2014-04-14 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob3991750539f18ee7501123eb5261a542baf625a4
1 2014-04-14  Richard Biener  <rguenther@suse.de>
3         PR middle-end/55022
4         * fold-const.c (negate_expr_p): Don't negate directional rounding
5         division.
6         (fold_negate_expr): Likewise.
8 2014-04-14  Richard Biener  <rguenther@suse.de>
10         PR tree-optimization/59817
11         PR tree-optimization/60453
12         * graphite-scop-detection.c (graphite_can_represent_scev): Complete
13         recursion to catch all CHRECs in the scalar evolution and restrict
14         the predicate for the remains appropriately.
16 2014-04-12  Catherine Moore  <clm@codesourcery.com>
18         * config/mips/constraints.md: Add new register constraint "kb".
19         * config/mips/mips.md (*mov<mode>_internal): Use constraint "kb".
20         (*movhi_internal): Likewise.
21         (*movqi_internal): Likewise.
22         * config/mips/mips.h (M16_STORE_REGS): New register class.
23         (REG_CLASS_NAMES): Add M16_STORE_REGS.
24         (REG_CLASS_CONTENTS): Likewise.
25         * config/mips/mips.c (mips_regno_to_class): Add M16_STORE_REGS.
27 2014-04-11  Tobias Burnus  <burnus@net-b.de>
29         PR c/60194
30         * doc/invoke.texi (-Wformat-signedness): Document it.
31         (Wformat=2): Mention that this enables -Wformat-signedness.
33 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
35         * common/config/epiphany/epiphany-common.c
36         (epiphany_option_optimization_table): Enable section anchors by
37         default at -O1 or higher.
38         * config/epiphany/epiphany.c (TARGET_MAX_ANCHOR_OFFSET): Define.
39         (TARGET_MIN_ANCHOR_OFFSET): Likewise.
40         (epiphany_rtx_costs) <SET>: For binary operators, the set as such
41         carries no extra cost.
42         (epiphany_legitimate_address_p): For BLKmode, apply SImode check.
43         * config/epiphany/epiphany.h (ASM_OUTPUT_DEF): Define.
44         * config/epiphany/predicates.md (memclob_operand): New predicate.
45         * config/epiphany/epiphany.md (stack_adjust_add, stack_adjust_str):
46         Use memclob_operand predicate and X constraint for operand 3.
48 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
50         * config/epiphany/epiphany.c (epiphany_rtx_cost): Compare
51         with CC_N_NE / CC_C_LTU / CC_C_GTU carries no extra cost for
52         its operands.
54 2014-04-11  Joern Rennecke  <joern.rennecke@embecosm.com>
56         PR rtl-optimization/60651
57         * mode-switching.c (optimize_mode_switching): Make sure to emit
58         sets of a lower numbered entity before sets of a higher numbered
59         entity to a mode of the same or lower priority.
60         When creating a seginfo for a basic block that starts with a code
61         label, move the insertion point past the code label.
62         (new_seginfo): Document and enforce requirement that
63         NOTE_INSN_BASIC_BLOCK only appears for empty blocks.
64         * doc/tm.texi.in: Document ordering constraint for emitted mode sets.
65         * doc/tm.texi: Regenerate.
67 2014-01-11  Joern Rennecke  <joern.rennecke@embecosm.com>
69         PR target/60811
70         * config/arc/arc.c (arc_save_restore): Fix assert typo.
72 2013-04-11  Jakub Jelinek  <jakub@redhat.com>
74         * BASE-VER: Set to 4.10.0.
76 2014-04-11  Tobias Burnus  <burnus@net-b.de>
78         PR other/59055
79         * doc/bugreport.texi (Bugs): Remove nodes pointing to the
80         nirvana.
81         * doc/gcc.texi (Service): Update description in the @menu
82         * doc/invoke.texi (Option Summary): Remove misplaced and
83         duplicated @menu.
85 2014-04-11  Steve Ellcey  <sellcey@mips.com>
86             Jakub Jelinek  <jakub@redhat.com>
88         PR middle-end/60556
89         * expr.c (convert_move): Use emit_store_flag_force instead of
90         emit_store_flag.  Pass lowpart_mode instead of VOIDmode as 5th
91         argument to it.
93 2014-04-11  Richard Biener  <rguenther@suse.de>
95         PR middle-end/60797
96         * varasm.c (assemble_alias): Avoid endless error reporting
97         recursion by setting TREE_ASM_WRITTEN.
99 2014-04-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
101         * config/s390/s390.md: Add a splitter for NOT rtx.
103 2014-04-11  Jakub Jelinek  <jakub@redhat.com>
105         PR rtl-optimization/60663
106         * cse.c (cse_insn): Set src_volatile on ASM_OPERANDS in
107         PARALLEL.
109 2014-04-10  Jan Hubicka  <hubicka@ucw.cz>
110             Jakub Jelinek  <jakub@redhat.com>
112         PR lto/60567
113         * ipa.c (function_and_variable_visibility): Copy forced_by_abi flag from
114         decl_node to node.
116 2014-04-10  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
118         PR debug/60655
119         * config/arm/arm.c (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Define
120         (arm_const_not_ok_for_debug_p): Reject MINUS with SYM_REF's
121         ameliorating the cases where it can be.
123 2014-04-09  David Edelsohn  <dje.gcc@gmail.com>
125         Revert
126         2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
128         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
129         (loadsync_<mode>): Change mode.
130         (load_quadpti, store_quadpti): New.
131         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
132         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
133         * config/rs6000/predicates.md (quad_memory_operand): !TARGET_SYNC_TI.
135 2014-04-09  Cong Hou  <congh@google.com>
137         PR testsuite/60773
138         * doc/sourcebuild.texi (vect_widen_mult_si_to_di_pattern): Add
139         documentation.
141 2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
143         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
144         instead of vnor to exploit possible fusion opportunity in the
145         future.
146         (altivec_expand_vec_perm_const_le): Likewise.
148 2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
150         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
151         (loadsync_<mode>): Change mode.
152         (load_quadpti, store_quadpti): New.
153         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
154         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
156 2014-04-08  Richard Sandiford  <rdsandiford@googlemail.com>
158         PR target/60763
159         * config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
160         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
161         Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.
163 2014-04-08  Richard Biener  <rguenther@suse.de>
165         PR middle-end/60706
166         * tree-pretty-print.c (pp_double_int): For HWI32 hosts with
167         a 64bit widest int print double-int similar to on HWI64 hosts.
169 2014-04-08  Richard Biener  <rguenther@suse.de>
171         PR tree-optimization/60785
172         * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
173         default defs properly.
175 2014-04-08  Nathan Sidwell  <nathan@codesourcery.com>
177         * doc/invoke (Wnon-virtual-dtor): Update to match implementation.
178         (Weffc++): Likewise.
180 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
182         * ipa-devirt.c (maybe_record_node): When node is not recorded,
183         set completep to false rather than true.
185 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
187         PR target/60504
188         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
189         ARM_TARGET2_DWARF_FORMAT.
191 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
193         PR target/60609
194         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
195         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
196         ADDR_DIFF_VEC.
198 2014-04-07  Richard Biener  <rguenther@suse.de>
200         PR tree-optimization/60766
201         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
202         (may_eliminate_iv): Convert cand_value_at result to desired type.
204 2014-04-07  Jason Merrill  <jason@redhat.com>
206         PR c++/60731
207         * common.opt (-fno-gnu-unique): Add.
208         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
210 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
212         * haifa-sched.c: Fix outdated function reference and minor
213         grammar errors in introductory comment.
215 2014-04-07  Richard Biener  <rguenther@suse.de>
217         PR middle-end/60750
218         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
219         for noreturn calls.
220         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
222 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
224         PR debug/55794
225         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
226         size accounting for thunks.
227         (pa_asm_output_mi_thunk): Use final_start_function() and
228         final_end_function() to output function start and end directives.
230 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
232         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
233         specific ISA/ feature information. Remove short_sp and errata_skip ds.
234         Add avr_device_specific_features enum to have device specific info.
235         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
236         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
237         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
238         updated device specific info.
239         * config/avr/avr-mcus.def: Merge device specific details to
240         dev_attribute field.
241         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
242         errata_skip.
243         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
244         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
245         assembler if RMW isa supported by current device.
246         * config/avr/genmultilib.awk: Update as device info structure changed.
247         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
249 2014-04-04  Cong Hou  <congh@google.com>
251         PR tree-optimization/60656
252         * tree-vect-stmts.c (supportable_widening_operation):
253         Fix a bug that elements in a vector with vect_used_by_reduction
254         property are incorrectly reordered when the operation on it is not
255         consistant with the one in reduction operation.
257 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
259         PR rtl-optimization/60155
260         * gcse.c (record_set_data): New function.
261         (single_set_gcse): New function.
262         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
263         (hoist_code): Likewise.
264         (get_pressure_class_and_nregs): Likewise.
266 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
268         * explow.c (probe_stack_range): Emit a final optimization blockage.
270 2014-04-04  Anthony Green  <green@moxielogic.com>
272         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
273         typos.
275 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
277         PR ipa/59626
278         * lto-cgraph.c (input_overwrite_node): Check that partitioning
279         flags are set only during streaming.
280         * ipa.c (process_references, walk_polymorphic_call_targets,
281         symtab_remove_unreachable_nodes): Drop bodies of always inline
282         after early inlining.
283         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
285 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
286         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
288         PR debug/60655
289         * dwarf2out.c (const_ok_for_output_1): Reject expressions
290         containing a NOT.
292 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
294         PR bootstrap/60743
295         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
296         duration.
297         (cortex_a53_fdivd): Likewise.
299 2014-04-04  Martin Jambor  <mjambor@suse.cz>
301         PR ipa/60640
302         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
303         Adjust all callers.
304         * cgraph.c (clone_of_p): Also return true if thunks match.
305         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
306         cgraph_function_or_thunk_node and an obsolete comment.
307         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
308         file.
309         (build_function_decl_skip_args): Likewise.
310         (set_new_clone_decl_and_node_flags): New function.
311         (duplicate_thunk_for_node): Likewise.
312         (redirect_edge_duplicating_thunks): Likewise.
313         (cgraph_clone_node): New parameter args_to_skip, pass it to
314         redirect_edge_duplicating_thunks which is called instead of
315         cgraph_redirect_edge_callee.
316         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
317         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
319 2014-04-04  Jeff Law  <law@redhat.com>
321         PR target/60657
322         * config/arm/predicates.md (const_int_I_operand): New predicate.
323         (const_int_M_operand): Similarly.
324         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
325         const_int_operand.
326         (insv_t2, extv_reg, extzv_t2): Likewise.
327         (load_multiple_with_writeback): Similarly for const_int_I_operand.
328         (pop_multiple_with_writeback_and_return): Likewise.
329         (vfp_pop_multiple_with_writeback): Likewise
331 2014-04-04  Richard Biener  <rguenther@suse.de>
333         PR ipa/60746
334         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
335         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
336         non-GIMPLE_LABELs.
337         * gimplify.h (gimple_add_tmp_var_fn): Declare.
338         * gimplify.c (gimple_add_tmp_var_fn): New function.
339         * gimple-expr.h (create_tmp_reg_fn): Declare.
340         * gimple-expr.c (create_tmp_reg_fn): New function.
341         * gimple-low.c (record_vars_into): Don't change cfun.
342         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
343         code generation without cfun.
345 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
347         PR bootstrap/60719
348         * Makefile.in (install-driver): Fix shell scripting.
350 2014-04-03  Cong Hou  <congh@google.com>
352         PR tree-optimization/60505
353         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
354         threshold of number of iterations below which no vectorization will be
355         done.
356         * tree-vect-loop.c (new_loop_vec_info):
357         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
358         * tree-vect-loop.c (vect_analyze_loop_operations):
359         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
360         * tree-vect-loop.c (vect_transform_loop):
361         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
362         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
363         of iterations of the loop and see if we should build the epilogue.
365 2014-04-03  Richard Biener  <rguenther@suse.de>
367         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
368         member.
369         (streamer_tree_cache_create): Adjust.
370         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
371         to allow optional nodes array.
372         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
373         (streamer_tree_cache_append): Likewise.
374         (streamer_tree_cache_create): Create nodes array optionally
375         as specified by parameter.
376         * lto-streamer-out.c (create_output_block): Avoid maintaining
377         the node array in the writer cache.
378         (DFS_write_tree): Remove assertion.
379         (produce_asm_for_decls): Free the out decl state hash table
380         early.
381         * lto-streamer-in.c (lto_data_in_create): Adjust for
382         streamer_tree_cache_create prototype change.
384 2014-04-03  Richard Biener  <rguenther@suse.de>
386         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
387         set TREE_CHAIN to NULL_TREE.
389 2014-04-03  Richard Biener  <rguenther@suse.de>
391         PR tree-optimization/60740
392         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
393         over all GIMPLE_COND operands.
395 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
397         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
398         (Weffc++): Remove Scott's numbering, merge lists and reference
399         Wnon-virtual-dtor.
401         c-family/
403         cp/
404         * class.c (accessible_nvdtor_p): New.
405         (check_bases): Don't check base destructor here ...
406         (check_bases_and_members): ... check them here.  Trigger on
407         Wnon-virtual-dtor flag.
408         (finish_struct_1): Use accessible_nvdtor_p.
410         testsuite/
411         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
412         * g++.dg/warn/Wnvdtor-2.C: New.
413         * g++.dg/warn/Wnvdtor-3.C: New.
414         * g++.dg/warn/Wnvdtor-4.C: New.
415         * g++.dg/warn/Weff1.C: Delete.
416         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
417         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
419 2014-04-03  Nick Clifton  <nickc@redhat.com>
421         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
422         properly.
424 2014-04-03  Martin Jambor  <mjambor@suse.cz>
426         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
427         mention gcc_unreachable before failing.
428         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
429         removed symbols.
431 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
433         PR ipa/60659
434         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
435         code and instead mark the context inconsistent.
436         (possible_polymorphic_call_targets): For inconsistent contexts
437         return empty complete list.
439 2014-04-02  Anthony Green  <green@moxielogic.com>
441         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
442         (extendqisi2, extendhisi2): Define.
443         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
444         (WCHAR_TYPE): Change to unsigned int.
446 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
448         PR tree-optimization/60733
449         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
450         insertion point for PHI candidates to be the end of the feeding
451         block for the PHI argument.
453 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
455         PR rtl-optimization/60650
456         * lra-constraints.c (process_alt_operands): Decrease reject for
457         earlyclobber matching.
459 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
461         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
463 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
465         * config/spu/spu.c (pad_bb): Do not crash when the last
466         insn is CODE_FOR_blockage.
468 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
470         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
471         lies outside the target mode.
473 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
475         PR target/60735
476         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
477         software floating point or no floating point registers, do not
478         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
479         in GPRs that occurs after we tested for GPRs that would never be
480         true.
482         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
483         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
484         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
485         specifically allow DDmode, since that does not use the SPE SIMD
486         instructions.
488 2014-04-02  Richard Biener  <rguenther@suse.de>
490         PR middle-end/60729
491         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
492         MODE_INTs.  Properly use negv_optab.
493         (expand_abs): Likewise.
495 2014-04-02  Richard Biener  <rguenther@suse.de>
497         PR bootstrap/60719
498         * Makefile.in (install-driver): Guard extra installs with special
499         names properly.
501 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
503         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
504         Document vec_vgbbd.
506 2014-04-01  Richard Henderson  <rth@redhat.com>
508         PR target/60704
509         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
510         alternative enabled before register allocation.
512 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
514         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
515         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
516         typo.
517         (nios2_large_got_address): Remove unneeded 'sym' parameter.
518         (nios2_got_address): Update nios2_large_got_address call site.
519         (nios2_delegitimize_address): New function.
520         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
521         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
522         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
524 2014-04-01  Martin Husemann  <martin@duskware.de>
526         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
527         for -mabi=32.
529 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
531         PR rtl-optimization/60604
532         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
533         check from register_operand.
534         (register_operand): Redefine in terms of general_operand.
535         (nonmemory_operand): Use register_operand for the non-constant cases.
537 2014-04-01  Richard Biener  <rguenther@suse.de>
539         * gimple.h (struct gimple_statement_base): Align subcode to
540         16 bits.
542 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
544         * doc/invoke.texi (mapp-regs): Clarify.
546 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
548         * config/i386/avx512fintrin.h (__v32hi): Define type.
549         (__v64qi): Likewise.
550         (_mm512_set1_epi8): Define.
551         (_mm512_set1_epi16): Define.
552         (_mm512_set4_epi32): Define.
553         (_mm512_set4_epi64): Define.
554         (_mm512_set4_pd): Define.
555         (_mm512_set4_ps): Define.
556         (_mm512_setr4_epi64): Define.
557         (_mm512_setr4_epi32): Define.
558         (_mm512_setr4_pd): Define.
559         (_mm512_setr4_ps): Define.
560         (_mm512_setzero_epi32): Define.
562 2014-03-31  Martin Jambor  <mjambor@suse.cz>
564         PR middle-end/60647
565         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
566         callsite_arguments_match_p.  Updated all callers.  Also check types of
567         corresponding formal parameters and actual arguments.
568         (not_all_callers_have_enough_arguments_p) Renamed to
569         some_callers_have_mismatched_arguments_p.
571 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
573         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
575 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
577         PR target/60034
578         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
579         section anchor.
581 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
583         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
584         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
585         Split out
586         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
587         Use FMAMODE_NOVF512 mode iterator.
588         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
589         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
590         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
591         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
592         Split out
593         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
594         Use VF_128_256 mode iterator.
595         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
596         Ditto.
598 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
600         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
601         static chain if needed.
603 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
605         PR target/60697
606         * lra-constraints.c (index_part_to_reg): New.
607         (process_address): Use it.
609 2014-03-27  Jeff Law  <law@redhat.com>
610             Jakub Jelinek  <jakub@redhat.com>
612         PR target/60648
613         * expr.c (do_tablejump): Use simplify_gen_binary rather than
614         gen_rtx_{PLUS,MULT} to build up the address expression.
616         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
617         creating non-canonical RTL.
619 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
621         PR ipa/60243
622         * ipa-inline.c (want_inline_small_function_p): Short circuit large
623         functions; reorganize to make cheap checks first.
624         (inline_small_functions): Do not estimate growth when dumping;
625         it is expensive.
626         * ipa-inline.h (inline_summary): Add min_size.
627         (growth_likely_positive): New function.
628         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
629         (set_cond_stmt_execution_predicate): Cleanup.
630         (estimate_edge_size_and_time): Compute min_size.
631         (estimate_calls_size_and_time): Likewise.
632         (estimate_node_size_and_time): Likewise.
633         (inline_update_overall_summary): Update min_size.
634         (do_estimate_edge_time): Likewise.
635         (do_estimate_edge_size): Update.
636         (do_estimate_edge_hints): Update.
637         (growth_likely_positive): New function.
639 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
641         PR target/60693
642         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
643         also if addr has VOIDmode.
645 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
647         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
648         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
649         Declare extern.
650         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
651         instructions as well as AdvancedSIMD loads.
653 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
655         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
656         Use crypto_aese type.
657         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
658         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
659         crypto_aese, crypto_aesmc.  Move to types.md.
660         * config/arm/types.md (crypto_aes): Split into crypto_aese,
661         crypto_aesmc.
662         * config/arm/iterators.md (crypto_type): Likewise.
664 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
666         * cgraph.c: Include expr.h and tree-dfa.h.
667         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
668         remove LHS.
670 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
672         PR target/60675
673         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
674         regs from checking multi-reg pseudos.
676 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
678         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
680 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
682         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
683         if it would clobber the stack pointer, even temporarily.
685 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
687         * mode-switching.c: Make small adjustments to the top comment.
689 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
691         * config/rs6000/constraints.md (wD constraint): New constraint to
692         match the constant integer to get the top DImode/DFmode out of a
693         vector in a VSX register.
695         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
696         match the constant integer to get the top DImode/DFmode out of a
697         vector in a VSX register.
699         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
700         for ISA 2.07.
702         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
703         vbpermq builtins.
705         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
706         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
708         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
709         Optimize vec_extract of 64-bit values, where the value being
710         extracted is in the top word, where we can use scalar
711         instructions.  Add direct move and store support.  Combine the big
712         endian/little endian vector select load support into a single insn.
713         (vsx_extract_<mode>_internal1): Likewise.
714         (vsx_extract_<mode>_internal2): Likewise.
715         (vsx_extract_<mode>_load): Likewise.
716         (vsx_extract_<mode>_store): Likewise.
717         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
718         combined into vsx_extract_<mode>_load.
719         (vsx_extract_<mode>_one_le): Likewise.
721         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
722         define the top 64-bit vector element.
724         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
725         constraint.
727         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
728         Document vec_vbpermq builtin.
730         PR target/60672
731         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
732         enable use of xxsldwi and xxpermdi builtin functions.
733         (vec_xxpermdi): Likewise.
735         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
736         Document use of vec_xxsldwi and vec_xxpermdi builtins.
738 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
740         PR rtl-optimization/60650
741         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
742         first_p.  Use it.
743         (find_spills_for): New.
744         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
745         Spill all pseudos on the second iteration.
747 2014-03-27  Marek Polacek  <polacek@redhat.com>
749         PR c/50347
750         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
751         types.
753 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
755         * config/s390/s390.c (s390_can_use_return_insn): Check for
756         call-saved FPRs on 31 bit.
758 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
760         PR middle-end/60682
761         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
762         if they need regimplification, just drop them instead of
763         calling gimple_regimplify_operands on them.
765 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
767         PR target/60580
768         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
769         (aarch64_frame_pointer_required): Adjust logic.
