026
3179         * tree-inline.c (copy_forbidden): Fail for
3180         __attribute__((optimize (0))) functions.
3182 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3184         * varpool.c: Include pointer-set.h.
3185         (varpool_remove_unreferenced_decls): Variables in other partitions
3186         will not be output; be however careful to not lose information
3187         about partitioning.
3189 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3191         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3192         lookup in the vtable constructor.
3194 2014-02-07  Jeff Law  <law@redhat.com>
3196         PR target/40977
3197         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3198         define_insn_and_split.
3200         * ipa-inline.c (inline_small_functions): Fix typos.
3202 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3204         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3205         (s390_can_use_return_insn): Declare.
3206         * config/s390/s390.h (EPILOGUE_USES): Define.
3207         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3208         instructions.
3209         (s390_chunkify_start): Handle return JUMP_LABELs.
3210         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3211         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3212         (s390_can_use_return_insn): New functions.
3213         (s390_fix_long_loop_prediction): Handle conditional returns.
3214         (TARGET_SET_UP_BY_PROLOGUE): Define.
3215         * config/s390/s390.md (ANY_RETURN): New code iterator.
3216         (*creturn, *csimple_return, return, simple_return): New patterns.
3218 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3220         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3221         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3222         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3223         REG_CFA_RESTORE list when deciding not to restore a register.
3225 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3227         * config/s390/s390.c: Include tree-pass.h and context.h.
3228         (s390_early_mach): New function, split out from...
3229         (s390_emit_prologue): ...here.
3230         (pass_data_s390_early_mach): New pass structure.
3231         (pass_s390_early_mach): New class.
3232         (s390_option_override): Create and register early_mach pass.
3233         Move to end of file.
3235 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3237         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3238         to match for the exit block.
3240 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3242         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3243         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3244         Reject misaligned operands.
3246 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3248         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3250 2014-02-07  Richard Biener  <rguenther@suse.de>
3252         PR middle-end/60092
3253         * gimple-low.c (lower_builtin_posix_memalign): New function.
3254         (lower_stmt): Call it to lower posix_memalign in a way
3255         to make alignment info accessible.
3257 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3259         PR c++/60082
3260         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3261         __builtin_setjmp_receiver.
3263 2014-02-07  Richard Biener  <rguenther@suse.de>
3265         PR middle-end/60092
3266         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3267         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3268         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3269         Handle BUILT_IN_POSIX_MEMALIGN.
3270         (find_func_clobbers): Likewise.
3271         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3272         (call_may_clobber_ref_p_1): Likewise.
3274 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3276         PR ipa/59918
3277         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3278         sanity check.
3280 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3282         PR ipa/59469
3283         * lto-cgraph.c (lto_output_node): Use
3284         symtab_get_symbol_partitioning_class.
3285         (lto_output_varpool_node): likewise.
3286         (symtab_get_symbol_partitioning_class): Move here from
3287         lto/lto-partition.c
3288         * cgraph.h (symbol_partitioning_class): Likewise.
3289         (symtab_get_symbol_partitioning_class): Declare.
3291 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3293         * ggc.h (ggc_internal_cleared_alloc): New macro.
3294         * vec.h (vec_safe_copy): Handle memory stats.
3295         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3296         * target-globals.c (save_target_globals): Likewise.
3298 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3300         PR target/60077
3301         * expr.c (emit_move_resolve_push): Export; be bit more selective
3302         on when to clear alias set.
3303         * expr.h (emit_move_resolve_push): Declare.
3304         * function.h (struct function): Add tail_call_marked.
3305         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3306         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3307         * config/i386/i386.md (TImode move expander): De not call
3308         ix86_expand_push.
3309         (FP push expanders): Preserve memory attributes.
3310         * config/i386/sse.md (push<mode>1): Remove.
3311         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3312         (ix86_expand_push): Remove.
3313         * config/i386/mmx.md (push<mode>1): Remove.
3315 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3317         PR rtl-optimization/60030
3318         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3319         lopart with paradoxical subreg before shifting it up by hprec.
3321 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3323         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3324         Remove extra newline at end of file.
3325         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3326         (arm_issue_rate): Handle cortexa57.
3327         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3328         (cortex-a57.cortex-a53): Likewise.
3330 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3332         PR target/59575
3333         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3334         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3335         bitmask.
3336         (arm_expand_prologue): Adjust all callers.
3337         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3338         info, registers also at the lowest numbered registers side.  Use
3339         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3340         XEXP.
3342         PR debug/59992
3343         * var-tracking.c (adjust_mems): Before adding a SET to
3344         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3346 2014-02-06  Alan Modra  <amodra@gmail.com>
3348         PR target/60032
3349         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3350         change SDmode to DDmode when lra_in_progress.
3352 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3354         PR middle-end/59150
3355         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3356         free_data_ref on the dr first, and before goto again also set dr
3357         to the next dr.  For simd_lane_access, free old datarefs[i] before
3358         overwriting it.  For get_vectype_for_scalar_type failure, don't
3359         free_data_ref if simd_lane_access.
3361         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3363         PR target/60062
3364         * tree.h (opts_for_fn): New inline function.
3365         (opt_for_fn): Define.
3366         * config/i386/i386.c (ix86_function_regparm): Use
3367         opt_for_fn (decl, optimize) instead of optimize.
3369 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3371         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3372         for SYMBOL_REF in large memory model.
3374 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3376         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3377         and crypto support.
3378         (cortex-a57): Likewise.
3379         (cortex-a57.cortex-a53): Likewise.
3381 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3382             Kugan Vivekanandarajah  <kuganv@linaro.org>
3384         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3385         unaligned_access.
3386         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3388 2014-02-06  Richard Biener  <rguenther@suse.de>
3390         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3391         set_loop_copy and initialize_original_copy_tables.
3393 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3395         * config/aarch64/aarch64-simd.md
3396         (aarch64_ashr_simddi): Change QI to SI.
3398 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3399             Jakub Jelinek  <jakub@redhat.com>
3401         PR middle-end/60013
3402         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3403         of the dataflow.
3405 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3407         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3408         CODE_FOR_altivec_vpku[hw]um to
3409         CODE_FOR_altivec_vpku[hw]um_direct.
3410         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3411         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3412         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3413         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3415 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3417         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3418         generation for -maltivec=be.
3419         (altivec_vsumsws): Simplify redundant test.
3421 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3423         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3424         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3425         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3426         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3427         gen_altivec_vpkuwum.
3428         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3429         BYTES_BIG_ENDIAN.
3430         (altivec_vpks<VI_char>ss): Likewise.
3431         (altivec_vpks<VI_char>us): Likewise.
3432         (altivec_vpku<VI_char>us): Likewise.
3433         (altivec_vpku<VI_char>um): Likewise.
3434         (altivec_vpku<VI_char>um_direct): New (copy of
3435         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3436         internal use).
3437         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3438         target is little endian and -maltivec=be is not specified.
3439         (*altivec_vupkhs<VU_char>_direct): New (copy of
3440         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3441         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3442         target is little endian and -maltivec=be is not specified.
3443         (*altivec_vupkls<VU_char>_direct): New (copy of
3444         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3445         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3446         little endian and -maltivec=be is not specified.
3447         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3448         little endian and -maltivec=be is not specified.
3450 2014-02-05  Richard Henderson  <rth@redhat.com>
3452         PR debug/52727
3453         * combine-stack-adj.c: Revert r206943.
3454         * sched-int.h (struct deps_desc): Add last_args_size.
3455         * sched-deps.c (init_deps): Initialize it.
3456         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3457         contain REG_ARGS_SIZE notes.
3459 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3461         * lto-cgraph.c (asm_nodes_output): Make global.
3462         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3463         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3464         (driver_handle_option): Handle OPT_fwpa.
3466 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3468         PR ipa/59947
3469         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3470         a comment typo and formatting issue.  If odr_hash hasn't been
3471         created, return vNULL and set *completep to false.
3473         PR middle-end/57499
3474         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3475         bb with no successors.
3477 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3479         PR target/59718
3480         * doc/invoke.texi (-march): Clarify documentation for ARM.
3481         (-mtune): Likewise.
3482         (-mcpu): Likewise.
3484 2014-02-05  Richard Biener  <rguenther@suse.de>
3486         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3487         when not vectorizing because of too many alias checks.
3488         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3489         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3491 2014-02-05  Nick Clifton  <nickc@redhat.com>
3493         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3494         accept extended registers in any mode when compiling for the MN10300.
3496 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3498         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3499         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3500         sanitization attributes.
3501         (can_inline_edge_p): Likewise.
3502         (sanitize_attrs_match_for_inline_p): New function.
3504 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3506         * ipa-prop.c (detect_type_change): Shor circuit testing of
3507         type changes on THIS pointer.
3509 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3511         PR target/59777
3512         * config/pa/pa.c (legitimize_tls_address): Return original address
3513         if not passed a SYMBOL_REF rtx.
3514         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3515         addresses.
3516         (pa_emit_move_sequence): Simplify TLS source operands.
3517         (pa_legitimate_constant_p): Reject all TLS constants.
3518         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3519         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3521 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3523         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3524         groups when we know they are controlled by LTO.
3525         * varasm.c (default_binds_local_p_1): If object is in other partition,
3526         it will be resolved locally.
3528 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3530         * config/host-linux.c (linux_gt_pch_use_address): Don't
3531         use SSIZE_MAX because it is not always defined.
3533 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3535         PR bootstrap/59913
3536         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3537         threshold for pseudo splitting.
3538         (update_ebb_live_info): Process call argument hard registers and
3539         hard registers from insn definition too.
3540         (max_small_class_regs_num): New constant.
3541         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3542         reloads_num only for small register classes.  Don't split for
3543         outputs of jumps.
3545 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3547         PR ipa/60058
3548         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3549         is non-null.
3551 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3553         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3554         visibility is safe.
3556 2014-02-04  Marek Polacek  <polacek@redhat.com>
3558         * gdbinit.in (pel): Define.
3560 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3562         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3563         behavior.
3565 2014-02-04  Richard Biener  <rguenther@suse.de>
3567         PR lto/59723
3568         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3569         in function context local.
3570         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3571         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3572         similar to LTO_imported_decl_ref.
3574 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3576         PR tree-optimization/60002
3577         * cgraphclones.c (build_function_decl_skip_args): Clear
3578         DECL_LANG_SPECIFIC.
3580         PR tree-optimization/60023
3581         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3582         false to gsi_replace.
3583         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3584         has been in some EH region and vec_stmt could throw, add
3585         vec_stmt into the same EH region.
3586         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3587         has no lhs, ignore it.
3588         * internal-fn.c (expand_MASK_LOAD): Likewise.
3590         PR ipa/60026
3591         * tree-inline.c (copy_forbidden): Fail for
3592         __attribute__((optimize (0))) functions.
