PR middle-end/60556
[official-gcc.git] / gcc / ChangeLog
blob116d6c4d6cc636db6dff82f7e8ac4de21f580b5e
1 2014-04-11  Steve Ellcey  <sellcey@mips.com>
2             Jakub Jelinek  <jakub@redhat.com>
4         PR middle-end/60556
5         * expr.c (convert_move): Use emit_store_flag_force instead of
6         emit_store_flag.  Pass lowpart_mode instead of VOIDmode as 5th
7         argument to it.
9 2014-04-11  Richard Biener  <rguenther@suse.de>
11         PR middle-end/60797
12         * varasm.c (assemble_alias): Avoid endless error reporting
13         recursion by setting TREE_ASM_WRITTEN.
15 2014-04-11  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
17         * config/s390/s390.md: Add a splitter for NOT rtx.
19 2014-04-11  Jakub Jelinek  <jakub@redhat.com>
21         PR rtl-optimization/60663
22         * cse.c (cse_insn): Set src_volatile on ASM_OPERANDS in
23         PARALLEL.
25 2014-04-10  Jan Hubicka  <hubicka@ucw.cz>
26             Jakub Jelinek  <jakub@redhat.com>
28         PR lto/60567
29         * ipa.c (function_and_variable_visibility): Copy forced_by_abi flag from
30         decl_node to node.
32 2014-04-10  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34         PR debug/60655
35         * config/arm/arm.c (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Define
36         (arm_const_not_ok_for_debug_p): Reject MINUS with SYM_REF's
37         ameliorating the cases where it can be.
39 2014-04-09  David Edelsohn  <dje.gcc@gmail.com>
41         Revert
42         2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
44         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
45         (loadsync_<mode>): Change mode.
46         (load_quadpti, store_quadpti): New.
47         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
48         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
49         * config/rs6000/predicates.md (quad_memory_operand): !TARGET_SYNC_TI.
51 2014-04-09  Cong Hou  <congh@google.com>
53         PR testsuite/60773
54         * doc/sourcebuild.texi (vect_widen_mult_si_to_di_pattern): Add
55         documentation.
57 2014-04-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
59         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Use vnand
60         instead of vnor to exploit possible fusion opportunity in the
61         future.
62         (altivec_expand_vec_perm_const_le): Likewise.
64 2014-04-08  Pat Haugen  <pthaugen@us.ibm.com>
66         * config/rs6000/sync.md (AINT mode_iterator): Move definition.
67         (loadsync_<mode>): Change mode.
68         (load_quadpti, store_quadpti): New.
69         (atomic_load<mode>, atomic_store<mode>): Add support for TI mode.
70         * config/rs6000/rs6000.md (unspec enum): Add UNSPEC_LSQ.
72 2014-04-08  Richard Sandiford  <rdsandiford@googlemail.com>
74         PR target/60763
75         * config/rs6000/vsx.md (vsx_xscvdpspn_scalar): Change input to DImode.
76         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Update accordingly.
77         Use gen_rtx_REG rather than simplify_gen_subreg for op0_di.
79 2014-04-08  Richard Biener  <rguenther@suse.de>
81         PR middle-end/60706
82         * tree-pretty-print.c (pp_double_int): For HWI32 hosts with
83         a 64bit widest int print double-int similar to on HWI64 hosts.
85 2014-04-08  Richard Biener  <rguenther@suse.de>
87         PR tree-optimization/60785
88         * graphite-sese-to-poly.c (rewrite_phi_out_of_ssa): Treat
89         default defs properly.
91 2014-04-08  Nathan Sidwell  <nathan@codesourcery.com>
93         * doc/invoke (Wnon-virtual-dtor): Update to match implementation.
94         (Weffc++): Likewise.
96 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
98         * ipa-devirt.c (maybe_record_node): When node is not recorded,
99         set completep to false rather than true.
101 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
103         PR target/60504
104         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
105         ARM_TARGET2_DWARF_FORMAT.
107 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
109         PR target/60609
110         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
111         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
112         ADDR_DIFF_VEC.
114 2014-04-07  Richard Biener  <rguenther@suse.de>
116         PR tree-optimization/60766
117         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
118         (may_eliminate_iv): Convert cand_value_at result to desired type.
120 2014-04-07  Jason Merrill  <jason@redhat.com>
122         PR c++/60731
123         * common.opt (-fno-gnu-unique): Add.
124         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
126 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
128         * haifa-sched.c: Fix outdated function reference and minor
129         grammar errors in introductory comment.
131 2014-04-07  Richard Biener  <rguenther@suse.de>
133         PR middle-end/60750
134         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
135         for noreturn calls.
136         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
138 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
140         PR debug/55794
141         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
142         size accounting for thunks.
143         (pa_asm_output_mi_thunk): Use final_start_function() and
144         final_end_function() to output function start and end directives.
146 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
148         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
149         specific ISA/ feature information. Remove short_sp and errata_skip ds.
150         Add avr_device_specific_features enum to have device specific info.
151         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
152         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
153         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
154         updated device specific info.
155         * config/avr/avr-mcus.def: Merge device specific details to
156         dev_attribute field.
157         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
158         errata_skip.
159         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
160         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
161         assembler if RMW isa supported by current device.
162         * config/avr/genmultilib.awk: Update as device info structure changed.
163         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
165 2014-04-04  Cong Hou  <congh@google.com>
167         PR tree-optimization/60656
168         * tree-vect-stmts.c (supportable_widening_operation):
169         Fix a bug that elements in a vector with vect_used_by_reduction
170         property are incorrectly reordered when the operation on it is not
171         consistant with the one in reduction operation.
173 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
175         PR rtl-optimization/60155
176         * gcse.c (record_set_data): New function.
177         (single_set_gcse): New function.
178         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
179         (hoist_code): Likewise.
180         (get_pressure_class_and_nregs): Likewise.
182 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
184         * explow.c (probe_stack_range): Emit a final optimization blockage.
186 2014-04-04  Anthony Green  <green@moxielogic.com>
188         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
189         typos.
191 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
193         PR ipa/59626
194         * lto-cgraph.c (input_overwrite_node): Check that partitioning
195         flags are set only during streaming.
196         * ipa.c (process_references, walk_polymorphic_call_targets,
197         symtab_remove_unreachable_nodes): Drop bodies of always inline
198         after early inlining.
199         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
201 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
202         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
204         PR debug/60655
205         * dwarf2out.c (const_ok_for_output_1): Reject expressions
206         containing a NOT.
208 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
210         PR bootstrap/60743
211         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
212         duration.
213         (cortex_a53_fdivd): Likewise.
215 2014-04-04  Martin Jambor  <mjambor@suse.cz>
217         PR ipa/60640
218         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
219         Adjust all callers.
220         * cgraph.c (clone_of_p): Also return true if thunks match.
221         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
222         cgraph_function_or_thunk_node and an obsolete comment.
223         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
224         file.
225         (build_function_decl_skip_args): Likewise.
226         (set_new_clone_decl_and_node_flags): New function.
227         (duplicate_thunk_for_node): Likewise.
228         (redirect_edge_duplicating_thunks): Likewise.
229         (cgraph_clone_node): New parameter args_to_skip, pass it to
230         redirect_edge_duplicating_thunks which is called instead of
231         cgraph_redirect_edge_callee.
232         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
233         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
235 2014-04-04  Jeff Law  <law@redhat.com>
237         PR target/60657
238         * config/arm/predicates.md (const_int_I_operand): New predicate.
239         (const_int_M_operand): Similarly.
240         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
241         const_int_operand.
242         (insv_t2, extv_reg, extzv_t2): Likewise.
243         (load_multiple_with_writeback): Similarly for const_int_I_operand.
244         (pop_multiple_with_writeback_and_return): Likewise.
245         (vfp_pop_multiple_with_writeback): Likewise
247 2014-04-04  Richard Biener  <rguenther@suse.de>
249         PR ipa/60746
250         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
251         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
252         non-GIMPLE_LABELs.
253         * gimplify.h (gimple_add_tmp_var_fn): Declare.
254         * gimplify.c (gimple_add_tmp_var_fn): New function.
255         * gimple-expr.h (create_tmp_reg_fn): Declare.
256         * gimple-expr.c (create_tmp_reg_fn): New function.
257         * gimple-low.c (record_vars_into): Don't change cfun.
258         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
259         code generation without cfun.
261 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
263         PR bootstrap/60719
264         * Makefile.in (install-driver): Fix shell scripting.
266 2014-04-03  Cong Hou  <congh@google.com>
268         PR tree-optimization/60505
269         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
270         threshold of number of iterations below which no vectorization will be
271         done.
272         * tree-vect-loop.c (new_loop_vec_info):
273         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
274         * tree-vect-loop.c (vect_analyze_loop_operations):
275         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
276         * tree-vect-loop.c (vect_transform_loop):
277         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
278         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
279         of iterations of the loop and see if we should build the epilogue.
281 2014-04-03  Richard Biener  <rguenther@suse.de>
283         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
284         member.
285         (streamer_tree_cache_create): Adjust.
286         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
287         to allow optional nodes array.
288         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
289         (streamer_tree_cache_append): Likewise.
290         (streamer_tree_cache_create): Create nodes array optionally
291         as specified by parameter.
292         * lto-streamer-out.c (create_output_block): Avoid maintaining
293         the node array in the writer cache.
294         (DFS_write_tree): Remove assertion.
295         (produce_asm_for_decls): Free the out decl state hash table
296         early.
297         * lto-streamer-in.c (lto_data_in_create): Adjust for
298         streamer_tree_cache_create prototype change.
300 2014-04-03  Richard Biener  <rguenther@suse.de>
302         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
303         set TREE_CHAIN to NULL_TREE.
305 2014-04-03  Richard Biener  <rguenther@suse.de>
307         PR tree-optimization/60740
308         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
309         over all GIMPLE_COND operands.
311 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
313         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
314         (Weffc++): Remove Scott's numbering, merge lists and reference
315         Wnon-virtual-dtor.
317         c-family/
319         cp/
320         * class.c (accessible_nvdtor_p): New.
321         (check_bases): Don't check base destructor here ...
322         (check_bases_and_members): ... check them here.  Trigger on
323         Wnon-virtual-dtor flag.
324         (finish_struct_1): Use accessible_nvdtor_p.
326         testsuite/
327         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
328         * g++.dg/warn/Wnvdtor-2.C: New.
329         * g++.dg/warn/Wnvdtor-3.C: New.
330         * g++.dg/warn/Wnvdtor-4.C: New.
331         * g++.dg/warn/Weff1.C: Delete.
332         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
333         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
335 2014-04-03  Nick Clifton  <nickc@redhat.com>
337         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
338         properly.
340 2014-04-03  Martin Jambor  <mjambor@suse.cz>
342         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
343         mention gcc_unreachable before failing.
344         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
345         removed symbols.
347 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
349         PR ipa/60659
350         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
351         code and instead mark the context inconsistent.
352         (possible_polymorphic_call_targets): For inconsistent contexts
353         return empty complete list.
355 2014-04-02  Anthony Green  <green@moxielogic.com>
357         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
358         (extendqisi2, extendhisi2): Define.
359         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
360         (WCHAR_TYPE): Change to unsigned int.
362 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
364         PR tree-optimization/60733
365         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
366         insertion point for PHI candidates to be the end of the feeding
367         block for the PHI argument.
369 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
371         PR rtl-optimization/60650
372         * lra-constraints.c (process_alt_operands): Decrease reject for
373         earlyclobber matching.
375 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
377         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
379 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
381         * config/spu/spu.c (pad_bb): Do not crash when the last
382         insn is CODE_FOR_blockage.
384 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
386         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
387         lies outside the target mode.
389 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
391         PR target/60735
392         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
393         software floating point or no floating point registers, do not
394         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
395         in GPRs that occurs after we tested for GPRs that would never be
396         true.
398         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
399         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
400         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
401         specifically allow DDmode, since that does not use the SPE SIMD
402         instructions.
404 2014-04-02  Richard Biener  <rguenther@suse.de>
406         PR middle-end/60729
407         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
408         MODE_INTs.  Properly use negv_optab.
409         (expand_abs): Likewise.
411 2014-04-02  Richard Biener  <rguenther@suse.de>
413         PR bootstrap/60719
414         * Makefile.in (install-driver): Guard extra installs with special
415         names properly.
417 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
419         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
420         Document vec_vgbbd.
422 2014-04-01  Richard Henderson  <rth@redhat.com>
424         PR target/60704
425         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
426         alternative enabled before register allocation.
428 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
430         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
431         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
432         typo.
433         (nios2_large_got_address): Remove unneeded 'sym' parameter.
434         (nios2_got_address): Update nios2_large_got_address call site.
435         (nios2_delegitimize_address): New function.
436         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
437         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
438         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
440 2014-04-01  Martin Husemann  <martin@duskware.de>
442         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
443         for -mabi=32.
445 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
447         PR rtl-optimization/60604
448         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
449         check from register_operand.
450         (register_operand): Redefine in terms of general_operand.
451         (nonmemory_operand): Use register_operand for the non-constant cases.
453 2014-04-01  Richard Biener  <rguenther@suse.de>
455         * gimple.h (struct gimple_statement_base): Align subcode to
456         16 bits.
458 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
460         * doc/invoke.texi (mapp-regs): Clarify.
462 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
464         * config/i386/avx512fintrin.h (__v32hi): Define type.
465         (__v64qi): Likewise.
466         (_mm512_set1_epi8): Define.
467         (_mm512_set1_epi16): Define.
468         (_mm512_set4_epi32): Define.
469         (_mm512_set4_epi64): Define.
470         (_mm512_set4_pd): Define.
471         (_mm512_set4_ps): Define.
472         (_mm512_setr4_epi64): Define.
473         (_mm512_setr4_epi32): Define.
474         (_mm512_setr4_pd): Define.
475         (_mm512_setr4_ps): Define.
476         (_mm512_setzero_epi32): Define.
478 2014-03-31  Martin Jambor  <mjambor@suse.cz>
480         PR middle-end/60647
481         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
482         callsite_arguments_match_p.  Updated all callers.  Also check types of
483         corresponding formal parameters and actual arguments.
484         (not_all_callers_have_enough_arguments_p) Renamed to
485         some_callers_have_mismatched_arguments_p.
487 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
489         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
491 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
493         PR target/60034
494         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
495         section anchor.
497 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
499         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
500         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
501         Split out
502         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
503         Use FMAMODE_NOVF512 mode iterator.
504         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
505         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
506         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
507         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
508         Split out
509         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
510         Use VF_128_256 mode iterator.
511         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
512         Ditto.
514 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
516         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
517         static chain if needed.
519 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
521         PR target/60697
522         * lra-constraints.c (index_part_to_reg): New.
523         (process_address): Use it.
525 2014-03-27  Jeff Law  <law@redhat.com>
526             Jakub Jelinek  <jakub@redhat.com>
528         PR target/60648
529         * expr.c (do_tablejump): Use simplify_gen_binary rather than
530         gen_rtx_{PLUS,MULT} to build up the address expression.
532         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
533         creating non-canonical RTL.
535 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
537         PR ipa/60243
538         * ipa-inline.c (want_inline_small_function_p): Short circuit large
539         functions; reorganize to make cheap checks first.
540         (inline_small_functions): Do not estimate growth when dumping;
541         it is expensive.
542         * ipa-inline.h (inline_summary): Add min_size.
543         (growth_likely_positive): New function.
544         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
545         (set_cond_stmt_execution_predicate): Cleanup.
546         (estimate_edge_size_and_time): Compute min_size.
547         (estimate_calls_size_and_time): Likewise.
548         (estimate_node_size_and_time): Likewise.
549         (inline_update_overall_summary): Update min_size.
550         (do_estimate_edge_time): Likewise.
551         (do_estimate_edge_size): Update.
552         (do_estimate_edge_hints): Update.
553         (growth_likely_positive): New function.
555 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
557         PR target/60693
558         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
559         also if addr has VOIDmode.
561 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
563         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
564         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
565         Declare extern.
566         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
567         instructions as well as AdvancedSIMD loads.
569 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
571         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
572         Use crypto_aese type.
573         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
574         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
575         crypto_aese, crypto_aesmc.  Move to types.md.
576         * config/arm/types.md (crypto_aes): Split into crypto_aese,
577         crypto_aesmc.
578         * config/arm/iterators.md (crypto_type): Likewise.
580 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
582         * cgraph.c: Include expr.h and tree-dfa.h.
583         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
584         remove LHS.
586 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
588         PR target/60675
589         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
590         regs from checking multi-reg pseudos.
592 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
594         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
596 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
598         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
599         if it would clobber the stack pointer, even temporarily.
601 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
603         * mode-switching.c: Make small adjustments to the top comment.
605 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
607         * config/rs6000/constraints.md (wD constraint): New constraint to
608         match the constant integer to get the top DImode/DFmode out of a
609         vector in a VSX register.
611         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
612         match the constant integer to get the top DImode/DFmode out of a
613         vector in a VSX register.
615         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
616         for ISA 2.07.
618         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
619         vbpermq builtins.
621         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
622         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
624         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
625         Optimize vec_extract of 64-bit values, where the value being
626         extracted is in the top word, where we can use scalar
627         instructions.  Add direct move and store support.  Combine the big
628         endian/little endian vector select load support into a single insn.
629         (vsx_extract_<mode>_internal1): Likewise.
630         (vsx_extract_<mode>_internal2): Likewise.
631         (vsx_extract_<mode>_load): Likewise.
632         (vsx_extract_<mode>_store): Likewise.
633         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
634         combined into vsx_extract_<mode>_load.
635         (vsx_extract_<mode>_one_le): Likewise.
637         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
638         define the top 64-bit vector element.
640         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
641         constraint.
643         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
644         Document vec_vbpermq builtin.
646         PR target/60672
647         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
648         enable use of xxsldwi and xxpermdi builtin functions.
649         (vec_xxpermdi): Likewise.
651         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
652         Document use of vec_xxsldwi and vec_xxpermdi builtins.
654 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
656         PR rtl-optimization/60650
657         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
658         first_p.  Use it.
659         (find_spills_for): New.
660         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
661         Spill all pseudos on the second iteration.
663 2014-03-27  Marek Polacek  <polacek@redhat.com>
665         PR c/50347
666         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
667         types.
669 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
671         * config/s390/s390.c (s390_can_use_return_insn): Check for
672         call-saved FPRs on 31 bit.
674 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
676         PR middle-end/60682
677         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
678         if they need regimplification, just drop them instead of
679         calling gimple_regimplify_operands on them.
681 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
683         PR target/60580
684         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
685         (aarch64_frame_pointer_required): Adjust logic.
686         (aarch64_can_eliminate): Adjust logic.
687         (aarch64_override_options_after_change): Adjust logic.
689 2014-03-27  Dehao Chen  <dehao@google.com>
691         * ipa-inline.c (early_inliner): Update node's inline info.
693 2014-03-26  Dehao Chen  <dehao@google.com>
695         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
696         compiler inserted conditional jumps for NAN float check.
698 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
700         * ubsan.h (ubsan_create_data): Change second argument's type
701         to const location_t *.
702         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
703         _("<unknown>").
704         (ubsan_create_data): Change second argument to const location_t *PLOC.
705         Create Loc field whenever PLOC is non-NULL.
706         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
707         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
708         callers.
710         PR other/59545
711         * real.c (real_to_integer2): Change type of low to UHWI.
713 2014-03-26  Tobias Burnus  <burnus@net-b.de>
715         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
716         (CILK_SELF_SPECS): New define.
717         (driver_self_specs): Use it.
719 2014-03-26  Richard Biener  <rguenther@suse.de>
721         * tree-pretty-print.c (percent_K_format): Implement special
722         case for LTO and its stripped down BLOCK tree.
724 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
726         PR sanitizer/60636
727         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
729         * tree-vrp.c (simplify_internal_call_using_ranges): If only
730         one range is range_int_cst_p, but not both, at least optimize
731         addition/subtraction of 0 and multiplication by 0 or 1.
732         * gimple-fold.c (gimple_fold_call): Fold
733         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
734         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
735         INTEGER_CSTs, try to fold at least x * 0 and y - y.
737 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
739         PR rtl-optimization/60452
740         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
741         <case REG>: Return 1 for invalid offsets from the frame pointer.
743 2014-03-26  Marek Polacek  <polacek@redhat.com>
745         PR c/37428
746         * doc/extend.texi (C Extensions): Mention variable-length arrays in
747         a structure/union.
749 2014-03-26  Marek Polacek  <polacek@redhat.com>
751         PR c/39525
752         * doc/extend.texi (Designated Inits): Describe what happens to omitted
753         field members.
755 2014-03-26  Marek Polacek  <polacek@redhat.com>
757         PR other/59545
758         * ira-color.c (update_conflict_hard_regno_costs): Perform the
759         multiplication in unsigned type.
761 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
763         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
765 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
767         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
769 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
771         PR ipa/60315
772         * cif-code.def (UNREACHABLE) New code.
773         * ipa-inline.c (inline_small_functions): Skip edges to
774         __builtlin_unreachable.
775         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
776         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
777         predicate to __bulitin_unreachable.
