* ipa-devirt.c (maybe_record_node): When node is not recorded,
[official-gcc.git] / gcc / ChangeLog
blob34a0d5e8f07cc168487f6ff99b176aa89d8ef247
1 2014-04-07  Jan Hubicka  <hubcika@ucw.cz>
3         * ipa-devirt.c (maybe_record_node): When node is not recorded,
4         set completep to false rather than true.
6 2014-04-07  Douglas B Rupp  <rupp@adacore.com>
8         PR target/60504
9         * config/arm/arm.h (ASM_PREFERRED_EH_DATA_FORMAT): Expose from
10         ARM_TARGET2_DWARF_FORMAT.
12 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
14         PR target/60609
15         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
16         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
17         ADDR_DIFF_VEC.
19 2014-04-07  Richard Biener  <rguenther@suse.de>
21         PR tree-optimization/60766
22         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an unsigned type.
23         (may_eliminate_iv): Convert cand_value_at result to desired type.
25 2014-04-07  Jason Merrill  <jason@redhat.com>
27         PR c++/60731
28         * common.opt (-fno-gnu-unique): Add.
29         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
31 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
33         * haifa-sched.c: Fix outdated function reference and minor
34         grammar errors in introductory comment.
36 2014-04-07  Richard Biener  <rguenther@suse.de>
38         PR middle-end/60750
39         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
40         for noreturn calls.
41         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
43 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
45         PR debug/55794
46         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
47         size accounting for thunks.
48         (pa_asm_output_mi_thunk): Use final_start_function() and
49         final_end_function() to output function start and end directives.
51 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
53         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
54         specific ISA/ feature information. Remove short_sp and errata_skip ds.
55         Add avr_device_specific_features enum to have device specific info.
56         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
57         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
58         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
59         updated device specific info.
60         * config/avr/avr-mcus.def: Merge device specific details to
61         dev_attribute field.
62         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
63         errata_skip.
64         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
65         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
66         assembler if RMW isa supported by current device.
67         * config/avr/genmultilib.awk: Update as device info structure changed.
68         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
70 2014-04-04  Cong Hou  <congh@google.com>
72         PR tree-optimization/60656
73         * tree-vect-stmts.c (supportable_widening_operation):
74         Fix a bug that elements in a vector with vect_used_by_reduction
75         property are incorrectly reordered when the operation on it is not
76         consistant with the one in reduction operation.
78 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
80         PR rtl-optimization/60155
81         * gcse.c (record_set_data): New function.
82         (single_set_gcse): New function.
83         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
84         (hoist_code): Likewise.
85         (get_pressure_class_and_nregs): Likewise.
87 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
89         * explow.c (probe_stack_range): Emit a final optimization blockage.
91 2014-04-04  Anthony Green  <green@moxielogic.com>
93         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
94         typos.
96 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
98         PR ipa/59626
99         * lto-cgraph.c (input_overwrite_node): Check that partitioning
100         flags are set only during streaming.
101         * ipa.c (process_references, walk_polymorphic_call_targets,
102         symtab_remove_unreachable_nodes): Drop bodies of always inline
103         after early inlining.
104         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
106 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
107         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
109         PR debug/60655
110         * dwarf2out.c (const_ok_for_output_1): Reject expressions
111         containing a NOT.
113 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
115         PR bootstrap/60743
116         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
117         duration.
118         (cortex_a53_fdivd): Likewise.
120 2014-04-04  Martin Jambor  <mjambor@suse.cz>
122         PR ipa/60640
123         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
124         Adjust all callers.
125         * cgraph.c (clone_of_p): Also return true if thunks match.
126         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
127         cgraph_function_or_thunk_node and an obsolete comment.
128         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
129         file.
130         (build_function_decl_skip_args): Likewise.
131         (set_new_clone_decl_and_node_flags): New function.
132         (duplicate_thunk_for_node): Likewise.
133         (redirect_edge_duplicating_thunks): Likewise.
134         (cgraph_clone_node): New parameter args_to_skip, pass it to
135         redirect_edge_duplicating_thunks which is called instead of
136         cgraph_redirect_edge_callee.
137         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
138         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
140 2014-04-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
142         PR fortran/60191
143         * fortran/trans-types.c (gfc_get_function_type): In case of recursion
144         build a variadic function type with empty argument list instead of a
145         stdarg-like function type with incomplete argument list.
147 2014-04-04  Jeff Law  <law@redhat.com>
149         PR target/60657
150         * config/arm/predicates.md (const_int_I_operand): New predicate.
151         (const_int_M_operand): Similarly.
152         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
153         const_int_operand.
154         (insv_t2, extv_reg, extzv_t2): Likewise.
155         (load_multiple_with_writeback): Similarly for const_int_I_operand.
156         (pop_multiple_with_writeback_and_return): Likewise.
157         (vfp_pop_multiple_with_writeback): Likewise
159 2014-04-04  Richard Biener  <rguenther@suse.de>
161         PR ipa/60746
162         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
163         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
164         non-GIMPLE_LABELs.
165         * gimplify.h (gimple_add_tmp_var_fn): Declare.
166         * gimplify.c (gimple_add_tmp_var_fn): New function.
167         * gimple-expr.h (create_tmp_reg_fn): Declare.
168         * gimple-expr.c (create_tmp_reg_fn): New function.
169         * gimple-low.c (record_vars_into): Don't change cfun.
170         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
171         code generation without cfun.
173 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
175         PR bootstrap/60719
176         * Makefile.in (install-driver): Fix shell scripting.
178 2014-04-03  Cong Hou  <congh@google.com>
180         PR tree-optimization/60505
181         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
182         threshold of number of iterations below which no vectorization will be
183         done.
184         * tree-vect-loop.c (new_loop_vec_info):
185         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
186         * tree-vect-loop.c (vect_analyze_loop_operations):
187         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
188         * tree-vect-loop.c (vect_transform_loop):
189         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
190         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
191         of iterations of the loop and see if we should build the epilogue.
193 2014-04-03  Richard Biener  <rguenther@suse.de>
195         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
196         member.
197         (streamer_tree_cache_create): Adjust.
198         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
199         to allow optional nodes array.
200         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
201         (streamer_tree_cache_append): Likewise.
202         (streamer_tree_cache_create): Create nodes array optionally
203         as specified by parameter.
204         * lto-streamer-out.c (create_output_block): Avoid maintaining
205         the node array in the writer cache.
206         (DFS_write_tree): Remove assertion.
207         (produce_asm_for_decls): Free the out decl state hash table
208         early.
209         * lto-streamer-in.c (lto_data_in_create): Adjust for
210         streamer_tree_cache_create prototype change.
212 2014-04-03  Richard Biener  <rguenther@suse.de>
214         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
215         set TREE_CHAIN to NULL_TREE.
217 2014-04-03  Richard Biener  <rguenther@suse.de>
219         PR tree-optimization/60740
220         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
221         over all GIMPLE_COND operands.
223 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
225         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
226         (Weffc++): Remove Scott's numbering, merge lists and reference
227         Wnon-virtual-dtor.
229         c-family/
231         cp/
232         * class.c (accessible_nvdtor_p): New.
233         (check_bases): Don't check base destructor here ...
234         (check_bases_and_members): ... check them here.  Trigger on
235         Wnon-virtual-dtor flag.
236         (finish_struct_1): Use accessible_nvdtor_p.
238         testsuite/
239         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
240         * g++.dg/warn/Wnvdtor-2.C: New.
241         * g++.dg/warn/Wnvdtor-3.C: New.
242         * g++.dg/warn/Wnvdtor-4.C: New.
243         * g++.dg/warn/Weff1.C: Delete.
244         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
245         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
247 2014-04-03  Nick Clifton  <nickc@redhat.com>
249         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
250         properly.
252 2014-04-03  Martin Jambor  <mjambor@suse.cz>
254         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
255         mention gcc_unreachable before failing.
256         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
257         removed symbols.
259 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
261         PR ipa/60659
262         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
263         code and instead mark the context inconsistent.
264         (possible_polymorphic_call_targets): For inconsistent contexts
265         return empty complete list.
267 2014-04-02  Anthony Green  <green@moxielogic.com>
269         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
270         (extendqisi2, extendhisi2): Define.
271         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
272         (WCHAR_TYPE): Change to unsigned int.
274 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
276         PR tree-optimization/60733
277         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
278         insertion point for PHI candidates to be the end of the feeding
279         block for the PHI argument.
281 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
283         PR rtl-optimization/60650
284         * lra-constraints.c (process_alt_operands): Decrease reject for
285         earlyclobber matching.
287 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
289         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
291 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
293         * config/spu/spu.c (pad_bb): Do not crash when the last
294         insn is CODE_FOR_blockage.
296 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
298         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
299         lies outside the target mode.
301 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
303         PR target/60735
304         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
305         software floating point or no floating point registers, do not
306         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
307         in GPRs that occurs after we tested for GPRs that would never be
308         true.