3594         PR other/58712
3595         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3596         is set, copy one less argument.
3597         (expand_simd_clones): Don't subtract clone_info->inbranch
3598         from simd_clone_struct_alloc argument.
3600         PR rtl-optimization/57915
3601         * recog.c (simplify_while_replacing): If all unary/binary/relational
3602         operation arguments are constant, attempt to simplify those.
3604         PR middle-end/59261
3605         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3606         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3608 2014-02-04  Richard Biener  <rguenther@suse.de>
3610         PR tree-optimization/60012
3611         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3612         TBAA disambiguation to all DDRs.
3614 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3616         PR target/59788
3617         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3618         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3620 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3622         PR ipa/59882
3623         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3625 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3627         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3628         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3630 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3632         PR ipa/59831
3633         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3634         to figure out targets of polymorphic calls with known decl.
3635         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3636         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3637         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3638         (get_polymorphic_call_info): ... here.
3639         (get_polymorphic_call_info_from_invariant): New function.
3641 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3643         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3644         lookup via vtable pointer; check for type consistency
3645         and turn inconsitent facts into UNREACHABLE.
3646         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3647         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3648         type inconsistent querries; return UNREACHABLE instead.
3650 2014-02-03  Richard Henderson  <rth@twiddle.net>
3652         PR tree-opt/59924
3653         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3654         already processed this node.
3655         (normalize_one_pred_1): Pass along mark_set.
3656         (normalize_one_pred): Create and destroy a pointer_set_t.
3657         (normalize_one_pred_chain): Likewise.
3659 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3661         PR gcov-profile/58602
3662         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3664 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3666         PR ipa/59831
3667         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3668         -fno-devirtualize; try to devirtualize by the knowledge of
3669         virtual table pointer given by aggregate propagation.
3670         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3671         (ipa_print_node_jump_functions): Dump also offset that
3672         is relevant for polymorphic calls.
3673         (determine_known_aggregate_parts): Add arg_type parameter; use it
3674         instead of determining the type from pointer type.
3675         (ipa_compute_jump_functions_for_edge): Update call of
3676         determine_known_aggregate_parts.
3677         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3678         (gimple_get_virt_method_for_binfo): ... here; simplify using
3679         vtable_pointer_value_to_vtable.
3680         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3681         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3682         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3683         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3684         POINTER_PLUS_EXPR.
3685         (vtable_pointer_value_to_binfo): ... here.
3686         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3688 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3690         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3691         redef of outer loop index variable.
3693 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3695         PR c++/53017
3696         PR c++/59211
3697         * doc/extend.texi (Function Attributes): Typo.
3699 2014-02-03  Cong Hou  <congh@google.com>
3701         PR tree-optimization/60000
3702         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3703         if the vectorized statement is a store.  A store statement can only
3704         appear at the end of pattern statements.
3706 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3708         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3709         (ix86_option_override_internal): Default long double to 64-bit for
3710         32-bit Bionic and to 128-bit for 64-bit Bionic.
3712         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3713         TARGET_LONG_DOUBLE_128 is true.
3714         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3716         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3717         (mlong-double-64): Negate -mlong-double-128.
3718         (mlong-double-128): New option.
3720         * config/i386/i386-c.c (ix86_target_macros): Define
3721         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3723         * doc/invoke.texi: Document -mlong-double-128.
3725 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3727         PR rtl-optimization/60024
3728         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3730 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3732         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3734 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3736         PR rtl-optimization/57662
3737         * sel-sched.c (code_motion_path_driver): Do not mark already not
3738         existing blocks in the visiting bitmap.
3740 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3742         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3743         on the insn being emitted.
3745 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3746             Will Deacon  <will.deacon@arm.com>
3748         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3750 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3752         * config/arm/arm-tables.opt: Regenerate.
3754 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3756         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3757         for vector types other than V16QImode.
3758         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3759         define_expand, and call altivec_expand_vec_perm_le when producing
3760         code with little endian element order.
3761         (*altivec_vperm_<mode>_internal): New insn having previous
3762         behavior of altivec_vperm_<mode>.
3763         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3764         altivec_expand_vec_perm_le when producing code with little endian
3765         element order.
3766         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3767         behavior of altivec_vperm_<mode>_uns.
3769 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3771         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3772         (altivec_vsumsws): Add handling for -maltivec=be with a little
3773         endian target.
3774         (altivec_vsumsws_direct): New.
3775         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3776         gen_altivec_vsumsws.
3778 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3780         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3781         vtable_pointer_value_to_binfo): New functions.
3782         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3783         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3785 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3787         * config/nios2/nios2.md (load_got_register): Initialize GOT
3788         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3789         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3791 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3793         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3794         preserverd by passthrough, do not propagate the type.
3796 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3798         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3799         (mips_atomic_assign_expand_fenv): New function.
3800         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3802 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3804         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3805         (__builtin_mips_set_fcsr): Likewise.
3806         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3807         MIPS_USI_FTYPE_VOID.
3808         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3809         (mips16_expand_set_fcsr): Likewise.
3810         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3811         (mips16_set_fcsr_stub): Likewise.
3812         (mips16_get_fcsr_one_only_stub): New class.
3813         (mips16_set_fcsr_one_only_stub): Likewise.
3814         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3815         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3816         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3817         (hard_float): New availability predicate.
3818         (mips_builtins): Add get_fcsr and set_fcsr.
3819         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3820         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3821         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3822         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3823         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3824         patterns.
3826 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3828         * config/mips/mips.c (mips_one_only_stub): New class.
3829         (mips_need_mips16_rdhwr_p): Replace with...
3830         (mips16_rdhwr_stub): ...this new variable.
3831         (mips16_stub_call_address): New function.
3832         (mips16_rdhwr_one_only_stub): New class.
3833         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3834         (mips_output_mips16_rdhwr): Delete.
3835         (mips_finish_stub): New function.
3836         (mips_code_end): Use it to handle rdhwr stubs.
3838 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3840         PR target/60017
3841         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3842         when calculating size of integer atomic types.
3844 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3846         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3848 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3850         PR tree-optimization/60003
3851         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3852         * profile.c (branch_prob): Use gimple_call_builtin_p
3853         to check for BUILT_IN_SETJMP_RECEIVER.
3854         * tree-inline.c (copy_bb): Call notice_special_calls.
3856 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3858         PR bootstrap/59985
3859         * lra-constraints.c (process_alt_operands): Update reload_sum only
3860         on the first pass.
3862 2014-01-31  Richard Henderson  <rth@redhat.com>
3864         PR middle-end/60004
3865         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3866         until after else_eh is processed.
3868 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3870         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3871         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3872         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3873         in smmintrin.h, remove them.
3874         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3875         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3876         * config/i386/i386.md (ROUND_SAE): Fix value.
3877         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3878         (const48_operand): New.
3879         * config/i386/subst.md (round), (round_expand): Use
3880         const_4_or_8_to_11_operand.
3881         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3883 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3885         * config/i386/constraints.md (Yk): Swap meaning with k.
3886         * config/i386/i386.md (movhi_internal): Change Yk to k.
3887         (movqi_internal): Ditto.
3888         (*k<logic><mode>): Ditto.
3889         (*andhi_1): Ditto.
3890         (*andqi_1): Ditto.
3891         (kandn<mode>): Ditto.
3892         (*<code>hi_1): Ditto.
3893         (*<code>qi_1): Ditto.
3894         (kxnor<mode>): Ditto.
3895         (kortestzhi): Ditto.
3896         (kortestchi): Ditto.
3897         (kunpckhi): Ditto.
3898         (*one_cmplhi2_1): Ditto.
3899         (*one_cmplqi2_1): Ditto.
3900         * config/i386/sse.md (): Change k to Yk.
3901         (avx512f_load<mode>_mask): Ditto.
3902         (avx512f_blendm<mode>): Ditto.
3903         (avx512f_store<mode>_mask): Ditto.
3904         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3905         (avx512f_storedqu<mode>_mask): Ditto.
3906         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3907         Ditto.
3908         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3909         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3910         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3911         (avx512f_maskcmp<mode>3): Ditto.
3912         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3913         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3914         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3915         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3916         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3917         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3918         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3919         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3920         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3921         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3922         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3923         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3924         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3925         (vec_extract_lo_<mode>_maskm): Ditto.
3926         (vec_extract_hi_<mode>_maskm): Ditto.
3927         (avx512f_vternlog<mode>_mask): Ditto.
3928         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3929         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3930         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3931         (avx512f_<code>v8div16qi2_mask): Ditto.
3932         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3933         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3934         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3935         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3936         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3937         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3938         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3939         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3940         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3941         (avx512cd_maskb_vec_dupv8di): Ditto.
3942         (avx512cd_maskw_vec_dupv16si): Ditto.
3943         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3944         (avx512f_vpermi2var<mode>3_mask): Ditto.
3945         (avx512f_vpermi2var<mode>3_mask): Ditto.
3946         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3947         (*avx512f_gathersi<mode>): Ditto.
3948         (*avx512f_gathersi<mode>_2): Ditto.
3949         (*avx512f_gatherdi<mode>): Ditto.
3950         (*avx512f_gatherdi<mode>_2): Ditto.
3951         (*avx512f_scattersi<mode>): Ditto.
3952         (*avx512f_scatterdi<mode>): Ditto.
3953         (avx512f_compress<mode>_mask): Ditto.
3954         (avx512f_compressstore<mode>_mask): Ditto.
3955         (avx512f_expand<mode>_mask): Ditto.
3956         * config/i386/subst.md (mask): Change k to Yk.
3957         (mask_scalar_merge): Ditto.
3958         (sd): Ditto.
3960 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3962         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3964 2014-01-31  Richard Biener  <rguenther@suse.de>
3966         PR middle-end/59990
3967         * builtins.c (fold_builtin_memory_op): Make sure to not
3968         use a floating-point mode or a boolean or enumeral type for
3969         the copy operation.
3971 2014-01-30  DJ Delorie  <dj@redhat.com>
3973         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3974         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3975         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3976         whenever main() has an epilogue.
3978 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3980         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3981         unused variable "field".
3982         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3983         (vsx_mergeh_<mode>): Likewise.
3984         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3985         (altivec_vmrghh): Likewise.
3986         (altivec_vmrghw): Likewise.
3987         (altivec_vmrglb): Likewise.
3988         (altivec_vmrglh): Likewise.
3989         (altivec_vmrglw): Likewise.
3990         (altivec_vspltb): Add missing uses.
3991         (altivec_vsplth): Likewise.
3992         (altivec_vspltw): Likewise.
3993         (altivec_vspltsf): Likewise.
3995 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3997         PR target/59923
3998         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3999         frame related instructions.
4001 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
4003         PR rtl-optimization/59959
4004         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
4005         any reload of register whose subreg is invalid.