778         (set_cond_stmt_execution_predicate): Fix issue when
779         invert_tree_comparison returns ERROR_MARK.
780         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
781         propagate to inline clones.
782         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
783         to unreachable.
784         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
785         * cgraphclones.c (cgraph_clone_node): If call destination is already
786         ureachable, do not redirect it back.
787         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
788         unreachable.
790 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
792         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
793         Do not modify inline clones.
795 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
797         * config/i386/i386.md (general_sext_operand): New mode attr.
798         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
799         don't generate (sign_extend (const_int)).
800         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
801         operands[2].  Use We constraint instead of <i> and
802         <general_sext_operand> predicate instead of <general_operand>.
803         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
804         * config/i386/constraints.md (We): New constraint.
805         * config/i386/predicates.md (x86_64_sext_operand,
806         sext_operand): New predicates.
808 2014-03-25  Martin Jambor  <mjambor@suse.cz>
810         PR ipa/60600
811         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
812         inconsistent devirtualizations to __builtin_unreachable.
814 2014-03-25  Marek Polacek  <polacek@redhat.com>
816         PR c/35449
817         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
819 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
821         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
822         order of elements for big-endian.
824 2014-03-25  Richard Biener  <rguenther@suse.de>
826         PR middle-end/60635
827         * gimplify-me.c (gimple_regimplify_operands): Update the
828         re-gimplifed stmt.
830 2014-03-25  Martin Jambor  <mjambor@suse.cz>
832         PR ipa/59176
833         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
834         (lto_output_varpool_node): Likewise.
835         (input_overwrite_node): Likewise.
836         (input_varpool_node): Likewise.
838 2014-03-25  Richard Biener  <rguenther@suse.de>
840         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
841         (run_gcc): Likewise.
843 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
845         * combine.c (simplify_compare_const): Add MODE argument.
846         Handle mode_width 0 as very large mode_width.
847         (try_combine, simplify_comparison): Adjust callers.
849         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
850         type to avoid signed integer overflow.
851         * explow.c (plus_constant): Likewise.
853 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
855         * doc/generic.texi: Correct typos.
857 2014-03-24  Tobias Burnus  <burnus@net-b.de>
859         * doc/invoke.texi (-flto): Expand section about
860         using static libraries with LTO.
862 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
864         PR rtl-optimization/60501
865         * optabs.def (addptr3_optab): New optab.
866         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
867         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
868         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
870         * lra.c (emit_add3_insn): Use the addptr pattern if available.
872         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
874 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
876         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
877         _mm512_set1_pd.
879         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
880         (_mm256_undefined_ps): Define.
881         (_mm256_undefined_pd): Define.
882         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
883         (_mm_undefined_pd): Define.
884         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
885         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
886         (_mm512_undefined_ps): Define.
887         (_mm512_undefined_pd): Define.
888         Use _mm*_undefined_*.
889         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
891 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
893         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
894         (lshr_simd): DI mode added.
895         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
896         (aarch64_ushr_simddi): Likewise.
897         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
898         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
899         (vshrd_n_u64): Likewise.
901 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
903         * Makefile.in (s-macro_list): Depend on cc1.
905 2014-03-23  Teresa Johnson  <tejohnson@google.com>
907         * ipa-utils.c (ipa_print_order): Use specified dump file.
909 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
911         PR rtl-optimization/60601
912         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
914         * gcc.c (eval_spec_function): Initialize save_growing_value.
916 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
918         PR sanitizer/60613
919         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
920         code == MINUS_EXPR, never swap op0 with op1.
922         * toplev.c (init_local_tick): Avoid signed integer multiplication
923         overflow.
924         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
925         shift by first operand's bitsize.
927 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
929         PR target/60610
930         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
931         redefine to 1 or 0.
932         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
933         TARGET_ISA_64BIT_P(x).
935 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
937         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
938         pattern for vector nor instead of subtract from splat(-1).
939         (altivec_expand_vec_perm_const_le): Likewise.
941 2014-03-21  Richard Henderson  <rth@twiddle.net>
943         PR target/60598
944         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
945         related insns after epilogue_completed.
947 2014-03-21  Martin Jambor  <mjambor@suse.cz>
949         PR ipa/59176
950         * cgraph.h (symtab_node): New flag body_removed.
951         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
952         when removing bodies.
953         * symtab.c (dump_symtab_base): Dump body_removed flag.
954         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
955         had their bodies removed.
957 2014-03-21  Martin Jambor  <mjambor@suse.cz>
959         PR ipa/60419
960         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
961         in the border.
963 2014-03-21  Richard Biener  <rguenther@suse.de>
965         PR tree-optimization/60577
966         * tree-core.h (struct tree_base): Document nothrow_flag use
967         in VAR_DECL_NONALIASED.
968         * tree.h (VAR_DECL_NONALIASED): New.
969         (may_be_aliased): Adjust.
970         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
972 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
974         * expr.c (expand_expr_real_1): Remove outdated comment.
976 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
978         PR middle-end/60597
979         * ira.c (adjust_cleared_regs): Call copy_rtx on
980         *reg_equiv[REGNO (loc)].src_p before passing it to
981         simplify_replace_fn_rtx.
983         PR target/60568
984         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
985         into CONST, put pic register as first operand of PLUS.  Use
986         gen_const_mem for both 32-bit and 64-bit PIC got loads.
988 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
990         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
992 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
994         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
995         around for store forwarding issue in the FPU on the UT699.
996         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
997         loads and operations if -mfix-ut699 is specified.
998         (divtf3_hq): Tweak attribute.
999         (sqrttf2_hq): Likewise.
1001 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
1003         * calls.c (store_one_arg): Remove incorrect const qualification on the
1004         type of the temporary.
1005         * cfgexpand.c (expand_return): Likewise.
1006         * expr.c (expand_constructor): Likewise.
1007         (expand_expr_real_1): Likewise.
1009 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
1011         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
1012         of parts.
1014 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
1016         PR target/60039
1017         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
1019 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
1021         * config/arm/aarch-common-protos.h
1022         (alu_cost_table): Fix spelling of "extend".
1023         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
1025 2014-03-19  Richard Biener  <rguenther@suse.de>
1027         PR middle-end/60553
1028         * tree-core.h (tree_type_common): Re-order pointer members
1029         to reduce recursion depth during GC walks.
1031 2014-03-19  Marek Polacek  <polacek@redhat.com>
1033         PR sanitizer/60569
1034         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
1035         before accessing it.
1037 2014-03-19  Richard Biener  <rguenther@suse.de>
1039         PR lto/59543
1040         * lto-streamer-in.c (input_function): In WPA stage do not drop
1041         debug stmts.
1043 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
1045         PR tree-optimization/60559
1046         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
1047         with build_zero_cst assignment.
1049 2014-03-18  Kai Tietz  <ktietz@redhat.com>
1051         PR rtl-optimization/56356
1052         * sdbout.c (sdbout_parms): Verify that parms'
1053         incoming argument is valid.
1054         (sdbout_reg_parms): Likewise.
1056 2014-03-18  Richard Henderson  <rth@redhat.com>
1058         PR target/60562
1059         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
1060         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
1061         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
1063 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
1065         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
1066         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
1067         Italicize plugin event names in description.  Explain that
1068         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
1069         Remind that no GCC functions should be called after PLUGIN_FINISH.
1070         Explain what pragmas with expansion are.
1072 2014-03-18  Martin Liska  <mliska@suse.cz>
1074         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
1075         gimple call statement is update.
1076         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
1077         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
1079 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1081         PR sanitizer/60557
1082         * ubsan.c (ubsan_instrument_unreachable): Call
1083         initialize_sanitizer_builtins.
1084         (ubsan_pass): Likewise.
1086         PR sanitizer/60535
1087         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1088         varpool_finalize_decl instead of rest_of_decl_compilation.
1090 2014-03-18  Richard Biener  <rguenther@suse.de>
1092         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1093         by using bitmap_and_compl instead of bitmap_and_compl_into.
1094         (df_rd_transfer_function): Likewise.
1096 2014-03-18  Richard Biener  <rguenther@suse.de>
1098         * doc/lto.texi (fresolution): Fix typo.
1100 2014-03-18  Richard Biener  <rguenther@suse.de>
1102         * doc/invoke.texi (flto): Update for changes in 4.9.
1104 2014-03-18  Richard Biener  <rguenther@suse.de>
1106         * doc/loop.texi: Remove section on the removed lambda framework.
1107         Update loop docs with recent changes in preserving loop structure.
1109 2014-03-18  Richard Biener  <rguenther@suse.de>
1111         * doc/lto.texi (-fresolution): Document.
1113 2014-03-18  Richard Biener  <rguenther@suse.de>
1115         * doc/contrib.texi: Adjust my name.
1117 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1119         PR ipa/58721
1120         * internal-fn.c: Include diagnostic-core.h.
1121         (expand_BUILTIN_EXPECT): New function.
1122         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1123         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1124         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1125         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1126         IFN_BUILTIN_EXPECT.
1127         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1128         Revert 3 argument __builtin_expect code.
1129         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1130         * gimple-fold.c (gimple_fold_call): Likewise.
1131         * tree.h (fold_builtin_expect): New prototype.
1132         * builtins.c (build_builtin_expect_predicate): Add predictor
1133         argument, if non-NULL, create 3 argument __builtin_expect.
1134         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1135         pass it through to build_builtin_expect_predicate.
1136         (fold_builtin_2): Adjust caller.
1137         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1138         * internal-fn.def (BUILTIN_EXPECT): New.
1140 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1142         PR ipa/58721
1143         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1144         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1145         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1147 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1149         PR ipa/58721
1150         * predict.c (combine_predictions_for_bb): Fix up formatting.
1151         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1152         fill what it points to if non-NULL.
1153         (tree_predict_by_opcode): Adjust caller, use the predictor.
1154         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1156 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1158         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1159         proper constant for the store mode.
1161 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1163         * symtab.c (change_decl_assembler_name): Fix transparent alias
1164         chain construction.
1166 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1168         * config/aarch64/aarch64.c: Correct the comments about the
1169         aarch64 stack layout.
1171 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1173         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1174         check for GF_OMP_FOR_KIND_FOR.
1176 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1178         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1179         ymm and zmm register names.
1181 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1183         PR target/60516
1184         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1185         note creation for the 2010-08-31 changes.
1187 2014-03-17  Marek Polacek  <polacek@redhat.com>
1189         PR middle-end/60534
1190         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1191         as -fno-tree-loop-vectorize.
1192         (expand_omp_simd): Likewise.
1194 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1196         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1197         (eligible_for_call_delay): New prototype.
1198         * config/sparc/sparc.c (tls_call_delay): Rename into...
1199         (eligible_for_call_delay): ...this.  Return false if the instruction
1200         cannot be put in the delay slot of a branch.
1201         (eligible_for_restore_insn): Simplify.
1202         (eligible_for_return_delay): Return false if the instruction cannot be
1203         put in the delay slot of a branch and simplify.
1204         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1205         put in the delay slot of a branch.
1206         * config/sparc/sparc.md (fix_ut699): New attribute.
1207         (tls_call_delay): Delete.
1208         (in_call_delay): Reimplement.
1209         (eligible_for_sibcall_delay): Rename into...
1210         (in_sibcall_delay): ...this.
1211         (eligible_for_return_delay): Rename into...
1212         (in_return_delay): ...this.
1213         (in_branch_delay): Reimplement.
1214         (in_uncond_branch_delay): Delete.
1215         (in_annul_branch_delay): Delete.
1217 2014-03-14  Richard Henderson  <rth@redhat.com>
1219         PR target/60525
1220         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1221         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1222         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1223         (floathi splitters): Remove.
1224         (float<SWI48x>xf2): New pattern.
1225         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1226         code that tried to handle DImode for 32-bit, but which was excluded
1227         by the pattern's condition.  Drop allocation of stack temporary.
1228         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1229         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1230         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1231         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1232         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1233         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1234         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1235         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1236         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1237         (*float<SWI48x><X87MODEF>2_i387): Remove.
1238         (all float _with_temp splitters): Remove.
1239         (*float<SWI48x><MODEF>2_i387): New pattern.
1240         (*float<SWI48><MODEF>2_sse): New pattern.
1241         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1242         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1244 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1245             Marek Polacek  <polacek@redhat.com>
1247         PR middle-end/60484
1248         * common.opt (dump_base_name_prefixed): New Variable.
1249         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1250         if x_dump_base_name_prefixed is already set, set it at the end.
1252 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1254         PR rtl-optimization/60508
1255         * lra-constraints.c (get_reload_reg): Add new parameter
1256         in_subreg_p.
1257         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1258         Pass the new parameter values.
1260 2014-03-14  Richard Biener  <rguenther@suse.de>
1262         * common.opt: Revert unintented changes from r205065.
1263         * opts.c: Likewise.
1265 2014-03-14  Richard Biener  <rguenther@suse.de>
1267         PR middle-end/60518
1268         * cfghooks.c (split_block): Properly adjust all loops the
1269         block was a latch of.
1271 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1273         PR lto/60461
1274         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1275         and simplify it.
1277 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1279         PR target/59396
1280         * config/avr/avr.c (avr_set_current_function): Pass function name
1281         through default_strip_name_encoding before sanity checking instead
1282         of skipping the first char of the assembler name.
1284 2014-03-13  Richard Henderson  <rth@redhat.com>
1286         PR debug/60438
1287         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1288         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1289         * config/i386/i386-protos.h: Likewise.
1290         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1291         in the expander instead of a splitter.
1292         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1293         any possibility of requiring a memory.
1294         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1295         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1296         (fp branch splitters): Update for ix86_split_fp_branch.
1297         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1298         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1299         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1300         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1301         (*fop_<MODEF>_3_i387): Likewise.
1302         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1303         (splitters for the fop_* register patterns): Remove.
1304         (fscalexf4_i387): Rename from *fscalexf4_i387.
1305         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1307 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1309         PR tree-optimization/59779
1310         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1311         type for bitsize and maxsize instead of HOST_WIDE_INT.
1313 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1315         PR rtl-optimization/57320
1316         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1317         the CFG after thread_prologue_and_epilogue_insns.
1319 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1321         PR rtl-optimization/57189
1322         * lra-constraints.c (process_alt_operands): Disfavor spilling
1323         vector pseudos.
1325 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1327         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1329 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1331         PR tree-optimization/59025
1332         PR middle-end/60418
1333         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1334         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1336 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1338         PR target/60486
1339         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1340         calls of avr_out_plus_1.
1342 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1344         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1345         BB's single pred and update the father loop's latch info later.
1347 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1349         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1350         (VEC_M): Likewise.
1351         (VEC_N): Likewise.
1352         (VEC_R): Likewise.
1353         (VEC_base): Likewise.
1354         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1355         registers, we need to swap double words in little endian mode.
1357         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1358         to be a container mode for 128-bit integer operations added in ISA
1359         2.07.  Unlike TImode and PTImode, the preferred register set is
1360         the Altivec/VMX registers for the 128-bit operations.
1362         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1363         declarations.
1364         (rs6000_split_128bit_ok_p): Likewise.
1366         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1367         macros for creating ISA 2.07 normal and overloaded builtin
1368         functions with 3 arguments.
1369         (BU_P8V_OVERLOAD_3): Likewise.
1370         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1371         for use as overloaded functions.
1372         (VPERM_1TI_UNS): Likewise.
1373         (VSEL_1TI): Likewise.
1374         (VSEL_1TI_UNS): Likewise.
1375         (ST_INTERNAL_1ti): Likewise.
1376         (LD_INTERNAL_1ti): Likewise.
1377         (XXSEL_1TI): Likewise.
1378         (XXSEL_1TI_UNS): Likewise.
1379         (VPERM_1TI): Likewise.
1380         (VPERM_1TI_UNS): Likewise.
1381         (XXPERMDI_1TI): Likewise.
1382         (SET_1TI): Likewise.
1383         (LXVD2X_V1TI): Likewise.
1384         (STXVD2X_V1TI): Likewise.
1385         (VEC_INIT_V1TI): Likewise.
1386         (VEC_SET_V1TI): Likewise.
1387         (VEC_EXT_V1TI): Likewise.
1388         (EQV_V1TI): Likewise.
1389         (NAND_V1TI): Likewise.
1390         (ORC_V1TI): Likewise.
1391         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1392         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1393         overloaded builtin.
1394         (VADDUQM): Likewise.
1395         (VSUBCUQ): Likewise.
1396         (VADDEUQM): Likewise.
1397         (VADDECUQ): Likewise.
1398         (VSUBEUQM): Likewise.
1399         (VSUBECUQ): Likewise.
1401         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1402         __int128_t and __uint128_t types.
1403         (__uint128_type): Likewise.
1404         (altivec_categorize_keyword): Add support for vector __int128_t,
1405         vector __uint128_t, vector __int128, and vector unsigned __int128
1406         as a container type for TImode operations that need to be done in
1407         VSX/Altivec registers.
1408         (rs6000_macro_to_expand): Likewise.
1409         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1410         to support 128-bit integer instructions vaddcuq, vadduqm,
1411         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1412         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1414         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1415         for V1TImode, and set up preferences to use VSX/Altivec registers.
1416         Setup VSX reload handlers.
1417         (rs6000_debug_reg_global): Likewise.
1418         (rs6000_init_hard_regno_mode_ok): Likewise.
1419         (rs6000_preferred_simd_mode): Likewise.
1420         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1421         (easy_altivec_constant): Likewise.
1422         (output_vec_const_move): Likewise.
1423         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1424         simple move.
1425         (rs6000_expand_vector_extract): Likewise.
1426         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1427         addressing.
1428         (rs6000_const_vec): Add support for V1TImode.
1429         (rs6000_emit_le_vsx_load): Swap double words when loading or
1430         storing TImode/V1TImode.
1431         (rs6000_emit_le_vsx_store): Likewise.
1432         (rs6000_emit_le_vsx_move): Likewise.
1433         (rs6000_emit_move): Add support for V1TImode.
1434         (altivec_expand_ld_builtin): Likewise.
1435         (altivec_expand_st_builtin): Likewise.
1436         (altivec_expand_vec_init_builtin): Likewise.
1437         (altivec_expand_builtin): Likewise.
1438         (rs6000_init_builtins): Add support for V1TImode type.  Add
1439         support for ISA 2.07 128-bit integer builtins.  Define type names
1440         for the VSX/Altivec vector types.
1441         (altivec_init_builtins): Add support for overloaded vector
1442         functions with V1TImode type.
1443         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1444         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1445         external function.
1446         (rs6000_split_128bit_ok_p): Likewise.
1447         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1448         __int128_t and vector __uint128_t.
1450         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1451         and mode attributes.
1452         (VSX_M): Likewise.
1453         (VSX_M2): Likewise.
1454         (VSm): Likewise.
1455         (VSs): Likewise.
1456         (VSr): Likewise.
1457         (VSv): Likewise.
1458         (VS_scalar): Likewise.
1459         (VS_double): Likewise.
1460         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1462         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1463         we support the ISA 2.07 128-bit integer arithmetic instructions.
1464         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1465         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1466         and TImode types for use with the builtin functions.
1467         (V1TI_type_node): Likewise.
1468         (unsigned_V1TI_type_node): Likewise.
1469         (intTI_type_internal_node): Likewise.
1470         (uintTI_type_internal_node): Likewise.
1472         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1473         128-bit builtin functions.
1474         (UNSPEC_VADDEUQM): Likewise.
1475         (UNSPEC_VADDECUQ): Likewise.
1476         (UNSPEC_VSUBCUQ): Likewise.
1477         (UNSPEC_VSUBEUQM): Likewise.
1478         (UNSPEC_VSUBECUQ): Likewise.
1479         (VM): Add V1TImode to vector mode iterators.
1480         (VM2): Likewise.
1481         (VI_unit): Likewise.
1482         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1483         (altivec_vaddcuq): Likewise.
1484         (altivec_vsubuqm): Likewise.
1485         (altivec_vsubcuq): Likewise.
1486         (altivec_vaddeuqm): Likewise.
1487         (altivec_vaddecuq): Likewise.
1488         (altivec_vsubeuqm): Likewise.
1489         (altivec_vsubecuq): Likewise.
1491         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1492         mode iterators.
1493         (BOOL_128): Likewise.
1494         (BOOL_REGS_OUTPUT): Likewise.
1495         (BOOL_REGS_OP1): Likewise.
1496         (BOOL_REGS_OP2): Likewise.
1497         (BOOL_REGS_UNARY): Likewise.
1498         (BOOL_REGS_AND_CR0): Likewise.
1500         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1501         128-bit integer builtin support.
1502         (vec_vadduqm): Likewise.
1503         (vec_vaddecuq): Likewise.
1504         (vec_vaddeuqm): Likewise.
1505         (vec_vsubecuq): Likewise.
1506         (vec_vsubeuqm): Likewise.
1507         (vec_vsubcuq): Likewise.
1508         (vec_vsubuqm): Likewise.
1510         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1511         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1512         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1513         128-bit integer add/subtract to ISA 2.07.
1515 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1517         * config/arc/arc.c (arc_predicate_delay_insns):
1518         Fix third argument passed to conditionalize_nonjump.
1520 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1522         * config/aarch64/aarch64-builtins.c
1523         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1524         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1525         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1526         instead of __builtin_lfloor.