310         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
311         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
312         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
313         specifically allow DDmode, since that does not use the SPE SIMD
314         instructions.
316 2014-04-02  Richard Biener  <rguenther@suse.de>
318         PR middle-end/60729
319         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
320         MODE_INTs.  Properly use negv_optab.
321         (expand_abs): Likewise.
323 2014-04-02  Richard Biener  <rguenther@suse.de>
325         PR bootstrap/60719
326         * Makefile.in (install-driver): Guard extra installs with special
327         names properly.
329 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
331         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
332         Document vec_vgbbd.
334 2014-04-01  Richard Henderson  <rth@redhat.com>
336         PR target/60704
337         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
338         alternative enabled before register allocation.
340 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
342         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
343         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
344         typo.
345         (nios2_large_got_address): Remove unneeded 'sym' parameter.
346         (nios2_got_address): Update nios2_large_got_address call site.
347         (nios2_delegitimize_address): New function.
348         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
349         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
350         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
352 2014-04-01  Martin Husemann  <martin@duskware.de>
354         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
355         for -mabi=32.
357 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
359         PR rtl-optimization/60604
360         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
361         check from register_operand.
362         (register_operand): Redefine in terms of general_operand.
363         (nonmemory_operand): Use register_operand for the non-constant cases.
365 2014-04-01  Richard Biener  <rguenther@suse.de>
367         * gimple.h (struct gimple_statement_base): Align subcode to
368         16 bits.
370 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
372         * doc/invoke.texi (mapp-regs): Clarify.
374 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
376         * config/i386/avx512fintrin.h (__v32hi): Define type.
377         (__v64qi): Likewise.
378         (_mm512_set1_epi8): Define.
379         (_mm512_set1_epi16): Define.
380         (_mm512_set4_epi32): Define.
381         (_mm512_set4_epi64): Define.
382         (_mm512_set4_pd): Define.
383         (_mm512_set4_ps): Define.
384         (_mm512_setr4_epi64): Define.
385         (_mm512_setr4_epi32): Define.
386         (_mm512_setr4_pd): Define.
387         (_mm512_setr4_ps): Define.
388         (_mm512_setzero_epi32): Define.
390 2014-03-31  Martin Jambor  <mjambor@suse.cz>
392         PR middle-end/60647
393         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
394         callsite_arguments_match_p.  Updated all callers.  Also check types of
395         corresponding formal parameters and actual arguments.
396         (not_all_callers_have_enough_arguments_p) Renamed to
397         some_callers_have_mismatched_arguments_p.
399 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
401         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
403 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
405         PR target/60034
406         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
407         section anchor.
409 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
411         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
412         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
413         Split out
414         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
415         Use FMAMODE_NOVF512 mode iterator.
416         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
417         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
418         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
419         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
420         Split out
421         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
422         Use VF_128_256 mode iterator.
423         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
424         Ditto.
426 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
428         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
429         static chain if needed.
431 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
433         PR target/60697
434         * lra-constraints.c (index_part_to_reg): New.
435         (process_address): Use it.
437 2014-03-27  Jeff Law  <law@redhat.com>
438             Jakub Jelinek  <jakub@redhat.com>
440         PR target/60648
441         * expr.c (do_tablejump): Use simplify_gen_binary rather than
442         gen_rtx_{PLUS,MULT} to build up the address expression.
444         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
445         creating non-canonical RTL.
447 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
449         PR ipa/60243
450         * ipa-inline.c (want_inline_small_function_p): Short circuit large
451         functions; reorganize to make cheap checks first.
452         (inline_small_functions): Do not estimate growth when dumping;
453         it is expensive.
454         * ipa-inline.h (inline_summary): Add min_size.
455         (growth_likely_positive): New function.
456         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
457         (set_cond_stmt_execution_predicate): Cleanup.
458         (estimate_edge_size_and_time): Compute min_size.
459         (estimate_calls_size_and_time): Likewise.
460         (estimate_node_size_and_time): Likewise.
461         (inline_update_overall_summary): Update min_size.
462         (do_estimate_edge_time): Likewise.
463         (do_estimate_edge_size): Update.
464         (do_estimate_edge_hints): Update.
465         (growth_likely_positive): New function.
467 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
469         PR target/60693
470         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
471         also if addr has VOIDmode.
473 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
475         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
476         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
477         Declare extern.
478         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
479         instructions as well as AdvancedSIMD loads.
481 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
483         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
484         Use crypto_aese type.
485         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
486         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
487         crypto_aese, crypto_aesmc.  Move to types.md.
488         * config/arm/types.md (crypto_aes): Split into crypto_aese,
489         crypto_aesmc.
490         * config/arm/iterators.md (crypto_type): Likewise.
492 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
494         * cgraph.c: Include expr.h and tree-dfa.h.
495         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
496         remove LHS.
498 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
500         PR target/60675
501         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
502         regs from checking multi-reg pseudos.
504 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
506         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
508 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
510         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
511         if it would clobber the stack pointer, even temporarily.
513 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
515         * mode-switching.c: Make small adjustments to the top comment.
517 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
519         * config/rs6000/constraints.md (wD constraint): New constraint to
520         match the constant integer to get the top DImode/DFmode out of a
521         vector in a VSX register.
523         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
524         match the constant integer to get the top DImode/DFmode out of a
525         vector in a VSX register.
527         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
528         for ISA 2.07.
530         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
531         vbpermq builtins.
533         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
534         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
536         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
537         Optimize vec_extract of 64-bit values, where the value being
538         extracted is in the top word, where we can use scalar
539         instructions.  Add direct move and store support.  Combine the big
540         endian/little endian vector select load support into a single insn.
541         (vsx_extract_<mode>_internal1): Likewise.
542         (vsx_extract_<mode>_internal2): Likewise.
543         (vsx_extract_<mode>_load): Likewise.
544         (vsx_extract_<mode>_store): Likewise.
545         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
546         combined into vsx_extract_<mode>_load.
547         (vsx_extract_<mode>_one_le): Likewise.
549         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
550         define the top 64-bit vector element.
552         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
553         constraint.
555         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
556         Document vec_vbpermq builtin.
558         PR target/60672
559         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
560         enable use of xxsldwi and xxpermdi builtin functions.
561         (vec_xxpermdi): Likewise.
563         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
564         Document use of vec_xxsldwi and vec_xxpermdi builtins.
566 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
568         PR rtl-optimization/60650
569         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
570         first_p.  Use it.
571         (find_spills_for): New.
572         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
573         Spill all pseudos on the second iteration.
575 2014-03-27  Marek Polacek  <polacek@redhat.com>
577         PR c/50347
578         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
579         types.
581 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
583         * config/s390/s390.c (s390_can_use_return_insn): Check for
584         call-saved FPRs on 31 bit.
586 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
588         PR middle-end/60682
589         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
590         if they need regimplification, just drop them instead of
591         calling gimple_regimplify_operands on them.
593 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
595         PR target/60580
596         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
597         (aarch64_frame_pointer_required): Adjust logic.
598         (aarch64_can_eliminate): Adjust logic.
599         (aarch64_override_options_after_change): Adjust logic.
601 2014-03-27  Dehao Chen  <dehao@google.com>
603         * ipa-inline.c (early_inliner): Update node's inline info.
605 2014-03-26  Dehao Chen  <dehao@google.com>
607         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
608         compiler inserted conditional jumps for NAN float check.
610 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
612         * ubsan.h (ubsan_create_data): Change second argument's type
613         to const location_t *.
614         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
615         _("<unknown>").
616         (ubsan_create_data): Change second argument to const location_t *PLOC.
617         Create Loc field whenever PLOC is non-NULL.
618         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
619         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
620         callers.
622         PR other/59545
623         * real.c (real_to_integer2): Change type of low to UHWI.
625 2014-03-26  Tobias Burnus  <burnus@net-b.de>
627         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
628         (CILK_SELF_SPECS): New define.
629         (driver_self_specs): Use it.
631 2014-03-26  Richard Biener  <rguenther@suse.de>
633         * tree-pretty-print.c (percent_K_format): Implement special
634         case for LTO and its stripped down BLOCK tree.
636 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
638         PR sanitizer/60636
639         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
641         * tree-vrp.c (simplify_internal_call_using_ranges): If only
642         one range is range_int_cst_p, but not both, at least optimize
643         addition/subtraction of 0 and multiplication by 0 or 1.
644         * gimple-fold.c (gimple_fold_call): Fold
645         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
646         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
647         INTEGER_CSTs, try to fold at least x * 0 and y - y.
649 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
651         PR rtl-optimization/60452
652         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
653         <case REG>: Return 1 for invalid offsets from the frame pointer.
655 2014-03-26  Marek Polacek  <polacek@redhat.com>
657         PR c/37428
658         * doc/extend.texi (C Extensions): Mention variable-length arrays in
659         a structure/union.
661 2014-03-26  Marek Polacek  <polacek@redhat.com>
663         PR c/39525
664         * doc/extend.texi (Designated Inits): Describe what happens to omitted
665         field members.