4007 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4009         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
4010         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
4011         Add missing return type - void.
4013 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4015         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
4016         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
4017         remove element index adjustment for endian (now handled in vsx.md
4018         and altivec.md).
4019         (altivec_expand_vec_perm_const): Use
4020         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
4021         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
4022         (vsx_xxspltw_<mode>): Adjust element index for little endian.
4023         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
4024         define_expand and a new define_insn *altivec_vspltb_internal;
4025         adjust for -maltivec=be on a little endian target.
4026         (altivec_vspltb_direct): New.
4027         (altivec_vsplth): Divide into a define_expand and a new
4028         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
4029         little endian target.
4030         (altivec_vsplth_direct): New.
4031         (altivec_vspltw): Divide into a define_expand and a new
4032         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
4033         little endian target.
4034         (altivec_vspltw_direct): New.
4035         (altivec_vspltsf): Divide into a define_expand and a new
4036         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
4037         a little endian target.
4039 2014-01-30  Richard Biener  <rguenther@suse.de>
4041         PR tree-optimization/59993
4042         * tree-ssa-forwprop.c (associate_pointerplus): Check we
4043         can propagate form the earlier stmt and avoid the transform
4044         when the intermediate result is needed.
4046 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
4048         * README.Portability: Fix typo.
4050 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
4052         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
4053         comparison_operator with ordered_comparison_operator.
4055 2014-01-30  Nick Clifton  <nickc@redhat.com>
4057         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
4058         Rename to mn10300_store_multiple_regs.
4059         * config/mn10300/mn10300.c: Likewise.
4060         * config/mn10300/mn10300.md (store_movm): Fix typo: call
4061         store_multiple_regs.
4062         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
4063         Call mn10300_store_multiple_regs.
4065 2014-01-30  Nick Clifton  <nickc@redhat.com>
4066             DJ Delorie  <dj@redhat.com>
4068         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
4069         %fp 2 to keep registers after it properly word-aligned.
4070         (rl78_alloc_physical_registers_umul): Handle the case where both
4071         input operands are the same.
4073 2014-01-30  Richard Biener  <rguenther@suse.de>
4075         PR tree-optimization/59903
4076         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4077         check properly.
4079 2014-01-30  Jason Merrill  <jason@redhat.com>
4081         PR c++/59633
4082         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
4084         PR c++/59645
4085         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
4087 2014-01-30  Richard Biener  <rguenther@suse.de>
4089         PR tree-optimization/59951
4090         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
4092 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
4094         PR target/59784
4095         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
4096         SFmode to DFmode case.
4098 2014-01-29  DJ Delorie  <dj@redhat.com>
4100         * config/msp430/msp430.opt (-minrt): New.
4101         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
4102         if -minrt given.
4103         (ENDFILE_SPEC): Likewise.
4105 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
4107         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
4108         (estimate_function_body_sizes): Use it.
4110 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
4112         PR c++/58561
4113         * dwarf2out.c (is_cxx_auto): New.
4114         (is_base_type): Use it.
4115         (gen_type_die_with_usage): Likewise.
4117 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4119         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
4120         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
4121         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
4122         -maltivec=be with LE targets.
4123         (vsx_mergeh_<mode>): Likewise.
4124         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
4125         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
4126         (altivec_vmrghb): Replace with define_expand and new
4127         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
4128         (altivec_vmrghb_direct): New define_insn.
4129         (altivec_vmrghh): Replace with define_expand and new
4130         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
4131         (altivec_vmrghh_direct): New define_insn.
4132         (altivec_vmrghw): Replace with define_expand and new
4133         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
4134         (altivec_vmrghw_direct): New define_insn.
4135         (*altivec_vmrghsf): Adjust for endianness.
4136         (altivec_vmrglb): Replace with define_expand and new
4137         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
4138         (altivec_vmrglb_direct): New define_insn.
4139         (altivec_vmrglh): Replace with define_expand and new
4140         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
4141         (altivec_vmrglh_direct): New define_insn.
4142         (altivec_vmrglw): Replace with define_expand and new
4143         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
4144         (altivec_vmrglw_direct): New define_insn.
4145         (*altivec_vmrglsf): Adjust for endianness.
4146         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4147         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4148         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4149         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4150         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4151         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4152         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4153         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4155 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4157         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4158         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4159         whitespace.
4161 2014-01-29  Richard Biener  <rguenther@suse.de>
4163         PR tree-optimization/58742
4164         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4165         associate_pointerplus_align.
4166         (associate_pointerplus_diff): New function.
4167         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4168         and associate_pointerplus_diff.
4170 2014-01-29  Richard Biener  <rguenther@suse.de>
4172         * lto-streamer.h (LTO_major_version): Bump to 3.
4173         (LTO_minor_version): Reset to 0.
4175 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4177         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4178         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4179         (arm_file_start): Generate correct asm header for armv7ve.
4180         * config/arm/bpabi.h: Add multilib support for armv7ve.
4181         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4182         and cortex-a15 to armv7ve.
4183         * config/arm/t-aprofile: Add multilib support for armv7ve.
4184         * doc/invoke.texi: Document -march=armv7ve.
4186 2014-01-29  Richard Biener  <rguenther@suse.de>
4188         PR tree-optimization/58742
4189         * tree-ssa-forwprop.c (associate_plusminus): Return true
4190         if we changed sth, defer EH cleanup to ...
4191         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4192         (simplify_mult): New function.
4194 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4196         PR middle-end/59917
4197         PR tree-optimization/59920
4198         * tree.c (build_common_builtin_nodes): Remove
4199         __builtin_setjmp_dispatcher initialization.
4200         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4201         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4202         instead of gsi_after_labels + manually skipping debug stmts.
4203         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4204         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4205         * tree-inline.c (copy_edges_for_bb): Remove
4206         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4207         argument.  Ignore computed_goto_p stmts.  Don't call
4208         make_abnormal_goto_edges.  If a call might need abnormal edges
4209         for non-local gotos, see if it already has an edge to
4210         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4211         with true argument, don't do anything then, otherwise add
4212         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4213         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4214         caller.
4215         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4216         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4217         (lower_stmt): Don't set data->calls_builtin_setjmp.
4218         (lower_builtin_setjmp): Adjust comment.
4219         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4220         * tree-cfg.c (found_computed_goto): Remove.
4221         (factor_computed_gotos): Remove.
4222         (make_goto_expr_edges): Return bool, true for computed gotos.
4223         Don't call make_abnormal_goto_edges.
4224         (build_gimple_cfg): Don't set found_computed_goto, don't call
4225         factor_computed_gotos.
4226         (computed_goto_p): No longer static.
4227         (make_blocks): Don't set found_computed_goto.
4228         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4229         (make_edges): If make_goto_expr_edges returns true, push bb
4230         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4231         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4232         vector.  Record mapping between bbs and OpenMP regions if there
4233         are any, adjust make_gimple_omp_edges caller.  Call
4234         handle_abnormal_edges.
4235         (make_abnormal_goto_edges): Remove.
4236         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4237         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4238         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4239         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4240         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4241         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4242         filling *region also set *region_idx to (*region)->entry->index.
4244         PR other/58712
4245         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4246         For REGs set ORIGINAL_REGNO.
4248 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4250         * doc/md.texi: Mention that a target shouldn't implement
4251         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4252         than hi/lo pair.
4254 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4256         PR tree-optimization/59594
4257         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4258         a copy of the datarefs vector rather than the vector itself.
4260 2014-01-28  Jason Merrill  <jason@redhat.com>
4262         PR c++/53756
4263         * dwarf2out.c (auto_die): New static.
4264         (gen_type_die_with_usage): Handle C++1y 'auto'.
4265         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4266         on definition.
4268 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4270         PR target/59672
4271         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4272         (SPEC_X32): Likewise.
4273         (SPEC_64): Likewise.
4274         * config/i386/i386.c (ix86_option_override_internal): Turn off
4275         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4276         for TARGET_16BIT.
4277         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4278         * config/i386/i386.h (TARGET_16BIT): New macro.
4279         (TARGET_16BIT_P): Likewise.
4280         * config/i386/i386.opt: Add m16.
4281         * doc/invoke.texi: Document -m16.
4283 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4285         PR preprocessor/59935
4286         * input.c (location_get_source_line): Bail out on when line number
4287         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4289 2014-01-28  Richard Biener  <rguenther@suse.de>
4291         PR tree-optimization/58742
4292         * tree-ssa-forwprop.c (associate_plusminus): Handle
4293         pointer subtraction of the form (T)(P + A) - (T)P.
4295 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4297         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4298         at const_int_cost.
4300 2014-01-28  Richard Biener  <rguenther@suse.de>
4302         Revert
4303         2014-01-28  Richard Biener  <rguenther@suse.de>
4305         PR rtl-optimization/45364
4306         PR rtl-optimization/59890
4307         * var-tracking.c (local_get_addr_clear_given_value): Handle
4308         already cleared slot.
4309         (val_reset): Handle not allocated local_get_addr_cache.
4310         (vt_find_locations): Use post-order on the inverted CFG.
4312 2014-01-28  Richard Biener  <rguenther@suse.de>
4314         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4316 2014-01-28  Richard Biener  <rguenther@suse.de>
4318         PR rtl-optimization/45364
4319         PR rtl-optimization/59890
4320         * var-tracking.c (local_get_addr_clear_given_value): Handle
4321         already cleared slot.
4322         (val_reset): Handle not allocated local_get_addr_cache.
4323         (vt_find_locations): Use post-order on the inverted CFG.
4325 2014-01-28  Alan Modra  <amodra@gmail.com>
4327         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4328         * configure.ac <recursive call for build != host>: Define
4329         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4330         and LD_FOR_BUILD too.
4331         * configure: Regenerate.
4333 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4335         * config/i386/i386.c (get_builtin_code_for_version): Separate
4336         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4337         Broadwell from Haswell.
4339 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4341         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4342         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4343         * config/mips/mips.c (mips_option_override): Change setting
4344         of TARGET_DSP.
4345         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4346         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4347         Change from Mask to Var.
4349 2014-01-27  Jeff Law  <law@redhat.com>
4351         * ipa-inline.c (inline_small_functions): Fix typo.
4353 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4355         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4356         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4357         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4358         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4359         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4360         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4361         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4362         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4363         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4364         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4365         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4366         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4367         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4368         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4369         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4370         (_mm512_storeu_epi64): Ditto.
4371         (_mm512_cmpge_epi32_mask): Ditto.
4372         (_mm512_cmpge_epu32_mask): Ditto.
4373         (_mm512_cmpge_epi64_mask): Ditto.
4374         (_mm512_cmpge_epu64_mask): Ditto.
4375         (_mm512_cmple_epi32_mask): Ditto.