1527         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1529 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1531         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1532         (tree_ssa_ifcombine_bb_1): New function.
1533         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1534         is an empty forwarder block to then_bb or vice versa and then_bb
1535         and else_bb are effectively swapped.
1537 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1539         PR target/60264
1540         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1541         REG_CFA_DEF_CFA note.
1542         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1543         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1545 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1547         PR tree-optimization/60454
1548         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1550 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1552         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1553         Do not define target_cpu_default2 to generic.
1554         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1555         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1556         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1558 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1559             Marc Glisse  <marc.glisse@inria.fr>
1561         PR tree-optimization/60502
1562         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1563         instead of build_low_bits_mask.
1565 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1567         PR middle-end/60482
1568         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1569         if there are multiple uses, but op doesn't live on E edge.
1570         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1571         clobber stmts before __builtin_unreachable.
1573 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1575         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1576         hard_frame_pointer_rtx.
1577         * cse.c (cse_insn): Remove volatile check.
1578         * cselib.c (cselib_process_insn): Likewise.
1579         * dse.c (scan_insn): Likewise.
1581 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1583         * config/arc/arc.c (conditionalize_nonjump): New function,
1584         broken out of ...
1585         (arc_ifcvt): ... this.
1586         (arc_predicate_delay_insns): Use it.
1588 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1590         * config/arc/predicates.md (extend_operand): During/after reload,
1591         allow const_int_operand.
1592         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1593         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1594         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1595         to "i".
1596         (umulsi3_highpart_i): Likewise.
1598 2014-03-11  Richard Biener  <rguenther@suse.de>
1600         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1601         Add asserts to guard possible wrong-code bugs.
1603 2014-03-11  Richard Biener  <rguenther@suse.de>
1605         PR tree-optimization/60429
1606         PR tree-optimization/60485
1607         * tree-ssa-structalias.c (set_union_with_increment): Properly
1608         take into account all fields that overlap the shifted vars.
1609         (do_sd_constraint): Likewise.
1610         (do_ds_constraint): Likewise.
1611         (get_constraint_for_ptr_offset): Likewise.
1613 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1615         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1616         (nios2_compute_frame_layout):
1617         Add calculation of cfun->machine->fp_save_offset.
1618         (nios2_expand_prologue): Correct setting of frame pointer register
1619         in prologue.
1620         (nios2_expand_epilogue): Update recovery of stack pointer from
1621         frame pointer accordingly.
1622         (nios2_initial_elimination_offset): Update calculation of offset
1623         for eliminating to HARD_FRAME_POINTER_REGNUM.
1625 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1627         PR ipa/60457
1628         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1629         cgraph_get_create_node on VAR_DECLs.
1631 2014-03-10  Richard Biener  <rguenther@suse.de>
1633         PR middle-end/60474
1634         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1636 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1638         * config/vms/vms.opt (vms_float_format): New variable.
1640 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1642         * doc/invoke.texi (-fcilkplus): Update implementation status.
1644 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1645             Richard Biener  <rguenther@suse.de>
1647         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1648         consistently accross all TUs.
1649         (run_gcc): Enable -fshort-double automatically at link at link-time
1650         and disallow override.
1652 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1654         PR target/58271
1655         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1656         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1657         if they can't be used.
1659 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1661         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1662         for Solaris 11/x86 ld.
1663         * configure: Regenerate.
1665 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1667         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1668         (LIB_TLS_SPEC): Save as ld_tls_libs.
1669         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1670         (HAVE_AS_IX86_TLSLDM): New test.
1671         * configure, config.in: Regenerate.
1672         * config/i386/i386.c (legitimize_tls_address): Fall back to
1673         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1674         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1675         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1676         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1678 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1680         * common.opt (fira-loop-pressure): Mark as optimization.
1682 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1684         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1685         an OpenMP mappable type.
1687 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1689         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1690         MULTILIB_OSDIRNAMES is not defined.
1692 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1693             Meador Inge  <meadori@codesourcery.com>
1695         PR target/58595
1696         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1697         (arm_legitimize_address): Call legitimize_tls_address for any
1698         arm_tls_referenced_p expression, handle constant addend.  Call it
1699         before testing for !TARGET_ARM.
1700         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1702 2014-03-06  Richard Biener  <rguenther@suse.de>
1704         PR middle-end/60445
1705         PR lto/60424
1706         PR lto/60427
1707         Revert
1708         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1710         * tree-streamer.c (record_common_node): Assert we don't record
1711         nodes with type double.
1712         (preload_common_node): Skip type double, complex double and double
1713         pointer since it is now frontend dependent due to fshort-double option.
1715 2014-03-06  Richard Biener  <rguenther@suse.de>
1717         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1718         or -fno-lto is specified and the linker has full plugin support.
1719         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1720         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1721         * lto-wrapper.c (merge_and_complain): Merge compile-time
1722         optimization levels.
1723         (run_gcc): And pass it through to the link options.
1725 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1727         PR debug/60381
1728         Revert:
1729         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1730         PR debug/59992
1731         * cselib.c (remove_useless_values): Skip to avoid quadratic
1732         behavior if the condition moved from...
1733         (cselib_process_insn): ... here holds.
1735 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1737         PR plugins/59335
1738         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1739         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1741         PR plugins/59335
1742         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1743         (TM_H): Add x86-tune.def.
1745 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1747         * config/aarch64/aarch64.c (generic_tunings):
1748         Use cortexa57_extra_costs.
1750 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1752         PR lto/60404
1753         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1754         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1755         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1756         cost for in_lto_p.
1758 2014-03-04  Heiher  <r@hev.cc>
1760         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1761         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1763 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1765         * config/i386/predicates.md (const2356_operand): Change to ...
1766         (const2367_operand): ... this.
1767         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1768         const2367_operand.
1769         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1770         (*avx512pf_scatterpf<mode>sf): Ditto.
1771         (avx512pf_scatterpf<mode>df): Ditto.
1772         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1773         (*avx512pf_scatterpf<mode>df): Ditto.
1774         * config/i386/i386.c (ix86_expand_builtin): Update
1775         incorrect hint operand error message.
1777 2014-03-04  Richard Biener  <rguenther@suse.de>
1779         * lto-section-in.c (lto_get_section_data): Fix const cast.
1781 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1783         * tree-streamer.c (record_common_node): Assert we don't record
1784         nodes with type double.
1785         (preload_common_node): Skip type double, complex double and double
1786         pointer since it is now frontend dependent due to fshort-double option.
1788 2014-03-04  Richard Biener  <rguenther@suse.de>
1790         PR lto/60405
1791         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1792         (lto_input_toplevel_asms): Likewise.
1793         * lto-section-in.c (lto_get_section_data): Instead do it here
1794         for every section.
1796 2014-03-04  Richard Biener  <rguenther@suse.de>
1798         PR tree-optimization/60382
1799         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1800         dead PHIs a reduction.
1802 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1804         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1805         hint value.
1806         (_mm_prefetch): Move out of GCC target("sse") pragma.
1807         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1808         GCC target("prfchw") pragma.
1809         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1810         for locality <= 2.
1811         * config/i386/i386.c (ix86_option_override_internal): Enable
1812         -mprfchw with -mprefetchwt1.
1814 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1816         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1817         Mark as varying.
1819 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1821         * opts.h (CL_PCH_IGNORE): Define.
1822         * targhooks.c (option_affects_pch_p):
1823         Return false for options that have CL_PCH_IGNORE set.
1824         * opt-functions.awk: Process PchIgnore.
1825         * doc/options.texi: Document PchIgnore.
1827         * config/arc/arc.opt (misize): Add PchIgnore property.
1829 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1831         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1832         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1833         constraint on constants to permit them being loaded into
1834         GENERAL_REGS or BASE_REGS.
1836 2014-03-03  Nick Clifton  <nickc@redhat.com>
1838         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1839         anti-cacnonical alternatives.
1840         (negandhi3_real): New pattern.
1841         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1843 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1845         * config/avr/avr-mcus.def: Remove atxmega16x1.
1846         * config/avr/avr-tables.opt: Regenerate.
1847         * config/avr/t-multilib: Regenerate.
1848         * doc/avr-mmcu.texi: Regenerate.
1850 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1851             Mircea Namolaru  <mircea.namolaru@inria.fr>
1853         PR tree-optimization/58028
1854         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1855         scalar dimensions.
1857 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1859         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1860         not handled by recognizers.
1862 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1864         PR middle-end/60175
1865         * function.c (expand_function_end): Don't emit
1866         clobber_return_register sequence if clobber_after is a BARRIER.
1867         * cfgexpand.c (construct_exit_block): Append instructions before
1868         return_label to prev_bb.
1870 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1872         * config/rs6000/constraints.md: Document reserved use of "wc".
1874 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1876         PR ipa/60150
1877         * ipa.c (function_and_variable_visibility): When dissolving comdat
1878         group, also set all symbols to local.
1880 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1882         PR ipa/60306
1884         Revert:
1885         2013-12-14   Jan Hubicka  <jh@suse.cz>
1886         PR middle-end/58477
1887         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1889 2014-03-02  Jon Beniston  <jon@beniston.com>
1891         PR bootstrap/48230
1892         PR bootstrap/50927
1893         PR bootstrap/52466
1894         PR target/46898
1895         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1896           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1897         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1898         (simple_return, *simple_return): New patterns
1899         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1900         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1902 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1904         * dwarf2out.c (gen_subprogram_die): Tidy.
1906 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1908         PR target/60071
1909         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1910         (*mov_t_msb_neg_negc): ... this new insn.
1912 2014-02-28  Jason Merrill  <jason@redhat.com>
1914         PR c++/58678
1915         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1916         function.
1918 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1920         PR c++/60314
1921         * dwarf2out.c (decltype_auto_die): New static.
1922         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1923         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1924         (is_cxx_auto): Likewise.
1926 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1928         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1929         we are not using general regs only.
1931 2014-02-28  Richard Biener  <rguenther@suse.de>
1933         PR target/60280
1934         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1935         previous fix and only allow to remove trivial pre-headers
1936         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1937         (remove_forwarder_block): Properly update the latch of a loop.
1939 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1941         PR debug/59992
1942         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1943         (cselib_preserved_hash_table): New.
1944         (preserve_constants_and_equivs): Move preserved vals to it.
1945         (cselib_find_slot): Look it up first.
1946         (cselib_init): Initialize it.
1947         (cselib_finish): Release it.
1948         (dump_cselib_table): Dump it.
1950 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1952         PR debug/59992
1953         * cselib.c (remove_useless_values): Skip to avoid quadratic
1954         behavior if the condition moved from...
1955         (cselib_process_insn): ... here holds.
1957 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1959         PR debug/57232
1960         * var-tracking.c (vt_initialize): Apply the same condition to
1961         preserve the CFA base value.
1963 2014-02-28  Joey Ye  <joey.ye@arm.com>
1965         PR target/PR60169
1966         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1967         if reload in progress or completed.
1969 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1971         PR middle-end/60147
1972         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1973         NAMELIST_DECL.
1975 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1977         * doc/tm.texi.in (Condition Code Status): Update documention for
1978         relative locations of cc0-setter and cc0-user.
1980 2014-02-27  Jeff Law  <law@redhat.com>
1982         PR rtl-optimization/52714
1983         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1984         into two independent simple sets, if I3 is a jump, ensure the
1985         pattern we place into I3 is a (set (pc) ...).
1987 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1988             Jeff Law  <law@redhat.com>
1990         PR rtl-optimization/49847
1991         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1992         are in different blocks.
1993         * doc/tm.texi (Condition Code Status): Update documention for
1994         relative locations of cc0-setter and cc0-user.
1996 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1998         PR target/59222
1999         * lra.c (lra_emit_add): Check SUBREG too.
2001 2014-02-27  Andreas Schwab  <schwab@suse.de>
2003         * config/m68k/m68k.c (m68k_option_override): Disable
2004         -flive-range-shrinkage for classic m68k.
2005         (m68k_override_options_after_change): Likewise.
2007 2014-02-27  Marek Polacek  <polacek@redhat.com>
2009         PR middle-end/59223
2010         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
2011         -Wmaybe-uninitialized.
2013 2014-02-27  Alan Modra  <amodra@gmail.com>
2015         PR target/57936
2016         * reload1.c (emit_input_reload_insns): When reload_override_in,
2017         set old to rl->in_reg when rl->in_reg is a subreg.
2019 2014-02-26  Richard Biener  <rguenther@suse.de>
2021         PR bootstrap/60343
2022         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
2024 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2026         * common/config/i386/predicates.md (const1256_operand): Remove.
2027         (const2356_operand): New.
2028         (const_1_to_2_operand): Remove.
2029         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2030         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2031         (*avx512pf_gatherpf<mode>sf): Ditto.
2032         (avx512pf_gatherpf<mode>df): Ditto.
2033         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2034         (*avx512pf_gatherpf<mode>df): Ditto.
2035         (avx512pf_scatterpf<mode>sf): Ditto.
2036         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2037         (*avx512pf_scatterpf<mode>sf): Ditto.
2038         (avx512pf_scatterpf<mode>df): Ditto.
2039         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2040         (*avx512pf_scatterpf<mode>df): Ditto.
2041         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
2043 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
2045         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
2046         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
2047         (_mm512_mask_testn_epi64_mask): Move to ...
2048         * config/i386/avx512cdintrin.h: Here.
2049         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
2050         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
2051         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
2052         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
2053         TARGET_AVX512F from TARGET_AVX512CD.
2055 2014-02-26  Richard Biener  <rguenther@suse.de>
2057         PR ipa/60327
2058         * ipa.c (walk_polymorphic_call_targets): Properly guard
2059         call to inline_update_overall_summary.
2061 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
2063         PR target/60280
2064         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
2065         and latches only if requested.  Fix latch if it is removed.
2066         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
2067         LOOPS_HAVE_PREHEADERS.
2069 2014-02-25  Andrew Pinski  <apinski@cavium.com>
2071         * builtins.c (expand_builtin_thread_pointer): Create a new target
2072         when the target is NULL.
2074 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
2076         PR rtl-optimization/60317
2077         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2078         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
2079         * lra-assigns.c: Include params.h.
2080         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
2081         other reload pseudos considerations.
2083 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2085         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
2086         to use canonical form for nor<mode>3.
2088 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2090         PR target/55426
2091         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2092         conversions.
2094 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2096         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2097         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2098         (ix86_handle_option): Handle OPT_mprefetchwt1.
2099         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2100         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2101         PREFETCHWT1 CPUID.
2102         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2103         OPTION_MASK_ISA_PREFETCHWT1.
2104         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2105         (PTA_PREFETCHWT1): New.
2106         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2107         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2108         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2109         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2110         (*prefetch_avx512pf_<mode>_: Change into ...
2111         (*prefetch_prefetchwt1_<mode>: This.
2112         * config/i386/i386.opt (mprefetchwt1): New.
2113         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2114         (_mm_prefetch): Handle intent to write.
2115         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2117 2014-02-25  Richard Biener  <rguenther@suse.de>
2119         PR middle-end/60291
2120         * emit-rtl.c (mem_attrs_htab): Remove.
2121         (mem_attrs_htab_hash): Likewise.
2122         (mem_attrs_htab_eq): Likewise.
2123         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2124         (init_emit_once): Do not allocate mem_attrs_htab.
2126 2014-02-25  Richard Biener  <rguenther@suse.de>
2128         PR lto/60319
2129         * lto-opts.c (lto_write_options): Output non-explicit conservative
2130         -fwrapv, -fno-trapv and -fno-strict-overflow.
2131         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2132         (run_gcc): And pass them through.
2134 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2136         * sel-sched.c (calculate_new_fences): New parameter ptime.
2137         Calculate it as a maximum over all fence cycles.
2138         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2139         Print the final schedule timing when sched_verbose.
2141 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2143         PR rtl-optimization/60292
2144         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2145         bit fot the fence instruction.
2147 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2149         * calls.h: Fix typo in comment.
2151 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2153         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2154         adjusting offsetable addresses.
2156 2014-02-24  Guozhi Wei  <carrot@google.com>
2158         * sparseset.h (sparseset_pop): Fix the wrong index.
2160 2014-02-24  Walter Lee  <walt@tilera.com>
2162         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2163         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2164         triplet.
2165         * common/config/tilegx/tilegx-common.c
2166         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2167         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2168         (LINK_SPEC): Ditto.
2169         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2170         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2171         (tilegx_gimplify_va_arg_expr): Handle big endian.
2172         (tilegx_expand_unaligned_load): Ditto.
2173         (tilegx_expand_unaligned_store): Ditto.
2174         (TARGET_RETURN_IN_MSB): New.
2175         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2176         (TARGET_ENDIAN_DEFAULT): New.
2177         (TARGET_BIG_ENDIAN): Handle big endian.
2178         (BYTES_BIG_ENDIAN): Ditto.
2179         (WORDS_BIG_ENDIAN): Ditto.
2180         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2181         (ENDIAN_SPEC): New.
2182         (EXTRA_SPECS): New.
2183         * config/tilegx/tilegx.md (extv): Handle big endian.
2184         (extzv): Ditto.
2185         (insn_st<n>): Ditto.
2186         (insn_st<n>_add<bitsuffix>): Ditto.
2187         (insn_stnt<n>): Ditto.
2188         (insn_stnt<n>_add<bitsuffix>):Ditto.
2189         (vec_interleave_highv8qi): Handle big endian.
2190         (vec_interleave_highv8qi_be): New.
2191         (vec_interleave_highv8qi_le): New.
2192         (insn_v1int_h): Handle big endian.
2193         (vec_interleave_lowv8qi): Handle big endian.
2194         (vec_interleave_lowv8qi_be): New.
2195         (vec_interleave_lowv8qi_le): New.
2196         (insn_v1int_l): Handle big endian.
2197         (vec_interleave_highv4hi): Handle big endian.
2198         (vec_interleave_highv4hi_be): New.
2199         (vec_interleave_highv4hi_le): New.
2200         (insn_v2int_h): Handle big endian.
2201         (vec_interleave_lowv4hi): Handle big endian.
2202         (vec_interleave_lowv4hi_be): New.
2203         (vec_interleave_lowv4hi_le): New.
2204         (insn_v2int_l): Handle big endian.
2205         (vec_interleave_highv2si): Handle big endian.
2206         (vec_interleave_highv2si_be): New.
2207         (vec_interleave_highv2si_le): New.
2208         (insn_v4int_h): Handle big endian.
2209         (vec_interleave_lowv2si): Handle big endian.
2210         (vec_interleave_lowv2si_be): New.
2211         (vec_interleave_lowv2si_le): New.
2212         (insn_v4int_l): Handle big endian.
2213         * config/tilegx/tilegx.opt (mbig-endian): New option.
2214         (mlittle-endian): New option.
2215         * doc/install.texi: Document tilegxbe-linux.
2216         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2218 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2220         PR ipa/60266
2221         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2222         there are no parameter descriptors.
2224 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2226         PR rtl-optimization/60268
2227         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2228         initialization to ...
2229         (sched_rgn_init): ... here.
2230         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2232 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2234         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2235         names.
2237 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2239         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2240         definition.
2242 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2244         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2245         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2247 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2249         * config/microblaze/predicates.md: Add cmp_op predicate.
2250         * config/microblaze/microblaze.md: Add branch_compare instruction
2251         which uses cmp_op predicate and emits cmp insn before branch.
2252         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2253         to microblaze_expand_conditional_branch and consolidate logic.
2254         (microblaze_expand_conditional_branch): emit branch_compare
2255         insn instead of handling cmp op separate from branch insn.
2257 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2259         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2260         to permit subregs.
2262 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2264         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2265         define_insn with define_expand and new define_insn
2266         *altivec_lve<VI_char>x_internal.
2267         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2268         and new define_insn *altivec_stve<VI_char>x_internal.
2269         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2270         prototype.
2271         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2272         lve*x built-ins.
2273         (altivec_expand_stvex_be): New function.
2275 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2277         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2278         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2279         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2280         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2282 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2284         PR target/60298
2285         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2286         instead of emit_move_insn.
2288 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2290         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2291         vspltw with vsldoi.
2292         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2293         gen_altivec_vsumsws.
2295 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2297         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2298         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2299         (altivec_lvxl_<mode>): New define_expand incorporating
2300         -maltivec=be semantics where needed.
2301         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2302         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2303         semantics where needed.
2304         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2305         (altivec_stvx_<mode>): New define_expand incorporating
2306         -maltivec=be semantics where needed.
2307         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2308         VM2 iterator instead of V4SI.
2309         (altivec_stvxl_<mode>): New define_expand incorporating
2310         -maltivec=be semantics where needed.
2311         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2312         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2313         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2314         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2315         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2316         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2317         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2318         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2319         ALTIVEC_BUILTIN_STVXL.
2320         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2321         (altivec_expand_stvx_be): Likewise.
2322         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2323         (altivec_expand_lvx_be): Likewise.
2324         (altivec_expand_stvx_be): Likewise.
2325         (altivec_expand_builtin): Add cases for
2326         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2327         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2328         (altivec_init_builtins): Add definitions for
2329         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2330         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2332 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2334         * doc/invoke.texi (mvirt, mno-virt): Document.