770         (aarch64_can_eliminate): Adjust logic.
771         (aarch64_override_options_after_change): Adjust logic.
773 2014-03-27  Dehao Chen  <dehao@google.com>
775         * ipa-inline.c (early_inliner): Update node's inline info.
777 2014-03-26  Dehao Chen  <dehao@google.com>
779         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
780         compiler inserted conditional jumps for NAN float check.
782 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
784         * ubsan.h (ubsan_create_data): Change second argument's type
785         to const location_t *.
786         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
787         _("<unknown>").
788         (ubsan_create_data): Change second argument to const location_t *PLOC.
789         Create Loc field whenever PLOC is non-NULL.
790         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
791         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
792         callers.
794         PR other/59545
795         * real.c (real_to_integer2): Change type of low to UHWI.
797 2014-03-26  Tobias Burnus  <burnus@net-b.de>
799         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
800         (CILK_SELF_SPECS): New define.
801         (driver_self_specs): Use it.
803 2014-03-26  Richard Biener  <rguenther@suse.de>
805         * tree-pretty-print.c (percent_K_format): Implement special
806         case for LTO and its stripped down BLOCK tree.
808 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
810         PR sanitizer/60636
811         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
813         * tree-vrp.c (simplify_internal_call_using_ranges): If only
814         one range is range_int_cst_p, but not both, at least optimize
815         addition/subtraction of 0 and multiplication by 0 or 1.
816         * gimple-fold.c (gimple_fold_call): Fold
817         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
818         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
819         INTEGER_CSTs, try to fold at least x * 0 and y - y.
821 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
823         PR rtl-optimization/60452
824         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
825         <case REG>: Return 1 for invalid offsets from the frame pointer.
827 2014-03-26  Marek Polacek  <polacek@redhat.com>
829         PR c/37428
830         * doc/extend.texi (C Extensions): Mention variable-length arrays in
831         a structure/union.
833 2014-03-26  Marek Polacek  <polacek@redhat.com>
835         PR c/39525
836         * doc/extend.texi (Designated Inits): Describe what happens to omitted
837         field members.
839 2014-03-26  Marek Polacek  <polacek@redhat.com>
841         PR other/59545
842         * ira-color.c (update_conflict_hard_regno_costs): Perform the
843         multiplication in unsigned type.
845 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
847         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
849 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
851         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
853 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
855         PR ipa/60315
856         * cif-code.def (UNREACHABLE) New code.
857         * ipa-inline.c (inline_small_functions): Skip edges to
858         __builtlin_unreachable.
859         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
860         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
861         predicate to __bulitin_unreachable.
862         (set_cond_stmt_execution_predicate): Fix issue when
863         invert_tree_comparison returns ERROR_MARK.
864         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
865         propagate to inline clones.
866         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
867         to unreachable.
868         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
869         * cgraphclones.c (cgraph_clone_node): If call destination is already
870         ureachable, do not redirect it back.
871         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
872         unreachable.
874 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
876         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
877         Do not modify inline clones.
879 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
881         * config/i386/i386.md (general_sext_operand): New mode attr.
882         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
883         don't generate (sign_extend (const_int)).
884         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
885         operands[2].  Use We constraint instead of <i> and
886         <general_sext_operand> predicate instead of <general_operand>.
887         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
888         * config/i386/constraints.md (We): New constraint.
889         * config/i386/predicates.md (x86_64_sext_operand,
890         sext_operand): New predicates.
892 2014-03-25  Martin Jambor  <mjambor@suse.cz>
894         PR ipa/60600
895         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
896         inconsistent devirtualizations to __builtin_unreachable.
898 2014-03-25  Marek Polacek  <polacek@redhat.com>
900         PR c/35449
901         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
903 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
905         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
906         order of elements for big-endian.
908 2014-03-25  Richard Biener  <rguenther@suse.de>
910         PR middle-end/60635
911         * gimplify-me.c (gimple_regimplify_operands): Update the
912         re-gimplifed stmt.
914 2014-03-25  Martin Jambor  <mjambor@suse.cz>
916         PR ipa/59176
917         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
918         (lto_output_varpool_node): Likewise.
919         (input_overwrite_node): Likewise.
920         (input_varpool_node): Likewise.
922 2014-03-25  Richard Biener  <rguenther@suse.de>
924         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
925         (run_gcc): Likewise.
927 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
929         * combine.c (simplify_compare_const): Add MODE argument.
930         Handle mode_width 0 as very large mode_width.
931         (try_combine, simplify_comparison): Adjust callers.
933         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
934         type to avoid signed integer overflow.
935         * explow.c (plus_constant): Likewise.
937 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
939         * doc/generic.texi: Correct typos.
941 2014-03-24  Tobias Burnus  <burnus@net-b.de>
943         * doc/invoke.texi (-flto): Expand section about
944         using static libraries with LTO.
946 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
948         PR rtl-optimization/60501
949         * optabs.def (addptr3_optab): New optab.
950         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
951         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
952         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
954         * lra.c (emit_add3_insn): Use the addptr pattern if available.
956         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
958 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
960         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
961         _mm512_set1_pd.
963         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
964         (_mm256_undefined_ps): Define.
965         (_mm256_undefined_pd): Define.
966         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
967         (_mm_undefined_pd): Define.
968         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
969         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
970         (_mm512_undefined_ps): Define.
971         (_mm512_undefined_pd): Define.
972         Use _mm*_undefined_*.
973         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
975 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
977         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
978         (lshr_simd): DI mode added.
979         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
980         (aarch64_ushr_simddi): Likewise.
981         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
982         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
983         (vshrd_n_u64): Likewise.
985 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
987         * Makefile.in (s-macro_list): Depend on cc1.
989 2014-03-23  Teresa Johnson  <tejohnson@google.com>
991         * ipa-utils.c (ipa_print_order): Use specified dump file.
993 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
995         PR rtl-optimization/60601
996         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
998         * gcc.c (eval_spec_function): Initialize save_growing_value.
1000 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
1002         PR sanitizer/60613
1003         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
1004         code == MINUS_EXPR, never swap op0 with op1.
1006         * toplev.c (init_local_tick): Avoid signed integer multiplication
1007         overflow.
1008         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
1009         shift by first operand's bitsize.
1011 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
1013         PR target/60610
1014         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
1015         redefine to 1 or 0.
1016         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
1017         TARGET_ISA_64BIT_P(x).
1019 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1021         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
1022         pattern for vector nor instead of subtract from splat(-1).
1023         (altivec_expand_vec_perm_const_le): Likewise.
1025 2014-03-21  Richard Henderson  <rth@twiddle.net>
1027         PR target/60598
1028         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
1029         related insns after epilogue_completed.
1031 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1033         PR ipa/59176
1034         * cgraph.h (symtab_node): New flag body_removed.
1035         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
1036         when removing bodies.
1037         * symtab.c (dump_symtab_base): Dump body_removed flag.
1038         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
1039         had their bodies removed.
1041 2014-03-21  Martin Jambor  <mjambor@suse.cz>
1043         PR ipa/60419
1044         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
1045         in the border.
1047 2014-03-21  Richard Biener  <rguenther@suse.de>
1049         PR tree-optimization/60577
1050         * tree-core.h (struct tree_base): Document nothrow_flag use
1051         in VAR_DECL_NONALIASED.
1052         * tree.h (VAR_DECL_NONALIASED): New.
1053         (may_be_aliased): Adjust.
1054         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
1056 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1058         * expr.c (expand_expr_real_1): Remove outdated comment.
1060 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
1062         PR middle-end/60597
1063         * ira.c (adjust_cleared_regs): Call copy_rtx on
1064         *reg_equiv[REGNO (loc)].src_p before passing it to
1065         simplify_replace_fn_rtx.
1067         PR target/60568
1068         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
1069         into CONST, put pic register as first operand of PLUS.  Use
1070         gen_const_mem for both 32-bit and 64-bit PIC got loads.
1072 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1074         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
1076 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1078         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
1079         around for store forwarding issue in the FPU on the UT699.
1080         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
1081         loads and operations if -mfix-ut699 is specified.
1082         (divtf3_hq): Tweak attribute.
1083         (sqrttf2_hq): Likewise.
1085 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1087         * calls.c (store_one_arg): Remove incorrect const qualification on the
1088         type of the temporary.
1089         * cfgexpand.c (expand_return): Likewise.
1090         * expr.c (expand_constructor): Likewise.
1091         (expand_expr_real_1): Likewise.
1093 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1095         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
1096         of parts.
1098 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
1100         PR target/60039
1101         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
1103 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
1105         * config/arm/aarch-common-protos.h
1106         (alu_cost_table): Fix spelling of "extend".
1107         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
1109 2014-03-19  Richard Biener  <rguenther@suse.de>
1111         PR middle-end/60553
1112         * tree-core.h (tree_type_common): Re-order pointer members
1113         to reduce recursion depth during GC walks.
1115 2014-03-19  Marek Polacek  <polacek@redhat.com>
1117         PR sanitizer/60569
1118         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
1119         before accessing it.
1121 2014-03-19  Richard Biener  <rguenther@suse.de>
1123         PR lto/59543
1124         * lto-streamer-in.c (input_function): In WPA stage do not drop
1125         debug stmts.
1127 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
1129         PR tree-optimization/60559
1130         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
1131         with build_zero_cst assignment.
1133 2014-03-18  Kai Tietz  <ktietz@redhat.com>
1135         PR rtl-optimization/56356
1136         * sdbout.c (sdbout_parms): Verify that parms'
1137         incoming argument is valid.
1138         (sdbout_reg_parms): Likewise.
1140 2014-03-18  Richard Henderson  <rth@redhat.com>
1142         PR target/60562
1143         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
1144         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
1145         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
1147 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
1149         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
1150         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
1151         Italicize plugin event names in description.  Explain that
1152         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
1153         Remind that no GCC functions should be called after PLUGIN_FINISH.
1154         Explain what pragmas with expansion are.
1156 2014-03-18  Martin Liska  <mliska@suse.cz>
1158         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
1159         gimple call statement is update.
1160         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
1161         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
1163 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1165         PR sanitizer/60557
1166         * ubsan.c (ubsan_instrument_unreachable): Call
1167         initialize_sanitizer_builtins.
1168         (ubsan_pass): Likewise.
1170         PR sanitizer/60535
1171         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1172         varpool_finalize_decl instead of rest_of_decl_compilation.
1174 2014-03-18  Richard Biener  <rguenther@suse.de>
1176         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1177         by using bitmap_and_compl instead of bitmap_and_compl_into.
1178         (df_rd_transfer_function): Likewise.
1180 2014-03-18  Richard Biener  <rguenther@suse.de>
1182         * doc/lto.texi (fresolution): Fix typo.
1184 2014-03-18  Richard Biener  <rguenther@suse.de>
1186         * doc/invoke.texi (flto): Update for changes in 4.9.
1188 2014-03-18  Richard Biener  <rguenther@suse.de>
1190         * doc/loop.texi: Remove section on the removed lambda framework.
1191         Update loop docs with recent changes in preserving loop structure.
1193 2014-03-18  Richard Biener  <rguenther@suse.de>
1195         * doc/lto.texi (-fresolution): Document.
1197 2014-03-18  Richard Biener  <rguenther@suse.de>
1199         * doc/contrib.texi: Adjust my name.
1201 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1203         PR ipa/58721
1204         * internal-fn.c: Include diagnostic-core.h.
1205         (expand_BUILTIN_EXPECT): New function.
1206         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1207         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1208         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1209         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1210         IFN_BUILTIN_EXPECT.
1211         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1212         Revert 3 argument __builtin_expect code.
1213         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1214         * gimple-fold.c (gimple_fold_call): Likewise.
1215         * tree.h (fold_builtin_expect): New prototype.
1216         * builtins.c (build_builtin_expect_predicate): Add predictor
1217         argument, if non-NULL, create 3 argument __builtin_expect.
1218         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1219         pass it through to build_builtin_expect_predicate.
1220         (fold_builtin_2): Adjust caller.
1221         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1222         * internal-fn.def (BUILTIN_EXPECT): New.
1224 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1226         PR ipa/58721
1227         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1228         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1229         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1231 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1233         PR ipa/58721
1234         * predict.c (combine_predictions_for_bb): Fix up formatting.
1235         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1236         fill what it points to if non-NULL.
1237         (tree_predict_by_opcode): Adjust caller, use the predictor.
1238         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1240 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1242         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1243         proper constant for the store mode.
1245 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1247         * symtab.c (change_decl_assembler_name): Fix transparent alias
1248         chain construction.
1250 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1252         * config/aarch64/aarch64.c: Correct the comments about the
1253         aarch64 stack layout.
1255 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1257         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1258         check for GF_OMP_FOR_KIND_FOR.
1260 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1262         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1263         ymm and zmm register names.
1265 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1267         PR target/60516
1268         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1269         note creation for the 2010-08-31 changes.
1271 2014-03-17  Marek Polacek  <polacek@redhat.com>
1273         PR middle-end/60534
1274         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1275         as -fno-tree-loop-vectorize.
1276         (expand_omp_simd): Likewise.
1278 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1280         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1281         (eligible_for_call_delay): New prototype.
1282         * config/sparc/sparc.c (tls_call_delay): Rename into...
1283         (eligible_for_call_delay): ...this.  Return false if the instruction
1284         cannot be put in the delay slot of a branch.
1285         (eligible_for_restore_insn): Simplify.
1286         (eligible_for_return_delay): Return false if the instruction cannot be
1287         put in the delay slot of a branch and simplify.
1288         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1289         put in the delay slot of a branch.
1290         * config/sparc/sparc.md (fix_ut699): New attribute.
1291         (tls_call_delay): Delete.
1292         (in_call_delay): Reimplement.
1293         (eligible_for_sibcall_delay): Rename into...
1294         (in_sibcall_delay): ...this.
1295         (eligible_for_return_delay): Rename into...
1296         (in_return_delay): ...this.
1297         (in_branch_delay): Reimplement.
1298         (in_uncond_branch_delay): Delete.
1299         (in_annul_branch_delay): Delete.
1301 2014-03-14  Richard Henderson  <rth@redhat.com>
1303         PR target/60525
1304         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1305         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1306         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1307         (floathi splitters): Remove.
1308         (float<SWI48x>xf2): New pattern.
1309         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1310         code that tried to handle DImode for 32-bit, but which was excluded
1311         by the pattern's condition.  Drop allocation of stack temporary.
1312         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1313         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1314         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1315         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1316         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1317         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1318         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1319         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1320         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1321         (*float<SWI48x><X87MODEF>2_i387): Remove.
1322         (all float _with_temp splitters): Remove.
1323         (*float<SWI48x><MODEF>2_i387): New pattern.
1324         (*float<SWI48><MODEF>2_sse): New pattern.
1325         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1326         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1328 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1329             Marek Polacek  <polacek@redhat.com>
1331         PR middle-end/60484
1332         * common.opt (dump_base_name_prefixed): New Variable.
1333         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1334         if x_dump_base_name_prefixed is already set, set it at the end.
1336 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1338         PR rtl-optimization/60508
1339         * lra-constraints.c (get_reload_reg): Add new parameter
1340         in_subreg_p.
1341         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1342         Pass the new parameter values.
1344 2014-03-14  Richard Biener  <rguenther@suse.de>
1346         * common.opt: Revert unintented changes from r205065.
1347         * opts.c: Likewise.
1349 2014-03-14  Richard Biener  <rguenther@suse.de>
1351         PR middle-end/60518
1352         * cfghooks.c (split_block): Properly adjust all loops the
1353         block was a latch of.
1355 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1357         PR lto/60461
1358         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1359         and simplify it.
1361 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1363         PR target/59396
1364         * config/avr/avr.c (avr_set_current_function): Pass function name
1365         through default_strip_name_encoding before sanity checking instead
1366         of skipping the first char of the assembler name.
1368 2014-03-13  Richard Henderson  <rth@redhat.com>
1370         PR debug/60438
1371         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1372         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1373         * config/i386/i386-protos.h: Likewise.
1374         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1375         in the expander instead of a splitter.
1376         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1377         any possibility of requiring a memory.
1378         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1379         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1380         (fp branch splitters): Update for ix86_split_fp_branch.
1381         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1382         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1383         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1384         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1385         (*fop_<MODEF>_3_i387): Likewise.
1386         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1387         (splitters for the fop_* register patterns): Remove.
1388         (fscalexf4_i387): Rename from *fscalexf4_i387.
1389         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1391 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1393         PR tree-optimization/59779
1394         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1395         type for bitsize and maxsize instead of HOST_WIDE_INT.
1397 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1399         PR rtl-optimization/57320
1400         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1401         the CFG after thread_prologue_and_epilogue_insns.
1403 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1405         PR rtl-optimization/57189
1406         * lra-constraints.c (process_alt_operands): Disfavor spilling
1407         vector pseudos.
1409 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1411         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1413 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1415         PR tree-optimization/59025
1416         PR middle-end/60418
1417         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1418         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1420 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1422         PR target/60486
1423         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1424         calls of avr_out_plus_1.
1426 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1428         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1429         BB's single pred and update the father loop's latch info later.
1431 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1433         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1434         (VEC_M): Likewise.
1435         (VEC_N): Likewise.
1436         (VEC_R): Likewise.
1437         (VEC_base): Likewise.
1438         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1439         registers, we need to swap double words in little endian mode.
1441         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1442         to be a container mode for 128-bit integer operations added in ISA
1443         2.07.  Unlike TImode and PTImode, the preferred register set is
1444         the Altivec/VMX registers for the 128-bit operations.
1446         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1447         declarations.
1448         (rs6000_split_128bit_ok_p): Likewise.
1450         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1451         macros for creating ISA 2.07 normal and overloaded builtin
1452         functions with 3 arguments.
1453         (BU_P8V_OVERLOAD_3): Likewise.
1454         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1455         for use as overloaded functions.
1456         (VPERM_1TI_UNS): Likewise.
1457         (VSEL_1TI): Likewise.
1458         (VSEL_1TI_UNS): Likewise.
1459         (ST_INTERNAL_1ti): Likewise.
1460         (LD_INTERNAL_1ti): Likewise.
1461         (XXSEL_1TI): Likewise.
1462         (XXSEL_1TI_UNS): Likewise.
1463         (VPERM_1TI): Likewise.
1464         (VPERM_1TI_UNS): Likewise.
1465         (XXPERMDI_1TI): Likewise.
1466         (SET_1TI): Likewise.
1467         (LXVD2X_V1TI): Likewise.
1468         (STXVD2X_V1TI): Likewise.
1469         (VEC_INIT_V1TI): Likewise.
1470         (VEC_SET_V1TI): Likewise.
1471         (VEC_EXT_V1TI): Likewise.
1472         (EQV_V1TI): Likewise.
1473         (NAND_V1TI): Likewise.
1474         (ORC_V1TI): Likewise.
1475         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1476         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1477         overloaded builtin.
1478         (VADDUQM): Likewise.
1479         (VSUBCUQ): Likewise.
1480         (VADDEUQM): Likewise.
1481         (VADDECUQ): Likewise.
1482         (VSUBEUQM): Likewise.
1483         (VSUBECUQ): Likewise.
1485         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1486         __int128_t and __uint128_t types.
1487         (__uint128_type): Likewise.
1488         (altivec_categorize_keyword): Add support for vector __int128_t,
1489         vector __uint128_t, vector __int128, and vector unsigned __int128
1490         as a container type for TImode operations that need to be done in
1491         VSX/Altivec registers.
1492         (rs6000_macro_to_expand): Likewise.
1493         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1494         to support 128-bit integer instructions vaddcuq, vadduqm,
1495         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1496         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1498         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1499         for V1TImode, and set up preferences to use VSX/Altivec registers.
1500         Setup VSX reload handlers.
1501         (rs6000_debug_reg_global): Likewise.
1502         (rs6000_init_hard_regno_mode_ok): Likewise.
1503         (rs6000_preferred_simd_mode): Likewise.
1504         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1505         (easy_altivec_constant): Likewise.
1506         (output_vec_const_move): Likewise.
1507         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1508         simple move.
1509         (rs6000_expand_vector_extract): Likewise.
1510         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1511         addressing.
1512         (rs6000_const_vec): Add support for V1TImode.
1513         (rs6000_emit_le_vsx_load): Swap double words when loading or
1514         storing TImode/V1TImode.
1515         (rs6000_emit_le_vsx_store): Likewise.
1516         (rs6000_emit_le_vsx_move): Likewise.
1517         (rs6000_emit_move): Add support for V1TImode.
1518         (altivec_expand_ld_builtin): Likewise.
1519         (altivec_expand_st_builtin): Likewise.
1520         (altivec_expand_vec_init_builtin): Likewise.
1521         (altivec_expand_builtin): Likewise.
1522         (rs6000_init_builtins): Add support for V1TImode type.  Add
1523         support for ISA 2.07 128-bit integer builtins.  Define type names
1524         for the VSX/Altivec vector types.
1525         (altivec_init_builtins): Add support for overloaded vector
1526         functions with V1TImode type.
1527         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1528         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1529         external function.
1530         (rs6000_split_128bit_ok_p): Likewise.
1531         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1532         __int128_t and vector __uint128_t.
1534         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1535         and mode attributes.
1536         (VSX_M): Likewise.
1537         (VSX_M2): Likewise.
1538         (VSm): Likewise.
1539         (VSs): Likewise.
1540         (VSr): Likewise.
1541         (VSv): Likewise.
1542         (VS_scalar): Likewise.
1543         (VS_double): Likewise.
1544         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1546         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1547         we support the ISA 2.07 128-bit integer arithmetic instructions.
1548         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1549         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1550         and TImode types for use with the builtin functions.