4376         (_mm512_cmple_epu32_mask): Ditto.
4377         (_mm512_cmple_epi64_mask): Ditto.
4378         (_mm512_cmple_epu64_mask): Ditto.
4379         (_mm512_cmplt_epi32_mask): Ditto.
4380         (_mm512_cmplt_epu32_mask): Ditto.
4381         (_mm512_cmplt_epi64_mask): Ditto.
4382         (_mm512_cmplt_epu64_mask): Ditto.
4383         (_mm512_cmpneq_epi32_mask): Ditto.
4384         (_mm512_cmpneq_epu32_mask): Ditto.
4385         (_mm512_cmpneq_epi64_mask): Ditto.
4386         (_mm512_cmpneq_epu64_mask): Ditto.
4387         (_mm512_expand_pd): Ditto.
4388         (_mm512_expand_ps): Ditto.
4389         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4390         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4391         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4392         * config/i386/i386.c (ix86_builtins): Add
4393         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4394         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4395         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4396         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4397         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4398         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4399         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4400         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4401         IX86_BUILTIN_PMOVUSQW512_MEM.
4402         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4403         __builtin_ia32_pmovsqd512mem_mask,
4404         __builtin_ia32_pmovqd512mem_mask,
4405         __builtin_ia32_pmovusqw512mem_mask,
4406         __builtin_ia32_pmovsqw512mem_mask,
4407         __builtin_ia32_pmovqw512mem_mask,
4408         __builtin_ia32_pmovusdw512mem_mask,
4409         __builtin_ia32_pmovsdw512mem_mask,
4410         __builtin_ia32_pmovdw512mem_mask,
4411         __builtin_ia32_pmovqb512mem_mask,
4412         __builtin_ia32_pmovusqb512mem_mask,
4413         __builtin_ia32_pmovsqb512mem_mask,
4414         __builtin_ia32_pmovusdb512mem_mask,
4415         __builtin_ia32_pmovsdb512mem_mask,
4416         __builtin_ia32_pmovdb512mem_mask.
4417         (bdesc_args): Add __builtin_ia32_expanddf512,
4418         __builtin_ia32_expandsf512.
4419         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4420         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4421         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4422         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4423         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4424         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4425         (avx512f_<code>v8div16qi2_mask_store): This.
4426         (avx512f_expand<mode>): New.
4428 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4430         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4431         New.
4432         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4433         (_mm512_prefetch_i32scatter_pd): Ditto.
4434         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4435         (_mm512_prefetch_i64scatter_pd): Ditto.
4436         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4437         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4438         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4439         (_mm512_prefetch_i32scatter_ps): Ditto.
4440         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4441         (_mm512_prefetch_i64scatter_ps): Ditto.
4442         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4443         * config/i386/i386-builtin-types.def: Define
4444         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4445         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4446         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4447         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4448         IX86_BUILTIN_SCATTERPFQPD.
4449         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4450         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4451         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4452         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4453         __builtin_ia32_scatterpfqps.
4454         (ix86_expand_builtin): Expand new built-ins.
4455         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4456         fix memory access data type.
4457         (*avx512pf_gatherpf<mode>_mask): Ditto.
4458         (*avx512pf_gatherpf<mode>): Ditto.
4459         (avx512pf_scatterpf<mode>): Ditto.
4460         (*avx512pf_scatterpf<mode>_mask): Ditto.
4461         (*avx512pf_scatterpf<mode>): Ditto.
4462         (GATHER_SCATTER_SF_MEM_MODE): New.
4463         (avx512pf_gatherpf<mode>df): Ditto.
4464         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4465         (*avx512pf_scatterpf<mode>df): Ditto.
4467 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4469         PR bootstrap/59934
4470         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4471         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4472         reached.
4474 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4476         * common/config/arm/arm-common.c
4477         (arm_rewrite_mcpu): Handle multiple names.
4478         * config/arm/arm.h
4479         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4481 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4483         * gimple-builder.h (create_gimple_tmp): Delete.
4485 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4487         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4488         words comparisons.
4490 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4492         * config/pa/pa.md (call): Generate indirect long calls to non-local
4493         functions when outputing 32-bit code.
4494         (call_value): Likewise except for special call to buggy powf function.
4496         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4497         portable runtime and PIC indirect calls.
4498         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4499         and PIC call sequences.  Use ldo instead of blr to set return register
4500         in PIC call sequence.
4502 2014-01-25  Walter Lee  <walt@tilera.com>
4504         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4505         avoid clobbering a live register.
4507 2014-01-25  Walter Lee  <walt@tilera.com>
4509         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4510         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4511         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4512         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4514 2014-01-25  Walter Lee  <walt@tilera.com>
4516         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4517         arguments on even registers.
4518         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4519         STACK_BOUNDARY.
4520         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4521         (BIGGEST_ALIGNMENT): Ditto.
4522         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4524 2014-01-25  Walter Lee  <walt@tilera.com>
4526         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4527         insns before bundling.
4528         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4530 2014-01-25  Walter Lee  <walt@tilera.com>
4532         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4533         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4534         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4536 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4538         * config/mips/constraints.md (kl): Delete.
4539         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4540         define expands, using...
4541         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4542         instructions for MIPS16.
4543         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4544         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4546 2014-01-25  Walter Lee  <walt@tilera.com>
4548         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4549         (clzdi2): Ditto.
4550         (ffsdi2): Ditto.
4552 2014-01-25  Walter Lee  <walt@tilera.com>
4554         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4555         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4557 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4559         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4560         Handle XOR.
4562 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4564         * print-rtl.c (in_call_function_usage): New var.
4565         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4566         EXPR_LIST mode as mode and not as reg note name.
4568         PR middle-end/59561
4569         * cfgloopmanip.c (copy_loop_info): If
4570         loop->warned_aggressive_loop_optimizations, make sure
4571         the flag is set in target loop too.
4573 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4575         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4576         flag_cilkplus.
4577         * builtins.def: Likewise.
4578         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4579         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4580         * ira.c (ira_setup_eliminable_regset): Likewise.
4581         * omp-low.c (gate_expand_omp): Likewise.
4582         (execute_lower_omp): Likewise.
4583         (diagnose_sb_0): Likewise.
4584         (gate_diagnose_omp_blocks): Likewise.
4585         (simd_clone_clauses_extract): Likewise.
4586         (gate): Likewise.
4588 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4590         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4591         correction for little endian...
4592         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4593         here.
4595 2014-01-24  Jeff Law  <law@redhat.com>
4597         PR tree-optimization/59919
4598         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4599         for non-returning calls.
4601 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4603         * common/config/aarch64/aarch64-common.c
4604         (aarch64_rewrite_mcpu): Handle multiple names.
4605         * config/aarch64/aarch64.h
4606         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4608 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4610         * input.c (add_file_to_cache_tab): Handle the case where fopen
4611         returns NULL.
4613 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4615         PR target/59929
4616         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4617         from push operand if code of push isn't PRE_DEC.
4619 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4621         PR target/59909
4622         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4623         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4624         it is only used for non-atomic loads/stores.
4626         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4627         -mquad-memory or -mquad-memory-atomic switches.
4629         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4630         -mquad-memory-atomic to ISA 2.07 support.
4632         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4633         to separate support of normal quad word memory operations (ldq, stq)
4634         from the atomic quad word memory operations.
4636         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4637         support to separate non-atomic quad word operations from atomic
4638         quad word operations.  Disable non-atomic quad word operations in
4639         little endian mode so that we don't have to swap words after the
4640         load and before the store.
4641         (quad_load_store_p): Add comment about atomic quad word support.
4642         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4643         options printed with -mdebug=reg.
4645         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4646         -mquad-memory-atomic as the test for whether we have quad word
4647         atomic instructions.
4648         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4649         or -mp8-vector are used, allow byte/half-word atomic operations.
4651         * config/rs6000/sync.md (load_lockedti): Insure that the address
4652         is a proper indexed or indirect address for the lqarx instruction.
4653         On little endian systems, swap the hi/lo registers after the lqarx
4654         instruction.
4655         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4656         insure the address is valid for the lqarx instruction.
4657         (store_conditionalti): Insure that the address is a proper indexed
4658         or indirect address for the stqcrx. instruction.  On little endian
4659         systems, swap the hi/lo registers before doing the stqcrx.
4660         instruction.
4661         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4662         insure the address is valid for the stqcrx. instruction.
4664         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4665         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4666         type of quad memory support is available.
4668 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4670         PR regression/59915
4671         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4672         there is a danger of looping.
4674 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4676         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4677         force flag_ira_loop_pressure if set via command line.
4679 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4681         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4682         (ashr_simd): New builtin handling DI mode.
4683         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4684         (aarch64_sshr_simddi): New match pattern.
4685         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4686         (vshrd_n_s64): Likewise.
4687         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4689 2014-01-23  Nick Clifton  <nickc@redhat.com>
4691         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4692         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4693         favour of mcu specific scripts.
4694         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4695         430x multilibs.
4697 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4698             Alex Velenko  <Alex.Velenko@arm.com>
4700         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4701         (vaddv_s16): Likewise.
4702         (vaddv_s32): Likewise.
4703         (vaddv_u8): Likewise.
4704         (vaddv_u16): Likewise.
4705         (vaddv_u32): Likewise.
4706         (vaddvq_s8): Likewise.
4707         (vaddvq_s16): Likewise.
4708         (vaddvq_s32): Likewise.
4709         (vaddvq_s64): Likewise.
4710         (vaddvq_u8): Likewise.
4711         (vaddvq_u16): Likewise.
4712         (vaddvq_u32): Likewise.
4713         (vaddvq_u64): Likewise.
4714         (vaddv_f32): Likewise.
4715         (vaddvq_f32): Likewise.
4716         (vaddvq_f64): Likewise.
4717         (vmaxv_f32): Likewise.
4718         (vmaxv_s8): Likewise.
4719         (vmaxv_s16): Likewise.
4720         (vmaxv_s32): Likewise.
4721         (vmaxv_u8): Likewise.
4722         (vmaxv_u16): Likewise.
4723         (vmaxv_u32): Likewise.
4724         (vmaxvq_f32): Likewise.
4725         (vmaxvq_f64): Likewise.
4726         (vmaxvq_s8): Likewise.
4727         (vmaxvq_s16): Likewise.
4728         (vmaxvq_s32): Likewise.
4729         (vmaxvq_u8): Likewise.
4730         (vmaxvq_u16): Likewise.
4731         (vmaxvq_u32): Likewise.
4732         (vmaxnmv_f32): Likewise.
4733         (vmaxnmvq_f32): Likewise.
4734         (vmaxnmvq_f64): Likewise.
4735         (vminv_f32): Likewise.
4736         (vminv_s8): Likewise.
4737         (vminv_s16): Likewise.