2335         * config/mips/mips.opt (mvirt): New option.
2336         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2338 2014-02-21  Richard Biener  <rguenther@suse.de>
2340         PR tree-optimization/60276
2341         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2342         (STMT_VINFO_MIN_NEG_DIST): New macro.
2343         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2344         STMT_VINFO_MIN_NEG_DIST.
2345         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2346         made for negative dependence distances still hold.
2348 2014-02-21  Richard Biener  <rguenther@suse.de>
2350         PR middle-end/60291
2351         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2352         DECL_INITIAL for globals not in the current function context.
2354 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2356         PR tree-optimization/56490
2357         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2358         * tree-ssa-uninit.c: Include params.h.
2359         (compute_control_dep_chain): Add num_calls argument, return false
2360         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2361         num_calls to recursive call.
2362         (find_predicates): Change dep_chain into normal array,
2363         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2364         variable and adjust compute_control_dep_chain caller.
2365         (find_def_preds): Likewise.
2367 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2369         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2370         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2372 2014-02-21  Nick Clifton  <nickc@redhat.com>
2374         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2375         (pushhi1): Likewise.
2376         (popqi1): Add mode to pre_dec.
2377         (pophi1): Likewise.
2379 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2381         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2382         mode for mask of V8SFmode permutation.
2384 2014-02-20  Richard Henderson  <rth@redhat.com>
2386         PR c++/60272
2387         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2388         a new pseudo for OLDVAL.
2390 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2392         PR target/57896
2393         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2394         gen_reg_rtx if d->testing_p.
2395         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2396         if d->testing_p and we will certainly return true.
2397         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2398         if d->testing_p.
2400 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2402         * emit-rtl.c (gen_reg_rtx): Assert that
2403         crtl->emit.regno_pointer_align_length is non-zero.
2405 2014-02-20  Richard Henderson  <rth@redhat.com>
2407         PR c++/60272
2408         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2409         on failure the store back into EXPECT.
2411 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2412             Sandra Loosemore  <sandra@codesourcery.com>
2414         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2415         * config/nios2/nios2.c (nios2_function_profiler): Add
2416         -fPIC (flag_pic == 2) support.
2417         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2418         (nios2_large_offset_p): New function.
2419         (nios2_unspec_reloc_p): Move up position, update to use
2420         nios2_large_offset_p.
2421         (nios2_unspec_address): Remove function.
2422         (nios2_unspec_offset): New function.
2423         (nios2_large_got_address): New function.
2424         (nios2_got_address): Add large offset support.
2425         (nios2_legitimize_tls_address): Update usage of removed and new
2426         functions.
2427         (nios2_symbol_binds_local_p): New function.
2428         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2429         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2430         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2431         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2432         processing for (const (unspec ...)).
2433         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2435 2014-02-20  Richard Biener  <rguenther@suse.de>
2437         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2438         doing the substitution.
2439         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2441 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2443         PR ipa/55260
2444         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2445         info when checking whether lattices are bottom.
2447 2014-02-20  Richard Biener  <rguenther@suse.de>
2449         PR middle-end/60221
2450         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2451         regions at -O0.
2453 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2455         PR ipa/58555
2456         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2457         parameter specifying the scaling.
2458         (inline_call): Update.
2459         (want_inline_recursively): Guard division by zero.
2460         (recursive_inlining): Update.
2461         * ipa-inline.h (clone_inlined_nodes): Update.
2463 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2465         PR target/60204
2466         * config/i386/i386.c (classify_argument): Pass structures of size
2467         64 bytes or less in register.
2469 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2470             Kirill Yukhin  <kirill.yukhin@intel.com>
2472         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2473         (_mm_rcp28_round_ss): Ditto.
2474         (_mm_rsqrt28_round_sd): Ditto.
2475         (_mm_rsqrt28_round_ss): Ditto.
2476         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2477         (_mm_rcp14_round_ss): Ditto.
2478         (_mm_rsqrt14_round_sd): Ditto.
2479         (_mm_rsqrt14_round_ss): Ditto.
2480         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2481         the first input operand, get rid of match_dup.
2482         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2483         attribute to sse.
2484         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2485         Ditto.
2486         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2487         operand as the first input operand, set type attribute.
2488         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2489         Set type attribute.
2490         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2491         operand as the first input operand, set type attribute.
2493 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2495         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2496         bit of zero.
2498 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2500         PR target/60207
2501         * config/i386/i386.c (construct_container): Remove TFmode check
2502         for X86_64_INTEGER_CLASS.
2504 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2506         PR target/59794
2507         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2508         only when -Wpsabi is enabled.
2510 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2512          PR target/59799
2513         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2514         passing arrays in registers are the same as for structs, so remove the
2515         special case for them.
2517 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2519         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2520         destination type, extract only the valid bits if the source type is not
2521         integral and has a different mode.
2523 2014-02-19  Richard Biener  <rguenther@suse.de>
2525         PR ipa/60243
2526         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2527         for all calls.
2529 2014-02-19  Richard Biener  <rguenther@suse.de>
2531         PR ipa/60243
2532         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2533         (ipa_modify_call_arguments): Emit an argument load explicitely and
2534         preserve virtual SSA form there and for the replacement call.
2535         Do not update SSA form nor free dominance info.
2537 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2539         * ipa.c (function_and_variable_visibility): Also clear WEAK
2540         flag when disolving COMDAT_GROUP.
2542 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2544         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2545         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2546         not devirtualizing.
2547         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2548         do more sanity checks.
2549         (detect_type_change): Return true when giving up early.
2550         (compute_complex_assign_jump_func): Fix type parameter of
2551         ipa_set_ancestor_jf.
2552         (compute_complex_ancestor_jump_func): Likewise.
2553         (update_jump_functions_after_inlining): Fix updating of
2554         ancestor function.
2555         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2557 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2559         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2560         inline clones when edge disappears.
2562 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2564         PR target/60203
2565         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2566         Split 64-bit moves into 2 patterns.  Do not allow the use of
2567         direct move for TDmode in little endian, since the decimal value
2568         has little endian bytes within a word, but the 64-bit pieces are
2569         ordered in a big endian fashion, and normal subreg's of TDmode are
2570         not allowed.
2571         (mov<mode>_64bit_dm): Likewise.
2572         (movtd_64bit_nodm): Likewise.
2574 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2576         PR tree-optimization/60174
2577         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2578         statement of an SSA_NAME that occurs in an abnormal PHI node.
2580 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2582         PR sanitizer/60142
2583         * final.c (SEEN_BB): Remove.
2584         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2585         (final_scan_insn): Don't force_source_line on second
2586         NOTE_INSN_BASIC_BLOCK.
2588 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2590         PR target/60205
2591         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2592         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2593         (type_natural_mode): Warn ABI change when %zmm register is not
2594         available for AVX512F vector value passing.
2596 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2598         PR target/60193
2599         * config/i386/i386.c (ix86_expand_prologue): Use value in
2600         rax register as displacement when restoring %r10 or %rax.
2601         Fix wrong offset when restoring both registers.
2603 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2605         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2606         assertion with conditional return.
2608 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2609             Uros Bizjak  <ubizjak@gmail.com>
2611         PR driver/60233
2612         * config/i386/driver-i386.c (host_detect_local_cpu): If
2613         YMM state is not saved by the OS, also clear has_f16c.  Move
2614         CPUID 0x80000001 handling before YMM state saving checking.
2616 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2618         PR rtl-optimization/58960
2619         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2620         factored out from ...
2621         (sched_init): ... here.
2622         (free_global_sched_pressure_data): New, factored out from ...
2623         (sched_finish): ... here.
2624         * sched-int.h (free_global_sched_pressure_data): Declare.
2625         * sched-rgn.c (nr_regions_initial): New static global.
2626         (haifa_find_rgns): Initialize it.
2627         (schedule_region): Disable sched-pressure for the newly
2628         generated regions.
2630 2014-02-17  Richard Biener  <rguenther@suse.de>
2632         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2633         release SSA defs of pattern stmts.
2635 2014-02-17  Richard Biener  <rguenther@suse.de>
2637         * tree-inline.c (expand_call_inline): Release the virtual
2638         operand defined by the call we are about to inline.
2640 2014-02-17  Richard Biener  <rguenther@suse.de>
2642         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2644 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2645             Ilya Tocar  <ilya.tocar@intel.com>
2647         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2648         arguments order in builtin.
2649         (_mm512_permutexvar_epi64): Ditto.
2650         (_mm512_mask_permutexvar_epi64): Ditto
2651         (_mm512_maskz_permutexvar_epi32): Ditto
2652         (_mm512_permutexvar_epi32): Ditto
2653         (_mm512_mask_permutexvar_epi32): Ditto
2655 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2657         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2658         (p8_vmrgow): Likewise.
2660 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2662         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2663         endian targets.
2665 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2667         PR target/60203
2668         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2669         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2670         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2671         using direct move instructions on ISA 2.07.  Also adjust
2672         instruction length for 64-bit.
2673         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2674         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2676 2014-02-15  Alan Modra  <amodra@gmail.com>
2678         PR target/58675
2679         PR target/57935
2680         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2681         find_replacement on parts of insn rtl that might be reloaded.
2683 2014-02-15  Richard Biener  <rguenther@suse.de>
2685         PR tree-optimization/60183
2686         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2687         (tree_ssa_phiprop): Calculate and free post-dominators.
2689 2014-02-14  Jeff Law  <law@redhat.com>
2691         PR rtl-optimization/60131
2692         * ree.c (get_extended_src_reg): New function.
2693         (combine_reaching_defs): Use it rather than assuming location of REG.
2694         (find_and_remove_re): Verify first operand of extension is
2695         a REG before adding the insns to the copy list.
2697 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2699         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2700         * configure: Regenerated.
2701         * config.in: Regenerated.
2702         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2703         instead of ASM_SHORT.
2705 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2706             Richard Earnshaw  <rearnsha@arm.com>
2708         PR rtl-optimization/59535
2709         * lra-constraints.c (process_alt_operands): Encourage alternative
2710         when unassigned pseudo class is superset of the alternative class.
2711         (inherit_reload_reg): Don't inherit when optimizing for code size.
2712         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2713         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2714         modes not less than 4 for Thumb1.
2716 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2718         PR pch/60010
2719         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2721 2014-02-14  Richard Biener  <rguenther@suse.de>
2723         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2724         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2725         Do not strip INDIRECT_REFs.
2727 2014-02-14  Richard Biener  <rguenther@suse.de>
2729         PR lto/60179
2730         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2731         DECL_FUNCTION_SPECIFIC_TARGET.
2732         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2733         * tree-streamer-out.c (pack_ts_target_option): Remove.
2734         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2735         (write_ts_function_decl_tree_pointers): Do not stream
2736         DECL_FUNCTION_SPECIFIC_TARGET.
2737         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2738         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2739         (lto_input_ts_function_decl_tree_pointers): Do not stream
2740         DECL_FUNCTION_SPECIFIC_TARGET.
2742 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2744         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2745         (get_initial_def_for_induction, vectorizable_induction): Ignore
2746         debug stmts when looking for exit_phi.
2747         (vectorizable_live_operation): Fix up condition.
2749 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2751         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2752         nreverse() because it changes the content of original tree list.
2754 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2756         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2757         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2759 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2761         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2762         GNU coding standards.
2764 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2766         PR debug/60152
2767         * dwarf2out.c (gen_subprogram_die): Don't call
2768         add_calling_convention_attribute if subr_die is old_die.
2770 2014-02-13  Sharad Singhai  <singhai@google.com>
2772         * doc/optinfo.texi: Fix order of nodes.
2774 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2776         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2777         operands[2], not operands[3].
2779 2014-02-13  Richard Biener  <rguenther@suse.de>
2781         PR bootstrap/59878
2782         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2783         mention the possibility of an in-tree build.
2784         (CLooG): Update recommended version to 0.18.1, mention the
2785         possibility of an in-tree build and clarify that the ISL
2786         bundled with CLooG does not work.
2788 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2790         PR target/43546
2791         * expr.c (compress_float_constant): If x is a hard register,
2792         extend into a pseudo and then move to x.
2794 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2796         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2797         caused by bad second argument to warning_at() with -mhotpatch and
2798         nested functions (e.g. with gfortran).
2800 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2802         * opts.c (option_name): Remove "enabled by default" rider.
2804 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2806         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2808 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2809             Uros Bizjak  <ubizjak@gmail.com>
2811         PR target/60151
2812         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2813         * configure: Regenerated.
2815 2014-02-12  Richard Biener  <rguenther@suse.de>
2817         * vec.c (vec_prefix::calculate_allocation): Move as
2818         inline variant to vec.h.
2819         (vec_prefix::calculate_allocation_1): New out-of-line version.
2820         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2821         (vec_prefix::m_has_auto_buf): Rename to ...
2822         (vec_prefix::m_using_auto_storage): ... this.
2823         (vec_prefix::calculate_allocation): Inline the easy cases
2824         and dispatch to calculate_allocation_1 which doesn't need the
2825         prefix address.
2826         (va_heap::reserve): Use gcc_checking_assert.
2827         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2828         m_using_auto_storage.
2829         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2830         member and adjust.
2831         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2832         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2833         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2835 2014-02-12  Richard Biener  <rguenther@suse.de>
2837         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2838         when we found a dependence.
2840 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2842         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2843         common code...
2844         (maybe_fold_stmt): ... into this new function.
2845         * omp-low.c (lower_omp): Update comment.
2847         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2848         last use.
2850         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2851         dereference.
2853 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2855         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2856         identifiers in comments.
2857         (cortexa53_extra_costs): Likewise.
2858         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2859         (cortexa7_extra_costs): Likewise.
2860         (cortexa12_extra_costs): Likewise.
2861         (cortexa15_extra_costs): Likewise.
2862         (v7m_extra_costs): Likewise.
2864 2014-02-12  Richard Biener  <rguenther@suse.de>
2866         PR middle-end/60092
2867         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2868         of posix_memalign being successful.
2869         (lower_stmt): Restrict lowering of posix_memalign to when
2870         -ftree-bit-ccp is enabled.
2872 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2874         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2875         arg_loc.
2876         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2878 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2880         PR rtl-optimization/60116
2881         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2882         other_insn once the combination has been validated.
2884 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2886         PR lto/59468
2887         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2888         and wrapper.
2889         * ipa-devirt.c: Include demangle.h
2890         (odr_violation_reported): New static variable.
2891         (add_type_duplicate): Update odr_violations.
2892         (maybe_record_node): Add completep parameter; update it.
2893         (record_target_from_binfo): Add COMPLETEP parameter;
2894         update it as needed.
2895         (possible_polymorphic_call_targets_1): Likewise.
2896         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2897         rename FINAL to COMPLETE.
2898         (record_targets_from_bases): Sanity check we found the binfo;
2899         fix COMPLETEP updating.
2900         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2901         parameter, fix computing of COMPLETEP.
2902         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2903         at LTO time do demangling.
2904         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2905         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2906         parameter.
2907         (gimple_get_virt_method_for_binfo): Likewise.
2908         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2909         gimple_get_virt_method_for_vtable): Update prototypes.
2911 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2913         PR target/49008
2914         * genautomata.c (add_presence_absence): Fix typo with
2915         {final_}presence_list.
2917 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2919         PR target/60137
2920         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2921         for VSX/Altivec vectors that land in GPR registers.
2923 2014-02-11  Richard Henderson  <rth@redhat.com>
2924             Jakub Jelinek  <jakub@redhat.com>
2926         PR debug/59776
2927         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2928         around drhs if type conversion to lacc->type is not useless.
2930 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2932         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2933         tuning struct.
2934         (cortex-a57.cortex-a53): Likewise.
2935         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2937 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2939         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2940         arm_restrict_it.
2942 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2944         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2945         add_options_for_arm_vfp3.
2947 2014-02-11  Jeff Law  <law@redhat.com>
2949         PR middle-end/54041
2950         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2951         object with an undesirable mode.
2953 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2955         PR libgomp/60107
2956         * config/i386/sol2-9.h: New file.
2957         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2958         *-*-solaris2.9*): Use it.
2960 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2962         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2963         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2965 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2967         * config/microblaze/microblaze.c: Extend mcpu version format
2969 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
2971         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2973 2014-02-10  Richard Henderson  <rth@redhat.com>
2975         PR target/59927
2976         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2977         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2978         ms-abi vs -mno-accumulate-outgoing-args.
2979         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2980         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2981         respect to ms-abi.
2983 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2985         PR middle-end/60080
2986         * cfgexpand.c (expand_asm_operands): Attach source location to
2987         ASM_INPUT rtx objects.
2988         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2990 2014-02-10  Nick Clifton  <nickc@redhat.com>
2992         * config/mn10300/mn10300.c (popcount): New function.
2993         (mn10300_expand_prologue): Include saved registers in stack usage
2994         count.
2996 2014-02-10  Jeff Law  <law@redhat.com>
2998         PR middle-end/52306
2999         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
3000         when changing the SET_DEST of a prior insn to avoid an input reload.
3002 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
3004         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
3005         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
3006         -mcall-openbsd, or -mcall-linux.
3007         (CC1_ENDIAN_BIG_SPEC): Remove.
3008         (CC1_ENDIAN_LITTLE_SPEC): Remove.
3009         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3010         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
3011         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
3012         and %cc1_endian_default.
3013         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
3015 2014-02-10  Richard Biener  <rguenther@suse.de>
3017         PR tree-optimization/60115
3018         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
3019         MEM_REF handling.  Properly verify that the accesses are not
3020         out of the objects bound.
3022 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3024         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
3025         coretex to cortex.
3027 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
3029         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
3030         proper constants and fix formatting.
3031         (possible_polymorphic_call_targets): Fix formatting.
3033 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
3034             Ilya Tocar  <ilya.tocar@intel.com>
3036         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
3037         (_mm512_loadu_epi32): Renamed into...
3038         (_mm512_loadu_si512): This.
3039         (_mm512_storeu_epi32): Renamed into...
3040         (_mm512_storeu_si512): This.
3041         (_mm512_maskz_ceil_ps): Removed.
3042         (_mm512_maskz_ceil_pd): Ditto.
3043         (_mm512_maskz_floor_ps): Ditto.
3044         (_mm512_maskz_floor_pd): Ditto.
3045         (_mm512_floor_round_ps): Ditto.
3046         (_mm512_floor_round_pd): Ditto.
3047         (_mm512_ceil_round_ps): Ditto.
3048         (_mm512_ceil_round_pd): Ditto.
3049         (_mm512_mask_floor_round_ps): Ditto.
3050         (_mm512_mask_floor_round_pd): Ditto.
3051         (_mm512_mask_ceil_round_ps): Ditto.
3052         (_mm512_mask_ceil_round_pd): Ditto.
3053         (_mm512_maskz_floor_round_ps): Ditto.
3054         (_mm512_maskz_floor_round_pd): Ditto.
3055         (_mm512_maskz_ceil_round_ps): Ditto.
3056         (_mm512_maskz_ceil_round_pd): Ditto.
3057         (_mm512_expand_pd): Ditto.
3058         (_mm512_expand_ps): Ditto.
3059         * config/i386/i386.c (ix86_builtins): Remove
3060         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
3061         (bdesc_args): Ditto.
3062         * config/i386/predicates.md (const1256_operand): New.
3063         (const_1_to_2_operand): Ditto.
3064         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
3065         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3066         (*avx512pf_gatherpf<mode>sf): Ditto.
3067         (avx512pf_gatherpf<mode>df): Ditto.
3068         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3069         (*avx512pf_gatherpf<mode>df): Ditto.
3070         (avx512pf_scatterpf<mode>sf): Ditto.
3071         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3072         (*avx512pf_scatterpf<mode>sf): Ditto.
3073         (avx512pf_scatterpf<mode>df): Ditto.
3074         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3075         (*avx512pf_scatterpf<mode>df): Ditto.
3076         (avx512f_expand<mode>): Removed.
3077         (<shift_insn><mode>3<mask_name>): Change predicate type.
3079 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3081         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
3082         not at the end of datarefs vector use ordered_remove to avoid
3083         reordering datarefs vector.
3085         PR c/59984
3086         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
3087         mark local addressable non-static vars as GOVD_PRIVATE
3088         instead of GOVD_LOCAL.
3089         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3090         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3091         than copying them.
3093         PR middle-end/60092
3094         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3095         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3096         assume_aligned or alloc_align attributes.
3097         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3098         arguments.  Handle also assume_aligned and alloc_align attributes.
3099         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3100         calls to functions with assume_aligned or alloc_align attributes.
3101         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3103 2014-02-08  Terry Guo  <terry.guo@arm.com>
3105         * doc/invoke.texi: Document ARM -march=armv7e-m.
3107 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3109         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3110         flag on __cilkrts_rethrow builtin.
3112         PR ipa/60026
3113         * ipa-cp.c (determine_versionability): Fail at -O0
3114         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3115         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3117         Revert:
3118         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3120         PR ipa/60026
3121         * tree-inline.c (copy_forbidden): Fail for
3122         __attribute__((optimize (0))) functions.
3124 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3126         * varpool.c: Include pointer-set.h.