1551         (V1TI_type_node): Likewise.
1552         (unsigned_V1TI_type_node): Likewise.
1553         (intTI_type_internal_node): Likewise.
1554         (uintTI_type_internal_node): Likewise.
1556         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1557         128-bit builtin functions.
1558         (UNSPEC_VADDEUQM): Likewise.
1559         (UNSPEC_VADDECUQ): Likewise.
1560         (UNSPEC_VSUBCUQ): Likewise.
1561         (UNSPEC_VSUBEUQM): Likewise.
1562         (UNSPEC_VSUBECUQ): Likewise.
1563         (VM): Add V1TImode to vector mode iterators.
1564         (VM2): Likewise.
1565         (VI_unit): Likewise.
1566         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1567         (altivec_vaddcuq): Likewise.
1568         (altivec_vsubuqm): Likewise.
1569         (altivec_vsubcuq): Likewise.
1570         (altivec_vaddeuqm): Likewise.
1571         (altivec_vaddecuq): Likewise.
1572         (altivec_vsubeuqm): Likewise.
1573         (altivec_vsubecuq): Likewise.
1575         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1576         mode iterators.
1577         (BOOL_128): Likewise.
1578         (BOOL_REGS_OUTPUT): Likewise.
1579         (BOOL_REGS_OP1): Likewise.
1580         (BOOL_REGS_OP2): Likewise.
1581         (BOOL_REGS_UNARY): Likewise.
1582         (BOOL_REGS_AND_CR0): Likewise.
1584         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1585         128-bit integer builtin support.
1586         (vec_vadduqm): Likewise.
1587         (vec_vaddecuq): Likewise.
1588         (vec_vaddeuqm): Likewise.
1589         (vec_vsubecuq): Likewise.
1590         (vec_vsubeuqm): Likewise.
1591         (vec_vsubcuq): Likewise.
1592         (vec_vsubuqm): Likewise.
1594         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1595         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1596         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1597         128-bit integer add/subtract to ISA 2.07.
1599 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1601         * config/arc/arc.c (arc_predicate_delay_insns):
1602         Fix third argument passed to conditionalize_nonjump.
1604 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1606         * config/aarch64/aarch64-builtins.c
1607         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1608         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1609         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1610         instead of __builtin_lfloor.
1611         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1613 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1615         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1616         (tree_ssa_ifcombine_bb_1): New function.
1617         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1618         is an empty forwarder block to then_bb or vice versa and then_bb
1619         and else_bb are effectively swapped.
1621 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1623         PR target/60264
1624         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1625         REG_CFA_DEF_CFA note.
1626         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1627         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1629 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1631         PR tree-optimization/60454
1632         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1634 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1636         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1637         Do not define target_cpu_default2 to generic.
1638         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1639         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1640         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1642 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1643             Marc Glisse  <marc.glisse@inria.fr>
1645         PR tree-optimization/60502
1646         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1647         instead of build_low_bits_mask.
1649 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1651         PR middle-end/60482
1652         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1653         if there are multiple uses, but op doesn't live on E edge.
1654         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1655         clobber stmts before __builtin_unreachable.
1657 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1659         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1660         hard_frame_pointer_rtx.
1661         * cse.c (cse_insn): Remove volatile check.
1662         * cselib.c (cselib_process_insn): Likewise.
1663         * dse.c (scan_insn): Likewise.
1665 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1667         * config/arc/arc.c (conditionalize_nonjump): New function,
1668         broken out of ...
1669         (arc_ifcvt): ... this.
1670         (arc_predicate_delay_insns): Use it.
1672 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1674         * config/arc/predicates.md (extend_operand): During/after reload,
1675         allow const_int_operand.
1676         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1677         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1678         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1679         to "i".
1680         (umulsi3_highpart_i): Likewise.
1682 2014-03-11  Richard Biener  <rguenther@suse.de>
1684         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1685         Add asserts to guard possible wrong-code bugs.
1687 2014-03-11  Richard Biener  <rguenther@suse.de>
1689         PR tree-optimization/60429
1690         PR tree-optimization/60485
1691         * tree-ssa-structalias.c (set_union_with_increment): Properly
1692         take into account all fields that overlap the shifted vars.
1693         (do_sd_constraint): Likewise.
1694         (do_ds_constraint): Likewise.
1695         (get_constraint_for_ptr_offset): Likewise.
1697 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1699         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1700         (nios2_compute_frame_layout):
1701         Add calculation of cfun->machine->fp_save_offset.
1702         (nios2_expand_prologue): Correct setting of frame pointer register
1703         in prologue.
1704         (nios2_expand_epilogue): Update recovery of stack pointer from
1705         frame pointer accordingly.
1706         (nios2_initial_elimination_offset): Update calculation of offset
1707         for eliminating to HARD_FRAME_POINTER_REGNUM.
1709 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1711         PR ipa/60457
1712         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1713         cgraph_get_create_node on VAR_DECLs.
1715 2014-03-10  Richard Biener  <rguenther@suse.de>
1717         PR middle-end/60474
1718         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1720 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1722         * config/vms/vms.opt (vms_float_format): New variable.
1724 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1726         * doc/invoke.texi (-fcilkplus): Update implementation status.
1728 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1729             Richard Biener  <rguenther@suse.de>
1731         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1732         consistently accross all TUs.
1733         (run_gcc): Enable -fshort-double automatically at link at link-time
1734         and disallow override.
1736 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1738         PR target/58271
1739         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1740         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1741         if they can't be used.
1743 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1745         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1746         for Solaris 11/x86 ld.
1747         * configure: Regenerate.
1749 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1751         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1752         (LIB_TLS_SPEC): Save as ld_tls_libs.
1753         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1754         (HAVE_AS_IX86_TLSLDM): New test.
1755         * configure, config.in: Regenerate.
1756         * config/i386/i386.c (legitimize_tls_address): Fall back to
1757         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1758         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1759         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1760         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1762 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1764         * common.opt (fira-loop-pressure): Mark as optimization.
1766 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1768         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1769         an OpenMP mappable type.
1771 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1773         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1774         MULTILIB_OSDIRNAMES is not defined.
1776 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1777             Meador Inge  <meadori@codesourcery.com>
1779         PR target/58595
1780         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1781         (arm_legitimize_address): Call legitimize_tls_address for any
1782         arm_tls_referenced_p expression, handle constant addend.  Call it
1783         before testing for !TARGET_ARM.
1784         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1786 2014-03-06  Richard Biener  <rguenther@suse.de>
1788         PR middle-end/60445
1789         PR lto/60424
1790         PR lto/60427
1791         Revert
1792         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1794         * tree-streamer.c (record_common_node): Assert we don't record
1795         nodes with type double.
1796         (preload_common_node): Skip type double, complex double and double
1797         pointer since it is now frontend dependent due to fshort-double option.
1799 2014-03-06  Richard Biener  <rguenther@suse.de>
1801         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1802         or -fno-lto is specified and the linker has full plugin support.
1803         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1804         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1805         * lto-wrapper.c (merge_and_complain): Merge compile-time
1806         optimization levels.
1807         (run_gcc): And pass it through to the link options.
1809 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1811         PR debug/60381
1812         Revert:
1813         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1814         PR debug/59992
1815         * cselib.c (remove_useless_values): Skip to avoid quadratic
1816         behavior if the condition moved from...
1817         (cselib_process_insn): ... here holds.
1819 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1821         PR plugins/59335
1822         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1823         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1825         PR plugins/59335
1826         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1827         (TM_H): Add x86-tune.def.
1829 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1831         * config/aarch64/aarch64.c (generic_tunings):
1832         Use cortexa57_extra_costs.
1834 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1836         PR lto/60404
1837         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1838         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1839         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1840         cost for in_lto_p.
1842 2014-03-04  Heiher  <r@hev.cc>
1844         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1845         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1847 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1849         * config/i386/predicates.md (const2356_operand): Change to ...
1850         (const2367_operand): ... this.
1851         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1852         const2367_operand.
1853         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1854         (*avx512pf_scatterpf<mode>sf): Ditto.
1855         (avx512pf_scatterpf<mode>df): Ditto.
1856         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1857         (*avx512pf_scatterpf<mode>df): Ditto.
1858         * config/i386/i386.c (ix86_expand_builtin): Update
1859         incorrect hint operand error message.
1861 2014-03-04  Richard Biener  <rguenther@suse.de>
1863         * lto-section-in.c (lto_get_section_data): Fix const cast.
1865 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1867         * tree-streamer.c (record_common_node): Assert we don't record
1868         nodes with type double.
1869         (preload_common_node): Skip type double, complex double and double
1870         pointer since it is now frontend dependent due to fshort-double option.
1872 2014-03-04  Richard Biener  <rguenther@suse.de>
1874         PR lto/60405
1875         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1876         (lto_input_toplevel_asms): Likewise.
1877         * lto-section-in.c (lto_get_section_data): Instead do it here
1878         for every section.
1880 2014-03-04  Richard Biener  <rguenther@suse.de>
1882         PR tree-optimization/60382
1883         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1884         dead PHIs a reduction.
1886 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1888         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1889         hint value.
1890         (_mm_prefetch): Move out of GCC target("sse") pragma.
1891         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1892         GCC target("prfchw") pragma.
1893         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1894         for locality <= 2.
1895         * config/i386/i386.c (ix86_option_override_internal): Enable
1896         -mprfchw with -mprefetchwt1.
1898 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1900         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1901         Mark as varying.
1903 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1905         * opts.h (CL_PCH_IGNORE): Define.
1906         * targhooks.c (option_affects_pch_p):
1907         Return false for options that have CL_PCH_IGNORE set.
1908         * opt-functions.awk: Process PchIgnore.
1909         * doc/options.texi: Document PchIgnore.
1911         * config/arc/arc.opt (misize): Add PchIgnore property.
1913 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1915         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1916         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1917         constraint on constants to permit them being loaded into
1918         GENERAL_REGS or BASE_REGS.
1920 2014-03-03  Nick Clifton  <nickc@redhat.com>
1922         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1923         anti-cacnonical alternatives.
1924         (negandhi3_real): New pattern.
1925         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1927 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1929         * config/avr/avr-mcus.def: Remove atxmega16x1.
1930         * config/avr/avr-tables.opt: Regenerate.
1931         * config/avr/t-multilib: Regenerate.
1932         * doc/avr-mmcu.texi: Regenerate.
1934 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1935             Mircea Namolaru  <mircea.namolaru@inria.fr>
1937         PR tree-optimization/58028
1938         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1939         scalar dimensions.
1941 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1943         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1944         not handled by recognizers.
1946 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1948         PR middle-end/60175
1949         * function.c (expand_function_end): Don't emit
1950         clobber_return_register sequence if clobber_after is a BARRIER.
1951         * cfgexpand.c (construct_exit_block): Append instructions before
1952         return_label to prev_bb.
1954 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1956         * config/rs6000/constraints.md: Document reserved use of "wc".
1958 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1960         PR ipa/60150
1961         * ipa.c (function_and_variable_visibility): When dissolving comdat
1962         group, also set all symbols to local.
1964 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1966         PR ipa/60306
1968         Revert:
1969         2013-12-14   Jan Hubicka  <jh@suse.cz>
1970         PR middle-end/58477
1971         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1973 2014-03-02  Jon Beniston  <jon@beniston.com>
1975         PR bootstrap/48230
1976         PR bootstrap/50927
1977         PR bootstrap/52466
1978         PR target/46898
1979         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1980           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1981         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1982         (simple_return, *simple_return): New patterns
1983         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1984         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1986 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1988         * dwarf2out.c (gen_subprogram_die): Tidy.
1990 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1992         PR target/60071
1993         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1994         (*mov_t_msb_neg_negc): ... this new insn.
1996 2014-02-28  Jason Merrill  <jason@redhat.com>
1998         PR c++/58678
1999         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
2000         function.
2002 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
2004         PR c++/60314
2005         * dwarf2out.c (decltype_auto_die): New static.
2006         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
2007         (gen_type_die_with_usage): Handle 'decltype(auto)'.
2008         (is_cxx_auto): Likewise.
2010 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
2012         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
2013         we are not using general regs only.
2015 2014-02-28  Richard Biener  <rguenther@suse.de>
2017         PR target/60280
2018         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
2019         previous fix and only allow to remove trivial pre-headers
2020         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
2021         (remove_forwarder_block): Properly update the latch of a loop.
2023 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2025         PR debug/59992
2026         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
2027         (cselib_preserved_hash_table): New.
2028         (preserve_constants_and_equivs): Move preserved vals to it.
2029         (cselib_find_slot): Look it up first.
2030         (cselib_init): Initialize it.
2031         (cselib_finish): Release it.
2032         (dump_cselib_table): Dump it.
2034 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2036         PR debug/59992
2037         * cselib.c (remove_useless_values): Skip to avoid quadratic
2038         behavior if the condition moved from...
2039         (cselib_process_insn): ... here holds.
2041 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
2043         PR debug/57232
2044         * var-tracking.c (vt_initialize): Apply the same condition to
2045         preserve the CFA base value.
2047 2014-02-28  Joey Ye  <joey.ye@arm.com>
2049         PR target/PR60169
2050         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
2051         if reload in progress or completed.
2053 2014-02-28  Tobias Burnus  <burnus@net-b.de>
2055         PR middle-end/60147
2056         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
2057         NAMELIST_DECL.
2059 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
2061         * doc/tm.texi.in (Condition Code Status): Update documention for
2062         relative locations of cc0-setter and cc0-user.
2064 2014-02-27  Jeff Law  <law@redhat.com>
2066         PR rtl-optimization/52714
2067         * combine.c (try_combine): When splitting an unrecognized PARALLEL
2068         into two independent simple sets, if I3 is a jump, ensure the
2069         pattern we place into I3 is a (set (pc) ...).
2071 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
2072             Jeff Law  <law@redhat.com>
2074         PR rtl-optimization/49847
2075         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
2076         are in different blocks.
2077         * doc/tm.texi (Condition Code Status): Update documention for
2078         relative locations of cc0-setter and cc0-user.
2080 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
2082         PR target/59222
2083         * lra.c (lra_emit_add): Check SUBREG too.
2085 2014-02-27  Andreas Schwab  <schwab@suse.de>
2087         * config/m68k/m68k.c (m68k_option_override): Disable
2088         -flive-range-shrinkage for classic m68k.
2089         (m68k_override_options_after_change): Likewise.
2091 2014-02-27  Marek Polacek  <polacek@redhat.com>
2093         PR middle-end/59223
2094         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
2095         -Wmaybe-uninitialized.
2097 2014-02-27  Alan Modra  <amodra@gmail.com>
2099         PR target/57936
2100         * reload1.c (emit_input_reload_insns): When reload_override_in,
2101         set old to rl->in_reg when rl->in_reg is a subreg.
2103 2014-02-26  Richard Biener  <rguenther@suse.de>
2105         PR bootstrap/60343
2106         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
2108 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2110         * common/config/i386/predicates.md (const1256_operand): Remove.
2111         (const2356_operand): New.
2112         (const_1_to_2_operand): Remove.
2113         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2114         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2115         (*avx512pf_gatherpf<mode>sf): Ditto.
2116         (avx512pf_gatherpf<mode>df): Ditto.
2117         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2118         (*avx512pf_gatherpf<mode>df): Ditto.
2119         (avx512pf_scatterpf<mode>sf): Ditto.
2120         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2121         (*avx512pf_scatterpf<mode>sf): Ditto.
2122         (avx512pf_scatterpf<mode>df): Ditto.
2123         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2124         (*avx512pf_scatterpf<mode>df): Ditto.
2125         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
2127 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
2129         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
2130         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
2131         (_mm512_mask_testn_epi64_mask): Move to ...
2132         * config/i386/avx512cdintrin.h: Here.
2133         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
2134         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
2135         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
2136         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
2137         TARGET_AVX512F from TARGET_AVX512CD.
2139 2014-02-26  Richard Biener  <rguenther@suse.de>
2141         PR ipa/60327
2142         * ipa.c (walk_polymorphic_call_targets): Properly guard
2143         call to inline_update_overall_summary.
2145 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
2147         PR target/60280
2148         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
2149         and latches only if requested.  Fix latch if it is removed.
2150         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
2151         LOOPS_HAVE_PREHEADERS.
2153 2014-02-25  Andrew Pinski  <apinski@cavium.com>
2155         * builtins.c (expand_builtin_thread_pointer): Create a new target
2156         when the target is NULL.
2158 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
2160         PR rtl-optimization/60317
2161         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2162         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2163         * lra-assigns.c: Include params.h.
2164         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
2165         other reload pseudos considerations.
2167 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2169         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
2170         to use canonical form for nor<mode>3.
2172 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2174         PR target/55426
2175         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2176         conversions.
2178 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2180         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2181         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2182         (ix86_handle_option): Handle OPT_mprefetchwt1.
2183         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2184         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2185         PREFETCHWT1 CPUID.
2186         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2187         OPTION_MASK_ISA_PREFETCHWT1.
2188         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2189         (PTA_PREFETCHWT1): New.
2190         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2191         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2192         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2193         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2194         (*prefetch_avx512pf_<mode>_: Change into ...
2195         (*prefetch_prefetchwt1_<mode>: This.
2196         * config/i386/i386.opt (mprefetchwt1): New.
2197         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2198         (_mm_prefetch): Handle intent to write.
2199         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2201 2014-02-25  Richard Biener  <rguenther@suse.de>
2203         PR middle-end/60291
2204         * emit-rtl.c (mem_attrs_htab): Remove.
2205         (mem_attrs_htab_hash): Likewise.
2206         (mem_attrs_htab_eq): Likewise.
2207         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2208         (init_emit_once): Do not allocate mem_attrs_htab.
2210 2014-02-25  Richard Biener  <rguenther@suse.de>
2212         PR lto/60319
2213         * lto-opts.c (lto_write_options): Output non-explicit conservative
2214         -fwrapv, -fno-trapv and -fno-strict-overflow.
2215         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2216         (run_gcc): And pass them through.
2218 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2220         * sel-sched.c (calculate_new_fences): New parameter ptime.
2221         Calculate it as a maximum over all fence cycles.
2222         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2223         Print the final schedule timing when sched_verbose.
2225 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2227         PR rtl-optimization/60292
2228         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2229         bit fot the fence instruction.
2231 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2233         * calls.h: Fix typo in comment.
2235 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2237         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2238         adjusting offsetable addresses.
2240 2014-02-24  Guozhi Wei  <carrot@google.com>
2242         * sparseset.h (sparseset_pop): Fix the wrong index.
2244 2014-02-24  Walter Lee  <walt@tilera.com>
2246         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2247         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2248         triplet.
2249         * common/config/tilegx/tilegx-common.c
2250         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2251         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2252         (LINK_SPEC): Ditto.
2253         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2254         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2255         (tilegx_gimplify_va_arg_expr): Handle big endian.
2256         (tilegx_expand_unaligned_load): Ditto.
2257         (tilegx_expand_unaligned_store): Ditto.
2258         (TARGET_RETURN_IN_MSB): New.
2259         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2260         (TARGET_ENDIAN_DEFAULT): New.
2261         (TARGET_BIG_ENDIAN): Handle big endian.
2262         (BYTES_BIG_ENDIAN): Ditto.
2263         (WORDS_BIG_ENDIAN): Ditto.
2264         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2265         (ENDIAN_SPEC): New.
2266         (EXTRA_SPECS): New.
2267         * config/tilegx/tilegx.md (extv): Handle big endian.
2268         (extzv): Ditto.
2269         (insn_st<n>): Ditto.
2270         (insn_st<n>_add<bitsuffix>): Ditto.
2271         (insn_stnt<n>): Ditto.
2272         (insn_stnt<n>_add<bitsuffix>):Ditto.
2273         (vec_interleave_highv8qi): Handle big endian.
2274         (vec_interleave_highv8qi_be): New.
2275         (vec_interleave_highv8qi_le): New.
2276         (insn_v1int_h): Handle big endian.
2277         (vec_interleave_lowv8qi): Handle big endian.
2278         (vec_interleave_lowv8qi_be): New.
2279         (vec_interleave_lowv8qi_le): New.
2280         (insn_v1int_l): Handle big endian.
2281         (vec_interleave_highv4hi): Handle big endian.
2282         (vec_interleave_highv4hi_be): New.
2283         (vec_interleave_highv4hi_le): New.
2284         (insn_v2int_h): Handle big endian.
2285         (vec_interleave_lowv4hi): Handle big endian.
2286         (vec_interleave_lowv4hi_be): New.
2287         (vec_interleave_lowv4hi_le): New.
2288         (insn_v2int_l): Handle big endian.
2289         (vec_interleave_highv2si): Handle big endian.
2290         (vec_interleave_highv2si_be): New.
2291         (vec_interleave_highv2si_le): New.
2292         (insn_v4int_h): Handle big endian.
2293         (vec_interleave_lowv2si): Handle big endian.
2294         (vec_interleave_lowv2si_be): New.
2295         (vec_interleave_lowv2si_le): New.
2296         (insn_v4int_l): Handle big endian.
2297         * config/tilegx/tilegx.opt (mbig-endian): New option.
2298         (mlittle-endian): New option.
2299         * doc/install.texi: Document tilegxbe-linux.
2300         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2302 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2304         PR ipa/60266
2305         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2306         there are no parameter descriptors.
2308 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2310         PR rtl-optimization/60268
2311         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2312         initialization to ...
2313         (sched_rgn_init): ... here.
2314         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2316 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2318         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2319         names.
2321 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2323         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2324         definition.
2326 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2328         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2329         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2331 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2333         * config/microblaze/predicates.md: Add cmp_op predicate.