4738         (vminv_s32): Likewise.
4739         (vminv_u8): Likewise.
4740         (vminv_u16): Likewise.
4741         (vminv_u32): Likewise.
4742         (vminvq_f32): Likewise.
4743         (vminvq_f64): Likewise.
4744         (vminvq_s8): Likewise.
4745         (vminvq_s16): Likewise.
4746         (vminvq_s32): Likewise.
4747         (vminvq_u8): Likewise.
4748         (vminvq_u16): Likewise.
4749         (vminvq_u32): Likewise.
4750         (vminnmv_f32): Likewise.
4751         (vminnmvq_f32): Likewise.
4752         (vminnmvq_f64): Likewise.
4754 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4756         * config/aarch64/aarch64-simd.md
4757         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4758         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4759         (*aarch64_mul3_elt<mode>): Likewise.
4760         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4761         (*aarch64_mul3_elt_to_64v2df): Likewise.
4762         (*aarch64_mla_elt<mode>): Likewise.
4763         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4764         (*aarch64_mls_elt<mode>): Likewise.
4765         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4766         (*aarch64_fma4_elt<mode>): Likewise.
4767         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4768         (*aarch64_fma4_elt_to_64v2df): Likewise.
4769         (*aarch64_fnma4_elt<mode>): Likewise.
4770         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4771         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4772         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4773         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4774         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4775         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4776         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4777         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4778         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4780 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4782         * config/aarch64/aarch64-simd.md
4783         (aarch64_be_checked_get_lane<mode>): New define_expand.
4784         * config/aarch64/aarch64-simd-builtins.def
4785         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4786         New builtin definition.
4787         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4788         Use new safe be builtin.
4790 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4792         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4793         New define_insn.
4794         (aarch64_be_st1<mode>): Likewise.
4795         (aarch_ld1<VALL:mode>): Define_expand modified.
4796         (aarch_st1<VALL:mode>): Likewise.
4797         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4798         (UNSPEC_ST1): Likewise.
4800 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
4802         * config/microblaze/microblaze.md: Add trap insn and attribute
4804 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4806         PR preprocessor/58580
4807         * input.h (location_get_source_line): Take an additional line_size
4808         parameter.
4809         (void diagnostics_file_cache_fini): Declare new function.
4810         * input.c (struct fcache): New type.
4811         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4812         New static constants.
4813         (diagnostic_file_cache_init, total_lines_num)
4814         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4815         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4816         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4817         (get_next_line, read_next_line, goto_next_line, read_line_num):
4818         New static function definitions.
4819         (diagnostic_file_cache_fini): New function.
4820         (location_get_source_line): Take an additional output line_len
4821         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4822         read_line_num.
4823         * diagnostic.c (diagnostic_finish): Call
4824         diagnostic_file_cache_fini.
4825         (adjust_line): Take an additional input parameter for the length
4826         of the line, rather than calculating it with strlen.
4827         (diagnostic_show_locus): Adjust the use of
4828         location_get_source_line and adjust_line with respect to their new
4829         signature.  While displaying a line now, do not stop at the first
4830         null byte.  Rather, display the zero byte as a space and keep
4831         going until we reach the size of the line.
4832         * Makefile.in: Add vec.o to OBJS-libcommon
4834 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4835             Ilya Tocar     <ilya.tocar@intel.com>
4837         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4838         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4839         (__builtin_ia32_kmov16): Ditto.
4840         * config/i386/i386.md (UNSPEC_KMOV): New.
4841         (kmovw): Ditto.
4843 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4845         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4846         (_mm512_storeu_si512): Ditto.
4848 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4850         PR target/52125
4851         * rtl.h (get_referenced_operands): Declare.
4852         * recog.c (get_referenced_operands): New function.
4853         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4854         operands have been referenced when recording LO_SUM references.
4856 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
4858         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4860 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4862         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4863         Enable for generic and recent AMD targets.
4865 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4867         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4868         ARG_SIZE note when adjustment was eliminated.
4870 2014-01-22  Jeff Law  <law@redhat.com>
4872         PR tree-optimization/59597
4873         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4874         in file.  Accept new argument REGISTERING and use it to modify
4875         dump output appropriately.
4876         (register_jump_thread): Corresponding changes.
4877         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4878         thread paths involving joiner blocks.  Add code to dump cancelled
4879         jump threading paths.
4881 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4883         PR rtl-optimization/59477
4884         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4885         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4887 2014-01-22  Tom Tromey  <tromey@redhat.com>
4889         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4890         PARAMS.
4891         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4893 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4895         PR rtl-optimization/59896
4896         * lra-constraints.c (process_alt_operands): Check unused note for
4897         matched operands of insn with no output reloads.
4899 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4901         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4902         (mips_move_from_gpr_cost): Likewise.
4904 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4906         PR rtl-optimization/59858
4907         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4908         ira_class_hard_regs_num.
4909         (process_alt_operands): Increase reject for dying matched operand.
4911 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4913         PR target/59003
4914         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4915         smaller than size, perform several stores or loads and stores
4916         at dst + count - size to store or copy all of size bytes, rather
4917         than just last modesize bytes.
4919 2014-01-20  DJ Delorie  <dj@redhat.com>
4921         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4922         that CLOBBERs are REGs before propogating their values.
4924 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4926         PR middle-end/59789
4927         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4928         (cgraph_inline_failed_type): New function.
4929         * cgraph.h (DEFCIFCODE): Add type.
4930         (cgraph_inline_failed_type_t): New enum.
4931         (cgraph_inline_failed_type): New prototype.
4932         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4933         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4934         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4935         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4936         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4937         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4938         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4939         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4940         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4941         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4942         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4943         OPTIMIZATION_MISMATCH.
4944         * tree-inline.c (expand_call_inline): Emit errors during
4945         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4947 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4949         PR target/59685
4950         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4951         mode attribute in insn output.
4953 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4955         * output.h (output_constant): Delete.
4956         * varasm.c (output_constant): Make private.
4958 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4960         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4962 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4964         PR middle-end/59860
4965         * tree.h (fold_builtin_strcat): New prototype.
4966         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4967         argument, if non-NULL, don't call c_strlen.  Optimize
4968         directly into __builtin_memcpy instead of __builtin_strcpy.
4969         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4970         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4972 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4974         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4975         for SImode_address_operand operands, having only a REG argument.
4977 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4979         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4980         loader name using mbig-endian.
4981         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4983 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4985         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4986         (-mtune): Likewise.
4987         (-mcpu): Likewise.
4989 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4991         * config/aarch64/aarch64-protos.h
4992         (aarch64_cannot_change_mode_class_ptr): Declare.
4993         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4994         aarch64_cannot_change_mode_class_ptr): New.
4995         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4996         backend hook aarch64_cannot_change_mode_class.
4998 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5000         * common/config/aarch64/aarch64-common.c
5001         (aarch64_handle_option): Don't handle any option order logic here.
5002         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
5003         selected_cpu, warn on architecture version mismatch.
5004         (aarch64_override_options): Fix parsing order for option strings.
5006 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5007             Iain Sandoe  <iain@codesourcery.com>
5009         PR bootstrap/59496
5010         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
5011         warning.  Amend comment to reflect current functionality.
5013 2014-01-20  Richard Biener  <rguenther@suse.de>
5015         PR middle-end/59860
5016         * builtins.c (fold_builtin_strcat): Remove case better handled
5017         by tree-ssa-strlen.c.
5019 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
5021         * config/aarch64/aarch64.opt
5022         (mcpu, march, mtune): Make case-insensitive.
5024 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5026         PR target/59880
5027         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5028         if operands[1] is a REG or ZERO_EXTEND of a REG.
5030 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
5032         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
5034 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
5036         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
5037         long non-pic millicode calls.
5039 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5041         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
5043 2014-01-19  Kito Cheng  <kito@0xlab.org>
5045         * builtins.c (expand_movstr): Check movstr expand done or fail.
5047 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5048             H.J. Lu  <hongjiu.lu@intel.com>
5050         PR target/59379
5051         * config/i386/i386.md (*lea<mode>): Zero-extend return register
5052         to DImode for zero-extended addresses.
5054 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
5056         PR rtl-optimization/57763
5057         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
5058         on the new indirect jump_insn and increment LABEL_NUSES (label).
5060 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
5062         PR bootstrap/59580
5063         PR bootstrap/59583
5064         * config.gcc (x86_archs): New variable.
5065         (x86_64_archs): Likewise.
5066         (x86_cpus): Likewise.
5067         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
5068         --with-arch/--with-cpu= options.
5069         Support --with-arch=/--with-cpu={nehalem,westmere,
5070         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
5072 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5074         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
5075         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
5077 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5079         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
5081 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
5083         PR target/58944
5084         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
5085         clear cpp_get_options (parse_in)->warn_unused_macros for
5086         ix86_target_macros_internal with cpp_define.
5088 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
5090         * jump.c (delete_related_insns): Keep (use (insn))s.
5091         * reorg.c (redundant_insn): Check for barriers too.
5093 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5095         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
5097 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
5099         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
5100         call to $$dyncall when TARGET_LONG_CALLS is true.
5102 2014-01-17  Jeff Law  <law@redhat.com>
5104         * ree.c (combine_set_extension): Temporarily disable test for
5105         changing number of hard registers.
5107 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5109         PR middle-end/58125
5110         * ipa-inline-analysis.c (inline_free_summary):
5111         Do not free summary of aliases.
5113 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5115         PR middle-end/59706
5116         * gimplify.c (gimplify_expr): Use create_tmp_var
5117         instead of create_tmp_var_raw.  If cond doesn't have
5118         integral type, don't add the IFN_ANNOTATE builtin at all.
5120 2014-01-17  Martin Jambor  <mjambor@suse.cz>
5122         PR ipa/59736
5123         * ipa-cp.c (prev_edge_clone): New variable.
5124         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
5125         Also resize prev_edge_clone vector.
5126         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
5127         (ipcp_edge_removal_hook): New function.
5128         (ipcp_driver): Register ipcp_edge_removal_hook.
5130 2014-01-17  Andrew Pinski  <apinski@cavium.com>
5131             Steve Ellcey  <sellcey@mips.com>
5133         PR target/59462
5134         * config/mips/mips.c (mips_print_operand): Check operand mode instead
5135         of operator mode.
5137 2014-01-17  Jeff Law  <law@redhat.com>
5139         PR middle-end/57904
5140         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
5141         so that pass_ccp runs first.
5143 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5145         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5146         (ix86_adjust_cost): Use !TARGET_XXX.
5147         (do_reorder_for_imul): Likewise.
5148         (swap_top_of_ready_list): Likewise.
5149         (ix86_sched_reorder): Likewise.
5151 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5153         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5154         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5155         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5156         (intel_memset): New.  Duplicate slm_memset.