3127         (varpool_remove_unreferenced_decls): Variables in other partitions
3128         will not be output; be however careful to not lose information
3129         about partitioning.
3131 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3133         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3134         lookup in the vtable constructor.
3136 2014-02-07  Jeff Law  <law@redhat.com>
3138         PR target/40977
3139         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3140         define_insn_and_split.
3142         * ipa-inline.c (inline_small_functions): Fix typos.
3144 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3146         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3147         (s390_can_use_return_insn): Declare.
3148         * config/s390/s390.h (EPILOGUE_USES): Define.
3149         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3150         instructions.
3151         (s390_chunkify_start): Handle return JUMP_LABELs.
3152         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3153         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3154         (s390_can_use_return_insn): New functions.
3155         (s390_fix_long_loop_prediction): Handle conditional returns.
3156         (TARGET_SET_UP_BY_PROLOGUE): Define.
3157         * config/s390/s390.md (ANY_RETURN): New code iterator.
3158         (*creturn, *csimple_return, return, simple_return): New patterns.
3160 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3162         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3163         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3164         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3165         REG_CFA_RESTORE list when deciding not to restore a register.
3167 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3169         * config/s390/s390.c: Include tree-pass.h and context.h.
3170         (s390_early_mach): New function, split out from...
3171         (s390_emit_prologue): ...here.
3172         (pass_data_s390_early_mach): New pass structure.
3173         (pass_s390_early_mach): New class.
3174         (s390_option_override): Create and register early_mach pass.
3175         Move to end of file.
3177 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3179         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3180         to match for the exit block.
3182 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3184         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3185         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3186         Reject misaligned operands.
3188 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3190         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3192 2014-02-07  Richard Biener  <rguenther@suse.de>
3194         PR middle-end/60092
3195         * gimple-low.c (lower_builtin_posix_memalign): New function.
3196         (lower_stmt): Call it to lower posix_memalign in a way
3197         to make alignment info accessible.
3199 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3201         PR c++/60082
3202         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3203         __builtin_setjmp_receiver.
3205 2014-02-07  Richard Biener  <rguenther@suse.de>
3207         PR middle-end/60092
3208         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3209         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3210         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3211         Handle BUILT_IN_POSIX_MEMALIGN.
3212         (find_func_clobbers): Likewise.
3213         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3214         (call_may_clobber_ref_p_1): Likewise.
3216 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3218         PR ipa/59918
3219         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3220         sanity check.
3222 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3224         PR ipa/59469
3225         * lto-cgraph.c (lto_output_node): Use
3226         symtab_get_symbol_partitioning_class.
3227         (lto_output_varpool_node): likewise.
3228         (symtab_get_symbol_partitioning_class): Move here from
3229         lto/lto-partition.c
3230         * cgraph.h (symbol_partitioning_class): Likewise.
3231         (symtab_get_symbol_partitioning_class): Declare.
3233 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3235         * ggc.h (ggc_internal_cleared_alloc): New macro.
3236         * vec.h (vec_safe_copy): Handle memory stats.
3237         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3238         * target-globals.c (save_target_globals): Likewise.
3240 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3242         PR target/60077
3243         * expr.c (emit_move_resolve_push): Export; be bit more selective
3244         on when to clear alias set.
3245         * expr.h (emit_move_resolve_push): Declare.
3246         * function.h (struct function): Add tail_call_marked.
3247         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3248         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3249         * config/i386/i386.md (TImode move expander): De not call
3250         ix86_expand_push.
3251         (FP push expanders): Preserve memory attributes.
3252         * config/i386/sse.md (push<mode>1): Remove.
3253         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3254         (ix86_expand_push): Remove.
3255         * config/i386/mmx.md (push<mode>1): Remove.
3257 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3259         PR rtl-optimization/60030
3260         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3261         lopart with paradoxical subreg before shifting it up by hprec.
3263 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3265         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3266         Remove extra newline at end of file.
3267         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3268         (arm_issue_rate): Handle cortexa57.
3269         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3270         (cortex-a57.cortex-a53): Likewise.
3272 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3274         PR target/59575
3275         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3276         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3277         bitmask.
3278         (arm_expand_prologue): Adjust all callers.
3279         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3280         info, registers also at the lowest numbered registers side.  Use
3281         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3282         XEXP.
3284         PR debug/59992
3285         * var-tracking.c (adjust_mems): Before adding a SET to
3286         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3288 2014-02-06  Alan Modra  <amodra@gmail.com>
3290         PR target/60032
3291         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3292         change SDmode to DDmode when lra_in_progress.
3294 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3296         PR middle-end/59150
3297         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3298         free_data_ref on the dr first, and before goto again also set dr
3299         to the next dr.  For simd_lane_access, free old datarefs[i] before
3300         overwriting it.  For get_vectype_for_scalar_type failure, don't
3301         free_data_ref if simd_lane_access.
3303         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3305         PR target/60062
3306         * tree.h (opts_for_fn): New inline function.
3307         (opt_for_fn): Define.
3308         * config/i386/i386.c (ix86_function_regparm): Use
3309         opt_for_fn (decl, optimize) instead of optimize.
3311 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3313         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3314         for SYMBOL_REF in large memory model.
3316 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3318         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3319         and crypto support.
3320         (cortex-a57): Likewise.
3321         (cortex-a57.cortex-a53): Likewise.
3323 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3324             Kugan Vivekanandarajah  <kuganv@linaro.org>
3326         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3327         unaligned_access.
3328         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3330 2014-02-06  Richard Biener  <rguenther@suse.de>
3332         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3333         set_loop_copy and initialize_original_copy_tables.
3335 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3337         * config/aarch64/aarch64-simd.md
3338         (aarch64_ashr_simddi): Change QI to SI.
3340 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3341             Jakub Jelinek  <jakub@redhat.com>
3343         PR middle-end/60013
3344         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3345         of the dataflow.
3347 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3349         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3350         CODE_FOR_altivec_vpku[hw]um to
3351         CODE_FOR_altivec_vpku[hw]um_direct.
3352         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3353         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3354         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3355         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3357 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3359         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3360         generation for -maltivec=be.
3361         (altivec_vsumsws): Simplify redundant test.
3363 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3365         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3366         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3367         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3368         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3369         gen_altivec_vpkuwum.
3370         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3371         BYTES_BIG_ENDIAN.
3372         (altivec_vpks<VI_char>ss): Likewise.
3373         (altivec_vpks<VI_char>us): Likewise.
3374         (altivec_vpku<VI_char>us): Likewise.
3375         (altivec_vpku<VI_char>um): Likewise.
3376         (altivec_vpku<VI_char>um_direct): New (copy of
3377         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3378         internal use).
3379         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3380         target is little endian and -maltivec=be is not specified.
3381         (*altivec_vupkhs<VU_char>_direct): New (copy of
3382         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3383         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3384         target is little endian and -maltivec=be is not specified.
3385         (*altivec_vupkls<VU_char>_direct): New (copy of
3386         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3387         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3388         little endian and -maltivec=be is not specified.
3389         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3390         little endian and -maltivec=be is not specified.
3392 2014-02-05  Richard Henderson  <rth@redhat.com>
3394         PR debug/52727
3395         * combine-stack-adj.c: Revert r206943.
3396         * sched-int.h (struct deps_desc): Add last_args_size.
3397         * sched-deps.c (init_deps): Initialize it.
3398         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3399         contain REG_ARGS_SIZE notes.
3401 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3403         * lto-cgraph.c (asm_nodes_output): Make global.
3404         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3405         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3406         (driver_handle_option): Handle OPT_fwpa.
3408 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3410         PR ipa/59947
3411         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3412         a comment typo and formatting issue.  If odr_hash hasn't been
3413         created, return vNULL and set *completep to false.
3415         PR middle-end/57499
3416         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3417         bb with no successors.
3419 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3421         PR target/59718
3422         * doc/invoke.texi (-march): Clarify documentation for ARM.
3423         (-mtune): Likewise.
3424         (-mcpu): Likewise.
3426 2014-02-05  Richard Biener  <rguenther@suse.de>
3428         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3429         when not vectorizing because of too many alias checks.
3430         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3431         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3433 2014-02-05  Nick Clifton  <nickc@redhat.com>
3435         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3436         accept extended registers in any mode when compiling for the MN10300.
3438 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3440         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3441         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3442         sanitization attributes.
3443         (can_inline_edge_p): Likewise.
3444         (sanitize_attrs_match_for_inline_p): New function.
3446 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3448         * ipa-prop.c (detect_type_change): Shor circuit testing of
3449         type changes on THIS pointer.
3451 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3453         PR target/59777
3454         * config/pa/pa.c (legitimize_tls_address): Return original address
3455         if not passed a SYMBOL_REF rtx.
3456         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3457         addresses.
3458         (pa_emit_move_sequence): Simplify TLS source operands.
3459         (pa_legitimate_constant_p): Reject all TLS constants.
3460         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3461         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3463 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3465         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3466         groups when we know they are controlled by LTO.
3467         * varasm.c (default_binds_local_p_1): If object is in other partition,
3468         it will be resolved locally.
3470 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3472         * config/host-linux.c (linux_gt_pch_use_address): Don't
3473         use SSIZE_MAX because it is not always defined.
3475 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3477         PR bootstrap/59913
3478         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3479         threshold for pseudo splitting.
3480         (update_ebb_live_info): Process call argument hard registers and
3481         hard registers from insn definition too.
3482         (max_small_class_regs_num): New constant.
3483         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3484         reloads_num only for small register classes.  Don't split for
3485         outputs of jumps.
3487 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3489         PR ipa/60058
3490         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3491         is non-null.
3493 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3495         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3496         visibility is safe.
3498 2014-02-04  Marek Polacek  <polacek@redhat.com>
3500         * gdbinit.in (pel): Define.
3502 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3504         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3505         behavior.
3507 2014-02-04  Richard Biener  <rguenther@suse.de>
3509         PR lto/59723
3510         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3511         in function context local.
3512         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3513         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3514         similar to LTO_imported_decl_ref.
3516 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3518         PR tree-optimization/60002
3519         * cgraphclones.c (build_function_decl_skip_args): Clear
3520         DECL_LANG_SPECIFIC.
3522         PR tree-optimization/60023
3523         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3524         false to gsi_replace.
3525         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3526         has been in some EH region and vec_stmt could throw, add
3527         vec_stmt into the same EH region.
3528         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3529         has no lhs, ignore it.
3530         * internal-fn.c (expand_MASK_LOAD): Likewise.
3532         PR ipa/60026
3533         * tree-inline.c (copy_forbidden): Fail for
3534         __attribute__((optimize (0))) functions.
3536         PR other/58712
3537         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3538         is set, copy one less argument.
3539         (expand_simd_clones): Don't subtract clone_info->inbranch
3540         from simd_clone_struct_alloc argument.
3542         PR rtl-optimization/57915
3543         * recog.c (simplify_while_replacing): If all unary/binary/relational
3544         operation arguments are constant, attempt to simplify those.
3546         PR middle-end/59261
3547         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3548         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3550 2014-02-04  Richard Biener  <rguenther@suse.de>
3552         PR tree-optimization/60012
3553         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3554         TBAA disambiguation to all DDRs.
3556 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3558         PR target/59788
3559         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3560         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3562 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3564         PR ipa/59882
3565         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3567 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3569         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3570         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3572 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3574         PR ipa/59831
3575         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3576         to figure out targets of polymorphic calls with known decl.
3577         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3578         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3579         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3580         (get_polymorphic_call_info): ... here.
3581         (get_polymorphic_call_info_from_invariant): New function.
3583 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3585         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3586         lookup via vtable pointer; check for type consistency
3587         and turn inconsitent facts into UNREACHABLE.
3588         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3589         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3590         type inconsistent querries; return UNREACHABLE instead.
3592 2014-02-03  Richard Henderson  <rth@twiddle.net>
3594         PR tree-opt/59924
3595         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3596         already processed this node.
3597         (normalize_one_pred_1): Pass along mark_set.
3598         (normalize_one_pred): Create and destroy a pointer_set_t.
3599         (normalize_one_pred_chain): Likewise.
3601 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3603         PR gcov-profile/58602
3604         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3606 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3608         PR ipa/59831
3609         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3610         -fno-devirtualize; try to devirtualize by the knowledge of
3611         virtual table pointer given by aggregate propagation.
3612         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3613         (ipa_print_node_jump_functions): Dump also offset that
3614         is relevant for polymorphic calls.
3615         (determine_known_aggregate_parts): Add arg_type parameter; use it
3616         instead of determining the type from pointer type.
3617         (ipa_compute_jump_functions_for_edge): Update call of
3618         determine_known_aggregate_parts.
3619         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3620         (gimple_get_virt_method_for_binfo): ... here; simplify using
3621         vtable_pointer_value_to_vtable.
3622         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3623         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3624         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3625         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3626         POINTER_PLUS_EXPR.
3627         (vtable_pointer_value_to_binfo): ... here.
3628         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3630 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3632         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3633         redef of outer loop index variable.
3635 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3637         PR c++/53017
3638         PR c++/59211
3639         * doc/extend.texi (Function Attributes): Typo.
3641 2014-02-03  Cong Hou  <congh@google.com>
3643         PR tree-optimization/60000
3644         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3645         if the vectorized statement is a store.  A store statement can only
3646         appear at the end of pattern statements.
3648 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3650         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3651         (ix86_option_override_internal): Default long double to 64-bit for
3652         32-bit Bionic and to 128-bit for 64-bit Bionic.
3654         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3655         TARGET_LONG_DOUBLE_128 is true.
3656         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3658         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3659         (mlong-double-64): Negate -mlong-double-128.
3660         (mlong-double-128): New option.
3662         * config/i386/i386-c.c (ix86_target_macros): Define
3663         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3665         * doc/invoke.texi: Document -mlong-double-128.
3667 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3669         PR rtl-optimization/60024
3670         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3672 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3674         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3676 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3678         PR rtl-optimization/57662
3679         * sel-sched.c (code_motion_path_driver): Do not mark already not
3680         existing blocks in the visiting bitmap.
3682 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3684         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3685         on the insn being emitted.
3687 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3688             Will Deacon  <will.deacon@arm.com>
3690         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3692 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3694         * config/arm/arm-tables.opt: Regenerate.
3696 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3698         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3699         for vector types other than V16QImode.
3700         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3701         define_expand, and call altivec_expand_vec_perm_le when producing
3702         code with little endian element order.
3703         (*altivec_vperm_<mode>_internal): New insn having previous
3704         behavior of altivec_vperm_<mode>.
3705         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3706         altivec_expand_vec_perm_le when producing code with little endian
3707         element order.
3708         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3709         behavior of altivec_vperm_<mode>_uns.
3711 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3713         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3714         (altivec_vsumsws): Add handling for -maltivec=be with a little
3715         endian target.
3716         (altivec_vsumsws_direct): New.
3717         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3718         gen_altivec_vsumsws.
3720 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3722         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3723         vtable_pointer_value_to_binfo): New functions.
3724         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3725         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3727 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3729         * config/nios2/nios2.md (load_got_register): Initialize GOT
3730         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3731         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3733 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3735         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3736         preserverd by passthrough, do not propagate the type.
3738 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3740         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3741         (mips_atomic_assign_expand_fenv): New function.
3742         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3744 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3746         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3747         (__builtin_mips_set_fcsr): Likewise.
3748         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3749         MIPS_USI_FTYPE_VOID.
3750         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3751         (mips16_expand_set_fcsr): Likewise.
3752         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3753         (mips16_set_fcsr_stub): Likewise.
3754         (mips16_get_fcsr_one_only_stub): New class.
3755         (mips16_set_fcsr_one_only_stub): Likewise.
3756         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3757         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3758         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3759         (hard_float): New availability predicate.
3760         (mips_builtins): Add get_fcsr and set_fcsr.
3761         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3762         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3763         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3764         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3765         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3766         patterns.
3768 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3770         * config/mips/mips.c (mips_one_only_stub): New class.
3771         (mips_need_mips16_rdhwr_p): Replace with...
3772         (mips16_rdhwr_stub): ...this new variable.
3773         (mips16_stub_call_address): New function.
3774         (mips16_rdhwr_one_only_stub): New class.
3775         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3776         (mips_output_mips16_rdhwr): Delete.
3777         (mips_finish_stub): New function.
3778         (mips_code_end): Use it to handle rdhwr stubs.
3780 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3782         PR target/60017
3783         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3784         when calculating size of integer atomic types.
3786 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3788         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3790 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3792         PR tree-optimization/60003
3793         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3794         * profile.c (branch_prob): Use gimple_call_builtin_p
3795         to check for BUILT_IN_SETJMP_RECEIVER.
3796         * tree-inline.c (copy_bb): Call notice_special_calls.
3798 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3800         PR bootstrap/59985
3801         * lra-constraints.c (process_alt_operands): Update reload_sum only
3802         on the first pass.
3804 2014-01-31  Richard Henderson  <rth@redhat.com>
3806         PR middle-end/60004
3807         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3808         until after else_eh is processed.
3810 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3812         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3813         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3814         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3815         in smmintrin.h, remove them.
3816         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3817         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3818         * config/i386/i386.md (ROUND_SAE): Fix value.
3819         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3820         (const48_operand): New.
3821         * config/i386/subst.md (round), (round_expand): Use
3822         const_4_or_8_to_11_operand.
3823         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3825 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3827         * config/i386/constraints.md (Yk): Swap meaning with k.
3828         * config/i386/i386.md (movhi_internal): Change Yk to k.
3829         (movqi_internal): Ditto.
3830         (*k<logic><mode>): Ditto.
3831         (*andhi_1): Ditto.
3832         (*andqi_1): Ditto.
3833         (kandn<mode>): Ditto.
3834         (*<code>hi_1): Ditto.
3835         (*<code>qi_1): Ditto.
3836         (kxnor<mode>): Ditto.
3837         (kortestzhi): Ditto.
3838         (kortestchi): Ditto.
3839         (kunpckhi): Ditto.
3840         (*one_cmplhi2_1): Ditto.
3841         (*one_cmplqi2_1): Ditto.
3842         * config/i386/sse.md (): Change k to Yk.
3843         (avx512f_load<mode>_mask): Ditto.
3844         (avx512f_blendm<mode>): Ditto.
3845         (avx512f_store<mode>_mask): Ditto.
3846         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3847         (avx512f_storedqu<mode>_mask): Ditto.
3848         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3849         Ditto.
3850         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3851         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3852         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3853         (avx512f_maskcmp<mode>3): Ditto.
3854         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3855         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3856         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3857         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3858         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3859         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3860         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3861         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3862         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3863         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3864         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3865         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3866         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3867         (vec_extract_lo_<mode>_maskm): Ditto.
3868         (vec_extract_hi_<mode>_maskm): Ditto.
3869         (avx512f_vternlog<mode>_mask): Ditto.
3870         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3871         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3872         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3873         (avx512f_<code>v8div16qi2_mask): Ditto.
3874         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3875         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3876         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3877         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3878         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3879         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3880         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3881         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3882         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3883         (avx512cd_maskb_vec_dupv8di): Ditto.
3884         (avx512cd_maskw_vec_dupv16si): Ditto.
3885         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3886         (avx512f_vpermi2var<mode>3_mask): Ditto.
3887         (avx512f_vpermi2var<mode>3_mask): Ditto.
3888         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3889         (*avx512f_gathersi<mode>): Ditto.
3890         (*avx512f_gathersi<mode>_2): Ditto.
3891         (*avx512f_gatherdi<mode>): Ditto.
3892         (*avx512f_gatherdi<mode>_2): Ditto.
3893         (*avx512f_scattersi<mode>): Ditto.
3894         (*avx512f_scatterdi<mode>): Ditto.
3895         (avx512f_compress<mode>_mask): Ditto.
3896         (avx512f_compressstore<mode>_mask): Ditto.
3897         (avx512f_expand<mode>_mask): Ditto.
3898         * config/i386/subst.md (mask): Change k to Yk.
3899         (mask_scalar_merge): Ditto.
3900         (sd): Ditto.
3902 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3904         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3906 2014-01-31  Richard Biener  <rguenther@suse.de>
3908         PR middle-end/59990
3909         * builtins.c (fold_builtin_memory_op): Make sure to not
3910         use a floating-point mode or a boolean or enumeral type for
3911         the copy operation.
3913 2014-01-30  DJ Delorie  <dj@redhat.com>
3915         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3916         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3917         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3918         whenever main() has an epilogue.
3920 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3922         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3923         unused variable "field".
3924         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3925         (vsx_mergeh_<mode>): Likewise.
3926         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3927         (altivec_vmrghh): Likewise.
3928         (altivec_vmrghw): Likewise.
3929         (altivec_vmrglb): Likewise.
3930         (altivec_vmrglh): Likewise.
3931         (altivec_vmrglw): Likewise.
3932         (altivec_vspltb): Add missing uses.
3933         (altivec_vsplth): Likewise.
3934         (altivec_vspltw): Likewise.
3935         (altivec_vspltsf): Likewise.
3937 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3939         PR target/59923
3940         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3941         frame related instructions.
3943 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3945         PR rtl-optimization/59959
3946         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3947         any reload of register whose subreg is invalid.
3949 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3951         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3952         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3953         Add missing return type - void.