2334         * config/microblaze/microblaze.md: Add branch_compare instruction
2335         which uses cmp_op predicate and emits cmp insn before branch.
2336         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2337         to microblaze_expand_conditional_branch and consolidate logic.
2338         (microblaze_expand_conditional_branch): emit branch_compare
2339         insn instead of handling cmp op separate from branch insn.
2341 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2343         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2344         to permit subregs.
2346 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2348         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2349         define_insn with define_expand and new define_insn
2350         *altivec_lve<VI_char>x_internal.
2351         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2352         and new define_insn *altivec_stve<VI_char>x_internal.
2353         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2354         prototype.
2355         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2356         lve*x built-ins.
2357         (altivec_expand_stvex_be): New function.
2359 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2361         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2362         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2363         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2364         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2366 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2368         PR target/60298
2369         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2370         instead of emit_move_insn.
2372 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2374         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2375         vspltw with vsldoi.
2376         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2377         gen_altivec_vsumsws.
2379 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2381         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2382         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2383         (altivec_lvxl_<mode>): New define_expand incorporating
2384         -maltivec=be semantics where needed.
2385         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2386         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2387         semantics where needed.
2388         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2389         (altivec_stvx_<mode>): New define_expand incorporating
2390         -maltivec=be semantics where needed.
2391         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2392         VM2 iterator instead of V4SI.
2393         (altivec_stvxl_<mode>): New define_expand incorporating
2394         -maltivec=be semantics where needed.
2395         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2396         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2397         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2398         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2399         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2400         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2401         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2402         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2403         ALTIVEC_BUILTIN_STVXL.
2404         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2405         (altivec_expand_stvx_be): Likewise.
2406         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2407         (altivec_expand_lvx_be): Likewise.
2408         (altivec_expand_stvx_be): Likewise.
2409         (altivec_expand_builtin): Add cases for
2410         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2411         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2412         (altivec_init_builtins): Add definitions for
2413         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2414         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2416 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2418         * doc/invoke.texi (mvirt, mno-virt): Document.
2419         * config/mips/mips.opt (mvirt): New option.
2420         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2422 2014-02-21  Richard Biener  <rguenther@suse.de>
2424         PR tree-optimization/60276
2425         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2426         (STMT_VINFO_MIN_NEG_DIST): New macro.
2427         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2428         STMT_VINFO_MIN_NEG_DIST.
2429         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2430         made for negative dependence distances still hold.
2432 2014-02-21  Richard Biener  <rguenther@suse.de>
2434         PR middle-end/60291
2435         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2436         DECL_INITIAL for globals not in the current function context.
2438 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2440         PR tree-optimization/56490
2441         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2442         * tree-ssa-uninit.c: Include params.h.
2443         (compute_control_dep_chain): Add num_calls argument, return false
2444         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2445         num_calls to recursive call.
2446         (find_predicates): Change dep_chain into normal array,
2447         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2448         variable and adjust compute_control_dep_chain caller.
2449         (find_def_preds): Likewise.
2451 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2453         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2454         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2456 2014-02-21  Nick Clifton  <nickc@redhat.com>
2458         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2459         (pushhi1): Likewise.
2460         (popqi1): Add mode to pre_dec.
2461         (pophi1): Likewise.
2463 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2465         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2466         mode for mask of V8SFmode permutation.
2468 2014-02-20  Richard Henderson  <rth@redhat.com>
2470         PR c++/60272
2471         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2472         a new pseudo for OLDVAL.
2474 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2476         PR target/57896
2477         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2478         gen_reg_rtx if d->testing_p.
2479         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2480         if d->testing_p and we will certainly return true.
2481         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2482         if d->testing_p.
2484 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2486         * emit-rtl.c (gen_reg_rtx): Assert that
2487         crtl->emit.regno_pointer_align_length is non-zero.
2489 2014-02-20  Richard Henderson  <rth@redhat.com>
2491         PR c++/60272
2492         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2493         on failure the store back into EXPECT.
2495 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2496             Sandra Loosemore  <sandra@codesourcery.com>
2498         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2499         * config/nios2/nios2.c (nios2_function_profiler): Add
2500         -fPIC (flag_pic == 2) support.
2501         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2502         (nios2_large_offset_p): New function.
2503         (nios2_unspec_reloc_p): Move up position, update to use
2504         nios2_large_offset_p.
2505         (nios2_unspec_address): Remove function.
2506         (nios2_unspec_offset): New function.
2507         (nios2_large_got_address): New function.
2508         (nios2_got_address): Add large offset support.
2509         (nios2_legitimize_tls_address): Update usage of removed and new
2510         functions.
2511         (nios2_symbol_binds_local_p): New function.
2512         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2513         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2514         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2515         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2516         processing for (const (unspec ...)).
2517         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2519 2014-02-20  Richard Biener  <rguenther@suse.de>
2521         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2522         doing the substitution.
2523         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2525 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2527         PR ipa/55260
2528         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2529         info when checking whether lattices are bottom.
2531 2014-02-20  Richard Biener  <rguenther@suse.de>
2533         PR middle-end/60221
2534         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2535         regions at -O0.
2537 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2539         PR ipa/58555
2540         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2541         parameter specifying the scaling.
2542         (inline_call): Update.
2543         (want_inline_recursively): Guard division by zero.
2544         (recursive_inlining): Update.
2545         * ipa-inline.h (clone_inlined_nodes): Update.
2547 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2549         PR target/60204
2550         * config/i386/i386.c (classify_argument): Pass structures of size
2551         64 bytes or less in register.
2553 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2554             Kirill Yukhin  <kirill.yukhin@intel.com>
2556         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2557         (_mm_rcp28_round_ss): Ditto.
2558         (_mm_rsqrt28_round_sd): Ditto.
2559         (_mm_rsqrt28_round_ss): Ditto.
2560         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2561         (_mm_rcp14_round_ss): Ditto.
2562         (_mm_rsqrt14_round_sd): Ditto.
2563         (_mm_rsqrt14_round_ss): Ditto.
2564         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2565         the first input operand, get rid of match_dup.
2566         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2567         attribute to sse.
2568         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2569         Ditto.
2570         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2571         operand as the first input operand, set type attribute.
2572         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2573         Set type attribute.
2574         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2575         operand as the first input operand, set type attribute.
2577 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2579         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2580         bit of zero.
2582 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2584         PR target/60207
2585         * config/i386/i386.c (construct_container): Remove TFmode check
2586         for X86_64_INTEGER_CLASS.
2588 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2590         PR target/59794
2591         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2592         only when -Wpsabi is enabled.
2594 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2596          PR target/59799
2597         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2598         passing arrays in registers are the same as for structs, so remove the
2599         special case for them.
2601 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2603         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2604         destination type, extract only the valid bits if the source type is not
2605         integral and has a different mode.
2607 2014-02-19  Richard Biener  <rguenther@suse.de>
2609         PR ipa/60243
2610         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2611         for all calls.
2613 2014-02-19  Richard Biener  <rguenther@suse.de>
2615         PR ipa/60243
2616         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2617         (ipa_modify_call_arguments): Emit an argument load explicitely and
2618         preserve virtual SSA form there and for the replacement call.
2619         Do not update SSA form nor free dominance info.
2621 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2623         * ipa.c (function_and_variable_visibility): Also clear WEAK
2624         flag when disolving COMDAT_GROUP.
2626 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2628         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2629         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2630         not devirtualizing.
2631         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2632         do more sanity checks.
2633         (detect_type_change): Return true when giving up early.
2634         (compute_complex_assign_jump_func): Fix type parameter of
2635         ipa_set_ancestor_jf.
2636         (compute_complex_ancestor_jump_func): Likewise.
2637         (update_jump_functions_after_inlining): Fix updating of
2638         ancestor function.
2639         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2641 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2643         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2644         inline clones when edge disappears.
2646 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2648         PR target/60203
2649         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2650         Split 64-bit moves into 2 patterns.  Do not allow the use of
2651         direct move for TDmode in little endian, since the decimal value
2652         has little endian bytes within a word, but the 64-bit pieces are
2653         ordered in a big endian fashion, and normal subreg's of TDmode are
2654         not allowed.
2655         (mov<mode>_64bit_dm): Likewise.
2656         (movtd_64bit_nodm): Likewise.
2658 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2660         PR tree-optimization/60174
2661         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2662         statement of an SSA_NAME that occurs in an abnormal PHI node.
2664 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2666         PR sanitizer/60142
2667         * final.c (SEEN_BB): Remove.
2668         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2669         (final_scan_insn): Don't force_source_line on second
2670         NOTE_INSN_BASIC_BLOCK.
2672 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2674         PR target/60205
2675         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2676         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2677         (type_natural_mode): Warn ABI change when %zmm register is not
2678         available for AVX512F vector value passing.
2680 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2682         PR target/60193
2683         * config/i386/i386.c (ix86_expand_prologue): Use value in
2684         rax register as displacement when restoring %r10 or %rax.
2685         Fix wrong offset when restoring both registers.
2687 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2689         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2690         assertion with conditional return.
2692 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2693             Uros Bizjak  <ubizjak@gmail.com>
2695         PR driver/60233
2696         * config/i386/driver-i386.c (host_detect_local_cpu): If
2697         YMM state is not saved by the OS, also clear has_f16c.  Move
2698         CPUID 0x80000001 handling before YMM state saving checking.
2700 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2702         PR rtl-optimization/58960
2703         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2704         factored out from ...
2705         (sched_init): ... here.
2706         (free_global_sched_pressure_data): New, factored out from ...
2707         (sched_finish): ... here.
2708         * sched-int.h (free_global_sched_pressure_data): Declare.
2709         * sched-rgn.c (nr_regions_initial): New static global.
2710         (haifa_find_rgns): Initialize it.
2711         (schedule_region): Disable sched-pressure for the newly
2712         generated regions.
2714 2014-02-17  Richard Biener  <rguenther@suse.de>
2716         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2717         release SSA defs of pattern stmts.
2719 2014-02-17  Richard Biener  <rguenther@suse.de>
2721         * tree-inline.c (expand_call_inline): Release the virtual
2722         operand defined by the call we are about to inline.
2724 2014-02-17  Richard Biener  <rguenther@suse.de>
2726         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2728 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2729             Ilya Tocar  <ilya.tocar@intel.com>
2731         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2732         arguments order in builtin.
2733         (_mm512_permutexvar_epi64): Ditto.
2734         (_mm512_mask_permutexvar_epi64): Ditto
2735         (_mm512_maskz_permutexvar_epi32): Ditto
2736         (_mm512_permutexvar_epi32): Ditto
2737         (_mm512_mask_permutexvar_epi32): Ditto
2739 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2741         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2742         (p8_vmrgow): Likewise.
2744 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2746         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2747         endian targets.
2749 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2751         PR target/60203
2752         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2753         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2754         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2755         using direct move instructions on ISA 2.07.  Also adjust
2756         instruction length for 64-bit.
2757         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2758         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2760 2014-02-15  Alan Modra  <amodra@gmail.com>
2762         PR target/58675
2763         PR target/57935
2764         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2765         find_replacement on parts of insn rtl that might be reloaded.
2767 2014-02-15  Richard Biener  <rguenther@suse.de>
2769         PR tree-optimization/60183
2770         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2771         (tree_ssa_phiprop): Calculate and free post-dominators.
2773 2014-02-14  Jeff Law  <law@redhat.com>
2775         PR rtl-optimization/60131
2776         * ree.c (get_extended_src_reg): New function.
2777         (combine_reaching_defs): Use it rather than assuming location of REG.
2778         (find_and_remove_re): Verify first operand of extension is
2779         a REG before adding the insns to the copy list.
2781 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2783         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2784         * configure: Regenerated.
2785         * config.in: Regenerated.
2786         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2787         instead of ASM_SHORT.
2789 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2790             Richard Earnshaw  <rearnsha@arm.com>
2792         PR rtl-optimization/59535
2793         * lra-constraints.c (process_alt_operands): Encourage alternative
2794         when unassigned pseudo class is superset of the alternative class.
2795         (inherit_reload_reg): Don't inherit when optimizing for code size.
2796         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2797         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2798         modes not less than 4 for Thumb1.
2800 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2802         PR pch/60010
2803         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2805 2014-02-14  Richard Biener  <rguenther@suse.de>
2807         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2808         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2809         Do not strip INDIRECT_REFs.
2811 2014-02-14  Richard Biener  <rguenther@suse.de>
2813         PR lto/60179
2814         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2815         DECL_FUNCTION_SPECIFIC_TARGET.
2816         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2817         * tree-streamer-out.c (pack_ts_target_option): Remove.
2818         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2819         (write_ts_function_decl_tree_pointers): Do not stream
2820         DECL_FUNCTION_SPECIFIC_TARGET.
2821         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2822         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2823         (lto_input_ts_function_decl_tree_pointers): Do not stream
2824         DECL_FUNCTION_SPECIFIC_TARGET.
2826 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2828         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2829         (get_initial_def_for_induction, vectorizable_induction): Ignore
2830         debug stmts when looking for exit_phi.
2831         (vectorizable_live_operation): Fix up condition.
2833 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2835         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2836         nreverse() because it changes the content of original tree list.
2838 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2840         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2841         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2843 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2845         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2846         GNU coding standards.
2848 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2850         PR debug/60152
2851         * dwarf2out.c (gen_subprogram_die): Don't call
2852         add_calling_convention_attribute if subr_die is old_die.
2854 2014-02-13  Sharad Singhai  <singhai@google.com>
2856         * doc/optinfo.texi: Fix order of nodes.
2858 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2860         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2861         operands[2], not operands[3].
2863 2014-02-13  Richard Biener  <rguenther@suse.de>
2865         PR bootstrap/59878
2866         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2867         mention the possibility of an in-tree build.
2868         (CLooG): Update recommended version to 0.18.1, mention the
2869         possibility of an in-tree build and clarify that the ISL
2870         bundled with CLooG does not work.
2872 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2874         PR target/43546
2875         * expr.c (compress_float_constant): If x is a hard register,
2876         extend into a pseudo and then move to x.
2878 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2880         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2881         caused by bad second argument to warning_at() with -mhotpatch and
2882         nested functions (e.g. with gfortran).
2884 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2886         * opts.c (option_name): Remove "enabled by default" rider.
2888 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2890         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2892 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2893             Uros Bizjak  <ubizjak@gmail.com>
2895         PR target/60151
2896         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2897         * configure: Regenerated.
2899 2014-02-12  Richard Biener  <rguenther@suse.de>
2901         * vec.c (vec_prefix::calculate_allocation): Move as
2902         inline variant to vec.h.
2903         (vec_prefix::calculate_allocation_1): New out-of-line version.
2904         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2905         (vec_prefix::m_has_auto_buf): Rename to ...
2906         (vec_prefix::m_using_auto_storage): ... this.
2907         (vec_prefix::calculate_allocation): Inline the easy cases
2908         and dispatch to calculate_allocation_1 which doesn't need the
2909         prefix address.
2910         (va_heap::reserve): Use gcc_checking_assert.
2911         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2912         m_using_auto_storage.
2913         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2914         member and adjust.
2915         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2916         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2917         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2919 2014-02-12  Richard Biener  <rguenther@suse.de>
2921         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2922         when we found a dependence.
2924 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2926         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2927         common code...
2928         (maybe_fold_stmt): ... into this new function.
2929         * omp-low.c (lower_omp): Update comment.
2931         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2932         last use.
2934         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2935         dereference.
2937 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2939         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2940         identifiers in comments.
2941         (cortexa53_extra_costs): Likewise.
2942         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2943         (cortexa7_extra_costs): Likewise.
2944         (cortexa12_extra_costs): Likewise.
2945         (cortexa15_extra_costs): Likewise.
2946         (v7m_extra_costs): Likewise.
2948 2014-02-12  Richard Biener  <rguenther@suse.de>
2950         PR middle-end/60092
2951         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2952         of posix_memalign being successful.
2953         (lower_stmt): Restrict lowering of posix_memalign to when
2954         -ftree-bit-ccp is enabled.
2956 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2958         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2959         arg_loc.
2960         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2962 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2964         PR rtl-optimization/60116
2965         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2966         other_insn once the combination has been validated.
2968 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2970         PR lto/59468
2971         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2972         and wrapper.
2973         * ipa-devirt.c: Include demangle.h
2974         (odr_violation_reported): New static variable.
2975         (add_type_duplicate): Update odr_violations.
2976         (maybe_record_node): Add completep parameter; update it.
2977         (record_target_from_binfo): Add COMPLETEP parameter;
2978         update it as needed.
2979         (possible_polymorphic_call_targets_1): Likewise.
2980         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2981         rename FINAL to COMPLETE.
2982         (record_targets_from_bases): Sanity check we found the binfo;
2983         fix COMPLETEP updating.
2984         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2985         parameter, fix computing of COMPLETEP.
2986         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2987         at LTO time do demangling.
2988         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2989         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2990         parameter.
2991         (gimple_get_virt_method_for_binfo): Likewise.
2992         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2993         gimple_get_virt_method_for_vtable): Update prototypes.
2995 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2997         PR target/49008
2998         * genautomata.c (add_presence_absence): Fix typo with
2999         {final_}presence_list.
3001 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
3003         PR target/60137
3004         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
3005         for VSX/Altivec vectors that land in GPR registers.
3007 2014-02-11  Richard Henderson  <rth@redhat.com>
3008             Jakub Jelinek  <jakub@redhat.com>
3010         PR debug/59776
3011         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
3012         around drhs if type conversion to lacc->type is not useless.
3014 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3016         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
3017         tuning struct.
3018         (cortex-a57.cortex-a53): Likewise.
3019         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
3021 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3023         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
3024         arm_restrict_it.
3026 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
3028         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
3029         add_options_for_arm_vfp3.
3031 2014-02-11  Jeff Law  <law@redhat.com>
3033         PR middle-end/54041
3034         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
3035         object with an undesirable mode.
3037 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3039         PR libgomp/60107
3040         * config/i386/sol2-9.h: New file.
3041         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
3042         *-*-solaris2.9*): Use it.
3044 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3046         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
3047         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
3049 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
3051         * config/microblaze/microblaze.c: Extend mcpu version format
3053 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
3055         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
3057 2014-02-10  Richard Henderson  <rth@redhat.com>
3059         PR target/59927
3060         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
3061         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
3062         ms-abi vs -mno-accumulate-outgoing-args.
3063         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
3064         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
3065         respect to ms-abi.
3067 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3069         PR middle-end/60080
3070         * cfgexpand.c (expand_asm_operands): Attach source location to
3071         ASM_INPUT rtx objects.
3072         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
3074 2014-02-10  Nick Clifton  <nickc@redhat.com>
3076         * config/mn10300/mn10300.c (popcount): New function.
3077         (mn10300_expand_prologue): Include saved registers in stack usage
3078         count.
3080 2014-02-10  Jeff Law  <law@redhat.com>
3082         PR middle-end/52306
3083         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
3084         when changing the SET_DEST of a prior insn to avoid an input reload.
3086 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
3088         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
3089         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
3090         -mcall-openbsd, or -mcall-linux.
3091         (CC1_ENDIAN_BIG_SPEC): Remove.
3092         (CC1_ENDIAN_LITTLE_SPEC): Remove.
3093         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3094         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
3095         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
3096         and %cc1_endian_default.
3097         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3099 2014-02-10  Richard Biener  <rguenther@suse.de>
3101         PR tree-optimization/60115
3102         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3103         MEM_REF handling.  Properly verify that the accesses are not
3104         out of the objects bound.
3106 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3108         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
3109         coretex to cortex.
3111 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
3113         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
3114         proper constants and fix formatting.
3115         (possible_polymorphic_call_targets): Fix formatting.
3117 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
3118             Ilya Tocar  <ilya.tocar@intel.com>
3120         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
3121         (_mm512_loadu_epi32): Renamed into...
3122         (_mm512_loadu_si512): This.
3123         (_mm512_storeu_epi32): Renamed into...
3124         (_mm512_storeu_si512): This.
3125         (_mm512_maskz_ceil_ps): Removed.
3126         (_mm512_maskz_ceil_pd): Ditto.
3127         (_mm512_maskz_floor_ps): Ditto.
3128         (_mm512_maskz_floor_pd): Ditto.
3129         (_mm512_floor_round_ps): Ditto.
3130         (_mm512_floor_round_pd): Ditto.
3131         (_mm512_ceil_round_ps): Ditto.
3132         (_mm512_ceil_round_pd): Ditto.
3133         (_mm512_mask_floor_round_ps): Ditto.
3134         (_mm512_mask_floor_round_pd): Ditto.
3135         (_mm512_mask_ceil_round_ps): Ditto.
3136         (_mm512_mask_ceil_round_pd): Ditto.
3137         (_mm512_maskz_floor_round_ps): Ditto.
3138         (_mm512_maskz_floor_round_pd): Ditto.
3139         (_mm512_maskz_ceil_round_ps): Ditto.
3140         (_mm512_maskz_ceil_round_pd): Ditto.
3141         (_mm512_expand_pd): Ditto.
3142         (_mm512_expand_ps): Ditto.
3143         * config/i386/i386.c (ix86_builtins): Remove
3144         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
3145         (bdesc_args): Ditto.
3146         * config/i386/predicates.md (const1256_operand): New.
3147         (const_1_to_2_operand): Ditto.
3148         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
3149         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3150         (*avx512pf_gatherpf<mode>sf): Ditto.
3151         (avx512pf_gatherpf<mode>df): Ditto.
3152         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3153         (*avx512pf_gatherpf<mode>df): Ditto.
3154         (avx512pf_scatterpf<mode>sf): Ditto.
3155         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3156         (*avx512pf_scatterpf<mode>sf): Ditto.