5157         (intel_cost): New.  Duplicate slm_cost.
5158         (m_INTEL): New macro.
5159         (processor_target_table): Add "intel".
5160         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5161         with PROCESSOR_INTEL for "intel".
5162         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5163         PROCESSOR_SILVERMONT.
5164         (ix86_issue_rate): Likewise.
5165         (ix86_adjust_cost): Likewise.
5166         (ia32_multipass_dfa_lookahead): Likewise.
5167         (swap_top_of_ready_list): Likewise.
5168         (ix86_sched_reorder): Likewise.
5169         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5170         instead of TARGET_OPT_AGU.
5171         * config/i386/i386.h (TARGET_INTEL): New.
5172         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5173         (processor_type): Add PROCESSOR_INTEL.
5174         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5175         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5177 2014-01-17  Marek Polacek  <polacek@redhat.com>
5179         PR c/58346
5180         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5181         size is zero.
5183 2014-01-17  Richard Biener  <rguenther@suse.de>
5185         PR tree-optimization/46590
5186         * opts.c (default_options_table): Add entries for
5187         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5188         all enabled at -O1 but not for -Og.
5189         * common.opt (fbranch-count-reg): Remove Init(1).
5190         (fmove-loop-invariants): Likewise.
5191         (ftree-pta): Likewise.
5193 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5195         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5196         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5197         decls to at least the GCC 4.8 used alignments.
5199         PR fortran/59440
5200         * tree-nested.c (convert_nonlocal_reference_stmt,
5201         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5202         of GIMPLE_BIND stmts, adjust associated decls.
5204 2014-01-17  Richard Biener  <rguenther@suse.de>
5206         PR tree-optimization/46590
5207         * vec.h (vec<>::bseach): New member function implementing
5208         binary search according to C89 bsearch.
5209         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5210         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5211         bitmap pointer again.  Make accesses_in_loop a flat array.
5212         (mem_ref_obstack): New global.
5213         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5214         (mark_ref_stored): Likewise.
5215         (ref_indep_loop_p_2): Likewise.
5216         (set_ref_stored_in_loop): New helper function.
5217         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5218         (memref_free): Adjust.
5219         (record_mem_ref_loc): Simplify.
5220         (gather_mem_refs_stmt): Adjust.
5221         (sort_locs_in_loop_postorder_cmp): New function.
5222         (analyze_memory_references): Sort accesses_in_loop after
5223         loop postorder number.
5224         (find_ref_loc_in_loop_cmp): New function.
5225         (for_all_locs_in_loop): Find relevant cluster of locs in
5226         accesses_in_loop and iterate without recursion.
5227         (execute_sm): Avoid uninit warning.
5228         (struct ref_always_accessed): Simplify.
5229         (ref_always_accessed::operator ()): Likewise.
5230         (ref_always_accessed_p): Likewise.
5231         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5232         loop postorder numbers here.
5233         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5234         numbers.
5236 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5238         PR c++/57945
5239         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5240         on decls for which assemble_alias has been called.
5242 2014-01-17  Nick Clifton  <nickc@redhat.com>
5244         * config/msp430/msp430.opt: (mcpu): New option.
5245         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5246         (msp430_option_override): Parse target_cpu.  If the MCU name
5247         matches a generic string, clear target_mcu.
5248         (msp430_attr): Allow numeric interrupt values up to 63.
5249         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5250         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5251         option.
5252         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5253         Add mcpu matches.
5254         * config/msp430/msp430.md (popm): Use %J rather than %I.
5255         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5256         (addhi_cy_i): Use immediate_operand for operand 2.
5257         * doc/invoke.texi: Document -mcpu option.
5259 2014-01-17  Richard Biener  <rguenther@suse.de>
5261         PR rtl-optimization/38518
5262         * df.h (df_analyze_loop): Declare.
5263         * df-core.c: Include cfgloop.h.
5264         (df_analyze_1): Split out main part of df_analyze.
5265         (df_analyze): Adjust.
5266         (loop_inverted_post_order_compute): New function.
5267         (loop_post_order_compute): Likewise.
5268         (df_analyze_loop): New function avoiding whole-function
5269         postorder computes.
5270         * loop-invariant.c (find_defs): Use df_analyze_loop.
5271         (find_invariants): Adjust.
5272         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5274 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5276         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5277         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5279 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5281         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5282         traversal when removing references.
5284 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5286         PR ipa/59775
5287         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5289 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5291         PR middle-end/56791
5292         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5293         pushing a reload for an autoinc when we had previously reloaded an
5294         inner part of the address.
5296 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5298         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5299         field.
5300         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5301         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5302         when not giving up or versioning for alias only because of
5303         loop->safelen.
5304         (vect_analyze_data_ref_dependences): Set to true.
5305         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5306         is a GIMPLE_PHI.
5307         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5308         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5309         to the condition.
5311         PR middle-end/58344
5312         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5314         PR target/59839
5315         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5316         operand 0 predicate for gathers, use a new pseudo as subtarget.
5318 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5320         PR middle-end/59609
5321         * lra-constraints.c (process_alt_operands): Add printing debug info.
5322         Check absence of input/output reloads for matched operands too.
5324 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5326         PR rtl-optimization/59835
5327         * ira.c (ira_init_register_move_cost): Increase cost for
5328         impossible modes.
5330 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5332         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5334 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5336         PR target/59780
5337         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5338         non-register objects.  Use gen_(high/low)part more consistently.
5339         Fix assertions.
5341 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5343         PR target/59844
5344         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5345         endian support, remove tests for WORDS_BIG_ENDIAN.
5346         (p8_mfvsrd_3_<mode>): Likewise.
5347         (reload_gpr_from_vsx<mode>): Likewise.
5348         (reload_gpr_from_vsxsf): Likewise.
5349         (p8_mfvsrd_4_disf): Likewise.
5351 2014-01-16  Richard Biener  <rguenther@suse.de>
5353         PR rtl-optimization/46590
5354         * lcm.c (compute_antinout_edge): Use postorder iteration.
5355         (compute_laterin): Use inverted postorder iteration.
5357 2014-01-16  Nick Clifton  <nickc@redhat.com>
5359         PR middle-end/28865
5360         * varasm.c (output_constant): Return the number of bytes actually
5361         emitted.
5362         (output_constructor_array_range): Update the field size with the
5363         number of bytes emitted by output_constant.
5364         (output_constructor_regular_field): Likewise.  Also do not
5365         complain if the total number of bytes emitted is now greater
5366         than the expected fieldpos.
5367         * output.h (output_constant): Update prototype and descriptive comment.
5369 2014-01-16  Marek Polacek  <polacek@redhat.com>
5371         PR middle-end/59827
5372         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5373         it is error_mark_node.
5375 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5377         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5378         VALID_AVX256_REG_OR_OI_MODE.
5380 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5382         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5383         current procedure should be profiled.
5385 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5387         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5388         of moving from/to the STACK_REG register class.
5390 2014-01-15  Richard Henderson  <rth@redhat.com>
5392         PR debug/54694
5393         * reginfo.c (global_regs_decl): Globalize.
5394         * rtl.h (global_regs_decl): Declare.
5395         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5396         reserved via global_regs.
5398 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5400         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5402 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5404         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5405         and vmulosh rather than call gen_vec_widen_smult_*.
5406         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5407         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5408         (vec_widen_smult_even_v16qi): Likewise.
5409         (vec_widen_umult_even_v8hi): Likewise.
5410         (vec_widen_smult_even_v8hi): Likewise.
5411         (vec_widen_umult_odd_v16qi): Likewise.
5412         (vec_widen_smult_odd_v16qi): Likewise.
5413         (vec_widen_umult_odd_v8hi): Likewise.
5414         (vec_widen_smult_odd_v8hi): Likewise.
5415         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5416         vmuloub rather than call gen_vec_widen_umult_*.
5417         (vec_widen_umult_lo_v16qi): Likewise.
5418         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5419         vmulosb rather than call gen_vec_widen_smult_*.
5420         (vec_widen_smult_lo_v16qi): Likewise.
5421         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5422         rather than call gen_vec_widen_umult_*.
5423         (vec_widen_umult_lo_v8hi): Likewise.
5424         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5425         rather than call gen_vec_widen_smult_*.
5426         (vec_widen_smult_lo_v8hi): Likewise.
5428 2014-01-15  Jeff Law  <law@redhat.com>
5430         PR tree-optimization/59747
5431         * ree.c (find_and_remove_re): Properly handle case where a second
5432         eliminated extension requires widening a copy created for elimination
5433         of a prior extension.
5434         (combine_set_extension): Ensure that the number of hard regs needed
5435         for a destination register does not change when we widen it.
5437 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5439         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5440         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5441         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5442         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5443         (avr-*-rtems*): Likewise.
5444         (bfin*-rtems*): Likewise.
5445         (moxie-*-rtems*): Likewise.
5446         (h8300-*-rtems*): Likewise.
5447         (i[34567]86-*-rtems*): Likewise.
5448         (lm32-*-rtems*): Likewise.
5449         (m32r-*-rtems*): Likewise.
5450         (m68k-*-rtems*): Likewise.
5451         (microblaze*-*-rtems*): Likewise.
5452         (mips*-*-rtems*): Likewise.
5453         (powerpc-*-rtems*): Likewise.
5454         (sh-*-rtems*): Likewise.
5455         (sparc-*-rtems*): Likewise.
5456         (sparc64-*-rtems*): Likewise.
5457         (v850-*-rtems*): Likewise.
5458         (m32c-*-rtems*): Likewise.
5460 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5462         PR rtl-optimization/59511
5463         * ira.c (ira_init_register_move_cost): Use memory costs for some
5464         cases of register move cost calculations.
5465         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5466         instead of BB frequency.
5467         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5468         * lra-assigns.c (find_hard_regno_for): Ditto.
5470 2014-01-15  Richard Biener  <rguenther@suse.de>
5472         PR tree-optimization/59822
5473         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5474         (vectorizable_load): Use it to hoist defs of uses of invariant
5475         loads out of the loop.
5477 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5478             Kugan Vivekanandarajah  <kuganv@linaro.org>
5480         PR target/59695
5481         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5482         truncation.
5484 2014-01-15  Richard Biener  <rguenther@suse.de>
5486         PR rtl-optimization/59802
5487         * lcm.c (compute_available): Use inverted postorder to seed
5488         the initial worklist.
5490 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5492         PR target/59803
5493         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5494         ADDR_REGS for invalid symrefs in non-PIC code.
5496 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5498         PR other/58712
5499         * builtins.c (determine_block_size): Initialize *probable_max_size
5500         even if len_rtx is CONST_INT.
5502 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5504         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5505         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5506         (cortexa53_tunings): Likewise.
5507         (aarch64_sched_issue_rate): New function.