667 2014-03-26  Marek Polacek  <polacek@redhat.com>
669         PR other/59545
670         * ira-color.c (update_conflict_hard_regno_costs): Perform the
671         multiplication in unsigned type.
673 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
675         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
677 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
679         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
681 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
683         PR ipa/60315
684         * cif-code.def (UNREACHABLE) New code.
685         * ipa-inline.c (inline_small_functions): Skip edges to
686         __builtlin_unreachable.
687         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
688         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
689         predicate to __bulitin_unreachable.
690         (set_cond_stmt_execution_predicate): Fix issue when
691         invert_tree_comparison returns ERROR_MARK.
692         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
693         propagate to inline clones.
694         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
695         to unreachable.
696         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
697         * cgraphclones.c (cgraph_clone_node): If call destination is already
698         ureachable, do not redirect it back.
699         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
700         unreachable.
702 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
704         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
705         Do not modify inline clones.
707 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
709         * config/i386/i386.md (general_sext_operand): New mode attr.
710         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
711         don't generate (sign_extend (const_int)).
712         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
713         operands[2].  Use We constraint instead of <i> and
714         <general_sext_operand> predicate instead of <general_operand>.
715         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
716         * config/i386/constraints.md (We): New constraint.
717         * config/i386/predicates.md (x86_64_sext_operand,
718         sext_operand): New predicates.
720 2014-03-25  Martin Jambor  <mjambor@suse.cz>
722         PR ipa/60600
723         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
724         inconsistent devirtualizations to __builtin_unreachable.
726 2014-03-25  Marek Polacek  <polacek@redhat.com>
728         PR c/35449
729         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
731 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
733         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
734         order of elements for big-endian.
736 2014-03-25  Richard Biener  <rguenther@suse.de>
738         PR middle-end/60635
739         * gimplify-me.c (gimple_regimplify_operands): Update the
740         re-gimplifed stmt.
742 2014-03-25  Martin Jambor  <mjambor@suse.cz>
744         PR ipa/59176
745         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
746         (lto_output_varpool_node): Likewise.
747         (input_overwrite_node): Likewise.
748         (input_varpool_node): Likewise.
750 2014-03-25  Richard Biener  <rguenther@suse.de>
752         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
753         (run_gcc): Likewise.
755 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
757         * combine.c (simplify_compare_const): Add MODE argument.
758         Handle mode_width 0 as very large mode_width.
759         (try_combine, simplify_comparison): Adjust callers.
761         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
762         type to avoid signed integer overflow.
763         * explow.c (plus_constant): Likewise.
765 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
767         * doc/generic.texi: Correct typos.
769 2014-03-24  Tobias Burnus  <burnus@net-b.de>
771         * doc/invoke.texi (-flto): Expand section about
772         using static libraries with LTO.
774 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
776         PR rtl-optimization/60501
777         * optabs.def (addptr3_optab): New optab.
778         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
779         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
780         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
782         * lra.c (emit_add3_insn): Use the addptr pattern if available.
784         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
786 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
788         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
789         _mm512_set1_pd.
791         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
792         (_mm256_undefined_ps): Define.
793         (_mm256_undefined_pd): Define.
794         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
795         (_mm_undefined_pd): Define.
796         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
797         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
798         (_mm512_undefined_ps): Define.
799         (_mm512_undefined_pd): Define.
800         Use _mm*_undefined_*.
801         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
803 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
805         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
806         (lshr_simd): DI mode added.
807         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
808         (aarch64_ushr_simddi): Likewise.
809         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
810         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
811         (vshrd_n_u64): Likewise.
813 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
815         * Makefile.in (s-macro_list): Depend on cc1.
817 2014-03-23  Teresa Johnson  <tejohnson@google.com>
819         * ipa-utils.c (ipa_print_order): Use specified dump file.
821 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
823         PR rtl-optimization/60601
824         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
826         * gcc.c (eval_spec_function): Initialize save_growing_value.
828 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
830         PR sanitizer/60613
831         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
832         code == MINUS_EXPR, never swap op0 with op1.
834         * toplev.c (init_local_tick): Avoid signed integer multiplication
835         overflow.
836         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
837         shift by first operand's bitsize.
839 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
841         PR target/60610
842         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
843         redefine to 1 or 0.
844         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
845         TARGET_ISA_64BIT_P(x).
847 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
849         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
850         pattern for vector nor instead of subtract from splat(-1).
851         (altivec_expand_vec_perm_const_le): Likewise.
853 2014-03-21  Richard Henderson  <rth@twiddle.net>
855         PR target/60598
856         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
857         related insns after epilogue_completed.
859 2014-03-21  Martin Jambor  <mjambor@suse.cz>
861         PR ipa/59176
862         * cgraph.h (symtab_node): New flag body_removed.
863         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
864         when removing bodies.
865         * symtab.c (dump_symtab_base): Dump body_removed flag.
866         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
867         had their bodies removed.
869 2014-03-21  Martin Jambor  <mjambor@suse.cz>
871         PR ipa/60419
872         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
873         in the border.
875 2014-03-21  Richard Biener  <rguenther@suse.de>
877         PR tree-optimization/60577
878         * tree-core.h (struct tree_base): Document nothrow_flag use
879         in VAR_DECL_NONALIASED.
880         * tree.h (VAR_DECL_NONALIASED): New.
881         (may_be_aliased): Adjust.
882         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
884 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
886         * expr.c (expand_expr_real_1): Remove outdated comment.
888 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
890         PR middle-end/60597
891         * ira.c (adjust_cleared_regs): Call copy_rtx on
892         *reg_equiv[REGNO (loc)].src_p before passing it to
893         simplify_replace_fn_rtx.
895         PR target/60568
896         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
897         into CONST, put pic register as first operand of PLUS.  Use
898         gen_const_mem for both 32-bit and 64-bit PIC got loads.
900 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
902         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
904 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
906         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
907         around for store forwarding issue in the FPU on the UT699.
908         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
909         loads and operations if -mfix-ut699 is specified.
910         (divtf3_hq): Tweak attribute.
911         (sqrttf2_hq): Likewise.
913 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
915         * calls.c (store_one_arg): Remove incorrect const qualification on the
916         type of the temporary.
917         * cfgexpand.c (expand_return): Likewise.
918         * expr.c (expand_constructor): Likewise.
919         (expand_expr_real_1): Likewise.
921 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
923         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
924         of parts.
926 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
928         PR target/60039
929         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
931 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
933         * config/arm/aarch-common-protos.h
934         (alu_cost_table): Fix spelling of "extend".
935         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
937 2014-03-19  Richard Biener  <rguenther@suse.de>
939         PR middle-end/60553
940         * tree-core.h (tree_type_common): Re-order pointer members
941         to reduce recursion depth during GC walks.
943 2014-03-19  Marek Polacek  <polacek@redhat.com>
945         PR sanitizer/60569
946         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
947         before accessing it.
949 2014-03-19  Richard Biener  <rguenther@suse.de>
951         PR lto/59543
952         * lto-streamer-in.c (input_function): In WPA stage do not drop
953         debug stmts.
955 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
957         PR tree-optimization/60559
958         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
959         with build_zero_cst assignment.
961 2014-03-18  Kai Tietz  <ktietz@redhat.com>
963         PR rtl-optimization/56356
964         * sdbout.c (sdbout_parms): Verify that parms'
965         incoming argument is valid.
966         (sdbout_reg_parms): Likewise.
968 2014-03-18  Richard Henderson  <rth@redhat.com>
970         PR target/60562
971         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
972         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
973         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
975 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
977         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
978         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
979         Italicize plugin event names in description.  Explain that
980         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
981         Remind that no GCC functions should be called after PLUGIN_FINISH.
982         Explain what pragmas with expansion are.
984 2014-03-18  Martin Liska  <mliska@suse.cz>
986         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
987         gimple call statement is update.
988         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
989         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
991 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
993         PR sanitizer/60557
994         * ubsan.c (ubsan_instrument_unreachable): Call
995         initialize_sanitizer_builtins.
996         (ubsan_pass): Likewise.
998         PR sanitizer/60535
999         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
1000         varpool_finalize_decl instead of rest_of_decl_compilation.
1002 2014-03-18  Richard Biener  <rguenther@suse.de>
1004         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
1005         by using bitmap_and_compl instead of bitmap_and_compl_into.
1006         (df_rd_transfer_function): Likewise.
1008 2014-03-18  Richard Biener  <rguenther@suse.de>
1010         * doc/lto.texi (fresolution): Fix typo.
1012 2014-03-18  Richard Biener  <rguenther@suse.de>
1014         * doc/invoke.texi (flto): Update for changes in 4.9.
1016 2014-03-18  Richard Biener  <rguenther@suse.de>
1018         * doc/loop.texi: Remove section on the removed lambda framework.
1019         Update loop docs with recent changes in preserving loop structure.
1021 2014-03-18  Richard Biener  <rguenther@suse.de>
1023         * doc/lto.texi (-fresolution): Document.