3157         (avx512pf_scatterpf<mode>df): Ditto.
3158         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3159         (*avx512pf_scatterpf<mode>df): Ditto.
3160         (avx512f_expand<mode>): Removed.
3161         (<shift_insn><mode>3<mask_name>): Change predicate type.
3163 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3165         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
3166         not at the end of datarefs vector use ordered_remove to avoid
3167         reordering datarefs vector.
3169         PR c/59984
3170         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
3171         mark local addressable non-static vars as GOVD_PRIVATE
3172         instead of GOVD_LOCAL.
3173         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3174         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3175         than copying them.
3177         PR middle-end/60092
3178         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3179         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3180         assume_aligned or alloc_align attributes.
3181         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3182         arguments.  Handle also assume_aligned and alloc_align attributes.
3183         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3184         calls to functions with assume_aligned or alloc_align attributes.
3185         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3187 2014-02-08  Terry Guo  <terry.guo@arm.com>
3189         * doc/invoke.texi: Document ARM -march=armv7e-m.
3191 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3193         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3194         flag on __cilkrts_rethrow builtin.
3196         PR ipa/60026
3197         * ipa-cp.c (determine_versionability): Fail at -O0
3198         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3199         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3201         Revert:
3202         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3204         PR ipa/60026
3205         * tree-inline.c (copy_forbidden): Fail for
3206         __attribute__((optimize (0))) functions.
3208 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3210         * varpool.c: Include pointer-set.h.
3211         (varpool_remove_unreferenced_decls): Variables in other partitions
3212         will not be output; be however careful to not lose information
3213         about partitioning.
3215 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3217         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3218         lookup in the vtable constructor.
3220 2014-02-07  Jeff Law  <law@redhat.com>
3222         PR target/40977
3223         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3224         define_insn_and_split.
3226         * ipa-inline.c (inline_small_functions): Fix typos.
3228 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3230         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3231         (s390_can_use_return_insn): Declare.
3232         * config/s390/s390.h (EPILOGUE_USES): Define.
3233         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3234         instructions.
3235         (s390_chunkify_start): Handle return JUMP_LABELs.
3236         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3237         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3238         (s390_can_use_return_insn): New functions.
3239         (s390_fix_long_loop_prediction): Handle conditional returns.
3240         (TARGET_SET_UP_BY_PROLOGUE): Define.
3241         * config/s390/s390.md (ANY_RETURN): New code iterator.
3242         (*creturn, *csimple_return, return, simple_return): New patterns.
3244 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3246         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3247         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3248         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3249         REG_CFA_RESTORE list when deciding not to restore a register.
3251 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3253         * config/s390/s390.c: Include tree-pass.h and context.h.
3254         (s390_early_mach): New function, split out from...
3255         (s390_emit_prologue): ...here.
3256         (pass_data_s390_early_mach): New pass structure.
3257         (pass_s390_early_mach): New class.
3258         (s390_option_override): Create and register early_mach pass.
3259         Move to end of file.
3261 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3263         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3264         to match for the exit block.
3266 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3268         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3269         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3270         Reject misaligned operands.
3272 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3274         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3276 2014-02-07  Richard Biener  <rguenther@suse.de>
3278         PR middle-end/60092
3279         * gimple-low.c (lower_builtin_posix_memalign): New function.
3280         (lower_stmt): Call it to lower posix_memalign in a way
3281         to make alignment info accessible.
3283 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3285         PR c++/60082
3286         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3287         __builtin_setjmp_receiver.
3289 2014-02-07  Richard Biener  <rguenther@suse.de>
3291         PR middle-end/60092
3292         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3293         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3294         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3295         Handle BUILT_IN_POSIX_MEMALIGN.
3296         (find_func_clobbers): Likewise.
3297         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3298         (call_may_clobber_ref_p_1): Likewise.
3300 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3302         PR ipa/59918
3303         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3304         sanity check.
3306 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3308         PR ipa/59469
3309         * lto-cgraph.c (lto_output_node): Use
3310         symtab_get_symbol_partitioning_class.
3311         (lto_output_varpool_node): likewise.
3312         (symtab_get_symbol_partitioning_class): Move here from
3313         lto/lto-partition.c
3314         * cgraph.h (symbol_partitioning_class): Likewise.
3315         (symtab_get_symbol_partitioning_class): Declare.
3317 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3319         * ggc.h (ggc_internal_cleared_alloc): New macro.
3320         * vec.h (vec_safe_copy): Handle memory stats.
3321         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3322         * target-globals.c (save_target_globals): Likewise.
3324 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3326         PR target/60077
3327         * expr.c (emit_move_resolve_push): Export; be bit more selective
3328         on when to clear alias set.
3329         * expr.h (emit_move_resolve_push): Declare.
3330         * function.h (struct function): Add tail_call_marked.
3331         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3332         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3333         * config/i386/i386.md (TImode move expander): De not call
3334         ix86_expand_push.
3335         (FP push expanders): Preserve memory attributes.
3336         * config/i386/sse.md (push<mode>1): Remove.
3337         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3338         (ix86_expand_push): Remove.
3339         * config/i386/mmx.md (push<mode>1): Remove.
3341 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3343         PR rtl-optimization/60030
3344         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3345         lopart with paradoxical subreg before shifting it up by hprec.
3347 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3349         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3350         Remove extra newline at end of file.
3351         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3352         (arm_issue_rate): Handle cortexa57.
3353         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3354         (cortex-a57.cortex-a53): Likewise.
3356 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3358         PR target/59575
3359         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3360         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3361         bitmask.
3362         (arm_expand_prologue): Adjust all callers.
3363         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3364         info, registers also at the lowest numbered registers side.  Use
3365         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3366         XEXP.
3368         PR debug/59992
3369         * var-tracking.c (adjust_mems): Before adding a SET to
3370         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3372 2014-02-06  Alan Modra  <amodra@gmail.com>
3374         PR target/60032
3375         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3376         change SDmode to DDmode when lra_in_progress.
3378 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3380         PR middle-end/59150
3381         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3382         free_data_ref on the dr first, and before goto again also set dr
3383         to the next dr.  For simd_lane_access, free old datarefs[i] before
3384         overwriting it.  For get_vectype_for_scalar_type failure, don't
3385         free_data_ref if simd_lane_access.
3387         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3389         PR target/60062
3390         * tree.h (opts_for_fn): New inline function.
3391         (opt_for_fn): Define.
3392         * config/i386/i386.c (ix86_function_regparm): Use
3393         opt_for_fn (decl, optimize) instead of optimize.
3395 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3397         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3398         for SYMBOL_REF in large memory model.
3400 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3402         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3403         and crypto support.
3404         (cortex-a57): Likewise.
3405         (cortex-a57.cortex-a53): Likewise.
3407 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3408             Kugan Vivekanandarajah  <kuganv@linaro.org>
3410         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3411         unaligned_access.
3412         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3414 2014-02-06  Richard Biener  <rguenther@suse.de>
3416         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3417         set_loop_copy and initialize_original_copy_tables.
3419 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3421         * config/aarch64/aarch64-simd.md
3422         (aarch64_ashr_simddi): Change QI to SI.
3424 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3425             Jakub Jelinek  <jakub@redhat.com>
3427         PR middle-end/60013
3428         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3429         of the dataflow.
3431 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3433         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3434         CODE_FOR_altivec_vpku[hw]um to
3435         CODE_FOR_altivec_vpku[hw]um_direct.
3436         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3437         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3438         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3439         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3441 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3443         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3444         generation for -maltivec=be.
3445         (altivec_vsumsws): Simplify redundant test.
3447 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3449         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3450         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3451         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3452         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3453         gen_altivec_vpkuwum.
3454         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3455         BYTES_BIG_ENDIAN.
3456         (altivec_vpks<VI_char>ss): Likewise.
3457         (altivec_vpks<VI_char>us): Likewise.
3458         (altivec_vpku<VI_char>us): Likewise.
3459         (altivec_vpku<VI_char>um): Likewise.
3460         (altivec_vpku<VI_char>um_direct): New (copy of
3461         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3462         internal use).
3463         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3464         target is little endian and -maltivec=be is not specified.
3465         (*altivec_vupkhs<VU_char>_direct): New (copy of
3466         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3467         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3468         target is little endian and -maltivec=be is not specified.
3469         (*altivec_vupkls<VU_char>_direct): New (copy of
3470         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3471         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3472         little endian and -maltivec=be is not specified.
3473         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3474         little endian and -maltivec=be is not specified.
3476 2014-02-05  Richard Henderson  <rth@redhat.com>
3478         PR debug/52727
3479         * combine-stack-adj.c: Revert r206943.
3480         * sched-int.h (struct deps_desc): Add last_args_size.
3481         * sched-deps.c (init_deps): Initialize it.
3482         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3483         contain REG_ARGS_SIZE notes.
3485 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3487         * lto-cgraph.c (asm_nodes_output): Make global.
3488         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3489         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3490         (driver_handle_option): Handle OPT_fwpa.
3492 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3494         PR ipa/59947
3495         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3496         a comment typo and formatting issue.  If odr_hash hasn't been
3497         created, return vNULL and set *completep to false.
3499         PR middle-end/57499
3500         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3501         bb with no successors.
3503 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3505         PR target/59718
3506         * doc/invoke.texi (-march): Clarify documentation for ARM.
3507         (-mtune): Likewise.
3508         (-mcpu): Likewise.
3510 2014-02-05  Richard Biener  <rguenther@suse.de>
3512         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3513         when not vectorizing because of too many alias checks.
3514         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3515         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3517 2014-02-05  Nick Clifton  <nickc@redhat.com>
3519         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3520         accept extended registers in any mode when compiling for the MN10300.
3522 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3524         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3525         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3526         sanitization attributes.
3527         (can_inline_edge_p): Likewise.
3528         (sanitize_attrs_match_for_inline_p): New function.
3530 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3532         * ipa-prop.c (detect_type_change): Shor circuit testing of
3533         type changes on THIS pointer.
3535 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3537         PR target/59777
3538         * config/pa/pa.c (legitimize_tls_address): Return original address
3539         if not passed a SYMBOL_REF rtx.
3540         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3541         addresses.
3542         (pa_emit_move_sequence): Simplify TLS source operands.
3543         (pa_legitimate_constant_p): Reject all TLS constants.
3544         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3545         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3547 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3549         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3550         groups when we know they are controlled by LTO.
3551         * varasm.c (default_binds_local_p_1): If object is in other partition,
3552         it will be resolved locally.
3554 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3556         * config/host-linux.c (linux_gt_pch_use_address): Don't
3557         use SSIZE_MAX because it is not always defined.
3559 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3561         PR bootstrap/59913
3562         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3563         threshold for pseudo splitting.
3564         (update_ebb_live_info): Process call argument hard registers and
3565         hard registers from insn definition too.
3566         (max_small_class_regs_num): New constant.
3567         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3568         reloads_num only for small register classes.  Don't split for
3569         outputs of jumps.
3571 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3573         PR ipa/60058
3574         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3575         is non-null.
3577 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3579         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3580         visibility is safe.
3582 2014-02-04  Marek Polacek  <polacek@redhat.com>
3584         * gdbinit.in (pel): Define.
3586 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3588         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3589         behavior.
3591 2014-02-04  Richard Biener  <rguenther@suse.de>
3593         PR lto/59723
3594         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3595         in function context local.
3596         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3597         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3598         similar to LTO_imported_decl_ref.
3600 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3602         PR tree-optimization/60002
3603         * cgraphclones.c (build_function_decl_skip_args): Clear
3604         DECL_LANG_SPECIFIC.
3606         PR tree-optimization/60023
3607         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3608         false to gsi_replace.
3609         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3610         has been in some EH region and vec_stmt could throw, add
3611         vec_stmt into the same EH region.
3612         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3613         has no lhs, ignore it.
3614         * internal-fn.c (expand_MASK_LOAD): Likewise.
3616         PR ipa/60026
3617         * tree-inline.c (copy_forbidden): Fail for
3618         __attribute__((optimize (0))) functions.
3620         PR other/58712
3621         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3622         is set, copy one less argument.
3623         (expand_simd_clones): Don't subtract clone_info->inbranch
3624         from simd_clone_struct_alloc argument.
3626         PR rtl-optimization/57915
3627         * recog.c (simplify_while_replacing): If all unary/binary/relational
3628         operation arguments are constant, attempt to simplify those.
3630         PR middle-end/59261
3631         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3632         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3634 2014-02-04  Richard Biener  <rguenther@suse.de>
3636         PR tree-optimization/60012
3637         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3638         TBAA disambiguation to all DDRs.
3640 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3642         PR target/59788
3643         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3644         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3646 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3648         PR ipa/59882
3649         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3651 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3653         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3654         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3656 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3658         PR ipa/59831
3659         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3660         to figure out targets of polymorphic calls with known decl.
3661         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3662         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3663         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3664         (get_polymorphic_call_info): ... here.
3665         (get_polymorphic_call_info_from_invariant): New function.
3667 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3669         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3670         lookup via vtable pointer; check for type consistency
3671         and turn inconsitent facts into UNREACHABLE.
3672         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3673         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3674         type inconsistent querries; return UNREACHABLE instead.
3676 2014-02-03  Richard Henderson  <rth@twiddle.net>
3678         PR tree-opt/59924
3679         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3680         already processed this node.
3681         (normalize_one_pred_1): Pass along mark_set.
3682         (normalize_one_pred): Create and destroy a pointer_set_t.
3683         (normalize_one_pred_chain): Likewise.
3685 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3687         PR gcov-profile/58602
3688         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3690 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3692         PR ipa/59831
3693         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3694         -fno-devirtualize; try to devirtualize by the knowledge of
3695         virtual table pointer given by aggregate propagation.
3696         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3697         (ipa_print_node_jump_functions): Dump also offset that
3698         is relevant for polymorphic calls.
3699         (determine_known_aggregate_parts): Add arg_type parameter; use it
3700         instead of determining the type from pointer type.
3701         (ipa_compute_jump_functions_for_edge): Update call of
3702         determine_known_aggregate_parts.
3703         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3704         (gimple_get_virt_method_for_binfo): ... here; simplify using
3705         vtable_pointer_value_to_vtable.
3706         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3707         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3708         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3709         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3710         POINTER_PLUS_EXPR.
3711         (vtable_pointer_value_to_binfo): ... here.
3712         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3714 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3716         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3717         redef of outer loop index variable.
3719 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3721         PR c++/53017
3722         PR c++/59211
3723         * doc/extend.texi (Function Attributes): Typo.
3725 2014-02-03  Cong Hou  <congh@google.com>
3727         PR tree-optimization/60000
3728         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3729         if the vectorized statement is a store.  A store statement can only
3730         appear at the end of pattern statements.
3732 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3734         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3735         (ix86_option_override_internal): Default long double to 64-bit for
3736         32-bit Bionic and to 128-bit for 64-bit Bionic.
3738         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3739         TARGET_LONG_DOUBLE_128 is true.
3740         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3742         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3743         (mlong-double-64): Negate -mlong-double-128.
3744         (mlong-double-128): New option.
3746         * config/i386/i386-c.c (ix86_target_macros): Define
3747         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3749         * doc/invoke.texi: Document -mlong-double-128.
3751 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3753         PR rtl-optimization/60024
3754         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3756 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3758         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3760 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3762         PR rtl-optimization/57662
3763         * sel-sched.c (code_motion_path_driver): Do not mark already not
3764         existing blocks in the visiting bitmap.
3766 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3768         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3769         on the insn being emitted.
3771 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3772             Will Deacon  <will.deacon@arm.com>
3774         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3776 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3778         * config/arm/arm-tables.opt: Regenerate.
3780 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3782         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3783         for vector types other than V16QImode.
3784         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3785         define_expand, and call altivec_expand_vec_perm_le when producing
3786         code with little endian element order.
3787         (*altivec_vperm_<mode>_internal): New insn having previous
3788         behavior of altivec_vperm_<mode>.
3789         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3790         altivec_expand_vec_perm_le when producing code with little endian
3791         element order.
3792         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3793         behavior of altivec_vperm_<mode>_uns.
3795 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3797         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3798         (altivec_vsumsws): Add handling for -maltivec=be with a little
3799         endian target.
3800         (altivec_vsumsws_direct): New.
3801         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3802         gen_altivec_vsumsws.
3804 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3806         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3807         vtable_pointer_value_to_binfo): New functions.
3808         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3809         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3811 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3813         * config/nios2/nios2.md (load_got_register): Initialize GOT
3814         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3815         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3817 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3819         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3820         preserverd by passthrough, do not propagate the type.
3822 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3824         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3825         (mips_atomic_assign_expand_fenv): New function.
3826         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3828 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3830         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3831         (__builtin_mips_set_fcsr): Likewise.
3832         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3833         MIPS_USI_FTYPE_VOID.
3834         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3835         (mips16_expand_set_fcsr): Likewise.
3836         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3837         (mips16_set_fcsr_stub): Likewise.
3838         (mips16_get_fcsr_one_only_stub): New class.
3839         (mips16_set_fcsr_one_only_stub): Likewise.
3840         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3841         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3842         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3843         (hard_float): New availability predicate.
3844         (mips_builtins): Add get_fcsr and set_fcsr.
3845         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3846         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3847         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3848         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3849         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3850         patterns.
3852 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3854         * config/mips/mips.c (mips_one_only_stub): New class.
3855         (mips_need_mips16_rdhwr_p): Replace with...
3856         (mips16_rdhwr_stub): ...this new variable.
3857         (mips16_stub_call_address): New function.
3858         (mips16_rdhwr_one_only_stub): New class.
3859         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3860         (mips_output_mips16_rdhwr): Delete.
3861         (mips_finish_stub): New function.
3862         (mips_code_end): Use it to handle rdhwr stubs.
3864 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3866         PR target/60017
3867         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3868         when calculating size of integer atomic types.
3870 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3872         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3874 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3876         PR tree-optimization/60003
3877         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3878         * profile.c (branch_prob): Use gimple_call_builtin_p
3879         to check for BUILT_IN_SETJMP_RECEIVER.
3880         * tree-inline.c (copy_bb): Call notice_special_calls.
3882 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3884         PR bootstrap/59985
3885         * lra-constraints.c (process_alt_operands): Update reload_sum only
3886         on the first pass.
3888 2014-01-31  Richard Henderson  <rth@redhat.com>
3890         PR middle-end/60004
3891         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3892         until after else_eh is processed.
3894 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3896         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3897         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3898         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3899         in smmintrin.h, remove them.
3900         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3901         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3902         * config/i386/i386.md (ROUND_SAE): Fix value.
3903         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3904         (const48_operand): New.
3905         * config/i386/subst.md (round), (round_expand): Use
3906         const_4_or_8_to_11_operand.
3907         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3909 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3911         * config/i386/constraints.md (Yk): Swap meaning with k.
3912         * config/i386/i386.md (movhi_internal): Change Yk to k.
3913         (movqi_internal): Ditto.
3914         (*k<logic><mode>): Ditto.
3915         (*andhi_1): Ditto.
3916         (*andqi_1): Ditto.
3917         (kandn<mode>): Ditto.
3918         (*<code>hi_1): Ditto.
3919         (*<code>qi_1): Ditto.
3920         (kxnor<mode>): Ditto.
3921         (kortestzhi): Ditto.
3922         (kortestchi): Ditto.
3923         (kunpckhi): Ditto.
3924         (*one_cmplhi2_1): Ditto.
3925         (*one_cmplqi2_1): Ditto.
3926         * config/i386/sse.md (): Change k to Yk.
3927         (avx512f_load<mode>_mask): Ditto.
3928         (avx512f_blendm<mode>): Ditto.
3929         (avx512f_store<mode>_mask): Ditto.
3930         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3931         (avx512f_storedqu<mode>_mask): Ditto.
3932         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3933         Ditto.
3934         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3935         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3936         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3937         (avx512f_maskcmp<mode>3): Ditto.
3938         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3939         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3940         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3941         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3942         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3943         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3944         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3945         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3946         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3947         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3948         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3949         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3950         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3951         (vec_extract_lo_<mode>_maskm): Ditto.
3952         (vec_extract_hi_<mode>_maskm): Ditto.
3953         (avx512f_vternlog<mode>_mask): Ditto.
3954         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3955         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3956         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3957         (avx512f_<code>v8div16qi2_mask): Ditto.
3958         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3959         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3960         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3961         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3962         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3963         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3964         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3965         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3966         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3967         (avx512cd_maskb_vec_dupv8di): Ditto.
3968         (avx512cd_maskw_vec_dupv16si): Ditto.
3969         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3970         (avx512f_vpermi2var<mode>3_mask): Ditto.
3971         (avx512f_vpermi2var<mode>3_mask): Ditto.
3972         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3973         (*avx512f_gathersi<mode>): Ditto.
3974         (*avx512f_gathersi<mode>_2): Ditto.
3975         (*avx512f_gatherdi<mode>): Ditto.
3976         (*avx512f_gatherdi<mode>_2): Ditto.
3977         (*avx512f_scattersi<mode>): Ditto.
3978         (*avx512f_scatterdi<mode>): Ditto.
3979         (avx512f_compress<mode>_mask): Ditto.
3980         (avx512f_compressstore<mode>_mask): Ditto.
3981         (avx512f_expand<mode>_mask): Ditto.
3982         * config/i386/subst.md (mask): Change k to Yk.
3983         (mask_scalar_merge): Ditto.
3984         (sd): Ditto.
3986 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3988         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3990 2014-01-31  Richard Biener  <rguenther@suse.de>
3992         PR middle-end/59990
3993         * builtins.c (fold_builtin_memory_op): Make sure to not
3994         use a floating-point mode or a boolean or enumeral type for
3995         the copy operation.