5508         (TARGET_SCHED_ISSUE_RATE): Define.
5510 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5512         * ira-costs.c (find_costs_and_classes): Add missed
5513         ira_init_register_move_cost_if_necessary.
5515 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5517         PR target/59787
5518         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5520 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5522         PR target/59794
5523         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5524         to indicate if type is used for function return value.  Warn ABI
5525         change if the vector mode isn't available for function return value.
5526         (ix86_function_arg_advance): Pass false to type_natural_mode.
5527         (ix86_function_arg): Likewise.
5528         (ix86_gimplify_va_arg): Likewise.
5529         (function_arg_32): Don't warn ABI change.
5530         (ix86_function_value): Pass true to type_natural_mode.
5531         (ix86_return_in_memory): Likewise.
5532         (ix86_struct_value_rtx): Removed.
5533         (TARGET_STRUCT_VALUE_RTX): Likewise.
5535 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5537         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5538         converting a conditional jump into a conditional return.
5540 2014-01-14  Richard Biener  <rguenther@suse.de>
5542         PR tree-optimization/58921
5543         PR tree-optimization/59006
5544         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5545         hoisting invariant stmts.
5546         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5547         invariant loads on the preheader edge if possible.
5549 2014-01-14  Joey Ye  <joey.ye@arm.com>
5551         * doc/plugin.texi (Building GCC plugins): Update to C++.
5553 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5555         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5556         (_mm_rcp28_round_ss): Ditto.
5557         (_mm_rsqrt28_round_sd): Ditto.
5558         (_mm_rsqrt28_round_ss): Ditto.
5559         (_mm_rcp28_sd): Ditto.
5560         (_mm_rcp28_ss): Ditto.
5561         (_mm_rsqrt28_sd): Ditto.
5562         (_mm_rsqrt28_ss): Ditto.
5563         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5564         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5565         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5566         (IX86_BUILTIN_RCP28SD): Ditto.
5567         (IX86_BUILTIN_RCP28SS): Ditto.
5568         (IX86_BUILTIN_RSQRT28SD): Ditto.
5569         (IX86_BUILTIN_RSQRT28SS): Ditto.
5570         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5571         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5572         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5573         (ix86_expand_special_args_builtin): Expand new FTYPE.
5574         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5575         (srcp14<mode>): Make insn unary.
5576         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5577         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5578         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5579         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5580         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5581         Fix rounding: make it SAE only.
5582         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5583         Ditto.
5584         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5585         Ditto.
5586         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5587         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5588         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5589         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5590         (round_saeonly_mask_scalar_operand4): Ditto.
5591         (round_saeonly_mask_scalar_op3): Ditto.
5592         (round_saeonly_mask_scalar_op4): Ditto.
5594 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5596         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5597         Implement -maltivec=be for vec_insert and vec_extract.
5599 2014-01-10  DJ Delorie  <dj@redhat.com>
5601         * config/msp430/msp430.md (call_internal): Don't allow memory
5602         references with SP as the base register.
5603         (call_value_internal): Likewise.
5604         * config/msp430/constraints.md (Yc): New.  For memory references
5605         that don't use SP as a base register.
5607         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5608         "an integer without a # prefix"
5609         * config/msp430/msp430.md (epilogue_helper): Use it.
5611 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5613         PR target/59617
5614         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5615         AVX512F gather builtins.
5616         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5617         on gather decls with INTEGER_TYPE masktype.
5618         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5619         directly into the builtin rather than hoisting it before loop.
5621         PR tree-optimization/59387
5622         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5623         (scev_const_prop): If folded_casts and type has undefined overflow,
5624         use force_gimple_operand instead of force_gimple_operand_gsi and
5625         for each added stmt if it is assign with
5626         arith_code_with_undefined_signed_overflow, call
5627         rewrite_to_defined_overflow.
5628         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5629         gimple-fold.h instead.
5630         (arith_code_with_undefined_signed_overflow,
5631         rewrite_to_defined_overflow): Moved to ...
5632         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5633         rewrite_to_defined_overflow): ... here.  No longer static.
5634         Include gimplify-me.h.
5635         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5636         rewrite_to_defined_overflow): New prototypes.
5638 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5640         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5642 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5644         * builtins.c (get_object_alignment_2): Minor tweak.
5645         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5647 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5649         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5650         optimized non constant lengths.
5652 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5654         PR libgomp/59194
5655         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5656         load as __atomic_load_N if possible.
5658 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5660         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5661         target parameter.
5662         (rs6000_expand_builtin): Adjust call.
5664 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5666         PR target/58115
5667         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5668         * config/rs6000/rs6000.c: Include target-globals.h.
5669         (rs6000_set_current_function): Instead of doing target_reinit
5670         unconditionally, use save_target_globals_default_opts and
5671         restore_target_globals.
5673         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5674         FPSCR.
5675         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5676         (rs6000_expand_builtin): Handle mffs and mtfsf.
5677         (rs6000_init_builtins): Define mffs and mtfsf.
5678         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5679         (rs6000_mffs): New pattern.
5680         (rs6000_mtfsf): New pattern.
5682 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5684         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5685         Start narrowing with START.  Apply candidate-use pair
5686         and check overall cost in narrowing.
5687         (iv_ca_prune): Pass new argument.
5689 2014-01-10  Jeff Law  <law@redhat.com>
5691         PR middle-end/59743
5692         * ree.c (combine_reaching_defs): Ensure the defining statement
5693         occurs before the extension when optimizing extensions with
5694         different source and destination hard registers.
5696 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5698         PR ipa/58585
5699         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5700         vtables into the type inheritance graph.
5702 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5704         PR rtl-optimization/59754
5705         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5706         modes in the REGNO != REGNO case.
5708 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5710         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5712 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5714         PR tree-optimization/59745
5715         * tree-predcom.c (tree_predictive_commoning_loop): Call
5716         free_affine_expand_cache if giving up because components is NULL.
5718         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5719         GC in payload of target_globals struct instead of allocating them on
5720         the heap and the larger structs separately using GC.
5721         * target-globals.h (struct target_globals): Make regs, hard_regs,
5722         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5723         of GTY((skip)) and change type to void *.
5724         (reset_target_globals): Cast loads from those fields to corresponding
5725         types.
5727 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5729         PR plugins/59335
5730         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5731         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5732         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5734 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5736         PR target/59744
5737         * aarch64-modes.def (CC_Zmode): New flags mode.
5738         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5739         represents an equality.
5740         (aarch64_get_condition_code): Handle CC_Zmode.
5741         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5743 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5745         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5746         extraction in good case.
5748 2014-01-10  Richard Biener  <rguenther@suse.de>
5750         PR tree-optimization/59374
5751         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5752         checking after SLP discovery.  Mark stmts not participating
5753         in any SLP instance properly.
5755 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5757         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5758         when handling a SET rtx.
5760 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5762         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5763         (cortex-a57): Likewise.
5764         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5766 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5768         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5769         non-iwmmxt builtins.
5771 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5773         PR ipa/58252
5774         PR ipa/59226
5775         * ipa-devirt.c record_target_from_binfo): Take as argument
5776         stack of binfos and lookup matching one for virtual inheritance.
5777         (possible_polymorphic_call_targets_1): Update.
5779 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5781         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5782         kernel strings for Loongson-2E/2F/3A.
5784 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5786         PR middle-end/59670
5787         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5788         is_gimple_call before calling gimple_call_internal_p.
5790 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5792         * Makefile.in (TREE_FLOW_H): Remove.
5793         (TREE_SSA_H): Add file names from tree-flow.h.
5794         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5795         * tree.h: Remove tree-flow.h reference.
5796         * hash-table.h: Remove tree-flow.h reference.
5797         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5798         reference with tree-ssa-loop.h.
5800 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5802         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5803         default element-order behavior for -maltivec.
5804         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5805         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5806         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5807         when targeting big endian, at least for now.
5808         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5810 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5812         PR middle-end/47735
5813         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5814         var satisfies use_register_for_decl, just take into account type
5815         alignment, rather than decl alignment.
5817         PR tree-optimization/59622
5818         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5819         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5820         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5821         Don't devirtualize for inplace at all.  For targets.length () == 1,
5822         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5824 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5826         * config/i386/i386.md (cpu): Remove the unused btver1.
5828 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5830         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5832 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5834         PR target/58115
5835         * tree-core.h (struct target_globals): New forward declaration.
5836         (struct tree_target_option): Add globals field.
5837         * tree.h (TREE_TARGET_GLOBALS): Define.
5838         (prepare_target_option_nodes_for_pch): New prototype.
5839         * target-globals.h (struct target_globals): Define even if
5840         !SWITCHABLE_TARGET.
5841         * tree.c (prepare_target_option_node_for_pch,
5842         prepare_target_option_nodes_for_pch): New functions.
5843         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5844         * config/i386/i386.c: Include target-globals.h.
5845         (ix86_set_current_function): Instead of doing target_reinit
5846         unconditionally, use save_target_globals_default_opts and
5847         restore_target_globals.
5849 2014-01-09  Richard Biener  <rguenther@suse.de>
5851         PR tree-optimization/59715
5852         * tree-cfg.h (split_critical_edges): Declare.
5853         * tree-cfg.c (split_critical_edges): Export.
5854         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5856 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5858         * cfgexpand.c (expand_stack_vars): Optionally disable
5859         asan stack protection.
5860         (expand_used_vars): Likewise.
5861         (partition_stack_vars): Likewise.
5862         * asan.c (asan_emit_stack_protection): Optionally disable
5863         after return stack usage.
5864         (instrument_derefs): Optionally disable memory access instrumentation.
5865         (instrument_builtin_call): Likewise.
5866         (instrument_strlen_call): Likewise.
5867         (asan_protect_global): Optionally disable global variables protection.
5868         * doc/invoke.texi: Added doc for new options.
5869         * params.def: Added new options.
5870         * params.h: Likewise.
5872 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5874         PR rtl-optimization/59724
5875         * ifcvt.c (cond_exec_process_if_block): Don't call
5876         flow_find_head_matching_sequence with 0 longest_match.
5877         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5878         non-active insns if !stop_after.
5879         (try_head_merge_bb): Revert 2014-01-07 changes.
5881 2014-01-08  Jeff Law  <law@redhat.com>
5883         * ree.c (get_sub_rtx): New function, extracted from...
5884         (merge_def_and_ext): Here.
5885         (combine_reaching_defs): Use get_sub_rtx.
5887 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5889         * cgraph.h (varpool_variable_node): Do not choke on null node.
5891 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5893         * config/mips/mips.md (simple_return): Attempt to use JRC
5894         for microMIPS.
5895         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5897 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5899         PR rtl-optimization/59137
5900         * reorg.c (steal_delay_list_from_target): Call update_block for
5901         elided insns.
5902         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5904 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5906         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5907         two duplicate entries.