1025 2014-03-18  Richard Biener  <rguenther@suse.de>
1027         * doc/contrib.texi: Adjust my name.
1029 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1031         PR ipa/58721
1032         * internal-fn.c: Include diagnostic-core.h.
1033         (expand_BUILTIN_EXPECT): New function.
1034         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1035         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1036         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1037         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1038         IFN_BUILTIN_EXPECT.
1039         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1040         Revert 3 argument __builtin_expect code.
1041         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1042         * gimple-fold.c (gimple_fold_call): Likewise.
1043         * tree.h (fold_builtin_expect): New prototype.
1044         * builtins.c (build_builtin_expect_predicate): Add predictor
1045         argument, if non-NULL, create 3 argument __builtin_expect.
1046         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1047         pass it through to build_builtin_expect_predicate.
1048         (fold_builtin_2): Adjust caller.
1049         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1050         * internal-fn.def (BUILTIN_EXPECT): New.
1052 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1054         PR ipa/58721
1055         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1056         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1057         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1059 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1061         PR ipa/58721
1062         * predict.c (combine_predictions_for_bb): Fix up formatting.
1063         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1064         fill what it points to if non-NULL.
1065         (tree_predict_by_opcode): Adjust caller, use the predictor.
1066         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1068 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1070         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1071         proper constant for the store mode.
1073 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1075         * symtab.c (change_decl_assembler_name): Fix transparent alias
1076         chain construction.
1078 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1080         * config/aarch64/aarch64.c: Correct the comments about the
1081         aarch64 stack layout.
1083 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1085         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1086         check for GF_OMP_FOR_KIND_FOR.
1088 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1090         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1091         ymm and zmm register names.
1093 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1095         PR target/60516
1096         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1097         note creation for the 2010-08-31 changes.
1099 2014-03-17  Marek Polacek  <polacek@redhat.com>
1101         PR middle-end/60534
1102         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1103         as -fno-tree-loop-vectorize.
1104         (expand_omp_simd): Likewise.
1106 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1108         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1109         (eligible_for_call_delay): New prototype.
1110         * config/sparc/sparc.c (tls_call_delay): Rename into...
1111         (eligible_for_call_delay): ...this.  Return false if the instruction
1112         cannot be put in the delay slot of a branch.
1113         (eligible_for_restore_insn): Simplify.
1114         (eligible_for_return_delay): Return false if the instruction cannot be
1115         put in the delay slot of a branch and simplify.
1116         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1117         put in the delay slot of a branch.
1118         * config/sparc/sparc.md (fix_ut699): New attribute.
1119         (tls_call_delay): Delete.
1120         (in_call_delay): Reimplement.
1121         (eligible_for_sibcall_delay): Rename into...
1122         (in_sibcall_delay): ...this.
1123         (eligible_for_return_delay): Rename into...
1124         (in_return_delay): ...this.
1125         (in_branch_delay): Reimplement.
1126         (in_uncond_branch_delay): Delete.
1127         (in_annul_branch_delay): Delete.
1129 2014-03-14  Richard Henderson  <rth@redhat.com>
1131         PR target/60525
1132         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1133         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1134         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1135         (floathi splitters): Remove.
1136         (float<SWI48x>xf2): New pattern.
1137         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1138         code that tried to handle DImode for 32-bit, but which was excluded
1139         by the pattern's condition.  Drop allocation of stack temporary.
1140         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1141         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1142         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1143         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1144         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1145         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1146         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1147         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1148         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1149         (*float<SWI48x><X87MODEF>2_i387): Remove.
1150         (all float _with_temp splitters): Remove.
1151         (*float<SWI48x><MODEF>2_i387): New pattern.
1152         (*float<SWI48><MODEF>2_sse): New pattern.
1153         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1154         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1156 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1157             Marek Polacek  <polacek@redhat.com>
1159         PR middle-end/60484
1160         * common.opt (dump_base_name_prefixed): New Variable.
1161         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1162         if x_dump_base_name_prefixed is already set, set it at the end.
1164 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1166         PR rtl-optimization/60508
1167         * lra-constraints.c (get_reload_reg): Add new parameter
1168         in_subreg_p.
1169         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1170         Pass the new parameter values.
1172 2014-03-14  Richard Biener  <rguenther@suse.de>
1174         * common.opt: Revert unintented changes from r205065.
1175         * opts.c: Likewise.
1177 2014-03-14  Richard Biener  <rguenther@suse.de>
1179         PR middle-end/60518
1180         * cfghooks.c (split_block): Properly adjust all loops the
1181         block was a latch of.
1183 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1185         PR lto/60461
1186         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1187         and simplify it.
1189 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1191         PR target/59396
1192         * config/avr/avr.c (avr_set_current_function): Pass function name
1193         through default_strip_name_encoding before sanity checking instead
1194         of skipping the first char of the assembler name.
1196 2014-03-13  Richard Henderson  <rth@redhat.com>
1198         PR debug/60438
1199         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1200         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1201         * config/i386/i386-protos.h: Likewise.
1202         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1203         in the expander instead of a splitter.
1204         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1205         any possibility of requiring a memory.
1206         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1207         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1208         (fp branch splitters): Update for ix86_split_fp_branch.
1209         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1210         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1211         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1212         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1213         (*fop_<MODEF>_3_i387): Likewise.
1214         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1215         (splitters for the fop_* register patterns): Remove.
1216         (fscalexf4_i387): Rename from *fscalexf4_i387.
1217         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1219 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1221         PR tree-optimization/59779
1222         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1223         type for bitsize and maxsize instead of HOST_WIDE_INT.
1225 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1227         PR rtl-optimization/57320
1228         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1229         the CFG after thread_prologue_and_epilogue_insns.
1231 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1233         PR rtl-optimization/57189
1234         * lra-constraints.c (process_alt_operands): Disfavor spilling
1235         vector pseudos.
1237 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1239         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1241 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1243         PR tree-optimization/59025
1244         PR middle-end/60418
1245         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1246         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1248 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1250         PR target/60486
1251         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1252         calls of avr_out_plus_1.
1254 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1256         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1257         BB's single pred and update the father loop's latch info later.
1259 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1261         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1262         (VEC_M): Likewise.
1263         (VEC_N): Likewise.
1264         (VEC_R): Likewise.
1265         (VEC_base): Likewise.
1266         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1267         registers, we need to swap double words in little endian mode.
1269         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1270         to be a container mode for 128-bit integer operations added in ISA
1271         2.07.  Unlike TImode and PTImode, the preferred register set is
1272         the Altivec/VMX registers for the 128-bit operations.
1274         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1275         declarations.
1276         (rs6000_split_128bit_ok_p): Likewise.
1278         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1279         macros for creating ISA 2.07 normal and overloaded builtin
1280         functions with 3 arguments.
1281         (BU_P8V_OVERLOAD_3): Likewise.
1282         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1283         for use as overloaded functions.
1284         (VPERM_1TI_UNS): Likewise.
1285         (VSEL_1TI): Likewise.
1286         (VSEL_1TI_UNS): Likewise.
1287         (ST_INTERNAL_1ti): Likewise.
1288         (LD_INTERNAL_1ti): Likewise.
1289         (XXSEL_1TI): Likewise.
1290         (XXSEL_1TI_UNS): Likewise.
1291         (VPERM_1TI): Likewise.
1292         (VPERM_1TI_UNS): Likewise.
1293         (XXPERMDI_1TI): Likewise.
1294         (SET_1TI): Likewise.
1295         (LXVD2X_V1TI): Likewise.
1296         (STXVD2X_V1TI): Likewise.
1297         (VEC_INIT_V1TI): Likewise.
1298         (VEC_SET_V1TI): Likewise.
1299         (VEC_EXT_V1TI): Likewise.
1300         (EQV_V1TI): Likewise.
1301         (NAND_V1TI): Likewise.
1302         (ORC_V1TI): Likewise.
1303         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1304         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1305         overloaded builtin.
1306         (VADDUQM): Likewise.
1307         (VSUBCUQ): Likewise.
1308         (VADDEUQM): Likewise.
1309         (VADDECUQ): Likewise.
1310         (VSUBEUQM): Likewise.
1311         (VSUBECUQ): Likewise.
1313         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1314         __int128_t and __uint128_t types.
1315         (__uint128_type): Likewise.
1316         (altivec_categorize_keyword): Add support for vector __int128_t,
1317         vector __uint128_t, vector __int128, and vector unsigned __int128
1318         as a container type for TImode operations that need to be done in
1319         VSX/Altivec registers.
1320         (rs6000_macro_to_expand): Likewise.
1321         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1322         to support 128-bit integer instructions vaddcuq, vadduqm,
1323         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1324         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1326         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1327         for V1TImode, and set up preferences to use VSX/Altivec registers.
1328         Setup VSX reload handlers.
1329         (rs6000_debug_reg_global): Likewise.
1330         (rs6000_init_hard_regno_mode_ok): Likewise.