3997 2014-01-30  DJ Delorie  <dj@redhat.com>
3999         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
4000         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
4001         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
4002         whenever main() has an epilogue.
4004 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4006         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
4007         unused variable "field".
4008         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
4009         (vsx_mergeh_<mode>): Likewise.
4010         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
4011         (altivec_vmrghh): Likewise.
4012         (altivec_vmrghw): Likewise.
4013         (altivec_vmrglb): Likewise.
4014         (altivec_vmrglh): Likewise.
4015         (altivec_vmrglw): Likewise.
4016         (altivec_vspltb): Add missing uses.
4017         (altivec_vsplth): Likewise.
4018         (altivec_vspltw): Likewise.
4019         (altivec_vspltsf): Likewise.
4021 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4023         PR target/59923
4024         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
4025         frame related instructions.
4027 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
4029         PR rtl-optimization/59959
4030         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
4031         any reload of register whose subreg is invalid.
4033 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
4035         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
4036         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
4037         Add missing return type - void.
4039 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4041         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
4042         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
4043         remove element index adjustment for endian (now handled in vsx.md
4044         and altivec.md).
4045         (altivec_expand_vec_perm_const): Use
4046         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
4047         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
4048         (vsx_xxspltw_<mode>): Adjust element index for little endian.
4049         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
4050         define_expand and a new define_insn *altivec_vspltb_internal;
4051         adjust for -maltivec=be on a little endian target.
4052         (altivec_vspltb_direct): New.
4053         (altivec_vsplth): Divide into a define_expand and a new
4054         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
4055         little endian target.
4056         (altivec_vsplth_direct): New.
4057         (altivec_vspltw): Divide into a define_expand and a new
4058         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
4059         little endian target.
4060         (altivec_vspltw_direct): New.
4061         (altivec_vspltsf): Divide into a define_expand and a new
4062         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
4063         a little endian target.
4065 2014-01-30  Richard Biener  <rguenther@suse.de>
4067         PR tree-optimization/59993
4068         * tree-ssa-forwprop.c (associate_pointerplus): Check we
4069         can propagate form the earlier stmt and avoid the transform
4070         when the intermediate result is needed.
4072 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
4074         * README.Portability: Fix typo.
4076 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
4078         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
4079         comparison_operator with ordered_comparison_operator.
4081 2014-01-30  Nick Clifton  <nickc@redhat.com>
4083         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
4084         Rename to mn10300_store_multiple_regs.
4085         * config/mn10300/mn10300.c: Likewise.
4086         * config/mn10300/mn10300.md (store_movm): Fix typo: call
4087         store_multiple_regs.
4088         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
4089         Call mn10300_store_multiple_regs.
4091 2014-01-30  Nick Clifton  <nickc@redhat.com>
4092             DJ Delorie  <dj@redhat.com>
4094         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
4095         %fp 2 to keep registers after it properly word-aligned.
4096         (rl78_alloc_physical_registers_umul): Handle the case where both
4097         input operands are the same.
4099 2014-01-30  Richard Biener  <rguenther@suse.de>
4101         PR tree-optimization/59903
4102         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4103         check properly.
4105 2014-01-30  Jason Merrill  <jason@redhat.com>
4107         PR c++/59633
4108         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
4110         PR c++/59645
4111         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
4113 2014-01-30  Richard Biener  <rguenther@suse.de>
4115         PR tree-optimization/59951
4116         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
4118 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
4120         PR target/59784
4121         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
4122         SFmode to DFmode case.
4124 2014-01-29  DJ Delorie  <dj@redhat.com>
4126         * config/msp430/msp430.opt (-minrt): New.
4127         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
4128         if -minrt given.
4129         (ENDFILE_SPEC): Likewise.
4131 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
4133         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
4134         (estimate_function_body_sizes): Use it.
4136 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
4138         PR c++/58561
4139         * dwarf2out.c (is_cxx_auto): New.
4140         (is_base_type): Use it.
4141         (gen_type_die_with_usage): Likewise.
4143 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4145         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
4146         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
4147         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
4148         -maltivec=be with LE targets.
4149         (vsx_mergeh_<mode>): Likewise.
4150         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
4151         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
4152         (altivec_vmrghb): Replace with define_expand and new
4153         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
4154         (altivec_vmrghb_direct): New define_insn.
4155         (altivec_vmrghh): Replace with define_expand and new
4156         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
4157         (altivec_vmrghh_direct): New define_insn.
4158         (altivec_vmrghw): Replace with define_expand and new
4159         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
4160         (altivec_vmrghw_direct): New define_insn.
4161         (*altivec_vmrghsf): Adjust for endianness.
4162         (altivec_vmrglb): Replace with define_expand and new
4163         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
4164         (altivec_vmrglb_direct): New define_insn.
4165         (altivec_vmrglh): Replace with define_expand and new
4166         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
4167         (altivec_vmrglh_direct): New define_insn.
4168         (altivec_vmrglw): Replace with define_expand and new
4169         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
4170         (altivec_vmrglw_direct): New define_insn.
4171         (*altivec_vmrglsf): Adjust for endianness.
4172         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4173         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4174         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4175         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4176         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4177         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4178         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4179         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4181 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4183         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4184         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4185         whitespace.
4187 2014-01-29  Richard Biener  <rguenther@suse.de>
4189         PR tree-optimization/58742
4190         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4191         associate_pointerplus_align.
4192         (associate_pointerplus_diff): New function.
4193         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4194         and associate_pointerplus_diff.
4196 2014-01-29  Richard Biener  <rguenther@suse.de>
4198         * lto-streamer.h (LTO_major_version): Bump to 3.
4199         (LTO_minor_version): Reset to 0.
4201 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4203         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4204         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4205         (arm_file_start): Generate correct asm header for armv7ve.
4206         * config/arm/bpabi.h: Add multilib support for armv7ve.
4207         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4208         and cortex-a15 to armv7ve.
4209         * config/arm/t-aprofile: Add multilib support for armv7ve.
4210         * doc/invoke.texi: Document -march=armv7ve.
4212 2014-01-29  Richard Biener  <rguenther@suse.de>
4214         PR tree-optimization/58742
4215         * tree-ssa-forwprop.c (associate_plusminus): Return true
4216         if we changed sth, defer EH cleanup to ...
4217         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4218         (simplify_mult): New function.
4220 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4222         PR middle-end/59917
4223         PR tree-optimization/59920
4224         * tree.c (build_common_builtin_nodes): Remove
4225         __builtin_setjmp_dispatcher initialization.
4226         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4227         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4228         instead of gsi_after_labels + manually skipping debug stmts.
4229         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4230         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4231         * tree-inline.c (copy_edges_for_bb): Remove
4232         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4233         argument.  Ignore computed_goto_p stmts.  Don't call
4234         make_abnormal_goto_edges.  If a call might need abnormal edges
4235         for non-local gotos, see if it already has an edge to
4236         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4237         with true argument, don't do anything then, otherwise add
4238         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4239         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4240         caller.
4241         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4242         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4243         (lower_stmt): Don't set data->calls_builtin_setjmp.
4244         (lower_builtin_setjmp): Adjust comment.
4245         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4246         * tree-cfg.c (found_computed_goto): Remove.
4247         (factor_computed_gotos): Remove.
4248         (make_goto_expr_edges): Return bool, true for computed gotos.
4249         Don't call make_abnormal_goto_edges.
4250         (build_gimple_cfg): Don't set found_computed_goto, don't call
4251         factor_computed_gotos.
4252         (computed_goto_p): No longer static.
4253         (make_blocks): Don't set found_computed_goto.
4254         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4255         (make_edges): If make_goto_expr_edges returns true, push bb
4256         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4257         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4258         vector.  Record mapping between bbs and OpenMP regions if there
4259         are any, adjust make_gimple_omp_edges caller.  Call
4260         handle_abnormal_edges.
4261         (make_abnormal_goto_edges): Remove.
4262         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4263         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4264         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4265         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4266         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4267         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4268         filling *region also set *region_idx to (*region)->entry->index.
4270         PR other/58712
4271         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4272         For REGs set ORIGINAL_REGNO.
4274 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4276         * doc/md.texi: Mention that a target shouldn't implement
4277         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4278         than hi/lo pair.
4280 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4282         PR tree-optimization/59594
4283         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4284         a copy of the datarefs vector rather than the vector itself.
4286 2014-01-28  Jason Merrill  <jason@redhat.com>
4288         PR c++/53756
4289         * dwarf2out.c (auto_die): New static.
4290         (gen_type_die_with_usage): Handle C++1y 'auto'.
4291         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4292         on definition.
4294 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4296         PR target/59672
4297         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4298         (SPEC_X32): Likewise.
4299         (SPEC_64): Likewise.
4300         * config/i386/i386.c (ix86_option_override_internal): Turn off
4301         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4302         for TARGET_16BIT.
4303         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4304         * config/i386/i386.h (TARGET_16BIT): New macro.
4305         (TARGET_16BIT_P): Likewise.
4306         * config/i386/i386.opt: Add m16.
4307         * doc/invoke.texi: Document -m16.
4309 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4311         PR preprocessor/59935
4312         * input.c (location_get_source_line): Bail out on when line number
4313         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4315 2014-01-28  Richard Biener  <rguenther@suse.de>
4317         PR tree-optimization/58742
4318         * tree-ssa-forwprop.c (associate_plusminus): Handle
4319         pointer subtraction of the form (T)(P + A) - (T)P.
4321 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4323         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4324         at const_int_cost.
4326 2014-01-28  Richard Biener  <rguenther@suse.de>
4328         Revert
4329         2014-01-28  Richard Biener  <rguenther@suse.de>
4331         PR rtl-optimization/45364
4332         PR rtl-optimization/59890
4333         * var-tracking.c (local_get_addr_clear_given_value): Handle
4334         already cleared slot.
4335         (val_reset): Handle not allocated local_get_addr_cache.
4336         (vt_find_locations): Use post-order on the inverted CFG.
4338 2014-01-28  Richard Biener  <rguenther@suse.de>
4340         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4342 2014-01-28  Richard Biener  <rguenther@suse.de>
4344         PR rtl-optimization/45364
4345         PR rtl-optimization/59890
4346         * var-tracking.c (local_get_addr_clear_given_value): Handle
4347         already cleared slot.
4348         (val_reset): Handle not allocated local_get_addr_cache.
4349         (vt_find_locations): Use post-order on the inverted CFG.
4351 2014-01-28  Alan Modra  <amodra@gmail.com>
4353         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4354         * configure.ac <recursive call for build != host>: Define
4355         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4356         and LD_FOR_BUILD too.
4357         * configure: Regenerate.
4359 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4361         * config/i386/i386.c (get_builtin_code_for_version): Separate
4362         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4363         Broadwell from Haswell.
4365 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4367         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4368         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4369         * config/mips/mips.c (mips_option_override): Change setting
4370         of TARGET_DSP.
4371         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4372         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4373         Change from Mask to Var.
4375 2014-01-27  Jeff Law  <law@redhat.com>
4377         * ipa-inline.c (inline_small_functions): Fix typo.
4379 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4381         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4382         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4383         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4384         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4385         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4386         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4387         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4388         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4389         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4390         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4391         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4392         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4393         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4394         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4395         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4396         (_mm512_storeu_epi64): Ditto.
4397         (_mm512_cmpge_epi32_mask): Ditto.
4398         (_mm512_cmpge_epu32_mask): Ditto.
4399         (_mm512_cmpge_epi64_mask): Ditto.
4400         (_mm512_cmpge_epu64_mask): Ditto.
4401         (_mm512_cmple_epi32_mask): Ditto.
4402         (_mm512_cmple_epu32_mask): Ditto.
4403         (_mm512_cmple_epi64_mask): Ditto.
4404         (_mm512_cmple_epu64_mask): Ditto.
4405         (_mm512_cmplt_epi32_mask): Ditto.
4406         (_mm512_cmplt_epu32_mask): Ditto.
4407         (_mm512_cmplt_epi64_mask): Ditto.
4408         (_mm512_cmplt_epu64_mask): Ditto.
4409         (_mm512_cmpneq_epi32_mask): Ditto.
4410         (_mm512_cmpneq_epu32_mask): Ditto.
4411         (_mm512_cmpneq_epi64_mask): Ditto.
4412         (_mm512_cmpneq_epu64_mask): Ditto.
4413         (_mm512_expand_pd): Ditto.
4414         (_mm512_expand_ps): Ditto.
4415         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4416         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4417         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4418         * config/i386/i386.c (ix86_builtins): Add
4419         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4420         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4421         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4422         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4423         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4424         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4425         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4426         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4427         IX86_BUILTIN_PMOVUSQW512_MEM.
4428         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4429         __builtin_ia32_pmovsqd512mem_mask,
4430         __builtin_ia32_pmovqd512mem_mask,
4431         __builtin_ia32_pmovusqw512mem_mask,
4432         __builtin_ia32_pmovsqw512mem_mask,
4433         __builtin_ia32_pmovqw512mem_mask,
4434         __builtin_ia32_pmovusdw512mem_mask,
4435         __builtin_ia32_pmovsdw512mem_mask,
4436         __builtin_ia32_pmovdw512mem_mask,
4437         __builtin_ia32_pmovqb512mem_mask,
4438         __builtin_ia32_pmovusqb512mem_mask,
4439         __builtin_ia32_pmovsqb512mem_mask,
4440         __builtin_ia32_pmovusdb512mem_mask,
4441         __builtin_ia32_pmovsdb512mem_mask,
4442         __builtin_ia32_pmovdb512mem_mask.
4443         (bdesc_args): Add __builtin_ia32_expanddf512,
4444         __builtin_ia32_expandsf512.
4445         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4446         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4447         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4448         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4449         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4450         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4451         (avx512f_<code>v8div16qi2_mask_store): This.
4452         (avx512f_expand<mode>): New.
4454 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4456         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4457         New.
4458         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4459         (_mm512_prefetch_i32scatter_pd): Ditto.
4460         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4461         (_mm512_prefetch_i64scatter_pd): Ditto.
4462         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4463         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4464         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4465         (_mm512_prefetch_i32scatter_ps): Ditto.
4466         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4467         (_mm512_prefetch_i64scatter_ps): Ditto.
4468         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4469         * config/i386/i386-builtin-types.def: Define
4470         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4471         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4472         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4473         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4474         IX86_BUILTIN_SCATTERPFQPD.
4475         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4476         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4477         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4478         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4479         __builtin_ia32_scatterpfqps.
4480         (ix86_expand_builtin): Expand new built-ins.
4481         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4482         fix memory access data type.
4483         (*avx512pf_gatherpf<mode>_mask): Ditto.
4484         (*avx512pf_gatherpf<mode>): Ditto.
4485         (avx512pf_scatterpf<mode>): Ditto.
4486         (*avx512pf_scatterpf<mode>_mask): Ditto.
4487         (*avx512pf_scatterpf<mode>): Ditto.
4488         (GATHER_SCATTER_SF_MEM_MODE): New.
4489         (avx512pf_gatherpf<mode>df): Ditto.
4490         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4491         (*avx512pf_scatterpf<mode>df): Ditto.
4493 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4495         PR bootstrap/59934
4496         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4497         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4498         reached.
4500 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4502         * common/config/arm/arm-common.c
4503         (arm_rewrite_mcpu): Handle multiple names.
4504         * config/arm/arm.h
4505         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4507 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4509         * gimple-builder.h (create_gimple_tmp): Delete.
4511 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4513         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4514         words comparisons.
4516 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4518         * config/pa/pa.md (call): Generate indirect long calls to non-local
4519         functions when outputing 32-bit code.
4520         (call_value): Likewise except for special call to buggy powf function.
4522         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4523         portable runtime and PIC indirect calls.
4524         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4525         and PIC call sequences.  Use ldo instead of blr to set return register
4526         in PIC call sequence.
4528 2014-01-25  Walter Lee  <walt@tilera.com>
4530         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4531         avoid clobbering a live register.
4533 2014-01-25  Walter Lee  <walt@tilera.com>
4535         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4536         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4537         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4538         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4540 2014-01-25  Walter Lee  <walt@tilera.com>
4542         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4543         arguments on even registers.
4544         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4545         STACK_BOUNDARY.
4546         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4547         (BIGGEST_ALIGNMENT): Ditto.
4548         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4550 2014-01-25  Walter Lee  <walt@tilera.com>
4552         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4553         insns before bundling.
4554         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4556 2014-01-25  Walter Lee  <walt@tilera.com>
4558         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4559         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4560         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4562 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4564         * config/mips/constraints.md (kl): Delete.
4565         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4566         define expands, using...
4567         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4568         instructions for MIPS16.
4569         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4570         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4572 2014-01-25  Walter Lee  <walt@tilera.com>
4574         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4575         (clzdi2): Ditto.
4576         (ffsdi2): Ditto.
4578 2014-01-25  Walter Lee  <walt@tilera.com>
4580         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4581         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4583 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4585         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4586         Handle XOR.
4588 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4590         * print-rtl.c (in_call_function_usage): New var.
4591         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4592         EXPR_LIST mode as mode and not as reg note name.
4594         PR middle-end/59561
4595         * cfgloopmanip.c (copy_loop_info): If
4596         loop->warned_aggressive_loop_optimizations, make sure
4597         the flag is set in target loop too.
4599 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4601         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4602         flag_cilkplus.
4603         * builtins.def: Likewise.
4604         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4605         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4606         * ira.c (ira_setup_eliminable_regset): Likewise.
4607         * omp-low.c (gate_expand_omp): Likewise.
4608         (execute_lower_omp): Likewise.
4609         (diagnose_sb_0): Likewise.
4610         (gate_diagnose_omp_blocks): Likewise.
4611         (simd_clone_clauses_extract): Likewise.
4612         (gate): Likewise.
4614 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4616         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4617         correction for little endian...
4618         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4619         here.
4621 2014-01-24  Jeff Law  <law@redhat.com>
4623         PR tree-optimization/59919
4624         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4625         for non-returning calls.
4627 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4629         * common/config/aarch64/aarch64-common.c
4630         (aarch64_rewrite_mcpu): Handle multiple names.
4631         * config/aarch64/aarch64.h
4632         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4634 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4636         * input.c (add_file_to_cache_tab): Handle the case where fopen
4637         returns NULL.
4639 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4641         PR target/59929
4642         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4643         from push operand if code of push isn't PRE_DEC.
4645 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4647         PR target/59909
4648         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4649         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4650         it is only used for non-atomic loads/stores.
4652         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4653         -mquad-memory or -mquad-memory-atomic switches.
4655         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4656         -mquad-memory-atomic to ISA 2.07 support.
4658         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4659         to separate support of normal quad word memory operations (ldq, stq)
4660         from the atomic quad word memory operations.
4662         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4663         support to separate non-atomic quad word operations from atomic
4664         quad word operations.  Disable non-atomic quad word operations in
4665         little endian mode so that we don't have to swap words after the
4666         load and before the store.
4667         (quad_load_store_p): Add comment about atomic quad word support.
4668         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4669         options printed with -mdebug=reg.
4671         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4672         -mquad-memory-atomic as the test for whether we have quad word
4673         atomic instructions.
4674         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4675         or -mp8-vector are used, allow byte/half-word atomic operations.
4677         * config/rs6000/sync.md (load_lockedti): Insure that the address
4678         is a proper indexed or indirect address for the lqarx instruction.
4679         On little endian systems, swap the hi/lo registers after the lqarx
4680         instruction.
4681         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4682         insure the address is valid for the lqarx instruction.
4683         (store_conditionalti): Insure that the address is a proper indexed
4684         or indirect address for the stqcrx. instruction.  On little endian
4685         systems, swap the hi/lo registers before doing the stqcrx.
4686         instruction.
4687         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4688         insure the address is valid for the stqcrx. instruction.
4690         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4691         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4692         type of quad memory support is available.
4694 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4696         PR regression/59915
4697         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4698         there is a danger of looping.
4700 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4702         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4703         force flag_ira_loop_pressure if set via command line.
4705 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4707         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4708         (ashr_simd): New builtin handling DI mode.
4709         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4710         (aarch64_sshr_simddi): New match pattern.
4711         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4712         (vshrd_n_s64): Likewise.
4713         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4715 2014-01-23  Nick Clifton  <nickc@redhat.com>
4717         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4718         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4719         favour of mcu specific scripts.
4720         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4721         430x multilibs.
4723 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4724             Alex Velenko  <Alex.Velenko@arm.com>
4726         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4727         (vaddv_s16): Likewise.
4728         (vaddv_s32): Likewise.
4729         (vaddv_u8): Likewise.
4730         (vaddv_u16): Likewise.
4731         (vaddv_u32): Likewise.
4732         (vaddvq_s8): Likewise.
4733         (vaddvq_s16): Likewise.
4734         (vaddvq_s32): Likewise.
4735         (vaddvq_s64): Likewise.
4736         (vaddvq_u8): Likewise.
4737         (vaddvq_u16): Likewise.
4738         (vaddvq_u32): Likewise.
4739         (vaddvq_u64): Likewise.
4740         (vaddv_f32): Likewise.
4741         (vaddvq_f32): Likewise.
4742         (vaddvq_f64): Likewise.
4743         (vmaxv_f32): Likewise.
4744         (vmaxv_s8): Likewise.
4745         (vmaxv_s16): Likewise.
4746         (vmaxv_s32): Likewise.
4747         (vmaxv_u8): Likewise.
4748         (vmaxv_u16): Likewise.
4749         (vmaxv_u32): Likewise.
4750         (vmaxvq_f32): Likewise.
4751         (vmaxvq_f64): Likewise.