3955 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3957         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3958         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3959         remove element index adjustment for endian (now handled in vsx.md
3960         and altivec.md).
3961         (altivec_expand_vec_perm_const): Use
3962         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3963         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3964         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3965         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3966         define_expand and a new define_insn *altivec_vspltb_internal;
3967         adjust for -maltivec=be on a little endian target.
3968         (altivec_vspltb_direct): New.
3969         (altivec_vsplth): Divide into a define_expand and a new
3970         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3971         little endian target.
3972         (altivec_vsplth_direct): New.
3973         (altivec_vspltw): Divide into a define_expand and a new
3974         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3975         little endian target.
3976         (altivec_vspltw_direct): New.
3977         (altivec_vspltsf): Divide into a define_expand and a new
3978         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3979         a little endian target.
3981 2014-01-30  Richard Biener  <rguenther@suse.de>
3983         PR tree-optimization/59993
3984         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3985         can propagate form the earlier stmt and avoid the transform
3986         when the intermediate result is needed.
3988 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3990         * README.Portability: Fix typo.
3992 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
3994         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3995         comparison_operator with ordered_comparison_operator.
3997 2014-01-30  Nick Clifton  <nickc@redhat.com>
3999         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
4000         Rename to mn10300_store_multiple_regs.
4001         * config/mn10300/mn10300.c: Likewise.
4002         * config/mn10300/mn10300.md (store_movm): Fix typo: call
4003         store_multiple_regs.
4004         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
4005         Call mn10300_store_multiple_regs.
4007 2014-01-30  Nick Clifton  <nickc@redhat.com>
4008             DJ Delorie  <dj@redhat.com>
4010         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
4011         %fp 2 to keep registers after it properly word-aligned.
4012         (rl78_alloc_physical_registers_umul): Handle the case where both
4013         input operands are the same.
4015 2014-01-30  Richard Biener  <rguenther@suse.de>
4017         PR tree-optimization/59903
4018         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
4019         check properly.
4021 2014-01-30  Jason Merrill  <jason@redhat.com>
4023         PR c++/59633
4024         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
4026         PR c++/59645
4027         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
4029 2014-01-30  Richard Biener  <rguenther@suse.de>
4031         PR tree-optimization/59951
4032         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
4034 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
4036         PR target/59784
4037         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
4038         SFmode to DFmode case.
4040 2014-01-29  DJ Delorie  <dj@redhat.com>
4042         * config/msp430/msp430.opt (-minrt): New.
4043         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
4044         if -minrt given.
4045         (ENDFILE_SPEC): Likewise.
4047 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
4049         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
4050         (estimate_function_body_sizes): Use it.
4052 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
4054         PR c++/58561
4055         * dwarf2out.c (is_cxx_auto): New.
4056         (is_base_type): Use it.
4057         (gen_type_die_with_usage): Likewise.
4059 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4061         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
4062         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
4063         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
4064         -maltivec=be with LE targets.
4065         (vsx_mergeh_<mode>): Likewise.
4066         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
4067         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
4068         (altivec_vmrghb): Replace with define_expand and new
4069         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
4070         (altivec_vmrghb_direct): New define_insn.
4071         (altivec_vmrghh): Replace with define_expand and new
4072         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
4073         (altivec_vmrghh_direct): New define_insn.
4074         (altivec_vmrghw): Replace with define_expand and new
4075         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
4076         (altivec_vmrghw_direct): New define_insn.
4077         (*altivec_vmrghsf): Adjust for endianness.
4078         (altivec_vmrglb): Replace with define_expand and new
4079         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
4080         (altivec_vmrglb_direct): New define_insn.
4081         (altivec_vmrglh): Replace with define_expand and new
4082         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
4083         (altivec_vmrglh_direct): New define_insn.
4084         (altivec_vmrglw): Replace with define_expand and new
4085         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
4086         (altivec_vmrglw_direct): New define_insn.
4087         (*altivec_vmrglsf): Adjust for endianness.
4088         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4089         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4090         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4091         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4092         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4093         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4094         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4095         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4097 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4099         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4100         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4101         whitespace.
4103 2014-01-29  Richard Biener  <rguenther@suse.de>
4105         PR tree-optimization/58742
4106         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4107         associate_pointerplus_align.
4108         (associate_pointerplus_diff): New function.
4109         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4110         and associate_pointerplus_diff.
4112 2014-01-29  Richard Biener  <rguenther@suse.de>
4114         * lto-streamer.h (LTO_major_version): Bump to 3.
4115         (LTO_minor_version): Reset to 0.
4117 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4119         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4120         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4121         (arm_file_start): Generate correct asm header for armv7ve.
4122         * config/arm/bpabi.h: Add multilib support for armv7ve.
4123         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4124         and cortex-a15 to armv7ve.
4125         * config/arm/t-aprofile: Add multilib support for armv7ve.
4126         * doc/invoke.texi: Document -march=armv7ve.
4128 2014-01-29  Richard Biener  <rguenther@suse.de>
4130         PR tree-optimization/58742
4131         * tree-ssa-forwprop.c (associate_plusminus): Return true
4132         if we changed sth, defer EH cleanup to ...
4133         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4134         (simplify_mult): New function.
4136 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4138         PR middle-end/59917
4139         PR tree-optimization/59920
4140         * tree.c (build_common_builtin_nodes): Remove
4141         __builtin_setjmp_dispatcher initialization.
4142         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4143         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4144         instead of gsi_after_labels + manually skipping debug stmts.
4145         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4146         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4147         * tree-inline.c (copy_edges_for_bb): Remove
4148         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4149         argument.  Ignore computed_goto_p stmts.  Don't call
4150         make_abnormal_goto_edges.  If a call might need abnormal edges
4151         for non-local gotos, see if it already has an edge to
4152         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4153         with true argument, don't do anything then, otherwise add
4154         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4155         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4156         caller.
4157         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4158         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4159         (lower_stmt): Don't set data->calls_builtin_setjmp.
4160         (lower_builtin_setjmp): Adjust comment.
4161         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4162         * tree-cfg.c (found_computed_goto): Remove.
4163         (factor_computed_gotos): Remove.
4164         (make_goto_expr_edges): Return bool, true for computed gotos.
4165         Don't call make_abnormal_goto_edges.
4166         (build_gimple_cfg): Don't set found_computed_goto, don't call
4167         factor_computed_gotos.
4168         (computed_goto_p): No longer static.
4169         (make_blocks): Don't set found_computed_goto.
4170         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4171         (make_edges): If make_goto_expr_edges returns true, push bb
4172         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4173         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4174         vector.  Record mapping between bbs and OpenMP regions if there
4175         are any, adjust make_gimple_omp_edges caller.  Call
4176         handle_abnormal_edges.
4177         (make_abnormal_goto_edges): Remove.
4178         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4179         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4180         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4181         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4182         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4183         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4184         filling *region also set *region_idx to (*region)->entry->index.
4186         PR other/58712
4187         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4188         For REGs set ORIGINAL_REGNO.
4190 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4192         * doc/md.texi: Mention that a target shouldn't implement
4193         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4194         than hi/lo pair.
4196 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4198         PR tree-optimization/59594
4199         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4200         a copy of the datarefs vector rather than the vector itself.
4202 2014-01-28  Jason Merrill  <jason@redhat.com>
4204         PR c++/53756
4205         * dwarf2out.c (auto_die): New static.
4206         (gen_type_die_with_usage): Handle C++1y 'auto'.
4207         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4208         on definition.
4210 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4212         PR target/59672
4213         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4214         (SPEC_X32): Likewise.
4215         (SPEC_64): Likewise.
4216         * config/i386/i386.c (ix86_option_override_internal): Turn off
4217         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4218         for TARGET_16BIT.
4219         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4220         * config/i386/i386.h (TARGET_16BIT): New macro.
4221         (TARGET_16BIT_P): Likewise.
4222         * config/i386/i386.opt: Add m16.
4223         * doc/invoke.texi: Document -m16.
4225 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4227         PR preprocessor/59935
4228         * input.c (location_get_source_line): Bail out on when line number
4229         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4231 2014-01-28  Richard Biener  <rguenther@suse.de>
4233         PR tree-optimization/58742
4234         * tree-ssa-forwprop.c (associate_plusminus): Handle
4235         pointer subtraction of the form (T)(P + A) - (T)P.
4237 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4239         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4240         at const_int_cost.
4242 2014-01-28  Richard Biener  <rguenther@suse.de>
4244         Revert
4245         2014-01-28  Richard Biener  <rguenther@suse.de>
4247         PR rtl-optimization/45364
4248         PR rtl-optimization/59890
4249         * var-tracking.c (local_get_addr_clear_given_value): Handle
4250         already cleared slot.
4251         (val_reset): Handle not allocated local_get_addr_cache.
4252         (vt_find_locations): Use post-order on the inverted CFG.
4254 2014-01-28  Richard Biener  <rguenther@suse.de>
4256         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4258 2014-01-28  Richard Biener  <rguenther@suse.de>
4260         PR rtl-optimization/45364
4261         PR rtl-optimization/59890
4262         * var-tracking.c (local_get_addr_clear_given_value): Handle
4263         already cleared slot.
4264         (val_reset): Handle not allocated local_get_addr_cache.
4265         (vt_find_locations): Use post-order on the inverted CFG.
4267 2014-01-28  Alan Modra  <amodra@gmail.com>
4269         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4270         * configure.ac <recursive call for build != host>: Define
4271         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4272         and LD_FOR_BUILD too.
4273         * configure: Regenerate.
4275 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4277         * config/i386/i386.c (get_builtin_code_for_version): Separate
4278         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4279         Broadwell from Haswell.
4281 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4283         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4284         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4285         * config/mips/mips.c (mips_option_override): Change setting
4286         of TARGET_DSP.
4287         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4288         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4289         Change from Mask to Var.
4291 2014-01-27  Jeff Law  <law@redhat.com>
4293         * ipa-inline.c (inline_small_functions): Fix typo.
4295 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4297         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4298         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4299         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4300         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4301         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4302         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4303         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4304         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4305         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4306         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4307         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4308         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4309         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4310         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4311         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4312         (_mm512_storeu_epi64): Ditto.
4313         (_mm512_cmpge_epi32_mask): Ditto.
4314         (_mm512_cmpge_epu32_mask): Ditto.
4315         (_mm512_cmpge_epi64_mask): Ditto.
4316         (_mm512_cmpge_epu64_mask): Ditto.
4317         (_mm512_cmple_epi32_mask): Ditto.
4318         (_mm512_cmple_epu32_mask): Ditto.
4319         (_mm512_cmple_epi64_mask): Ditto.
4320         (_mm512_cmple_epu64_mask): Ditto.
4321         (_mm512_cmplt_epi32_mask): Ditto.
4322         (_mm512_cmplt_epu32_mask): Ditto.
4323         (_mm512_cmplt_epi64_mask): Ditto.
4324         (_mm512_cmplt_epu64_mask): Ditto.
4325         (_mm512_cmpneq_epi32_mask): Ditto.
4326         (_mm512_cmpneq_epu32_mask): Ditto.
4327         (_mm512_cmpneq_epi64_mask): Ditto.
4328         (_mm512_cmpneq_epu64_mask): Ditto.
4329         (_mm512_expand_pd): Ditto.
4330         (_mm512_expand_ps): Ditto.
4331         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4332         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4333         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4334         * config/i386/i386.c (ix86_builtins): Add
4335         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4336         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4337         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4338         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4339         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4340         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4341         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4342         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4343         IX86_BUILTIN_PMOVUSQW512_MEM.
4344         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4345         __builtin_ia32_pmovsqd512mem_mask,
4346         __builtin_ia32_pmovqd512mem_mask,
4347         __builtin_ia32_pmovusqw512mem_mask,
4348         __builtin_ia32_pmovsqw512mem_mask,
4349         __builtin_ia32_pmovqw512mem_mask,
4350         __builtin_ia32_pmovusdw512mem_mask,
4351         __builtin_ia32_pmovsdw512mem_mask,
4352         __builtin_ia32_pmovdw512mem_mask,
4353         __builtin_ia32_pmovqb512mem_mask,
4354         __builtin_ia32_pmovusqb512mem_mask,
4355         __builtin_ia32_pmovsqb512mem_mask,
4356         __builtin_ia32_pmovusdb512mem_mask,
4357         __builtin_ia32_pmovsdb512mem_mask,
4358         __builtin_ia32_pmovdb512mem_mask.
4359         (bdesc_args): Add __builtin_ia32_expanddf512,
4360         __builtin_ia32_expandsf512.
4361         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4362         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4363         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4364         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4365         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4366         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4367         (avx512f_<code>v8div16qi2_mask_store): This.
4368         (avx512f_expand<mode>): New.
4370 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4372         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4373         New.
4374         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4375         (_mm512_prefetch_i32scatter_pd): Ditto.
4376         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4377         (_mm512_prefetch_i64scatter_pd): Ditto.
4378         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4379         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4380         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4381         (_mm512_prefetch_i32scatter_ps): Ditto.
4382         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4383         (_mm512_prefetch_i64scatter_ps): Ditto.
4384         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4385         * config/i386/i386-builtin-types.def: Define
4386         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4387         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4388         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4389         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4390         IX86_BUILTIN_SCATTERPFQPD.
4391         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4392         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4393         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4394         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4395         __builtin_ia32_scatterpfqps.
4396         (ix86_expand_builtin): Expand new built-ins.
4397         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4398         fix memory access data type.
4399         (*avx512pf_gatherpf<mode>_mask): Ditto.
4400         (*avx512pf_gatherpf<mode>): Ditto.
4401         (avx512pf_scatterpf<mode>): Ditto.
4402         (*avx512pf_scatterpf<mode>_mask): Ditto.
4403         (*avx512pf_scatterpf<mode>): Ditto.
4404         (GATHER_SCATTER_SF_MEM_MODE): New.
4405         (avx512pf_gatherpf<mode>df): Ditto.
4406         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4407         (*avx512pf_scatterpf<mode>df): Ditto.
4409 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4411         PR bootstrap/59934
4412         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4413         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4414         reached.
4416 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4418         * common/config/arm/arm-common.c
4419         (arm_rewrite_mcpu): Handle multiple names.
4420         * config/arm/arm.h
4421         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4423 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4425         * gimple-builder.h (create_gimple_tmp): Delete.
4427 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4429         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4430         words comparisons.
4432 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4434         * config/pa/pa.md (call): Generate indirect long calls to non-local
4435         functions when outputing 32-bit code.
4436         (call_value): Likewise except for special call to buggy powf function.
4438         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4439         portable runtime and PIC indirect calls.
4440         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4441         and PIC call sequences.  Use ldo instead of blr to set return register
4442         in PIC call sequence.
4444 2014-01-25  Walter Lee  <walt@tilera.com>
4446         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4447         avoid clobbering a live register.
4449 2014-01-25  Walter Lee  <walt@tilera.com>
4451         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4452         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4453         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4454         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4456 2014-01-25  Walter Lee  <walt@tilera.com>
4458         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4459         arguments on even registers.
4460         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4461         STACK_BOUNDARY.
4462         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4463         (BIGGEST_ALIGNMENT): Ditto.
4464         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4466 2014-01-25  Walter Lee  <walt@tilera.com>
4468         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4469         insns before bundling.
4470         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4472 2014-01-25  Walter Lee  <walt@tilera.com>
4474         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4475         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4476         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4478 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4480         * config/mips/constraints.md (kl): Delete.
4481         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4482         define expands, using...
4483         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4484         instructions for MIPS16.
4485         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4486         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4488 2014-01-25  Walter Lee  <walt@tilera.com>
4490         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4491         (clzdi2): Ditto.
4492         (ffsdi2): Ditto.
4494 2014-01-25  Walter Lee  <walt@tilera.com>
4496         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4497         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4499 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4501         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4502         Handle XOR.
4504 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4506         * print-rtl.c (in_call_function_usage): New var.
4507         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4508         EXPR_LIST mode as mode and not as reg note name.
4510         PR middle-end/59561
4511         * cfgloopmanip.c (copy_loop_info): If
4512         loop->warned_aggressive_loop_optimizations, make sure
4513         the flag is set in target loop too.
4515 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4517         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4518         flag_cilkplus.
4519         * builtins.def: Likewise.
4520         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4521         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4522         * ira.c (ira_setup_eliminable_regset): Likewise.
4523         * omp-low.c (gate_expand_omp): Likewise.
4524         (execute_lower_omp): Likewise.
4525         (diagnose_sb_0): Likewise.
4526         (gate_diagnose_omp_blocks): Likewise.
4527         (simd_clone_clauses_extract): Likewise.
4528         (gate): Likewise.
4530 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4532         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4533         correction for little endian...
4534         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4535         here.
4537 2014-01-24  Jeff Law  <law@redhat.com>
4539         PR tree-optimization/59919
4540         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4541         for non-returning calls.
4543 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4545         * common/config/aarch64/aarch64-common.c
4546         (aarch64_rewrite_mcpu): Handle multiple names.
4547         * config/aarch64/aarch64.h
4548         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4550 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4552         * input.c (add_file_to_cache_tab): Handle the case where fopen
4553         returns NULL.
4555 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4557         PR target/59929
4558         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4559         from push operand if code of push isn't PRE_DEC.
4561 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4563         PR target/59909
4564         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4565         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4566         it is only used for non-atomic loads/stores.
4568         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4569         -mquad-memory or -mquad-memory-atomic switches.
4571         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4572         -mquad-memory-atomic to ISA 2.07 support.
4574         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4575         to separate support of normal quad word memory operations (ldq, stq)
4576         from the atomic quad word memory operations.
4578         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4579         support to separate non-atomic quad word operations from atomic
4580         quad word operations.  Disable non-atomic quad word operations in
4581         little endian mode so that we don't have to swap words after the
4582         load and before the store.
4583         (quad_load_store_p): Add comment about atomic quad word support.
4584         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4585         options printed with -mdebug=reg.
4587         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4588         -mquad-memory-atomic as the test for whether we have quad word
4589         atomic instructions.
4590         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4591         or -mp8-vector are used, allow byte/half-word atomic operations.
4593         * config/rs6000/sync.md (load_lockedti): Insure that the address
4594         is a proper indexed or indirect address for the lqarx instruction.
4595         On little endian systems, swap the hi/lo registers after the lqarx
4596         instruction.
4597         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4598         insure the address is valid for the lqarx instruction.
4599         (store_conditionalti): Insure that the address is a proper indexed
4600         or indirect address for the stqcrx. instruction.  On little endian
4601         systems, swap the hi/lo registers before doing the stqcrx.
4602         instruction.
4603         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4604         insure the address is valid for the stqcrx. instruction.
4606         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4607         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4608         type of quad memory support is available.
4610 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4612         PR regression/59915
4613         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4614         there is a danger of looping.
4616 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4618         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4619         force flag_ira_loop_pressure if set via command line.
4621 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4623         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4624         (ashr_simd): New builtin handling DI mode.
4625         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4626         (aarch64_sshr_simddi): New match pattern.
4627         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4628         (vshrd_n_s64): Likewise.
4629         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4631 2014-01-23  Nick Clifton  <nickc@redhat.com>
4633         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4634         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4635         favour of mcu specific scripts.
4636         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4637         430x multilibs.
4639 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4640             Alex Velenko  <Alex.Velenko@arm.com>
4642         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4643         (vaddv_s16): Likewise.
4644         (vaddv_s32): Likewise.
4645         (vaddv_u8): Likewise.
4646         (vaddv_u16): Likewise.
4647         (vaddv_u32): Likewise.
4648         (vaddvq_s8): Likewise.
4649         (vaddvq_s16): Likewise.
4650         (vaddvq_s32): Likewise.
4651         (vaddvq_s64): Likewise.
4652         (vaddvq_u8): Likewise.
4653         (vaddvq_u16): Likewise.
4654         (vaddvq_u32): Likewise.
4655         (vaddvq_u64): Likewise.
4656         (vaddv_f32): Likewise.
4657         (vaddvq_f32): Likewise.
4658         (vaddvq_f64): Likewise.
4659         (vmaxv_f32): Likewise.
4660         (vmaxv_s8): Likewise.
4661         (vmaxv_s16): Likewise.
4662         (vmaxv_s32): Likewise.
4663         (vmaxv_u8): Likewise.
4664         (vmaxv_u16): Likewise.
4665         (vmaxv_u32): Likewise.
4666         (vmaxvq_f32): Likewise.
4667         (vmaxvq_f64): Likewise.
4668         (vmaxvq_s8): Likewise.
4669         (vmaxvq_s16): Likewise.
4670         (vmaxvq_s32): Likewise.
4671         (vmaxvq_u8): Likewise.
4672         (vmaxvq_u16): Likewise.
4673         (vmaxvq_u32): Likewise.
4674         (vmaxnmv_f32): Likewise.
4675         (vmaxnmvq_f32): Likewise.
4676         (vmaxnmvq_f64): Likewise.
4677         (vminv_f32): Likewise.
4678         (vminv_s8): Likewise.
4679         (vminv_s16): Likewise.
4680         (vminv_s32): Likewise.
4681         (vminv_u8): Likewise.
4682         (vminv_u16): Likewise.
4683         (vminv_u32): Likewise.