1331         (rs6000_preferred_simd_mode): Likewise.
1332         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1333         (easy_altivec_constant): Likewise.
1334         (output_vec_const_move): Likewise.
1335         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1336         simple move.
1337         (rs6000_expand_vector_extract): Likewise.
1338         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1339         addressing.
1340         (rs6000_const_vec): Add support for V1TImode.
1341         (rs6000_emit_le_vsx_load): Swap double words when loading or
1342         storing TImode/V1TImode.
1343         (rs6000_emit_le_vsx_store): Likewise.
1344         (rs6000_emit_le_vsx_move): Likewise.
1345         (rs6000_emit_move): Add support for V1TImode.
1346         (altivec_expand_ld_builtin): Likewise.
1347         (altivec_expand_st_builtin): Likewise.
1348         (altivec_expand_vec_init_builtin): Likewise.
1349         (altivec_expand_builtin): Likewise.
1350         (rs6000_init_builtins): Add support for V1TImode type.  Add
1351         support for ISA 2.07 128-bit integer builtins.  Define type names
1352         for the VSX/Altivec vector types.
1353         (altivec_init_builtins): Add support for overloaded vector
1354         functions with V1TImode type.
1355         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1356         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1357         external function.
1358         (rs6000_split_128bit_ok_p): Likewise.
1359         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1360         __int128_t and vector __uint128_t.
1362         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1363         and mode attributes.
1364         (VSX_M): Likewise.
1365         (VSX_M2): Likewise.
1366         (VSm): Likewise.
1367         (VSs): Likewise.
1368         (VSr): Likewise.
1369         (VSv): Likewise.
1370         (VS_scalar): Likewise.
1371         (VS_double): Likewise.
1372         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1374         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1375         we support the ISA 2.07 128-bit integer arithmetic instructions.
1376         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1377         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1378         and TImode types for use with the builtin functions.
1379         (V1TI_type_node): Likewise.
1380         (unsigned_V1TI_type_node): Likewise.
1381         (intTI_type_internal_node): Likewise.
1382         (uintTI_type_internal_node): Likewise.
1384         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1385         128-bit builtin functions.
1386         (UNSPEC_VADDEUQM): Likewise.
1387         (UNSPEC_VADDECUQ): Likewise.
1388         (UNSPEC_VSUBCUQ): Likewise.
1389         (UNSPEC_VSUBEUQM): Likewise.
1390         (UNSPEC_VSUBECUQ): Likewise.
1391         (VM): Add V1TImode to vector mode iterators.
1392         (VM2): Likewise.
1393         (VI_unit): Likewise.
1394         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1395         (altivec_vaddcuq): Likewise.
1396         (altivec_vsubuqm): Likewise.
1397         (altivec_vsubcuq): Likewise.
1398         (altivec_vaddeuqm): Likewise.
1399         (altivec_vaddecuq): Likewise.
1400         (altivec_vsubeuqm): Likewise.
1401         (altivec_vsubecuq): Likewise.
1403         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1404         mode iterators.
1405         (BOOL_128): Likewise.
1406         (BOOL_REGS_OUTPUT): Likewise.
1407         (BOOL_REGS_OP1): Likewise.
1408         (BOOL_REGS_OP2): Likewise.
1409         (BOOL_REGS_UNARY): Likewise.
1410         (BOOL_REGS_AND_CR0): Likewise.
1412         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1413         128-bit integer builtin support.
1414         (vec_vadduqm): Likewise.
1415         (vec_vaddecuq): Likewise.
1416         (vec_vaddeuqm): Likewise.
1417         (vec_vsubecuq): Likewise.
1418         (vec_vsubeuqm): Likewise.
1419         (vec_vsubcuq): Likewise.
1420         (vec_vsubuqm): Likewise.
1422         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1423         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1424         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1425         128-bit integer add/subtract to ISA 2.07.
1427 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1429         * config/arc/arc.c (arc_predicate_delay_insns):
1430         Fix third argument passed to conditionalize_nonjump.
1432 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1434         * config/aarch64/aarch64-builtins.c
1435         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1436         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1437         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1438         instead of __builtin_lfloor.
1439         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1441 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1443         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1444         (tree_ssa_ifcombine_bb_1): New function.
1445         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1446         is an empty forwarder block to then_bb or vice versa and then_bb
1447         and else_bb are effectively swapped.
1449 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1451         PR target/60264
1452         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1453         REG_CFA_DEF_CFA note.
1454         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1455         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1457 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1459         PR tree-optimization/60454
1460         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1462 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1464         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1465         Do not define target_cpu_default2 to generic.
1466         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1467         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1468         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1470 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1471             Marc Glisse  <marc.glisse@inria.fr>
1473         PR tree-optimization/60502
1474         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1475         instead of build_low_bits_mask.
1477 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1479         PR middle-end/60482
1480         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1481         if there are multiple uses, but op doesn't live on E edge.
1482         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1483         clobber stmts before __builtin_unreachable.
1485 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1487         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1488         hard_frame_pointer_rtx.
1489         * cse.c (cse_insn): Remove volatile check.
1490         * cselib.c (cselib_process_insn): Likewise.
1491         * dse.c (scan_insn): Likewise.
1493 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1495         * config/arc/arc.c (conditionalize_nonjump): New function,
1496         broken out of ...
1497         (arc_ifcvt): ... this.
1498         (arc_predicate_delay_insns): Use it.
1500 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1502         * config/arc/predicates.md (extend_operand): During/after reload,
1503         allow const_int_operand.
1504         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1505         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1506         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1507         to "i".
1508         (umulsi3_highpart_i): Likewise.
1510 2014-03-11  Richard Biener  <rguenther@suse.de>
1512         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1513         Add asserts to guard possible wrong-code bugs.
1515 2014-03-11  Richard Biener  <rguenther@suse.de>
1517         PR tree-optimization/60429
1518         PR tree-optimization/60485
1519         * tree-ssa-structalias.c (set_union_with_increment): Properly
1520         take into account all fields that overlap the shifted vars.
1521         (do_sd_constraint): Likewise.
1522         (do_ds_constraint): Likewise.
1523         (get_constraint_for_ptr_offset): Likewise.
1525 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1527         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1528         (nios2_compute_frame_layout):
1529         Add calculation of cfun->machine->fp_save_offset.
1530         (nios2_expand_prologue): Correct setting of frame pointer register
1531         in prologue.
1532         (nios2_expand_epilogue): Update recovery of stack pointer from
1533         frame pointer accordingly.
1534         (nios2_initial_elimination_offset): Update calculation of offset
1535         for eliminating to HARD_FRAME_POINTER_REGNUM.
1537 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1539         PR ipa/60457
1540         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1541         cgraph_get_create_node on VAR_DECLs.
1543 2014-03-10  Richard Biener  <rguenther@suse.de>
1545         PR middle-end/60474
1546         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1548 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1550         * config/vms/vms.opt (vms_float_format): New variable.
1552 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1554         * doc/invoke.texi (-fcilkplus): Update implementation status.
1556 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1557             Richard Biener  <rguenther@suse.de>
1559         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1560         consistently accross all TUs.
1561         (run_gcc): Enable -fshort-double automatically at link at link-time
1562         and disallow override.
1564 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1566         PR target/58271
1567         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1568         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1569         if they can't be used.
1571 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1573         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1574         for Solaris 11/x86 ld.
1575         * configure: Regenerate.
1577 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1579         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1580         (LIB_TLS_SPEC): Save as ld_tls_libs.
1581         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1582         (HAVE_AS_IX86_TLSLDM): New test.
1583         * configure, config.in: Regenerate.
1584         * config/i386/i386.c (legitimize_tls_address): Fall back to
1585         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1586         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1587         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1588         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1590 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1592         * common.opt (fira-loop-pressure): Mark as optimization.
1594 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1596         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1597         an OpenMP mappable type.
1599 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1601         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1602         MULTILIB_OSDIRNAMES is not defined.
1604 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1605             Meador Inge  <meadori@codesourcery.com>
1607         PR target/58595
1608         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1609         (arm_legitimize_address): Call legitimize_tls_address for any
1610         arm_tls_referenced_p expression, handle constant addend.  Call it
1611         before testing for !TARGET_ARM.
1612         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1614 2014-03-06  Richard Biener  <rguenther@suse.de>
1616         PR middle-end/60445
1617         PR lto/60424
1618         PR lto/60427
1619         Revert
1620         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1622         * tree-streamer.c (record_common_node): Assert we don't record
1623         nodes with type double.
1624         (preload_common_node): Skip type double, complex double and double
1625         pointer since it is now frontend dependent due to fshort-double option.
1627 2014-03-06  Richard Biener  <rguenther@suse.de>
1629         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1630         or -fno-lto is specified and the linker has full plugin support.
1631         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1632         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1633         * lto-wrapper.c (merge_and_complain): Merge compile-time
1634         optimization levels.
1635         (run_gcc): And pass it through to the link options.