5909 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5911         Revert:
5912         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5914         * config/mips/mips.c (mips_truncated_op_cost): New function.
5915         (mips_rtx_costs): Adjust test for BADDU.
5916         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5918         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5920         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5921         (*baddu_si): ...this new pattern.
5923 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5925         PR ipa/59722
5926         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5928 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5930         PR middle-end/57748
5931         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5932         inner_reference_p.
5933         (expand_expr, expand_normal): Adjust.
5934         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5935         inner_reference_p. Use inner_reference_p to expand inner references.
5936         (store_expr): Adjust.
5937         * cfgexpand.c (expand_call_stmt): Adjust.
5939 2014-01-08  Rong Xu  <xur@google.com>
5941         * gcov-io.c (gcov_var): Move from gcov-io.h.
5942         (gcov_position): Ditto.
5943         (gcov_is_error): Ditto.
5944         (gcov_rewrite): Ditto.
5945         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5946         only part to libgcc/libgcov.h.
5948 2014-01-08  Marek Polacek  <polacek@redhat.com>
5950         PR middle-end/59669
5951         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5953 2014-01-08  Marek Polacek  <polacek@redhat.com>
5955         PR sanitizer/59667
5956         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5958 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5960         PR rtl-optimization/59649
5961         * stor-layout.c (get_mode_bounds): For BImode return
5962         0 and STORE_FLAG_VALUE.
5964 2014-01-08  Richard Biener  <rguenther@suse.de>
5966         PR middle-end/59630
5967         * gimple.h (is_gimple_builtin_call): Remove.
5968         (gimple_builtin_call_types_compatible_p): New.
5969         (gimple_call_builtin_p): New overload.
5970         * gimple.c (is_gimple_builtin_call): Remove.
5971         (validate_call): Rename to ...
5972         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5973         check return types.
5974         (validate_type): New static function.
5975         (gimple_call_builtin_p): New overload and adjust.
5976         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5977         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5978         (gimple_fold_stmt_to_constant_1): Likewise.
5979         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5981 2014-01-08  Richard Biener  <rguenther@suse.de>
5983         PR middle-end/59471
5984         * gimplify.c (gimplify_expr): Gimplify register-register type
5985         VIEW_CONVERT_EXPRs to separate stmts.
5987 2014-01-07  Jeff Law  <law@redhat.com>
5989         PR middle-end/53623
5990         * ree.c (combine_set_extension): Handle case where source
5991         and destination registers in an extension insn are different.
5992         (combine_reaching_defs): Allow source and destination registers
5993         in extension to be different under limited circumstances.
5994         (add_removable_extension): Remove restriction that the
5995         source and destination registers in the extension are the same.
5996         (find_and_remove_re): Emit a copy from the extension's
5997         destination to its source after the defining insn if
5998         the source and destination registers are different.
6000         PR middle-end/59285
6001         * ifcvt.c (merge_if_block): If we are merging a block with more than
6002         one successor with a block with no successors, remove any BARRIER
6003         after the second block.
6005 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
6007         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
6009 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
6011         PR target/59652
6012         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
6013         for 14-bit register offsets when INT14_OK_STRICT is false.
6015 2014-01-07  Roland Stigge  <stigge@antcom.de>
6016             Michael Meissner  <meissner@linux.vnet.ibm.com>
6018         PR 57386/target
6019         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
6020         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
6022 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
6024         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
6025         -mcpu.
6027 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
6029         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
6030         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
6031         rtx is const0_rtx or not.
6033 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
6035         PR target/58115
6036         * target-globals.c (save_target_globals): Remove this_fn_optab
6037         handling.
6038         * toplev.c: Include optabs.h.
6039         (target_reinit): Temporarily restore the global options if another
6040         set of options are in force.
6042 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
6044         PR rtl-optimization/58668
6045         * cfgcleanup.c (flow_find_cross_jump): Don't count
6046         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
6047         to determine what is counted.
6048         (flow_find_head_matching_sequence): Use active_insn_p to determine
6049         what is counted.
6050         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
6051         counting change.
6052         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
6053         determine what is counted.
6055         PR tree-optimization/59643
6056         * tree-predcom.c (split_data_refs_to_components): If one dr is
6057         read and one write, determine_offset fails and the write isn't
6058         in the bad component, just put the read into the bad component.
6060 2014-01-07  Mike Stump  <mikestump@comcast.net>
6061             Jakub Jelinek  <jakub@redhat.com>
6063         PR pch/59436
6064         * tree-core.h (struct tree_optimization_option): Change optabs
6065         type from unsigned char * to void *.
6066         * optabs.c (init_tree_optimization_optabs): Adjust
6067         TREE_OPTIMIZATION_OPTABS initialization.
6069 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
6071         PR target/59644
6072         * config/i386/i386.h (struct machine_function): Add
6073         no_drap_save_restore field.
6074         * config/i386/i386.c (ix86_save_reg): Use
6075         !cfun->machine->no_drap_save_restore instead of
6076         crtl->stack_realign_needed.
6077         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
6078         this function clears frame_pointer_needed.  Set
6079         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
6080         and DRAP reg is needed.
6082 2014-01-06  Marek Polacek  <polacek@redhat.com>
6084         PR c/57773
6085         * doc/implement-c.texi: Mention that other integer types are
6086         permitted as bit-field types in strictly conforming mode.
6088 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
6090         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
6091         is newly allocated.
6093 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
6095         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
6097 2014-01-06  Martin Jambor  <mjambor@suse.cz>
6099         PR ipa/59008
6100         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
6101         to int.
6102         * ipa-prop.c (ipa_print_node_params): Fix indentation.
6104 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
6106         PR debug/59350
6107         PR debug/59510
6108         * var-tracking.c (add_stores): Preserve the value of the source even if
6109         we don't record the store.
6111 2014-01-06  Terry Guo  <terry.guo@arm.com>
6113         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
6115 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
6117         PR bootstrap/59541
6118         * config/darwin.c (darwin_function_section): Adjust return values to
6119         correspond to optimisation changes made in r206070.
6121 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
6123         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
6124         from prefetch_block tune setting.
6125         (nocona_cost): Correct size of prefetch block to 64.
6127 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
6129         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
6130         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
6131         used to save the static chain register in the computation of the offset
6132         from which the FP registers need to be restored.
6134 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
6136         PR tree-optimization/59519
6137         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
6138         ICE if get_current_def (current_new_name) is already non-NULL, as long
6139         as it is a phi result of some other phi in *new_exit_bb that has
6140         the same argument.
6142         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
6143         or vmovdqu* for misaligned_operand.
6144         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
6145         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6146         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6147         aligned_mem for AVX512F masked aligned load and store builtins and for
6148         non-temporal moves.
6150 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6152         PR tree-optimization/59651
6153         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6154         Address range for negative step should be added by TYPE_SIZE_UNIT.
6156 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6158         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6159         overlapping registers also for registers other than the stack pointer.
6161 2014-01-03  Marek Polacek  <polacek@redhat.com>
6163         PR other/59661
6164         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6165         __builtin_FILE.
6167 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6169         PR target/59625
6170         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6171         asm goto as jump.
6173         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6174         (push splitter): Use <P:MODE_SIZE> instead of
6175         GET_MODE_SIZE (<P:MODE>mode).
6176         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6177         (mov -1, reg peephole2): Likewise.
6178         * config/i386/sse.md (*mov<mode>_internal,
6179         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6180         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6181         *<code><mode>3, *andnot<mode>3<mask_name>,
6182         <mask_codefor><code><mode>3<mask_name>): Likewise.
6183         * config/i386/subst.md (mask_mode512bit_condition,
6184         sd_mask_mode512bit_condition): Likewise.
6186 2014-01-02  Xinliang David Li  <davidxl@google.com>
6188         PR tree-optimization/59303
6189         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6190         (dump_predicates): Better output format.
6191         (pred_equal_p): New function.
6192         (is_neq_relop_p): Ditto.
6193         (is_neq_zero_form_p): Ditto.
6194         (pred_expr_equal_p): Ditto.
6195         (pred_neg_p): Ditto.
6196         (simplify_pred): Ditto.
6197         (simplify_preds_2): Ditto.
6198         (simplify_preds_3): Ditto.
6199         (simplify_preds_4): Ditto.
6200         (simplify_preds): Ditto.
6201         (push_pred): Ditto.
6202         (push_to_worklist): Ditto.
6203         (get_pred_info_from_cmp): Ditto.
6204         (is_degenerated_phi): Ditto.
6205         (normalize_one_pred_1): Ditto.
6206         (normalize_one_pred): Ditto.
6207         (normalize_one_pred_chain): Ditto.
6208         (normalize_preds): Ditto.
6209         (normalize_cond_1): Remove function.
6210         (normalize_cond): Ditto.
6211         (is_gcond_subset_of): Ditto.
6212         (is_subset_of_any): Ditto.
6213         (is_or_set_subset_of): Ditto.
6214         (is_and_set_subset_of): Ditto.
6215         (is_norm_cond_subset_of): Ditto.
6216         (pred_chain_length_cmp): Ditto.
6217         (convert_control_dep_chain_into_preds): Type change.
6218         (find_predicates): Ditto.
6219         (find_def_preds): Ditto.
6220         (destroy_predicates_vecs): Ditto.
6221         (find_matching_predicates_in_rest_chains): Ditto.
6222         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6223         (is_pred_expr_subset): Ditto.
6224         (is_pred_chain_subset_of): Ditto.
6225         (is_included_in): Ditto.
6226         (is_superset_of): Ditto.
6228 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6230         Update copyright years.
6232 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6234         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6235         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6236         config/arc/arc.md, config/arc/arc.opt,
6237         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6238         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6239         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6240         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6241         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6242         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6243         vtable-verify.c, vtable-verify.h: Use the standard form for the
6244         copyright notice.
6246 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6248         * gcc.c (process_command): Update copyright notice dates.
6249         * gcov-dump.c: Ditto.
6250         * gcov.c: Ditto.
6251         * doc/cpp.texi: Bump @copying's copyright year.
6252         * doc/cppinternals.texi: Ditto.
6253         * doc/gcc.texi: Ditto.
6254         * doc/gccint.texi: Ditto.
6255         * doc/gcov.texi: Ditto.
6256         * doc/install.texi: Ditto.
6257         * doc/invoke.texi: Ditto.
6259 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6261         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6263 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6265         * config/i386/sse.md (*mov<mode>_internal): Guard
6266         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6268         PR rtl-optimization/59647
6269         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6270         new_rtx into UNSIGNED_FLOAT rtxes.
6272 Copyright (C) 2014 Free Software Foundation, Inc.
6274 Copying and distribution of this file, with or without modification,
6275 are permitted in any medium without royalty provided the copyright
6276 notice and this notice are preserved.