4684         (vminvq_f32): Likewise.
4685         (vminvq_f64): Likewise.
4686         (vminvq_s8): Likewise.
4687         (vminvq_s16): Likewise.
4688         (vminvq_s32): Likewise.
4689         (vminvq_u8): Likewise.
4690         (vminvq_u16): Likewise.
4691         (vminvq_u32): Likewise.
4692         (vminnmv_f32): Likewise.
4693         (vminnmvq_f32): Likewise.
4694         (vminnmvq_f64): Likewise.
4696 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4698         * config/aarch64/aarch64-simd.md
4699         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4700         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4701         (*aarch64_mul3_elt<mode>): Likewise.
4702         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4703         (*aarch64_mul3_elt_to_64v2df): Likewise.
4704         (*aarch64_mla_elt<mode>): Likewise.
4705         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4706         (*aarch64_mls_elt<mode>): Likewise.
4707         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4708         (*aarch64_fma4_elt<mode>): Likewise.
4709         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4710         (*aarch64_fma4_elt_to_64v2df): Likewise.
4711         (*aarch64_fnma4_elt<mode>): Likewise.
4712         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4713         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4714         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4715         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4716         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4717         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4718         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4719         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4720         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4722 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4724         * config/aarch64/aarch64-simd.md
4725         (aarch64_be_checked_get_lane<mode>): New define_expand.
4726         * config/aarch64/aarch64-simd-builtins.def
4727         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4728         New builtin definition.
4729         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4730         Use new safe be builtin.
4732 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4734         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4735         New define_insn.
4736         (aarch64_be_st1<mode>): Likewise.
4737         (aarch_ld1<VALL:mode>): Define_expand modified.
4738         (aarch_st1<VALL:mode>): Likewise.
4739         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4740         (UNSPEC_ST1): Likewise.
4742 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
4744         * config/microblaze/microblaze.md: Add trap insn and attribute
4746 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4748         PR preprocessor/58580
4749         * input.h (location_get_source_line): Take an additional line_size
4750         parameter.
4751         (void diagnostics_file_cache_fini): Declare new function.
4752         * input.c (struct fcache): New type.
4753         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4754         New static constants.
4755         (diagnostic_file_cache_init, total_lines_num)
4756         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4757         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4758         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4759         (get_next_line, read_next_line, goto_next_line, read_line_num):
4760         New static function definitions.
4761         (diagnostic_file_cache_fini): New function.
4762         (location_get_source_line): Take an additional output line_len
4763         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4764         read_line_num.
4765         * diagnostic.c (diagnostic_finish): Call
4766         diagnostic_file_cache_fini.
4767         (adjust_line): Take an additional input parameter for the length
4768         of the line, rather than calculating it with strlen.
4769         (diagnostic_show_locus): Adjust the use of
4770         location_get_source_line and adjust_line with respect to their new
4771         signature.  While displaying a line now, do not stop at the first
4772         null byte.  Rather, display the zero byte as a space and keep
4773         going until we reach the size of the line.
4774         * Makefile.in: Add vec.o to OBJS-libcommon
4776 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4777             Ilya Tocar     <ilya.tocar@intel.com>
4779         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4780         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4781         (__builtin_ia32_kmov16): Ditto.
4782         * config/i386/i386.md (UNSPEC_KMOV): New.
4783         (kmovw): Ditto.
4785 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4787         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4788         (_mm512_storeu_si512): Ditto.
4790 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4792         PR target/52125
4793         * rtl.h (get_referenced_operands): Declare.
4794         * recog.c (get_referenced_operands): New function.
4795         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4796         operands have been referenced when recording LO_SUM references.
4798 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
4800         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4802 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4804         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4805         Enable for generic and recent AMD targets.
4807 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4809         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4810         ARG_SIZE note when adjustment was eliminated.
4812 2014-01-22  Jeff Law  <law@redhat.com>
4814         PR tree-optimization/59597
4815         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4816         in file.  Accept new argument REGISTERING and use it to modify
4817         dump output appropriately.
4818         (register_jump_thread): Corresponding changes.
4819         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4820         thread paths involving joiner blocks.  Add code to dump cancelled
4821         jump threading paths.
4823 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4825         PR rtl-optimization/59477
4826         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4827         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4829 2014-01-22  Tom Tromey  <tromey@redhat.com>
4831         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4832         PARAMS.
4833         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4835 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4837         PR rtl-optimization/59896
4838         * lra-constraints.c (process_alt_operands): Check unused note for
4839         matched operands of insn with no output reloads.
4841 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4843         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4844         (mips_move_from_gpr_cost): Likewise.
4846 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4848         PR rtl-optimization/59858
4849         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4850         ira_class_hard_regs_num.
4851         (process_alt_operands): Increase reject for dying matched operand.
4853 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4855         PR target/59003
4856         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4857         smaller than size, perform several stores or loads and stores
4858         at dst + count - size to store or copy all of size bytes, rather
4859         than just last modesize bytes.
4861 2014-01-20  DJ Delorie  <dj@redhat.com>
4863         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4864         that CLOBBERs are REGs before propogating their values.
4866 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4868         PR middle-end/59789
4869         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4870         (cgraph_inline_failed_type): New function.
4871         * cgraph.h (DEFCIFCODE): Add type.
4872         (cgraph_inline_failed_type_t): New enum.
4873         (cgraph_inline_failed_type): New prototype.
4874         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4875         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4876         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4877         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4878         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4879         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4880         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4881         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4882         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4883         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4884         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4885         OPTIMIZATION_MISMATCH.
4886         * tree-inline.c (expand_call_inline): Emit errors during
4887         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4889 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4891         PR target/59685
4892         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4893         mode attribute in insn output.
4895 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4897         * output.h (output_constant): Delete.
4898         * varasm.c (output_constant): Make private.
4900 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4902         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4904 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4906         PR middle-end/59860
4907         * tree.h (fold_builtin_strcat): New prototype.
4908         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4909         argument, if non-NULL, don't call c_strlen.  Optimize
4910         directly into __builtin_memcpy instead of __builtin_strcpy.
4911         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4912         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4914 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4916         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4917         for SImode_address_operand operands, having only a REG argument.
4919 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4921         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4922         loader name using mbig-endian.
4923         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4925 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4927         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4928         (-mtune): Likewise.
4929         (-mcpu): Likewise.
4931 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4933         * config/aarch64/aarch64-protos.h
4934         (aarch64_cannot_change_mode_class_ptr): Declare.
4935         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4936         aarch64_cannot_change_mode_class_ptr): New.
4937         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4938         backend hook aarch64_cannot_change_mode_class.
4940 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4942         * common/config/aarch64/aarch64-common.c
4943         (aarch64_handle_option): Don't handle any option order logic here.
4944         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4945         selected_cpu, warn on architecture version mismatch.
4946         (aarch64_override_options): Fix parsing order for option strings.
4948 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4949             Iain Sandoe  <iain@codesourcery.com>
4951         PR bootstrap/59496
4952         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4953         warning.  Amend comment to reflect current functionality.
4955 2014-01-20  Richard Biener  <rguenther@suse.de>
4957         PR middle-end/59860
4958         * builtins.c (fold_builtin_strcat): Remove case better handled
4959         by tree-ssa-strlen.c.
4961 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4963         * config/aarch64/aarch64.opt
4964         (mcpu, march, mtune): Make case-insensitive.
4966 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4968         PR target/59880
4969         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4970         if operands[1] is a REG or ZERO_EXTEND of a REG.
4972 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4974         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4976 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4978         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4979         long non-pic millicode calls.
4981 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4983         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4985 2014-01-19  Kito Cheng  <kito@0xlab.org>
4987         * builtins.c (expand_movstr): Check movstr expand done or fail.
4989 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4990             H.J. Lu  <hongjiu.lu@intel.com>
4992         PR target/59379
4993         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4994         to DImode for zero-extended addresses.
4996 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4998         PR rtl-optimization/57763
4999         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
5000         on the new indirect jump_insn and increment LABEL_NUSES (label).
5002 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
5004         PR bootstrap/59580
5005         PR bootstrap/59583
5006         * config.gcc (x86_archs): New variable.
5007         (x86_64_archs): Likewise.
5008         (x86_cpus): Likewise.
5009         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
5010         --with-arch/--with-cpu= options.
5011         Support --with-arch=/--with-cpu={nehalem,westmere,
5012         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
5014 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5016         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
5017         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
5019 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
5021         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
5023 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
5025         PR target/58944
5026         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
5027         clear cpp_get_options (parse_in)->warn_unused_macros for
5028         ix86_target_macros_internal with cpp_define.
5030 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
5032         * jump.c (delete_related_insns): Keep (use (insn))s.
5033         * reorg.c (redundant_insn): Check for barriers too.
5035 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5037         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
5039 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
5041         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
5042         call to $$dyncall when TARGET_LONG_CALLS is true.
5044 2014-01-17  Jeff Law  <law@redhat.com>
5046         * ree.c (combine_set_extension): Temporarily disable test for
5047         changing number of hard registers.
5049 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5051         PR middle-end/58125
5052         * ipa-inline-analysis.c (inline_free_summary):
5053         Do not free summary of aliases.
5055 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5057         PR middle-end/59706
5058         * gimplify.c (gimplify_expr): Use create_tmp_var
5059         instead of create_tmp_var_raw.  If cond doesn't have
5060         integral type, don't add the IFN_ANNOTATE builtin at all.
5062 2014-01-17  Martin Jambor  <mjambor@suse.cz>
5064         PR ipa/59736
5065         * ipa-cp.c (prev_edge_clone): New variable.
5066         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
5067         Also resize prev_edge_clone vector.
5068         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
5069         (ipcp_edge_removal_hook): New function.
5070         (ipcp_driver): Register ipcp_edge_removal_hook.
5072 2014-01-17  Andrew Pinski  <apinski@cavium.com>
5073             Steve Ellcey  <sellcey@mips.com>
5075         PR target/59462
5076         * config/mips/mips.c (mips_print_operand): Check operand mode instead
5077         of operator mode.
5079 2014-01-17  Jeff Law  <law@redhat.com>
5081         PR middle-end/57904
5082         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
5083         so that pass_ccp runs first.
5085 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5087         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5088         (ix86_adjust_cost): Use !TARGET_XXX.
5089         (do_reorder_for_imul): Likewise.
5090         (swap_top_of_ready_list): Likewise.
5091         (ix86_sched_reorder): Likewise.
5093 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5095         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5096         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5097         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5098         (intel_memset): New.  Duplicate slm_memset.
5099         (intel_cost): New.  Duplicate slm_cost.
5100         (m_INTEL): New macro.
5101         (processor_target_table): Add "intel".
5102         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5103         with PROCESSOR_INTEL for "intel".
5104         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5105         PROCESSOR_SILVERMONT.
5106         (ix86_issue_rate): Likewise.
5107         (ix86_adjust_cost): Likewise.
5108         (ia32_multipass_dfa_lookahead): Likewise.
5109         (swap_top_of_ready_list): Likewise.
5110         (ix86_sched_reorder): Likewise.
5111         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5112         instead of TARGET_OPT_AGU.
5113         * config/i386/i386.h (TARGET_INTEL): New.
5114         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5115         (processor_type): Add PROCESSOR_INTEL.
5116         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5117         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5119 2014-01-17  Marek Polacek  <polacek@redhat.com>
5121         PR c/58346
5122         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5123         size is zero.
5125 2014-01-17  Richard Biener  <rguenther@suse.de>
5127         PR tree-optimization/46590
5128         * opts.c (default_options_table): Add entries for
5129         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5130         all enabled at -O1 but not for -Og.
5131         * common.opt (fbranch-count-reg): Remove Init(1).
5132         (fmove-loop-invariants): Likewise.
5133         (ftree-pta): Likewise.
5135 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5137         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5138         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5139         decls to at least the GCC 4.8 used alignments.
5141         PR fortran/59440
5142         * tree-nested.c (convert_nonlocal_reference_stmt,
5143         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5144         of GIMPLE_BIND stmts, adjust associated decls.
5146 2014-01-17  Richard Biener  <rguenther@suse.de>
5148         PR tree-optimization/46590
5149         * vec.h (vec<>::bseach): New member function implementing
5150         binary search according to C89 bsearch.
5151         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5152         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5153         bitmap pointer again.  Make accesses_in_loop a flat array.
5154         (mem_ref_obstack): New global.
5155         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5156         (mark_ref_stored): Likewise.
5157         (ref_indep_loop_p_2): Likewise.
5158         (set_ref_stored_in_loop): New helper function.
5159         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5160         (memref_free): Adjust.
5161         (record_mem_ref_loc): Simplify.
5162         (gather_mem_refs_stmt): Adjust.
5163         (sort_locs_in_loop_postorder_cmp): New function.
5164         (analyze_memory_references): Sort accesses_in_loop after
5165         loop postorder number.
5166         (find_ref_loc_in_loop_cmp): New function.
5167         (for_all_locs_in_loop): Find relevant cluster of locs in
5168         accesses_in_loop and iterate without recursion.
5169         (execute_sm): Avoid uninit warning.
5170         (struct ref_always_accessed): Simplify.
5171         (ref_always_accessed::operator ()): Likewise.
5172         (ref_always_accessed_p): Likewise.
5173         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5174         loop postorder numbers here.
5175         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5176         numbers.
5178 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5180         PR c++/57945
5181         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5182         on decls for which assemble_alias has been called.
5184 2014-01-17  Nick Clifton  <nickc@redhat.com>
5186         * config/msp430/msp430.opt: (mcpu): New option.
5187         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5188         (msp430_option_override): Parse target_cpu.  If the MCU name
5189         matches a generic string, clear target_mcu.
5190         (msp430_attr): Allow numeric interrupt values up to 63.
5191         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5192         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5193         option.
5194         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5195         Add mcpu matches.
5196         * config/msp430/msp430.md (popm): Use %J rather than %I.
5197         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5198         (addhi_cy_i): Use immediate_operand for operand 2.
5199         * doc/invoke.texi: Document -mcpu option.
5201 2014-01-17  Richard Biener  <rguenther@suse.de>
5203         PR rtl-optimization/38518
5204         * df.h (df_analyze_loop): Declare.
5205         * df-core.c: Include cfgloop.h.
5206         (df_analyze_1): Split out main part of df_analyze.
5207         (df_analyze): Adjust.
5208         (loop_inverted_post_order_compute): New function.
5209         (loop_post_order_compute): Likewise.
5210         (df_analyze_loop): New function avoiding whole-function
5211         postorder computes.
5212         * loop-invariant.c (find_defs): Use df_analyze_loop.
5213         (find_invariants): Adjust.
5214         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5216 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5218         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5219         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5221 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5223         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5224         traversal when removing references.
5226 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5228         PR ipa/59775
5229         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5231 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5233         PR middle-end/56791
5234         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5235         pushing a reload for an autoinc when we had previously reloaded an
5236         inner part of the address.
5238 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5240         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5241         field.
5242         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5243         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5244         when not giving up or versioning for alias only because of
5245         loop->safelen.
5246         (vect_analyze_data_ref_dependences): Set to true.
5247         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5248         is a GIMPLE_PHI.
5249         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5250         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5251         to the condition.
5253         PR middle-end/58344
5254         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5256         PR target/59839
5257         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5258         operand 0 predicate for gathers, use a new pseudo as subtarget.
5260 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5262         PR middle-end/59609
5263         * lra-constraints.c (process_alt_operands): Add printing debug info.
5264         Check absence of input/output reloads for matched operands too.
5266 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5268         PR rtl-optimization/59835
5269         * ira.c (ira_init_register_move_cost): Increase cost for
5270         impossible modes.
5272 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5274         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5276 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5278         PR target/59780
5279         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5280         non-register objects.  Use gen_(high/low)part more consistently.
5281         Fix assertions.
5283 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5285         PR target/59844
5286         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5287         endian support, remove tests for WORDS_BIG_ENDIAN.
5288         (p8_mfvsrd_3_<mode>): Likewise.
5289         (reload_gpr_from_vsx<mode>): Likewise.
5290         (reload_gpr_from_vsxsf): Likewise.
5291         (p8_mfvsrd_4_disf): Likewise.
5293 2014-01-16  Richard Biener  <rguenther@suse.de>
5295         PR rtl-optimization/46590
5296         * lcm.c (compute_antinout_edge): Use postorder iteration.
5297         (compute_laterin): Use inverted postorder iteration.
5299 2014-01-16  Nick Clifton  <nickc@redhat.com>
5301         PR middle-end/28865
5302         * varasm.c (output_constant): Return the number of bytes actually
5303         emitted.
5304         (output_constructor_array_range): Update the field size with the
5305         number of bytes emitted by output_constant.
5306         (output_constructor_regular_field): Likewise.  Also do not
5307         complain if the total number of bytes emitted is now greater
5308         than the expected fieldpos.
5309         * output.h (output_constant): Update prototype and descriptive comment.
5311 2014-01-16  Marek Polacek  <polacek@redhat.com>
5313         PR middle-end/59827
5314         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5315         it is error_mark_node.
5317 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5319         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5320         VALID_AVX256_REG_OR_OI_MODE.
5322 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5324         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5325         current procedure should be profiled.
5327 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5329         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5330         of moving from/to the STACK_REG register class.
5332 2014-01-15  Richard Henderson  <rth@redhat.com>
5334         PR debug/54694
5335         * reginfo.c (global_regs_decl): Globalize.
5336         * rtl.h (global_regs_decl): Declare.
5337         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5338         reserved via global_regs.
5340 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5342         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5344 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5346         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5347         and vmulosh rather than call gen_vec_widen_smult_*.
5348         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5349         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5350         (vec_widen_smult_even_v16qi): Likewise.
5351         (vec_widen_umult_even_v8hi): Likewise.
5352         (vec_widen_smult_even_v8hi): Likewise.
5353         (vec_widen_umult_odd_v16qi): Likewise.
5354         (vec_widen_smult_odd_v16qi): Likewise.
5355         (vec_widen_umult_odd_v8hi): Likewise.
5356         (vec_widen_smult_odd_v8hi): Likewise.
5357         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5358         vmuloub rather than call gen_vec_widen_umult_*.
5359         (vec_widen_umult_lo_v16qi): Likewise.
5360         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5361         vmulosb rather than call gen_vec_widen_smult_*.
5362         (vec_widen_smult_lo_v16qi): Likewise.
5363         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5364         rather than call gen_vec_widen_umult_*.
5365         (vec_widen_umult_lo_v8hi): Likewise.
5366         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5367         rather than call gen_vec_widen_smult_*.
5368         (vec_widen_smult_lo_v8hi): Likewise.
5370 2014-01-15  Jeff Law  <law@redhat.com>
5372         PR tree-optimization/59747
5373         * ree.c (find_and_remove_re): Properly handle case where a second
5374         eliminated extension requires widening a copy created for elimination
5375         of a prior extension.
5376         (combine_set_extension): Ensure that the number of hard regs needed
5377         for a destination register does not change when we widen it.
5379 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5381         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5382         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5383         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5384         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5385         (avr-*-rtems*): Likewise.
5386         (bfin*-rtems*): Likewise.
5387         (moxie-*-rtems*): Likewise.
5388         (h8300-*-rtems*): Likewise.
5389         (i[34567]86-*-rtems*): Likewise.
5390         (lm32-*-rtems*): Likewise.
5391         (m32r-*-rtems*): Likewise.
5392         (m68k-*-rtems*): Likewise.
5393         (microblaze*-*-rtems*): Likewise.
5394         (mips*-*-rtems*): Likewise.
5395         (powerpc-*-rtems*): Likewise.
5396         (sh-*-rtems*): Likewise.
5397         (sparc-*-rtems*): Likewise.
5398         (sparc64-*-rtems*): Likewise.
5399         (v850-*-rtems*): Likewise.
5400         (m32c-*-rtems*): Likewise.
5402 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5404         PR rtl-optimization/59511
5405         * ira.c (ira_init_register_move_cost): Use memory costs for some
5406         cases of register move cost calculations.
5407         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5408         instead of BB frequency.
5409         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5410         * lra-assigns.c (find_hard_regno_for): Ditto.
5412 2014-01-15  Richard Biener  <rguenther@suse.de>
5414         PR tree-optimization/59822
5415         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5416         (vectorizable_load): Use it to hoist defs of uses of invariant
5417         loads out of the loop.
5419 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5420             Kugan Vivekanandarajah  <kuganv@linaro.org>
5422         PR target/59695
5423         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5424         truncation.
5426 2014-01-15  Richard Biener  <rguenther@suse.de>
5428         PR rtl-optimization/59802
5429         * lcm.c (compute_available): Use inverted postorder to seed
5430         the initial worklist.
5432 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5434         PR target/59803
5435         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5436         ADDR_REGS for invalid symrefs in non-PIC code.
5438 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5440         PR other/58712
5441         * builtins.c (determine_block_size): Initialize *probable_max_size
5442         even if len_rtx is CONST_INT.
5444 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5446         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5447         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5448         (cortexa53_tunings): Likewise.
5449         (aarch64_sched_issue_rate): New function.
5450         (TARGET_SCHED_ISSUE_RATE): Define.