1637 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1639         PR debug/60381
1640         Revert:
1641         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1642         PR debug/59992
1643         * cselib.c (remove_useless_values): Skip to avoid quadratic
1644         behavior if the condition moved from...
1645         (cselib_process_insn): ... here holds.
1647 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1649         PR plugins/59335
1650         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1651         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1653         PR plugins/59335
1654         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1655         (TM_H): Add x86-tune.def.
1657 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1659         * config/aarch64/aarch64.c (generic_tunings):
1660         Use cortexa57_extra_costs.
1662 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1664         PR lto/60404
1665         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1666         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1667         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1668         cost for in_lto_p.
1670 2014-03-04  Heiher  <r@hev.cc>
1672         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1673         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1675 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1677         * config/i386/predicates.md (const2356_operand): Change to ...
1678         (const2367_operand): ... this.
1679         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1680         const2367_operand.
1681         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1682         (*avx512pf_scatterpf<mode>sf): Ditto.
1683         (avx512pf_scatterpf<mode>df): Ditto.
1684         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1685         (*avx512pf_scatterpf<mode>df): Ditto.
1686         * config/i386/i386.c (ix86_expand_builtin): Update
1687         incorrect hint operand error message.
1689 2014-03-04  Richard Biener  <rguenther@suse.de>
1691         * lto-section-in.c (lto_get_section_data): Fix const cast.
1693 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1695         * tree-streamer.c (record_common_node): Assert we don't record
1696         nodes with type double.
1697         (preload_common_node): Skip type double, complex double and double
1698         pointer since it is now frontend dependent due to fshort-double option.
1700 2014-03-04  Richard Biener  <rguenther@suse.de>
1702         PR lto/60405
1703         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1704         (lto_input_toplevel_asms): Likewise.
1705         * lto-section-in.c (lto_get_section_data): Instead do it here
1706         for every section.
1708 2014-03-04  Richard Biener  <rguenther@suse.de>
1710         PR tree-optimization/60382
1711         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1712         dead PHIs a reduction.
1714 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1716         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1717         hint value.
1718         (_mm_prefetch): Move out of GCC target("sse") pragma.
1719         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1720         GCC target("prfchw") pragma.
1721         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1722         for locality <= 2.
1723         * config/i386/i386.c (ix86_option_override_internal): Enable
1724         -mprfchw with -mprefetchwt1.
1726 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1728         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1729         Mark as varying.
1731 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1733         * opts.h (CL_PCH_IGNORE): Define.
1734         * targhooks.c (option_affects_pch_p):
1735         Return false for options that have CL_PCH_IGNORE set.
1736         * opt-functions.awk: Process PchIgnore.
1737         * doc/options.texi: Document PchIgnore.
1739         * config/arc/arc.opt (misize): Add PchIgnore property.
1741 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1743         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1744         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1745         constraint on constants to permit them being loaded into
1746         GENERAL_REGS or BASE_REGS.
1748 2014-03-03  Nick Clifton  <nickc@redhat.com>
1750         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1751         anti-cacnonical alternatives.
1752         (negandhi3_real): New pattern.
1753         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1755 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1757         * config/avr/avr-mcus.def: Remove atxmega16x1.
1758         * config/avr/avr-tables.opt: Regenerate.
1759         * config/avr/t-multilib: Regenerate.
1760         * doc/avr-mmcu.texi: Regenerate.
1762 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1763             Mircea Namolaru  <mircea.namolaru@inria.fr>
1765         PR tree-optimization/58028
1766         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1767         scalar dimensions.
1769 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1771         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1772         not handled by recognizers.
1774 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1776         PR middle-end/60175
1777         * function.c (expand_function_end): Don't emit
1778         clobber_return_register sequence if clobber_after is a BARRIER.
1779         * cfgexpand.c (construct_exit_block): Append instructions before
1780         return_label to prev_bb.
1782 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1784         * config/rs6000/constraints.md: Document reserved use of "wc".
1786 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1788         PR ipa/60150
1789         * ipa.c (function_and_variable_visibility): When dissolving comdat
1790         group, also set all symbols to local.
1792 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1794         PR ipa/60306
1796         Revert:
1797         2013-12-14   Jan Hubicka  <jh@suse.cz>
1798         PR middle-end/58477
1799         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1801 2014-03-02  Jon Beniston  <jon@beniston.com>
1803         PR bootstrap/48230
1804         PR bootstrap/50927
1805         PR bootstrap/52466
1806         PR target/46898
1807         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1808           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1809         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1810         (simple_return, *simple_return): New patterns
1811         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1812         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1814 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1816         * dwarf2out.c (gen_subprogram_die): Tidy.
1818 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1820         PR target/60071
1821         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1822         (*mov_t_msb_neg_negc): ... this new insn.
1824 2014-02-28  Jason Merrill  <jason@redhat.com>
1826         PR c++/58678
1827         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1828         function.
1830 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1832         PR c++/60314
1833         * dwarf2out.c (decltype_auto_die): New static.
1834         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1835         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1836         (is_cxx_auto): Likewise.
1838 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1840         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1841         we are not using general regs only.
1843 2014-02-28  Richard Biener  <rguenther@suse.de>
1845         PR target/60280
1846         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1847         previous fix and only allow to remove trivial pre-headers
1848         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1849         (remove_forwarder_block): Properly update the latch of a loop.
1851 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1853         PR debug/59992
1854         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1855         (cselib_preserved_hash_table): New.
1856         (preserve_constants_and_equivs): Move preserved vals to it.
1857         (cselib_find_slot): Look it up first.
1858         (cselib_init): Initialize it.
1859         (cselib_finish): Release it.
1860         (dump_cselib_table): Dump it.
1862 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1864         PR debug/59992
1865         * cselib.c (remove_useless_values): Skip to avoid quadratic
1866         behavior if the condition moved from...
1867         (cselib_process_insn): ... here holds.
1869 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1871         PR debug/57232
1872         * var-tracking.c (vt_initialize): Apply the same condition to
1873         preserve the CFA base value.
1875 2014-02-28  Joey Ye  <joey.ye@arm.com>
1877         PR target/PR60169
1878         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1879         if reload in progress or completed.
1881 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1883         PR middle-end/60147
1884         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1885         NAMELIST_DECL.
1887 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1889         * doc/tm.texi.in (Condition Code Status): Update documention for
1890         relative locations of cc0-setter and cc0-user.
1892 2014-02-27  Jeff Law  <law@redhat.com>
1894         PR rtl-optimization/52714
1895         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1896         into two independent simple sets, if I3 is a jump, ensure the
1897         pattern we place into I3 is a (set (pc) ...).
1899 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1900             Jeff Law  <law@redhat.com>
1902         PR rtl-optimization/49847
1903         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1904         are in different blocks.
1905         * doc/tm.texi (Condition Code Status): Update documention for
1906         relative locations of cc0-setter and cc0-user.
1908 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1910         PR target/59222
1911         * lra.c (lra_emit_add): Check SUBREG too.
1913 2014-02-27  Andreas Schwab  <schwab@suse.de>
1915         * config/m68k/m68k.c (m68k_option_override): Disable
1916         -flive-range-shrinkage for classic m68k.
1917         (m68k_override_options_after_change): Likewise.
1919 2014-02-27  Marek Polacek  <polacek@redhat.com>
1921         PR middle-end/59223
1922         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1923         -Wmaybe-uninitialized.
1925 2014-02-27  Alan Modra  <amodra@gmail.com>
1927         PR target/57936
1928         * reload1.c (emit_input_reload_insns): When reload_override_in,
1929         set old to rl->in_reg when rl->in_reg is a subreg.
1931 2014-02-26  Richard Biener  <rguenther@suse.de>
1933         PR bootstrap/60343
1934         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1936 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1938         * common/config/i386/predicates.md (const1256_operand): Remove.
1939         (const2356_operand): New.
1940         (const_1_to_2_operand): Remove.
1941         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1942         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1943         (*avx512pf_gatherpf<mode>sf): Ditto.
1944         (avx512pf_gatherpf<mode>df): Ditto.
1945         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1946         (*avx512pf_gatherpf<mode>df): Ditto.
1947         (avx512pf_scatterpf<mode>sf): Ditto.
1948         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1949         (*avx512pf_scatterpf<mode>sf): Ditto.
1950         (avx512pf_scatterpf<mode>df): Ditto.
1951         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1952         (*avx512pf_scatterpf<mode>df): Ditto.
1953         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1955 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1957         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1958         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1959         (_mm512_mask_testn_epi64_mask): Move to ...
1960         * config/i386/avx512cdintrin.h: Here.
1961         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1962         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1963         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1964         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1965         TARGET_AVX512F from TARGET_AVX512CD.
1967 2014-02-26  Richard Biener  <rguenther@suse.de>
1969         PR ipa/60327
1970         * ipa.c (walk_polymorphic_call_targets): Properly guard
1971         call to inline_update_overall_summary.