4752         (vmaxvq_s8): Likewise.
4753         (vmaxvq_s16): Likewise.
4754         (vmaxvq_s32): Likewise.
4755         (vmaxvq_u8): Likewise.
4756         (vmaxvq_u16): Likewise.
4757         (vmaxvq_u32): Likewise.
4758         (vmaxnmv_f32): Likewise.
4759         (vmaxnmvq_f32): Likewise.
4760         (vmaxnmvq_f64): Likewise.
4761         (vminv_f32): Likewise.
4762         (vminv_s8): Likewise.
4763         (vminv_s16): Likewise.
4764         (vminv_s32): Likewise.
4765         (vminv_u8): Likewise.
4766         (vminv_u16): Likewise.
4767         (vminv_u32): Likewise.
4768         (vminvq_f32): Likewise.
4769         (vminvq_f64): Likewise.
4770         (vminvq_s8): Likewise.
4771         (vminvq_s16): Likewise.
4772         (vminvq_s32): Likewise.
4773         (vminvq_u8): Likewise.
4774         (vminvq_u16): Likewise.
4775         (vminvq_u32): Likewise.
4776         (vminnmv_f32): Likewise.
4777         (vminnmvq_f32): Likewise.
4778         (vminnmvq_f64): Likewise.
4780 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4782         * config/aarch64/aarch64-simd.md
4783         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4784         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4785         (*aarch64_mul3_elt<mode>): Likewise.
4786         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4787         (*aarch64_mul3_elt_to_64v2df): Likewise.
4788         (*aarch64_mla_elt<mode>): Likewise.
4789         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4790         (*aarch64_mls_elt<mode>): Likewise.
4791         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4792         (*aarch64_fma4_elt<mode>): Likewise.
4793         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4794         (*aarch64_fma4_elt_to_64v2df): Likewise.
4795         (*aarch64_fnma4_elt<mode>): Likewise.
4796         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4797         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4798         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4799         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4800         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4801         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4802         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4803         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4804         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4806 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4808         * config/aarch64/aarch64-simd.md
4809         (aarch64_be_checked_get_lane<mode>): New define_expand.
4810         * config/aarch64/aarch64-simd-builtins.def
4811         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4812         New builtin definition.
4813         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4814         Use new safe be builtin.
4816 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4818         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4819         New define_insn.
4820         (aarch64_be_st1<mode>): Likewise.
4821         (aarch_ld1<VALL:mode>): Define_expand modified.
4822         (aarch_st1<VALL:mode>): Likewise.
4823         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4824         (UNSPEC_ST1): Likewise.
4826 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
4828         * config/microblaze/microblaze.md: Add trap insn and attribute
4830 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4832         PR preprocessor/58580
4833         * input.h (location_get_source_line): Take an additional line_size
4834         parameter.
4835         (void diagnostics_file_cache_fini): Declare new function.
4836         * input.c (struct fcache): New type.
4837         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4838         New static constants.
4839         (diagnostic_file_cache_init, total_lines_num)
4840         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4841         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4842         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4843         (get_next_line, read_next_line, goto_next_line, read_line_num):
4844         New static function definitions.
4845         (diagnostic_file_cache_fini): New function.
4846         (location_get_source_line): Take an additional output line_len
4847         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4848         read_line_num.
4849         * diagnostic.c (diagnostic_finish): Call
4850         diagnostic_file_cache_fini.
4851         (adjust_line): Take an additional input parameter for the length
4852         of the line, rather than calculating it with strlen.
4853         (diagnostic_show_locus): Adjust the use of
4854         location_get_source_line and adjust_line with respect to their new
4855         signature.  While displaying a line now, do not stop at the first
4856         null byte.  Rather, display the zero byte as a space and keep
4857         going until we reach the size of the line.
4858         * Makefile.in: Add vec.o to OBJS-libcommon
4860 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4861             Ilya Tocar     <ilya.tocar@intel.com>
4863         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4864         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4865         (__builtin_ia32_kmov16): Ditto.
4866         * config/i386/i386.md (UNSPEC_KMOV): New.
4867         (kmovw): Ditto.
4869 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4871         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4872         (_mm512_storeu_si512): Ditto.
4874 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4876         PR target/52125
4877         * rtl.h (get_referenced_operands): Declare.
4878         * recog.c (get_referenced_operands): New function.
4879         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4880         operands have been referenced when recording LO_SUM references.
4882 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
4884         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4886 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4888         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4889         Enable for generic and recent AMD targets.
4891 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4893         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4894         ARG_SIZE note when adjustment was eliminated.
4896 2014-01-22  Jeff Law  <law@redhat.com>
4898         PR tree-optimization/59597
4899         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4900         in file.  Accept new argument REGISTERING and use it to modify
4901         dump output appropriately.
4902         (register_jump_thread): Corresponding changes.
4903         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4904         thread paths involving joiner blocks.  Add code to dump cancelled
4905         jump threading paths.
4907 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4909         PR rtl-optimization/59477
4910         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4911         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4913 2014-01-22  Tom Tromey  <tromey@redhat.com>
4915         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4916         PARAMS.
4917         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4919 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4921         PR rtl-optimization/59896
4922         * lra-constraints.c (process_alt_operands): Check unused note for
4923         matched operands of insn with no output reloads.
4925 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4927         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4928         (mips_move_from_gpr_cost): Likewise.
4930 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4932         PR rtl-optimization/59858
4933         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4934         ira_class_hard_regs_num.
4935         (process_alt_operands): Increase reject for dying matched operand.
4937 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4939         PR target/59003
4940         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4941         smaller than size, perform several stores or loads and stores
4942         at dst + count - size to store or copy all of size bytes, rather
4943         than just last modesize bytes.
4945 2014-01-20  DJ Delorie  <dj@redhat.com>
4947         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4948         that CLOBBERs are REGs before propogating their values.
4950 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4952         PR middle-end/59789
4953         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4954         (cgraph_inline_failed_type): New function.
4955         * cgraph.h (DEFCIFCODE): Add type.
4956         (cgraph_inline_failed_type_t): New enum.
4957         (cgraph_inline_failed_type): New prototype.
4958         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4959         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4960         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4961         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4962         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4963         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4964         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4965         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4966         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4967         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4968         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4969         OPTIMIZATION_MISMATCH.
4970         * tree-inline.c (expand_call_inline): Emit errors during
4971         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4973 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4975         PR target/59685
4976         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4977         mode attribute in insn output.
4979 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4981         * output.h (output_constant): Delete.
4982         * varasm.c (output_constant): Make private.
4984 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4986         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4988 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4990         PR middle-end/59860
4991         * tree.h (fold_builtin_strcat): New prototype.
4992         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4993         argument, if non-NULL, don't call c_strlen.  Optimize
4994         directly into __builtin_memcpy instead of __builtin_strcpy.
4995         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4996         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4998 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
5000         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5001         for SImode_address_operand operands, having only a REG argument.
5003 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
5005         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
5006         loader name using mbig-endian.
5007         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
5009 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5011         * doc/invoke.texi (-march): Clarify documentation for AArch64.
5012         (-mtune): Likewise.
5013         (-mcpu): Likewise.
5015 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
5017         * config/aarch64/aarch64-protos.h
5018         (aarch64_cannot_change_mode_class_ptr): Declare.
5019         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
5020         aarch64_cannot_change_mode_class_ptr): New.
5021         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
5022         backend hook aarch64_cannot_change_mode_class.
5024 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
5026         * common/config/aarch64/aarch64-common.c
5027         (aarch64_handle_option): Don't handle any option order logic here.
5028         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
5029         selected_cpu, warn on architecture version mismatch.
5030         (aarch64_override_options): Fix parsing order for option strings.
5032 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5033             Iain Sandoe  <iain@codesourcery.com>
5035         PR bootstrap/59496
5036         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
5037         warning.  Amend comment to reflect current functionality.
5039 2014-01-20  Richard Biener  <rguenther@suse.de>
5041         PR middle-end/59860
5042         * builtins.c (fold_builtin_strcat): Remove case better handled
5043         by tree-ssa-strlen.c.
5045 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
5047         * config/aarch64/aarch64.opt
5048         (mcpu, march, mtune): Make case-insensitive.
5050 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
5052         PR target/59880
5053         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
5054         if operands[1] is a REG or ZERO_EXTEND of a REG.
5056 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
5058         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
5060 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
5062         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
5063         long non-pic millicode calls.
5065 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5067         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
5069 2014-01-19  Kito Cheng  <kito@0xlab.org>
5071         * builtins.c (expand_movstr): Check movstr expand done or fail.
5073 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5074             H.J. Lu  <hongjiu.lu@intel.com>
5076         PR target/59379
5077         * config/i386/i386.md (*lea<mode>): Zero-extend return register
5078         to DImode for zero-extended addresses.
5080 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
5082         PR rtl-optimization/57763
5083         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
5084         on the new indirect jump_insn and increment LABEL_NUSES (label).
5086 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
5088         PR bootstrap/59580
5089         PR bootstrap/59583
5090         * config.gcc (x86_archs): New variable.
5091         (x86_64_archs): Likewise.
5092         (x86_cpus): Likewise.
5093         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
5094         --with-arch/--with-cpu= options.
5095         Support --with-arch=/--with-cpu={nehalem,westmere,
5096         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
5098 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5100         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
5101         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
5103 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5105         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
5107 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
5109         PR target/58944
5110         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
5111         clear cpp_get_options (parse_in)->warn_unused_macros for
5112         ix86_target_macros_internal with cpp_define.
5114 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
5116         * jump.c (delete_related_insns): Keep (use (insn))s.
5117         * reorg.c (redundant_insn): Check for barriers too.
5119 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5121         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
5123 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
5125         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
5126         call to $$dyncall when TARGET_LONG_CALLS is true.
5128 2014-01-17  Jeff Law  <law@redhat.com>
5130         * ree.c (combine_set_extension): Temporarily disable test for
5131         changing number of hard registers.
5133 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5135         PR middle-end/58125
5136         * ipa-inline-analysis.c (inline_free_summary):
5137         Do not free summary of aliases.
5139 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5141         PR middle-end/59706
5142         * gimplify.c (gimplify_expr): Use create_tmp_var
5143         instead of create_tmp_var_raw.  If cond doesn't have
5144         integral type, don't add the IFN_ANNOTATE builtin at all.
5146 2014-01-17  Martin Jambor  <mjambor@suse.cz>
5148         PR ipa/59736
5149         * ipa-cp.c (prev_edge_clone): New variable.
5150         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
5151         Also resize prev_edge_clone vector.
5152         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
5153         (ipcp_edge_removal_hook): New function.
5154         (ipcp_driver): Register ipcp_edge_removal_hook.
5156 2014-01-17  Andrew Pinski  <apinski@cavium.com>
5157             Steve Ellcey  <sellcey@mips.com>
5159         PR target/59462
5160         * config/mips/mips.c (mips_print_operand): Check operand mode instead
5161         of operator mode.
5163 2014-01-17  Jeff Law  <law@redhat.com>
5165         PR middle-end/57904
5166         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
5167         so that pass_ccp runs first.
5169 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5171         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5172         (ix86_adjust_cost): Use !TARGET_XXX.
5173         (do_reorder_for_imul): Likewise.
5174         (swap_top_of_ready_list): Likewise.
5175         (ix86_sched_reorder): Likewise.
5177 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5179         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5180         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5181         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5182         (intel_memset): New.  Duplicate slm_memset.
5183         (intel_cost): New.  Duplicate slm_cost.
5184         (m_INTEL): New macro.
5185         (processor_target_table): Add "intel".
5186         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5187         with PROCESSOR_INTEL for "intel".
5188         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5189         PROCESSOR_SILVERMONT.
5190         (ix86_issue_rate): Likewise.
5191         (ix86_adjust_cost): Likewise.
5192         (ia32_multipass_dfa_lookahead): Likewise.
5193         (swap_top_of_ready_list): Likewise.
5194         (ix86_sched_reorder): Likewise.
5195         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5196         instead of TARGET_OPT_AGU.
5197         * config/i386/i386.h (TARGET_INTEL): New.
5198         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5199         (processor_type): Add PROCESSOR_INTEL.
5200         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5201         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5203 2014-01-17  Marek Polacek  <polacek@redhat.com>
5205         PR c/58346
5206         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5207         size is zero.
5209 2014-01-17  Richard Biener  <rguenther@suse.de>
5211         PR tree-optimization/46590
5212         * opts.c (default_options_table): Add entries for
5213         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5214         all enabled at -O1 but not for -Og.
5215         * common.opt (fbranch-count-reg): Remove Init(1).
5216         (fmove-loop-invariants): Likewise.
5217         (ftree-pta): Likewise.
5219 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5221         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5222         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5223         decls to at least the GCC 4.8 used alignments.
5225         PR fortran/59440
5226         * tree-nested.c (convert_nonlocal_reference_stmt,
5227         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5228         of GIMPLE_BIND stmts, adjust associated decls.
5230 2014-01-17  Richard Biener  <rguenther@suse.de>
5232         PR tree-optimization/46590
5233         * vec.h (vec<>::bseach): New member function implementing
5234         binary search according to C89 bsearch.
5235         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5236         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5237         bitmap pointer again.  Make accesses_in_loop a flat array.
5238         (mem_ref_obstack): New global.
5239         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5240         (mark_ref_stored): Likewise.
5241         (ref_indep_loop_p_2): Likewise.
5242         (set_ref_stored_in_loop): New helper function.
5243         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5244         (memref_free): Adjust.
5245         (record_mem_ref_loc): Simplify.
5246         (gather_mem_refs_stmt): Adjust.
5247         (sort_locs_in_loop_postorder_cmp): New function.
5248         (analyze_memory_references): Sort accesses_in_loop after
5249         loop postorder number.
5250         (find_ref_loc_in_loop_cmp): New function.
5251         (for_all_locs_in_loop): Find relevant cluster of locs in
5252         accesses_in_loop and iterate without recursion.
5253         (execute_sm): Avoid uninit warning.
5254         (struct ref_always_accessed): Simplify.
5255         (ref_always_accessed::operator ()): Likewise.
5256         (ref_always_accessed_p): Likewise.
5257         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5258         loop postorder numbers here.
5259         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5260         numbers.
5262 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5264         PR c++/57945
5265         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5266         on decls for which assemble_alias has been called.
5268 2014-01-17  Nick Clifton  <nickc@redhat.com>
5270         * config/msp430/msp430.opt: (mcpu): New option.
5271         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5272         (msp430_option_override): Parse target_cpu.  If the MCU name
5273         matches a generic string, clear target_mcu.
5274         (msp430_attr): Allow numeric interrupt values up to 63.
5275         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5276         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5277         option.
5278         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5279         Add mcpu matches.
5280         * config/msp430/msp430.md (popm): Use %J rather than %I.
5281         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5282         (addhi_cy_i): Use immediate_operand for operand 2.
5283         * doc/invoke.texi: Document -mcpu option.
5285 2014-01-17  Richard Biener  <rguenther@suse.de>
5287         PR rtl-optimization/38518
5288         * df.h (df_analyze_loop): Declare.
5289         * df-core.c: Include cfgloop.h.
5290         (df_analyze_1): Split out main part of df_analyze.
5291         (df_analyze): Adjust.
5292         (loop_inverted_post_order_compute): New function.
5293         (loop_post_order_compute): Likewise.
5294         (df_analyze_loop): New function avoiding whole-function
5295         postorder computes.
5296         * loop-invariant.c (find_defs): Use df_analyze_loop.
5297         (find_invariants): Adjust.
5298         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5300 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5302         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5303         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5305 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5307         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5308         traversal when removing references.
5310 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5312         PR ipa/59775
5313         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5315 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5317         PR middle-end/56791
5318         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5319         pushing a reload for an autoinc when we had previously reloaded an
5320         inner part of the address.
5322 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5324         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5325         field.
5326         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5327         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5328         when not giving up or versioning for alias only because of
5329         loop->safelen.
5330         (vect_analyze_data_ref_dependences): Set to true.
5331         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5332         is a GIMPLE_PHI.
5333         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5334         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5335         to the condition.
5337         PR middle-end/58344
5338         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5340         PR target/59839
5341         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5342         operand 0 predicate for gathers, use a new pseudo as subtarget.
5344 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5346         PR middle-end/59609
5347         * lra-constraints.c (process_alt_operands): Add printing debug info.
5348         Check absence of input/output reloads for matched operands too.
5350 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5352         PR rtl-optimization/59835
5353         * ira.c (ira_init_register_move_cost): Increase cost for
5354         impossible modes.
5356 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5358         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5360 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5362         PR target/59780
5363         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5364         non-register objects.  Use gen_(high/low)part more consistently.
5365         Fix assertions.
5367 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5369         PR target/59844
5370         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5371         endian support, remove tests for WORDS_BIG_ENDIAN.
5372         (p8_mfvsrd_3_<mode>): Likewise.
5373         (reload_gpr_from_vsx<mode>): Likewise.
5374         (reload_gpr_from_vsxsf): Likewise.
5375         (p8_mfvsrd_4_disf): Likewise.
5377 2014-01-16  Richard Biener  <rguenther@suse.de>
5379         PR rtl-optimization/46590
5380         * lcm.c (compute_antinout_edge): Use postorder iteration.
5381         (compute_laterin): Use inverted postorder iteration.
5383 2014-01-16  Nick Clifton  <nickc@redhat.com>
5385         PR middle-end/28865
5386         * varasm.c (output_constant): Return the number of bytes actually
5387         emitted.
5388         (output_constructor_array_range): Update the field size with the
5389         number of bytes emitted by output_constant.
5390         (output_constructor_regular_field): Likewise.  Also do not
5391         complain if the total number of bytes emitted is now greater
5392         than the expected fieldpos.
5393         * output.h (output_constant): Update prototype and descriptive comment.
5395 2014-01-16  Marek Polacek  <polacek@redhat.com>
5397         PR middle-end/59827
5398         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5399         it is error_mark_node.
5401 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5403         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5404         VALID_AVX256_REG_OR_OI_MODE.
5406 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5408         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5409         current procedure should be profiled.
5411 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5413         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5414         of moving from/to the STACK_REG register class.
5416 2014-01-15  Richard Henderson  <rth@redhat.com>
5418         PR debug/54694
5419         * reginfo.c (global_regs_decl): Globalize.
5420         * rtl.h (global_regs_decl): Declare.
5421         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5422         reserved via global_regs.
5424 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5426         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5428 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5430         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5431         and vmulosh rather than call gen_vec_widen_smult_*.
5432         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5433         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5434         (vec_widen_smult_even_v16qi): Likewise.
5435         (vec_widen_umult_even_v8hi): Likewise.
5436         (vec_widen_smult_even_v8hi): Likewise.
5437         (vec_widen_umult_odd_v16qi): Likewise.
5438         (vec_widen_smult_odd_v16qi): Likewise.
5439         (vec_widen_umult_odd_v8hi): Likewise.
5440         (vec_widen_smult_odd_v8hi): Likewise.
5441         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5442         vmuloub rather than call gen_vec_widen_umult_*.
5443         (vec_widen_umult_lo_v16qi): Likewise.
5444         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5445         vmulosb rather than call gen_vec_widen_smult_*.
5446         (vec_widen_smult_lo_v16qi): Likewise.
5447         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5448         rather than call gen_vec_widen_umult_*.
5449         (vec_widen_umult_lo_v8hi): Likewise.
5450         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5451         rather than call gen_vec_widen_smult_*.
5452         (vec_widen_smult_lo_v8hi): Likewise.
5454 2014-01-15  Jeff Law  <law@redhat.com>
5456         PR tree-optimization/59747
5457         * ree.c (find_and_remove_re): Properly handle case where a second
5458         eliminated extension requires widening a copy created for elimination
5459         of a prior extension.
5460         (combine_set_extension): Ensure that the number of hard regs needed
5461         for a destination register does not change when we widen it.
5463 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5465         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5466         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5467         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5468         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5469         (avr-*-rtems*): Likewise.
5470         (bfin*-rtems*): Likewise.
5471         (moxie-*-rtems*): Likewise.
5472         (h8300-*-rtems*): Likewise.
5473         (i[34567]86-*-rtems*): Likewise.
5474         (lm32-*-rtems*): Likewise.
5475         (m32r-*-rtems*): Likewise.
5476         (m68k-*-rtems*): Likewise.
5477         (microblaze*-*-rtems*): Likewise.
5478         (mips*-*-rtems*): Likewise.
5479         (powerpc-*-rtems*): Likewise.
5480         (sh-*-rtems*): Likewise.
5481         (sparc-*-rtems*): Likewise.
5482         (sparc64-*-rtems*): Likewise.
5483         (v850-*-rtems*): Likewise.
5484         (m32c-*-rtems*): Likewise.
5486 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5488         PR rtl-optimization/59511
5489         * ira.c (ira_init_register_move_cost): Use memory costs for some
5490         cases of register move cost calculations.
5491         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5492         instead of BB frequency.
5493         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5494         * lra-assigns.c (find_hard_regno_for): Ditto.
5496 2014-01-15  Richard Biener  <rguenther@suse.de>
5498         PR tree-optimization/59822
5499         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5500         (vectorizable_load): Use it to hoist defs of uses of invariant
5501         loads out of the loop.
5503 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5504             Kugan Vivekanandarajah  <kuganv@linaro.org>
5506         PR target/59695
5507         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5508         truncation.
5510 2014-01-15  Richard Biener  <rguenther@suse.de>
5512         PR rtl-optimization/59802
5513         * lcm.c (compute_available): Use inverted postorder to seed
5514         the initial worklist.
5516 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5518         PR target/59803
5519         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5520         ADDR_REGS for invalid symrefs in non-PIC code.