5452 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5454         * ira-costs.c (find_costs_and_classes): Add missed
5455         ira_init_register_move_cost_if_necessary.
5457 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5459         PR target/59787
5460         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5462 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5464         PR target/59794
5465         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5466         to indicate if type is used for function return value.  Warn ABI
5467         change if the vector mode isn't available for function return value.
5468         (ix86_function_arg_advance): Pass false to type_natural_mode.
5469         (ix86_function_arg): Likewise.
5470         (ix86_gimplify_va_arg): Likewise.
5471         (function_arg_32): Don't warn ABI change.
5472         (ix86_function_value): Pass true to type_natural_mode.
5473         (ix86_return_in_memory): Likewise.
5474         (ix86_struct_value_rtx): Removed.
5475         (TARGET_STRUCT_VALUE_RTX): Likewise.
5477 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5479         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5480         converting a conditional jump into a conditional return.
5482 2014-01-14  Richard Biener  <rguenther@suse.de>
5484         PR tree-optimization/58921
5485         PR tree-optimization/59006
5486         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5487         hoisting invariant stmts.
5488         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5489         invariant loads on the preheader edge if possible.
5491 2014-01-14  Joey Ye  <joey.ye@arm.com>
5493         * doc/plugin.texi (Building GCC plugins): Update to C++.
5495 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5497         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5498         (_mm_rcp28_round_ss): Ditto.
5499         (_mm_rsqrt28_round_sd): Ditto.
5500         (_mm_rsqrt28_round_ss): Ditto.
5501         (_mm_rcp28_sd): Ditto.
5502         (_mm_rcp28_ss): Ditto.
5503         (_mm_rsqrt28_sd): Ditto.
5504         (_mm_rsqrt28_ss): Ditto.
5505         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5506         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5507         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5508         (IX86_BUILTIN_RCP28SD): Ditto.
5509         (IX86_BUILTIN_RCP28SS): Ditto.
5510         (IX86_BUILTIN_RSQRT28SD): Ditto.
5511         (IX86_BUILTIN_RSQRT28SS): Ditto.
5512         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5513         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5514         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5515         (ix86_expand_special_args_builtin): Expand new FTYPE.
5516         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5517         (srcp14<mode>): Make insn unary.
5518         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5519         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5520         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5521         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5522         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5523         Fix rounding: make it SAE only.
5524         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5525         Ditto.
5526         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5527         Ditto.
5528         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5529         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5530         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5531         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5532         (round_saeonly_mask_scalar_operand4): Ditto.
5533         (round_saeonly_mask_scalar_op3): Ditto.
5534         (round_saeonly_mask_scalar_op4): Ditto.
5536 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5538         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5539         Implement -maltivec=be for vec_insert and vec_extract.
5541 2014-01-10  DJ Delorie  <dj@redhat.com>
5543         * config/msp430/msp430.md (call_internal): Don't allow memory
5544         references with SP as the base register.
5545         (call_value_internal): Likewise.
5546         * config/msp430/constraints.md (Yc): New.  For memory references
5547         that don't use SP as a base register.
5549         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5550         "an integer without a # prefix"
5551         * config/msp430/msp430.md (epilogue_helper): Use it.
5553 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5555         PR target/59617
5556         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5557         AVX512F gather builtins.
5558         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5559         on gather decls with INTEGER_TYPE masktype.
5560         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5561         directly into the builtin rather than hoisting it before loop.
5563         PR tree-optimization/59387
5564         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5565         (scev_const_prop): If folded_casts and type has undefined overflow,
5566         use force_gimple_operand instead of force_gimple_operand_gsi and
5567         for each added stmt if it is assign with
5568         arith_code_with_undefined_signed_overflow, call
5569         rewrite_to_defined_overflow.
5570         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5571         gimple-fold.h instead.
5572         (arith_code_with_undefined_signed_overflow,
5573         rewrite_to_defined_overflow): Moved to ...
5574         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5575         rewrite_to_defined_overflow): ... here.  No longer static.
5576         Include gimplify-me.h.
5577         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5578         rewrite_to_defined_overflow): New prototypes.
5580 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5582         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5584 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5586         * builtins.c (get_object_alignment_2): Minor tweak.
5587         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5589 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5591         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5592         optimized non constant lengths.
5594 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5596         PR libgomp/59194
5597         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5598         load as __atomic_load_N if possible.
5600 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5602         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5603         target parameter.
5604         (rs6000_expand_builtin): Adjust call.
5606 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5608         PR target/58115
5609         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5610         * config/rs6000/rs6000.c: Include target-globals.h.
5611         (rs6000_set_current_function): Instead of doing target_reinit
5612         unconditionally, use save_target_globals_default_opts and
5613         restore_target_globals.
5615         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5616         FPSCR.
5617         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5618         (rs6000_expand_builtin): Handle mffs and mtfsf.
5619         (rs6000_init_builtins): Define mffs and mtfsf.
5620         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5621         (rs6000_mffs): New pattern.
5622         (rs6000_mtfsf): New pattern.
5624 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5626         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5627         Start narrowing with START.  Apply candidate-use pair
5628         and check overall cost in narrowing.
5629         (iv_ca_prune): Pass new argument.
5631 2014-01-10  Jeff Law  <law@redhat.com>
5633         PR middle-end/59743
5634         * ree.c (combine_reaching_defs): Ensure the defining statement
5635         occurs before the extension when optimizing extensions with
5636         different source and destination hard registers.
5638 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5640         PR ipa/58585
5641         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5642         vtables into the type inheritance graph.
5644 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5646         PR rtl-optimization/59754
5647         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5648         modes in the REGNO != REGNO case.
5650 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5652         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5654 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5656         PR tree-optimization/59745
5657         * tree-predcom.c (tree_predictive_commoning_loop): Call
5658         free_affine_expand_cache if giving up because components is NULL.
5660         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5661         GC in payload of target_globals struct instead of allocating them on
5662         the heap and the larger structs separately using GC.
5663         * target-globals.h (struct target_globals): Make regs, hard_regs,
5664         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5665         of GTY((skip)) and change type to void *.
5666         (reset_target_globals): Cast loads from those fields to corresponding
5667         types.
5669 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5671         PR plugins/59335
5672         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5673         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5674         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5676 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5678         PR target/59744
5679         * aarch64-modes.def (CC_Zmode): New flags mode.
5680         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5681         represents an equality.
5682         (aarch64_get_condition_code): Handle CC_Zmode.
5683         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5685 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5687         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5688         extraction in good case.
5690 2014-01-10  Richard Biener  <rguenther@suse.de>
5692         PR tree-optimization/59374
5693         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5694         checking after SLP discovery.  Mark stmts not participating
5695         in any SLP instance properly.
5697 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5699         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5700         when handling a SET rtx.
5702 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5704         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5705         (cortex-a57): Likewise.
5706         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5708 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5710         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5711         non-iwmmxt builtins.
5713 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5715         PR ipa/58252
5716         PR ipa/59226
5717         * ipa-devirt.c record_target_from_binfo): Take as argument
5718         stack of binfos and lookup matching one for virtual inheritance.
5719         (possible_polymorphic_call_targets_1): Update.
5721 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5723         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5724         kernel strings for Loongson-2E/2F/3A.
5726 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5728         PR middle-end/59670
5729         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5730         is_gimple_call before calling gimple_call_internal_p.
5732 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5734         * Makefile.in (TREE_FLOW_H): Remove.
5735         (TREE_SSA_H): Add file names from tree-flow.h.
5736         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5737         * tree.h: Remove tree-flow.h reference.
5738         * hash-table.h: Remove tree-flow.h reference.
5739         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5740         reference with tree-ssa-loop.h.
5742 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5744         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5745         default element-order behavior for -maltivec.
5746         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5747         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5748         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5749         when targeting big endian, at least for now.
5750         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5752 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5754         PR middle-end/47735
5755         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5756         var satisfies use_register_for_decl, just take into account type
5757         alignment, rather than decl alignment.
5759         PR tree-optimization/59622
5760         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5761         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5762         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5763         Don't devirtualize for inplace at all.  For targets.length () == 1,
5764         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5766 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5768         * config/i386/i386.md (cpu): Remove the unused btver1.
5770 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5772         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5774 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5776         PR target/58115
5777         * tree-core.h (struct target_globals): New forward declaration.
5778         (struct tree_target_option): Add globals field.
5779         * tree.h (TREE_TARGET_GLOBALS): Define.
5780         (prepare_target_option_nodes_for_pch): New prototype.
5781         * target-globals.h (struct target_globals): Define even if
5782         !SWITCHABLE_TARGET.
5783         * tree.c (prepare_target_option_node_for_pch,
5784         prepare_target_option_nodes_for_pch): New functions.
5785         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5786         * config/i386/i386.c: Include target-globals.h.
5787         (ix86_set_current_function): Instead of doing target_reinit
5788         unconditionally, use save_target_globals_default_opts and
5789         restore_target_globals.
5791 2014-01-09  Richard Biener  <rguenther@suse.de>
5793         PR tree-optimization/59715
5794         * tree-cfg.h (split_critical_edges): Declare.
5795         * tree-cfg.c (split_critical_edges): Export.
5796         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5798 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5800         * cfgexpand.c (expand_stack_vars): Optionally disable
5801         asan stack protection.
5802         (expand_used_vars): Likewise.
5803         (partition_stack_vars): Likewise.
5804         * asan.c (asan_emit_stack_protection): Optionally disable
5805         after return stack usage.
5806         (instrument_derefs): Optionally disable memory access instrumentation.
5807         (instrument_builtin_call): Likewise.
5808         (instrument_strlen_call): Likewise.
5809         (asan_protect_global): Optionally disable global variables protection.
5810         * doc/invoke.texi: Added doc for new options.
5811         * params.def: Added new options.
5812         * params.h: Likewise.
5814 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5816         PR rtl-optimization/59724
5817         * ifcvt.c (cond_exec_process_if_block): Don't call
5818         flow_find_head_matching_sequence with 0 longest_match.
5819         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5820         non-active insns if !stop_after.
5821         (try_head_merge_bb): Revert 2014-01-07 changes.
5823 2014-01-08  Jeff Law  <law@redhat.com>
5825         * ree.c (get_sub_rtx): New function, extracted from...
5826         (merge_def_and_ext): Here.
5827         (combine_reaching_defs): Use get_sub_rtx.
5829 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5831         * cgraph.h (varpool_variable_node): Do not choke on null node.
5833 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5835         * config/mips/mips.md (simple_return): Attempt to use JRC
5836         for microMIPS.
5837         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5839 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5841         PR rtl-optimization/59137
5842         * reorg.c (steal_delay_list_from_target): Call update_block for
5843         elided insns.
5844         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5846 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5848         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5849         two duplicate entries.
5851 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5853         Revert:
5854         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5856         * config/mips/mips.c (mips_truncated_op_cost): New function.
5857         (mips_rtx_costs): Adjust test for BADDU.
5858         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5860         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5862         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5863         (*baddu_si): ...this new pattern.
5865 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5867         PR ipa/59722
5868         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5870 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5872         PR middle-end/57748
5873         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5874         inner_reference_p.
5875         (expand_expr, expand_normal): Adjust.
5876         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5877         inner_reference_p. Use inner_reference_p to expand inner references.
5878         (store_expr): Adjust.
5879         * cfgexpand.c (expand_call_stmt): Adjust.
5881 2014-01-08  Rong Xu  <xur@google.com>
5883         * gcov-io.c (gcov_var): Move from gcov-io.h.
5884         (gcov_position): Ditto.
5885         (gcov_is_error): Ditto.
5886         (gcov_rewrite): Ditto.
5887         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5888         only part to libgcc/libgcov.h.
5890 2014-01-08  Marek Polacek  <polacek@redhat.com>
5892         PR middle-end/59669
5893         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5895 2014-01-08  Marek Polacek  <polacek@redhat.com>
5897         PR sanitizer/59667
5898         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5900 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5902         PR rtl-optimization/59649
5903         * stor-layout.c (get_mode_bounds): For BImode return
5904         0 and STORE_FLAG_VALUE.
5906 2014-01-08  Richard Biener  <rguenther@suse.de>
5908         PR middle-end/59630
5909         * gimple.h (is_gimple_builtin_call): Remove.
5910         (gimple_builtin_call_types_compatible_p): New.
5911         (gimple_call_builtin_p): New overload.
5912         * gimple.c (is_gimple_builtin_call): Remove.
5913         (validate_call): Rename to ...
5914         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5915         check return types.
5916         (validate_type): New static function.
5917         (gimple_call_builtin_p): New overload and adjust.
5918         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5919         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5920         (gimple_fold_stmt_to_constant_1): Likewise.
5921         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5923 2014-01-08  Richard Biener  <rguenther@suse.de>
5925         PR middle-end/59471
5926         * gimplify.c (gimplify_expr): Gimplify register-register type
5927         VIEW_CONVERT_EXPRs to separate stmts.
5929 2014-01-07  Jeff Law  <law@redhat.com>
5931         PR middle-end/53623
5932         * ree.c (combine_set_extension): Handle case where source
5933         and destination registers in an extension insn are different.
5934         (combine_reaching_defs): Allow source and destination registers
5935         in extension to be different under limited circumstances.
5936         (add_removable_extension): Remove restriction that the
5937         source and destination registers in the extension are the same.
5938         (find_and_remove_re): Emit a copy from the extension's
5939         destination to its source after the defining insn if
5940         the source and destination registers are different.
5942         PR middle-end/59285
5943         * ifcvt.c (merge_if_block): If we are merging a block with more than
5944         one successor with a block with no successors, remove any BARRIER
5945         after the second block.
5947 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
5949         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5951 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5953         PR target/59652
5954         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5955         for 14-bit register offsets when INT14_OK_STRICT is false.
5957 2014-01-07  Roland Stigge  <stigge@antcom.de>
5958             Michael Meissner  <meissner@linux.vnet.ibm.com>
5960         PR 57386/target
5961         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5962         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5964 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5966         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5967         -mcpu.
5969 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5971         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5972         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5973         rtx is const0_rtx or not.
5975 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5977         PR target/58115
5978         * target-globals.c (save_target_globals): Remove this_fn_optab
5979         handling.
5980         * toplev.c: Include optabs.h.
5981         (target_reinit): Temporarily restore the global options if another
5982         set of options are in force.
5984 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5986         PR rtl-optimization/58668
5987         * cfgcleanup.c (flow_find_cross_jump): Don't count
5988         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5989         to determine what is counted.
5990         (flow_find_head_matching_sequence): Use active_insn_p to determine
5991         what is counted.
5992         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5993         counting change.
5994         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5995         determine what is counted.
5997         PR tree-optimization/59643
5998         * tree-predcom.c (split_data_refs_to_components): If one dr is
5999         read and one write, determine_offset fails and the write isn't
6000         in the bad component, just put the read into the bad component.
6002 2014-01-07  Mike Stump  <mikestump@comcast.net>
6003             Jakub Jelinek  <jakub@redhat.com>
6005         PR pch/59436
6006         * tree-core.h (struct tree_optimization_option): Change optabs
6007         type from unsigned char * to void *.
6008         * optabs.c (init_tree_optimization_optabs): Adjust
6009         TREE_OPTIMIZATION_OPTABS initialization.
6011 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
6013         PR target/59644
6014         * config/i386/i386.h (struct machine_function): Add
6015         no_drap_save_restore field.
6016         * config/i386/i386.c (ix86_save_reg): Use
6017         !cfun->machine->no_drap_save_restore instead of
6018         crtl->stack_realign_needed.
6019         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
6020         this function clears frame_pointer_needed.  Set
6021         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
6022         and DRAP reg is needed.
6024 2014-01-06  Marek Polacek  <polacek@redhat.com>
6026         PR c/57773
6027         * doc/implement-c.texi: Mention that other integer types are
6028         permitted as bit-field types in strictly conforming mode.
6030 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
6032         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
6033         is newly allocated.
6035 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
6037         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
6039 2014-01-06  Martin Jambor  <mjambor@suse.cz>
6041         PR ipa/59008
6042         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
6043         to int.
6044         * ipa-prop.c (ipa_print_node_params): Fix indentation.
6046 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
6048         PR debug/59350
6049         PR debug/59510
6050         * var-tracking.c (add_stores): Preserve the value of the source even if
6051         we don't record the store.
6053 2014-01-06  Terry Guo  <terry.guo@arm.com>
6055         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
6057 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
6059         PR bootstrap/59541
6060         * config/darwin.c (darwin_function_section): Adjust return values to
6061         correspond to optimisation changes made in r206070.
6063 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
6065         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
6066         from prefetch_block tune setting.
6067         (nocona_cost): Correct size of prefetch block to 64.
6069 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
6071         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
6072         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
6073         used to save the static chain register in the computation of the offset
6074         from which the FP registers need to be restored.
6076 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
6078         PR tree-optimization/59519
6079         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
6080         ICE if get_current_def (current_new_name) is already non-NULL, as long
6081         as it is a phi result of some other phi in *new_exit_bb that has
6082         the same argument.
6084         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
6085         or vmovdqu* for misaligned_operand.
6086         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
6087         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6088         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6089         aligned_mem for AVX512F masked aligned load and store builtins and for
6090         non-temporal moves.
6092 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6094         PR tree-optimization/59651
6095         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6096         Address range for negative step should be added by TYPE_SIZE_UNIT.
6098 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6100         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6101         overlapping registers also for registers other than the stack pointer.
6103 2014-01-03  Marek Polacek  <polacek@redhat.com>
6105         PR other/59661
6106         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6107         __builtin_FILE.
6109 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6111         PR target/59625
6112         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6113         asm goto as jump.
6115         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6116         (push splitter): Use <P:MODE_SIZE> instead of
6117         GET_MODE_SIZE (<P:MODE>mode).
6118         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6119         (mov -1, reg peephole2): Likewise.
6120         * config/i386/sse.md (*mov<mode>_internal,
6121         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6122         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6123         *<code><mode>3, *andnot<mode>3<mask_name>,
6124         <mask_codefor><code><mode>3<mask_name>): Likewise.
6125         * config/i386/subst.md (mask_mode512bit_condition,
6126         sd_mask_mode512bit_condition): Likewise.
6128 2014-01-02  Xinliang David Li  <davidxl@google.com>
6130         PR tree-optimization/59303
6131         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6132         (dump_predicates): Better output format.
6133         (pred_equal_p): New function.
6134         (is_neq_relop_p): Ditto.
6135         (is_neq_zero_form_p): Ditto.
6136         (pred_expr_equal_p): Ditto.
6137         (pred_neg_p): Ditto.
6138         (simplify_pred): Ditto.
6139         (simplify_preds_2): Ditto.
6140         (simplify_preds_3): Ditto.
6141         (simplify_preds_4): Ditto.
6142         (simplify_preds): Ditto.
6143         (push_pred): Ditto.
6144         (push_to_worklist): Ditto.
6145         (get_pred_info_from_cmp): Ditto.
6146         (is_degenerated_phi): Ditto.
6147         (normalize_one_pred_1): Ditto.
6148         (normalize_one_pred): Ditto.
6149         (normalize_one_pred_chain): Ditto.
6150         (normalize_preds): Ditto.
6151         (normalize_cond_1): Remove function.
6152         (normalize_cond): Ditto.
6153         (is_gcond_subset_of): Ditto.
6154         (is_subset_of_any): Ditto.
6155         (is_or_set_subset_of): Ditto.
6156         (is_and_set_subset_of): Ditto.
6157         (is_norm_cond_subset_of): Ditto.
6158         (pred_chain_length_cmp): Ditto.
6159         (convert_control_dep_chain_into_preds): Type change.
6160         (find_predicates): Ditto.
6161         (find_def_preds): Ditto.
6162         (destroy_predicates_vecs): Ditto.
6163         (find_matching_predicates_in_rest_chains): Ditto.
6164         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6165         (is_pred_expr_subset): Ditto.
6166         (is_pred_chain_subset_of): Ditto.
6167         (is_included_in): Ditto.
6168         (is_superset_of): Ditto.
6170 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6172         Update copyright years.
6174 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6176         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6177         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6178         config/arc/arc.md, config/arc/arc.opt,
6179         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6180         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6181         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6182         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6183         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6184         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6185         vtable-verify.c, vtable-verify.h: Use the standard form for the
6186         copyright notice.
6188 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6190         * gcc.c (process_command): Update copyright notice dates.
6191         * gcov-dump.c: Ditto.
6192         * gcov.c: Ditto.
6193         * doc/cpp.texi: Bump @copying's copyright year.
6194         * doc/cppinternals.texi: Ditto.
6195         * doc/gcc.texi: Ditto.
6196         * doc/gccint.texi: Ditto.
6197         * doc/gcov.texi: Ditto.
6198         * doc/install.texi: Ditto.
6199         * doc/invoke.texi: Ditto.
6201 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6203         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6205 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6207         * config/i386/sse.md (*mov<mode>_internal): Guard
6208         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6210         PR rtl-optimization/59647
6211         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6212         new_rtx into UNSIGNED_FLOAT rtxes.
6214 Copyright (C) 2014 Free Software Foundation, Inc.
6216 Copying and distribution of this file, with or without modification,
6217 are permitted in any medium without royalty provided the copyright
6218 notice and this notice are preserved.