1973 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1975         PR target/60280
1976         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1977         and latches only if requested.  Fix latch if it is removed.
1978         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1979         LOOPS_HAVE_PREHEADERS.
1981 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1983         * builtins.c (expand_builtin_thread_pointer): Create a new target
1984         when the target is NULL.
1986 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1988         PR rtl-optimization/60317
1989         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1990         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1991         * lra-assigns.c: Include params.h.
1992         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1993         other reload pseudos considerations.
1995 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1997         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1998         to use canonical form for nor<mode>3.
2000 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2002         PR target/55426
2003         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
2004         conversions.
2006 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
2008         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2009         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2010         (ix86_handle_option): Handle OPT_mprefetchwt1.
2011         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2012         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2013         PREFETCHWT1 CPUID.
2014         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2015         OPTION_MASK_ISA_PREFETCHWT1.
2016         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2017         (PTA_PREFETCHWT1): New.
2018         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2019         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2020         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2021         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2022         (*prefetch_avx512pf_<mode>_: Change into ...
2023         (*prefetch_prefetchwt1_<mode>: This.
2024         * config/i386/i386.opt (mprefetchwt1): New.
2025         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2026         (_mm_prefetch): Handle intent to write.
2027         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2029 2014-02-25  Richard Biener  <rguenther@suse.de>
2031         PR middle-end/60291
2032         * emit-rtl.c (mem_attrs_htab): Remove.
2033         (mem_attrs_htab_hash): Likewise.
2034         (mem_attrs_htab_eq): Likewise.
2035         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2036         (init_emit_once): Do not allocate mem_attrs_htab.
2038 2014-02-25  Richard Biener  <rguenther@suse.de>
2040         PR lto/60319
2041         * lto-opts.c (lto_write_options): Output non-explicit conservative
2042         -fwrapv, -fno-trapv and -fno-strict-overflow.
2043         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2044         (run_gcc): And pass them through.
2046 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2048         * sel-sched.c (calculate_new_fences): New parameter ptime.
2049         Calculate it as a maximum over all fence cycles.
2050         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2051         Print the final schedule timing when sched_verbose.
2053 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2055         PR rtl-optimization/60292
2056         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2057         bit fot the fence instruction.
2059 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2061         * calls.h: Fix typo in comment.
2063 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2065         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2066         adjusting offsetable addresses.
2068 2014-02-24  Guozhi Wei  <carrot@google.com>
2070         * sparseset.h (sparseset_pop): Fix the wrong index.
2072 2014-02-24  Walter Lee  <walt@tilera.com>
2074         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2075         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2076         triplet.
2077         * common/config/tilegx/tilegx-common.c
2078         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2079         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2080         (LINK_SPEC): Ditto.
2081         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2082         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2083         (tilegx_gimplify_va_arg_expr): Handle big endian.
2084         (tilegx_expand_unaligned_load): Ditto.
2085         (tilegx_expand_unaligned_store): Ditto.
2086         (TARGET_RETURN_IN_MSB): New.
2087         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2088         (TARGET_ENDIAN_DEFAULT): New.
2089         (TARGET_BIG_ENDIAN): Handle big endian.
2090         (BYTES_BIG_ENDIAN): Ditto.
2091         (WORDS_BIG_ENDIAN): Ditto.
2092         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2093         (ENDIAN_SPEC): New.
2094         (EXTRA_SPECS): New.
2095         * config/tilegx/tilegx.md (extv): Handle big endian.
2096         (extzv): Ditto.
2097         (insn_st<n>): Ditto.
2098         (insn_st<n>_add<bitsuffix>): Ditto.
2099         (insn_stnt<n>): Ditto.
2100         (insn_stnt<n>_add<bitsuffix>):Ditto.
2101         (vec_interleave_highv8qi): Handle big endian.
2102         (vec_interleave_highv8qi_be): New.
2103         (vec_interleave_highv8qi_le): New.
2104         (insn_v1int_h): Handle big endian.
2105         (vec_interleave_lowv8qi): Handle big endian.
2106         (vec_interleave_lowv8qi_be): New.
2107         (vec_interleave_lowv8qi_le): New.
2108         (insn_v1int_l): Handle big endian.
2109         (vec_interleave_highv4hi): Handle big endian.
2110         (vec_interleave_highv4hi_be): New.
2111         (vec_interleave_highv4hi_le): New.
2112         (insn_v2int_h): Handle big endian.
2113         (vec_interleave_lowv4hi): Handle big endian.
2114         (vec_interleave_lowv4hi_be): New.
2115         (vec_interleave_lowv4hi_le): New.
2116         (insn_v2int_l): Handle big endian.
2117         (vec_interleave_highv2si): Handle big endian.
2118         (vec_interleave_highv2si_be): New.
2119         (vec_interleave_highv2si_le): New.
2120         (insn_v4int_h): Handle big endian.
2121         (vec_interleave_lowv2si): Handle big endian.
2122         (vec_interleave_lowv2si_be): New.
2123         (vec_interleave_lowv2si_le): New.
2124         (insn_v4int_l): Handle big endian.
2125         * config/tilegx/tilegx.opt (mbig-endian): New option.
2126         (mlittle-endian): New option.
2127         * doc/install.texi: Document tilegxbe-linux.
2128         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2130 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2132         PR ipa/60266
2133         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2134         there are no parameter descriptors.
2136 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2138         PR rtl-optimization/60268
2139         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2140         initialization to ...
2141         (sched_rgn_init): ... here.
2142         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2144 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2146         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2147         names.
2149 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2151         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2152         definition.
2154 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2156         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2157         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2159 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2161         * config/microblaze/predicates.md: Add cmp_op predicate.
2162         * config/microblaze/microblaze.md: Add branch_compare instruction
2163         which uses cmp_op predicate and emits cmp insn before branch.
2164         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2165         to microblaze_expand_conditional_branch and consolidate logic.
2166         (microblaze_expand_conditional_branch): emit branch_compare
2167         insn instead of handling cmp op separate from branch insn.
2169 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2171         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2172         to permit subregs.
2174 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2176         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2177         define_insn with define_expand and new define_insn
2178         *altivec_lve<VI_char>x_internal.
2179         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2180         and new define_insn *altivec_stve<VI_char>x_internal.
2181         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2182         prototype.
2183         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2184         lve*x built-ins.
2185         (altivec_expand_stvex_be): New function.
2187 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2189         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2190         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2191         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2192         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2194 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2196         PR target/60298
2197         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2198         instead of emit_move_insn.
2200 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2202         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2203         vspltw with vsldoi.
2204         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2205         gen_altivec_vsumsws.
2207 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2209         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2210         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2211         (altivec_lvxl_<mode>): New define_expand incorporating
2212         -maltivec=be semantics where needed.
2213         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2214         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2215         semantics where needed.
2216         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2217         (altivec_stvx_<mode>): New define_expand incorporating
2218         -maltivec=be semantics where needed.
2219         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2220         VM2 iterator instead of V4SI.
2221         (altivec_stvxl_<mode>): New define_expand incorporating
2222         -maltivec=be semantics where needed.
2223         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2224         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2225         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2226         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2227         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2228         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2229         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2230         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2231         ALTIVEC_BUILTIN_STVXL.
2232         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2233         (altivec_expand_stvx_be): Likewise.
2234         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2235         (altivec_expand_lvx_be): Likewise.
2236         (altivec_expand_stvx_be): Likewise.
2237         (altivec_expand_builtin): Add cases for
2238         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2239         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2240         (altivec_init_builtins): Add definitions for
2241         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2242         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2244 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2246         * doc/invoke.texi (mvirt, mno-virt): Document.
2247         * config/mips/mips.opt (mvirt): New option.
2248         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2250 2014-02-21  Richard Biener  <rguenther@suse.de>
2252         PR tree-optimization/60276
2253         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2254         (STMT_VINFO_MIN_NEG_DIST): New macro.
2255         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2256         STMT_VINFO_MIN_NEG_DIST.
2257         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2258         made for negative dependence distances still hold.
2260 2014-02-21  Richard Biener  <rguenther@suse.de>
2262         PR middle-end/60291
2263         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2264         DECL_INITIAL for globals not in the current function context.
2266 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2268         PR tree-optimization/56490
2269         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2270         * tree-ssa-uninit.c: Include params.h.
2271         (compute_control_dep_chain): Add num_calls argument, return false
2272         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2273         num_calls to recursive call.
2274         (find_predicates): Change dep_chain into normal array,
2275         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2276         variable and adjust compute_control_dep_chain caller.
2277         (find_def_preds): Likewise.
2279 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2281         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2282         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2284 2014-02-21  Nick Clifton  <nickc@redhat.com>
2286         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2287         (pushhi1): Likewise.
2288         (popqi1): Add mode to pre_dec.
2289         (pophi1): Likewise.
2291 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2293         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2294         mode for mask of V8SFmode permutation.
2296 2014-02-20  Richard Henderson  <rth@redhat.com>
2298         PR c++/60272
2299         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2300         a new pseudo for OLDVAL.