5522 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5524         PR other/58712
5525         * builtins.c (determine_block_size): Initialize *probable_max_size
5526         even if len_rtx is CONST_INT.
5528 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5530         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5531         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5532         (cortexa53_tunings): Likewise.
5533         (aarch64_sched_issue_rate): New function.
5534         (TARGET_SCHED_ISSUE_RATE): Define.
5536 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5538         * ira-costs.c (find_costs_and_classes): Add missed
5539         ira_init_register_move_cost_if_necessary.
5541 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5543         PR target/59787
5544         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5546 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5548         PR target/59794
5549         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5550         to indicate if type is used for function return value.  Warn ABI
5551         change if the vector mode isn't available for function return value.
5552         (ix86_function_arg_advance): Pass false to type_natural_mode.
5553         (ix86_function_arg): Likewise.
5554         (ix86_gimplify_va_arg): Likewise.
5555         (function_arg_32): Don't warn ABI change.
5556         (ix86_function_value): Pass true to type_natural_mode.
5557         (ix86_return_in_memory): Likewise.
5558         (ix86_struct_value_rtx): Removed.
5559         (TARGET_STRUCT_VALUE_RTX): Likewise.
5561 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5563         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5564         converting a conditional jump into a conditional return.
5566 2014-01-14  Richard Biener  <rguenther@suse.de>
5568         PR tree-optimization/58921
5569         PR tree-optimization/59006
5570         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5571         hoisting invariant stmts.
5572         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5573         invariant loads on the preheader edge if possible.
5575 2014-01-14  Joey Ye  <joey.ye@arm.com>
5577         * doc/plugin.texi (Building GCC plugins): Update to C++.
5579 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5581         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5582         (_mm_rcp28_round_ss): Ditto.
5583         (_mm_rsqrt28_round_sd): Ditto.
5584         (_mm_rsqrt28_round_ss): Ditto.
5585         (_mm_rcp28_sd): Ditto.
5586         (_mm_rcp28_ss): Ditto.
5587         (_mm_rsqrt28_sd): Ditto.
5588         (_mm_rsqrt28_ss): Ditto.
5589         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5590         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5591         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5592         (IX86_BUILTIN_RCP28SD): Ditto.
5593         (IX86_BUILTIN_RCP28SS): Ditto.
5594         (IX86_BUILTIN_RSQRT28SD): Ditto.
5595         (IX86_BUILTIN_RSQRT28SS): Ditto.
5596         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5597         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5598         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5599         (ix86_expand_special_args_builtin): Expand new FTYPE.
5600         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5601         (srcp14<mode>): Make insn unary.
5602         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5603         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5604         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5605         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5606         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5607         Fix rounding: make it SAE only.
5608         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5609         Ditto.
5610         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5611         Ditto.
5612         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5613         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5614         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5615         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5616         (round_saeonly_mask_scalar_operand4): Ditto.
5617         (round_saeonly_mask_scalar_op3): Ditto.
5618         (round_saeonly_mask_scalar_op4): Ditto.
5620 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5622         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5623         Implement -maltivec=be for vec_insert and vec_extract.
5625 2014-01-10  DJ Delorie  <dj@redhat.com>
5627         * config/msp430/msp430.md (call_internal): Don't allow memory
5628         references with SP as the base register.
5629         (call_value_internal): Likewise.
5630         * config/msp430/constraints.md (Yc): New.  For memory references
5631         that don't use SP as a base register.
5633         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5634         "an integer without a # prefix"
5635         * config/msp430/msp430.md (epilogue_helper): Use it.
5637 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5639         PR target/59617
5640         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5641         AVX512F gather builtins.
5642         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5643         on gather decls with INTEGER_TYPE masktype.
5644         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5645         directly into the builtin rather than hoisting it before loop.
5647         PR tree-optimization/59387
5648         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5649         (scev_const_prop): If folded_casts and type has undefined overflow,
5650         use force_gimple_operand instead of force_gimple_operand_gsi and
5651         for each added stmt if it is assign with
5652         arith_code_with_undefined_signed_overflow, call
5653         rewrite_to_defined_overflow.
5654         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5655         gimple-fold.h instead.
5656         (arith_code_with_undefined_signed_overflow,
5657         rewrite_to_defined_overflow): Moved to ...
5658         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5659         rewrite_to_defined_overflow): ... here.  No longer static.
5660         Include gimplify-me.h.
5661         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5662         rewrite_to_defined_overflow): New prototypes.
5664 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5666         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5668 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5670         * builtins.c (get_object_alignment_2): Minor tweak.
5671         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5673 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5675         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5676         optimized non constant lengths.
5678 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5680         PR libgomp/59194
5681         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5682         load as __atomic_load_N if possible.
5684 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5686         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5687         target parameter.
5688         (rs6000_expand_builtin): Adjust call.
5690 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5692         PR target/58115
5693         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5694         * config/rs6000/rs6000.c: Include target-globals.h.
5695         (rs6000_set_current_function): Instead of doing target_reinit
5696         unconditionally, use save_target_globals_default_opts and
5697         restore_target_globals.
5699         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5700         FPSCR.
5701         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5702         (rs6000_expand_builtin): Handle mffs and mtfsf.
5703         (rs6000_init_builtins): Define mffs and mtfsf.
5704         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5705         (rs6000_mffs): New pattern.
5706         (rs6000_mtfsf): New pattern.
5708 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5710         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5711         Start narrowing with START.  Apply candidate-use pair
5712         and check overall cost in narrowing.
5713         (iv_ca_prune): Pass new argument.
5715 2014-01-10  Jeff Law  <law@redhat.com>
5717         PR middle-end/59743
5718         * ree.c (combine_reaching_defs): Ensure the defining statement
5719         occurs before the extension when optimizing extensions with
5720         different source and destination hard registers.
5722 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5724         PR ipa/58585
5725         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5726         vtables into the type inheritance graph.
5728 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5730         PR rtl-optimization/59754
5731         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5732         modes in the REGNO != REGNO case.
5734 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5736         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5738 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5740         PR tree-optimization/59745
5741         * tree-predcom.c (tree_predictive_commoning_loop): Call
5742         free_affine_expand_cache if giving up because components is NULL.
5744         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5745         GC in payload of target_globals struct instead of allocating them on
5746         the heap and the larger structs separately using GC.
5747         * target-globals.h (struct target_globals): Make regs, hard_regs,
5748         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5749         of GTY((skip)) and change type to void *.
5750         (reset_target_globals): Cast loads from those fields to corresponding
5751         types.
5753 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5755         PR plugins/59335
5756         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5757         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5758         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5760 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5762         PR target/59744
5763         * aarch64-modes.def (CC_Zmode): New flags mode.
5764         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5765         represents an equality.
5766         (aarch64_get_condition_code): Handle CC_Zmode.
5767         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5769 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5771         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5772         extraction in good case.
5774 2014-01-10  Richard Biener  <rguenther@suse.de>
5776         PR tree-optimization/59374
5777         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5778         checking after SLP discovery.  Mark stmts not participating
5779         in any SLP instance properly.
5781 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5783         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5784         when handling a SET rtx.
5786 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5788         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5789         (cortex-a57): Likewise.
5790         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5792 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5794         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5795         non-iwmmxt builtins.
5797 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5799         PR ipa/58252
5800         PR ipa/59226
5801         * ipa-devirt.c record_target_from_binfo): Take as argument
5802         stack of binfos and lookup matching one for virtual inheritance.
5803         (possible_polymorphic_call_targets_1): Update.
5805 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5807         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5808         kernel strings for Loongson-2E/2F/3A.
5810 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5812         PR middle-end/59670
5813         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5814         is_gimple_call before calling gimple_call_internal_p.
5816 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5818         * Makefile.in (TREE_FLOW_H): Remove.
5819         (TREE_SSA_H): Add file names from tree-flow.h.
5820         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5821         * tree.h: Remove tree-flow.h reference.
5822         * hash-table.h: Remove tree-flow.h reference.
5823         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5824         reference with tree-ssa-loop.h.
5826 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5828         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5829         default element-order behavior for -maltivec.
5830         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5831         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5832         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5833         when targeting big endian, at least for now.
5834         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5836 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5838         PR middle-end/47735
5839         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5840         var satisfies use_register_for_decl, just take into account type
5841         alignment, rather than decl alignment.
5843         PR tree-optimization/59622
5844         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5845         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5846         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5847         Don't devirtualize for inplace at all.  For targets.length () == 1,
5848         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5850 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5852         * config/i386/i386.md (cpu): Remove the unused btver1.
5854 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5856         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5858 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5860         PR target/58115
5861         * tree-core.h (struct target_globals): New forward declaration.
5862         (struct tree_target_option): Add globals field.
5863         * tree.h (TREE_TARGET_GLOBALS): Define.
5864         (prepare_target_option_nodes_for_pch): New prototype.
5865         * target-globals.h (struct target_globals): Define even if
5866         !SWITCHABLE_TARGET.
5867         * tree.c (prepare_target_option_node_for_pch,
5868         prepare_target_option_nodes_for_pch): New functions.
5869         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5870         * config/i386/i386.c: Include target-globals.h.
5871         (ix86_set_current_function): Instead of doing target_reinit
5872         unconditionally, use save_target_globals_default_opts and
5873         restore_target_globals.
5875 2014-01-09  Richard Biener  <rguenther@suse.de>
5877         PR tree-optimization/59715
5878         * tree-cfg.h (split_critical_edges): Declare.
5879         * tree-cfg.c (split_critical_edges): Export.
5880         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5882 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5884         * cfgexpand.c (expand_stack_vars): Optionally disable
5885         asan stack protection.
5886         (expand_used_vars): Likewise.
5887         (partition_stack_vars): Likewise.
5888         * asan.c (asan_emit_stack_protection): Optionally disable
5889         after return stack usage.
5890         (instrument_derefs): Optionally disable memory access instrumentation.
5891         (instrument_builtin_call): Likewise.
5892         (instrument_strlen_call): Likewise.
5893         (asan_protect_global): Optionally disable global variables protection.
5894         * doc/invoke.texi: Added doc for new options.
5895         * params.def: Added new options.
5896         * params.h: Likewise.
5898 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5900         PR rtl-optimization/59724
5901         * ifcvt.c (cond_exec_process_if_block): Don't call
5902         flow_find_head_matching_sequence with 0 longest_match.
5903         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5904         non-active insns if !stop_after.
5905         (try_head_merge_bb): Revert 2014-01-07 changes.
5907 2014-01-08  Jeff Law  <law@redhat.com>
5909         * ree.c (get_sub_rtx): New function, extracted from...
5910         (merge_def_and_ext): Here.
5911         (combine_reaching_defs): Use get_sub_rtx.
5913 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5915         * cgraph.h (varpool_variable_node): Do not choke on null node.
5917 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5919         * config/mips/mips.md (simple_return): Attempt to use JRC
5920         for microMIPS.
5921         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5923 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5925         PR rtl-optimization/59137
5926         * reorg.c (steal_delay_list_from_target): Call update_block for
5927         elided insns.
5928         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5930 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5932         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5933         two duplicate entries.
5935 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5937         Revert:
5938         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5940         * config/mips/mips.c (mips_truncated_op_cost): New function.
5941         (mips_rtx_costs): Adjust test for BADDU.
5942         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5944         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5946         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5947         (*baddu_si): ...this new pattern.
5949 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5951         PR ipa/59722
5952         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5954 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5956         PR middle-end/57748
5957         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5958         inner_reference_p.
5959         (expand_expr, expand_normal): Adjust.
5960         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5961         inner_reference_p. Use inner_reference_p to expand inner references.
5962         (store_expr): Adjust.
5963         * cfgexpand.c (expand_call_stmt): Adjust.
5965 2014-01-08  Rong Xu  <xur@google.com>
5967         * gcov-io.c (gcov_var): Move from gcov-io.h.
5968         (gcov_position): Ditto.
5969         (gcov_is_error): Ditto.
5970         (gcov_rewrite): Ditto.
5971         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5972         only part to libgcc/libgcov.h.
5974 2014-01-08  Marek Polacek  <polacek@redhat.com>
5976         PR middle-end/59669
5977         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5979 2014-01-08  Marek Polacek  <polacek@redhat.com>
5981         PR sanitizer/59667
5982         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5984 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5986         PR rtl-optimization/59649
5987         * stor-layout.c (get_mode_bounds): For BImode return
5988         0 and STORE_FLAG_VALUE.
5990 2014-01-08  Richard Biener  <rguenther@suse.de>
5992         PR middle-end/59630
5993         * gimple.h (is_gimple_builtin_call): Remove.
5994         (gimple_builtin_call_types_compatible_p): New.
5995         (gimple_call_builtin_p): New overload.
5996         * gimple.c (is_gimple_builtin_call): Remove.
5997         (validate_call): Rename to ...
5998         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5999         check return types.
6000         (validate_type): New static function.
6001         (gimple_call_builtin_p): New overload and adjust.
6002         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
6003         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
6004         (gimple_fold_stmt_to_constant_1): Likewise.
6005         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
6007 2014-01-08  Richard Biener  <rguenther@suse.de>
6009         PR middle-end/59471
6010         * gimplify.c (gimplify_expr): Gimplify register-register type
6011         VIEW_CONVERT_EXPRs to separate stmts.
6013 2014-01-07  Jeff Law  <law@redhat.com>
6015         PR middle-end/53623
6016         * ree.c (combine_set_extension): Handle case where source
6017         and destination registers in an extension insn are different.
6018         (combine_reaching_defs): Allow source and destination registers
6019         in extension to be different under limited circumstances.
6020         (add_removable_extension): Remove restriction that the
6021         source and destination registers in the extension are the same.
6022         (find_and_remove_re): Emit a copy from the extension's
6023         destination to its source after the defining insn if
6024         the source and destination registers are different.
6026         PR middle-end/59285
6027         * ifcvt.c (merge_if_block): If we are merging a block with more than
6028         one successor with a block with no successors, remove any BARRIER
6029         after the second block.
6031 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
6033         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
6035 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
6037         PR target/59652
6038         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
6039         for 14-bit register offsets when INT14_OK_STRICT is false.
6041 2014-01-07  Roland Stigge  <stigge@antcom.de>
6042             Michael Meissner  <meissner@linux.vnet.ibm.com>
6044         PR 57386/target
6045         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
6046         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
6048 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
6050         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
6051         -mcpu.
6053 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
6055         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
6056         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
6057         rtx is const0_rtx or not.
6059 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
6061         PR target/58115
6062         * target-globals.c (save_target_globals): Remove this_fn_optab
6063         handling.
6064         * toplev.c: Include optabs.h.
6065         (target_reinit): Temporarily restore the global options if another
6066         set of options are in force.
6068 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
6070         PR rtl-optimization/58668
6071         * cfgcleanup.c (flow_find_cross_jump): Don't count
6072         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
6073         to determine what is counted.
6074         (flow_find_head_matching_sequence): Use active_insn_p to determine
6075         what is counted.
6076         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
6077         counting change.
6078         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
6079         determine what is counted.
6081         PR tree-optimization/59643
6082         * tree-predcom.c (split_data_refs_to_components): If one dr is
6083         read and one write, determine_offset fails and the write isn't
6084         in the bad component, just put the read into the bad component.
6086 2014-01-07  Mike Stump  <mikestump@comcast.net>
6087             Jakub Jelinek  <jakub@redhat.com>
6089         PR pch/59436
6090         * tree-core.h (struct tree_optimization_option): Change optabs
6091         type from unsigned char * to void *.
6092         * optabs.c (init_tree_optimization_optabs): Adjust
6093         TREE_OPTIMIZATION_OPTABS initialization.
6095 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
6097         PR target/59644
6098         * config/i386/i386.h (struct machine_function): Add
6099         no_drap_save_restore field.
6100         * config/i386/i386.c (ix86_save_reg): Use
6101         !cfun->machine->no_drap_save_restore instead of
6102         crtl->stack_realign_needed.
6103         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
6104         this function clears frame_pointer_needed.  Set
6105         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
6106         and DRAP reg is needed.
6108 2014-01-06  Marek Polacek  <polacek@redhat.com>
6110         PR c/57773
6111         * doc/implement-c.texi: Mention that other integer types are
6112         permitted as bit-field types in strictly conforming mode.
6114 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
6116         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
6117         is newly allocated.
6119 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
6121         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
6123 2014-01-06  Martin Jambor  <mjambor@suse.cz>
6125         PR ipa/59008
6126         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
6127         to int.
6128         * ipa-prop.c (ipa_print_node_params): Fix indentation.
6130 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
6132         PR debug/59350
6133         PR debug/59510
6134         * var-tracking.c (add_stores): Preserve the value of the source even if
6135         we don't record the store.
6137 2014-01-06  Terry Guo  <terry.guo@arm.com>
6139         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
6141 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
6143         PR bootstrap/59541
6144         * config/darwin.c (darwin_function_section): Adjust return values to
6145         correspond to optimisation changes made in r206070.
6147 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
6149         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
6150         from prefetch_block tune setting.
6151         (nocona_cost): Correct size of prefetch block to 64.
6153 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
6155         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
6156         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
6157         used to save the static chain register in the computation of the offset
6158         from which the FP registers need to be restored.
6160 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
6162         PR tree-optimization/59519
6163         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
6164         ICE if get_current_def (current_new_name) is already non-NULL, as long
6165         as it is a phi result of some other phi in *new_exit_bb that has
6166         the same argument.
6168         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
6169         or vmovdqu* for misaligned_operand.
6170         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
6171         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6172         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6173         aligned_mem for AVX512F masked aligned load and store builtins and for
6174         non-temporal moves.
6176 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6178         PR tree-optimization/59651
6179         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6180         Address range for negative step should be added by TYPE_SIZE_UNIT.
6182 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6184         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6185         overlapping registers also for registers other than the stack pointer.
6187 2014-01-03  Marek Polacek  <polacek@redhat.com>
6189         PR other/59661
6190         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6191         __builtin_FILE.
6193 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6195         PR target/59625
6196         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6197         asm goto as jump.
6199         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6200         (push splitter): Use <P:MODE_SIZE> instead of
6201         GET_MODE_SIZE (<P:MODE>mode).
6202         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6203         (mov -1, reg peephole2): Likewise.
6204         * config/i386/sse.md (*mov<mode>_internal,
6205         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6206         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6207         *<code><mode>3, *andnot<mode>3<mask_name>,
6208         <mask_codefor><code><mode>3<mask_name>): Likewise.
6209         * config/i386/subst.md (mask_mode512bit_condition,
6210         sd_mask_mode512bit_condition): Likewise.
6212 2014-01-02  Xinliang David Li  <davidxl@google.com>
6214         PR tree-optimization/59303
6215         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6216         (dump_predicates): Better output format.
6217         (pred_equal_p): New function.
6218         (is_neq_relop_p): Ditto.
6219         (is_neq_zero_form_p): Ditto.
6220         (pred_expr_equal_p): Ditto.
6221         (pred_neg_p): Ditto.
6222         (simplify_pred): Ditto.
6223         (simplify_preds_2): Ditto.
6224         (simplify_preds_3): Ditto.
6225         (simplify_preds_4): Ditto.
6226         (simplify_preds): Ditto.
6227         (push_pred): Ditto.
6228         (push_to_worklist): Ditto.
6229         (get_pred_info_from_cmp): Ditto.
6230         (is_degenerated_phi): Ditto.
6231         (normalize_one_pred_1): Ditto.
6232         (normalize_one_pred): Ditto.
6233         (normalize_one_pred_chain): Ditto.
6234         (normalize_preds): Ditto.
6235         (normalize_cond_1): Remove function.
6236         (normalize_cond): Ditto.
6237         (is_gcond_subset_of): Ditto.
6238         (is_subset_of_any): Ditto.
6239         (is_or_set_subset_of): Ditto.
6240         (is_and_set_subset_of): Ditto.
6241         (is_norm_cond_subset_of): Ditto.
6242         (pred_chain_length_cmp): Ditto.
6243         (convert_control_dep_chain_into_preds): Type change.
6244         (find_predicates): Ditto.
6245         (find_def_preds): Ditto.
6246         (destroy_predicates_vecs): Ditto.
6247         (find_matching_predicates_in_rest_chains): Ditto.
6248         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6249         (is_pred_expr_subset): Ditto.
6250         (is_pred_chain_subset_of): Ditto.
6251         (is_included_in): Ditto.
6252         (is_superset_of): Ditto.
6254 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6256         Update copyright years.
6258 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6260         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6261         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6262         config/arc/arc.md, config/arc/arc.opt,
6263         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6264         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6265         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6266         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6267         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6268         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6269         vtable-verify.c, vtable-verify.h: Use the standard form for the
6270         copyright notice.
6272 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6274         * gcc.c (process_command): Update copyright notice dates.
6275         * gcov-dump.c: Ditto.
6276         * gcov.c: Ditto.
6277         * doc/cpp.texi: Bump @copying's copyright year.
6278         * doc/cppinternals.texi: Ditto.
6279         * doc/gcc.texi: Ditto.
6280         * doc/gccint.texi: Ditto.
6281         * doc/gcov.texi: Ditto.
6282         * doc/install.texi: Ditto.
6283         * doc/invoke.texi: Ditto.
6285 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6287         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6289 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6291         * config/i386/sse.md (*mov<mode>_internal): Guard
6292         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6294         PR rtl-optimization/59647
6295         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6296         new_rtx into UNSIGNED_FLOAT rtxes.
6298 Copyright (C) 2014 Free Software Foundation, Inc.
6300 Copying and distribution of this file, with or without modification,
6301 are permitted in any medium without royalty provided the copyright
6302 notice and this notice are preserved.