2302 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2304         PR target/57896
2305         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2306         gen_reg_rtx if d->testing_p.
2307         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2308         if d->testing_p and we will certainly return true.
2309         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2310         if d->testing_p.
2312 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2314         * emit-rtl.c (gen_reg_rtx): Assert that
2315         crtl->emit.regno_pointer_align_length is non-zero.
2317 2014-02-20  Richard Henderson  <rth@redhat.com>
2319         PR c++/60272
2320         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2321         on failure the store back into EXPECT.
2323 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2324             Sandra Loosemore  <sandra@codesourcery.com>
2326         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2327         * config/nios2/nios2.c (nios2_function_profiler): Add
2328         -fPIC (flag_pic == 2) support.
2329         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2330         (nios2_large_offset_p): New function.
2331         (nios2_unspec_reloc_p): Move up position, update to use
2332         nios2_large_offset_p.
2333         (nios2_unspec_address): Remove function.
2334         (nios2_unspec_offset): New function.
2335         (nios2_large_got_address): New function.
2336         (nios2_got_address): Add large offset support.
2337         (nios2_legitimize_tls_address): Update usage of removed and new
2338         functions.
2339         (nios2_symbol_binds_local_p): New function.
2340         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2341         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2342         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2343         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2344         processing for (const (unspec ...)).
2345         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2347 2014-02-20  Richard Biener  <rguenther@suse.de>
2349         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2350         doing the substitution.
2351         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2353 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2355         PR ipa/55260
2356         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2357         info when checking whether lattices are bottom.
2359 2014-02-20  Richard Biener  <rguenther@suse.de>
2361         PR middle-end/60221
2362         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2363         regions at -O0.
2365 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2367         PR ipa/58555
2368         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2369         parameter specifying the scaling.
2370         (inline_call): Update.
2371         (want_inline_recursively): Guard division by zero.
2372         (recursive_inlining): Update.
2373         * ipa-inline.h (clone_inlined_nodes): Update.
2375 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2377         PR target/60204
2378         * config/i386/i386.c (classify_argument): Pass structures of size
2379         64 bytes or less in register.
2381 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2382             Kirill Yukhin  <kirill.yukhin@intel.com>
2384         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2385         (_mm_rcp28_round_ss): Ditto.
2386         (_mm_rsqrt28_round_sd): Ditto.
2387         (_mm_rsqrt28_round_ss): Ditto.
2388         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2389         (_mm_rcp14_round_ss): Ditto.
2390         (_mm_rsqrt14_round_sd): Ditto.
2391         (_mm_rsqrt14_round_ss): Ditto.
2392         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2393         the first input operand, get rid of match_dup.
2394         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2395         attribute to sse.
2396         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2397         Ditto.
2398         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2399         operand as the first input operand, set type attribute.
2400         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2401         Set type attribute.
2402         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2403         operand as the first input operand, set type attribute.
2405 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2407         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2408         bit of zero.
2410 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2412         PR target/60207
2413         * config/i386/i386.c (construct_container): Remove TFmode check
2414         for X86_64_INTEGER_CLASS.
2416 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2418         PR target/59794
2419         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2420         only when -Wpsabi is enabled.
2422 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2424          PR target/59799
2425         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2426         passing arrays in registers are the same as for structs, so remove the
2427         special case for them.
2429 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2431         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2432         destination type, extract only the valid bits if the source type is not
2433         integral and has a different mode.
2435 2014-02-19  Richard Biener  <rguenther@suse.de>
2437         PR ipa/60243
2438         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2439         for all calls.
2441 2014-02-19  Richard Biener  <rguenther@suse.de>
2443         PR ipa/60243
2444         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2445         (ipa_modify_call_arguments): Emit an argument load explicitely and
2446         preserve virtual SSA form there and for the replacement call.
2447         Do not update SSA form nor free dominance info.
2449 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2451         * ipa.c (function_and_variable_visibility): Also clear WEAK
2452         flag when disolving COMDAT_GROUP.
2454 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2456         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2457         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2458         not devirtualizing.
2459         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2460         do more sanity checks.
2461         (detect_type_change): Return true when giving up early.
2462         (compute_complex_assign_jump_func): Fix type parameter of
2463         ipa_set_ancestor_jf.
2464         (compute_complex_ancestor_jump_func): Likewise.
2465         (update_jump_functions_after_inlining): Fix updating of
2466         ancestor function.
2467         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2469 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2471         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2472         inline clones when edge disappears.
2474 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2476         PR target/60203
2477         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2478         Split 64-bit moves into 2 patterns.  Do not allow the use of
2479         direct move for TDmode in little endian, since the decimal value
2480         has little endian bytes within a word, but the 64-bit pieces are
2481         ordered in a big endian fashion, and normal subreg's of TDmode are
2482         not allowed.
2483         (mov<mode>_64bit_dm): Likewise.
2484         (movtd_64bit_nodm): Likewise.
2486 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2488         PR tree-optimization/60174
2489         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2490         statement of an SSA_NAME that occurs in an abnormal PHI node.
2492 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2494         PR sanitizer/60142
2495         * final.c (SEEN_BB): Remove.
2496         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2497         (final_scan_insn): Don't force_source_line on second
2498         NOTE_INSN_BASIC_BLOCK.
2500 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2502         PR target/60205
2503         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2504         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2505         (type_natural_mode): Warn ABI change when %zmm register is not
2506         available for AVX512F vector value passing.
2508 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2510         PR target/60193
2511         * config/i386/i386.c (ix86_expand_prologue): Use value in
2512         rax register as displacement when restoring %r10 or %rax.
2513         Fix wrong offset when restoring both registers.
2515 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2517         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2518         assertion with conditional return.
2520 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2521             Uros Bizjak  <ubizjak@gmail.com>
2523         PR driver/60233
2524         * config/i386/driver-i386.c (host_detect_local_cpu): If
2525         YMM state is not saved by the OS, also clear has_f16c.  Move
2526         CPUID 0x80000001 handling before YMM state saving checking.
2528 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2530         PR rtl-optimization/58960
2531         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2532         factored out from ...
2533         (sched_init): ... here.
2534         (free_global_sched_pressure_data): New, factored out from ...
2535         (sched_finish): ... here.
2536         * sched-int.h (free_global_sched_pressure_data): Declare.
2537         * sched-rgn.c (nr_regions_initial): New static global.
2538         (haifa_find_rgns): Initialize it.
2539         (schedule_region): Disable sched-pressure for the newly
2540         generated regions.
2542 2014-02-17  Richard Biener  <rguenther@suse.de>
2544         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2545         release SSA defs of pattern stmts.
2547 2014-02-17  Richard Biener  <rguenther@suse.de>
2549         * tree-inline.c (expand_call_inline): Release the virtual
2550         operand defined by the call we are about to inline.
2552 2014-02-17  Richard Biener  <rguenther@suse.de>
2554         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2556 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2557             Ilya Tocar  <ilya.tocar@intel.com>
2559         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2560         arguments order in builtin.
2561         (_mm512_permutexvar_epi64): Ditto.
2562         (_mm512_mask_permutexvar_epi64): Ditto
2563         (_mm512_maskz_permutexvar_epi32): Ditto
2564         (_mm512_permutexvar_epi32): Ditto
2565         (_mm512_mask_permutexvar_epi32): Ditto
2567 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2569         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2570         (p8_vmrgow): Likewise.
2572 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2574         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2575         endian targets.
2577 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2579         PR target/60203
2580         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2581         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2582         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2583         using direct move instructions on ISA 2.07.  Also adjust
2584         instruction length for 64-bit.
2585         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2586         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2588 2014-02-15  Alan Modra  <amodra@gmail.com>
2590         PR target/58675
2591         PR target/57935
2592         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2593         find_replacement on parts of insn rtl that might be reloaded.
2595 2014-02-15  Richard Biener  <rguenther@suse.de>
2597         PR tree-optimization/60183
2598         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2599         (tree_ssa_phiprop): Calculate and free post-dominators.
2601 2014-02-14  Jeff Law  <law@redhat.com>
2603         PR rtl-optimization/60131
2604         * ree.c (get_extended_src_reg): New function.
2605         (combine_reaching_defs): Use it rather than assuming location of REG.
2606         (find_and_remove_re): Verify first operand of extension is
2607         a REG before adding the insns to the copy list.
2609 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2611         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2612         * configure: Regenerated.
2613         * config.in: Regenerated.
2614         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2615         instead of ASM_SHORT.
2617 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2618             Richard Earnshaw  <rearnsha@arm.com>
2620         PR rtl-optimization/59535
2621         * lra-constraints.c (process_alt_operands): Encourage alternative
2622         when unassigned pseudo class is superset of the alternative class.
2623         (inherit_reload_reg): Don't inherit when optimizing for code size.
2624         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2625         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2626         modes not less than 4 for Thumb1.
2628 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2630         PR pch/60010
2631         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2633 2014-02-14  Richard Biener  <rguenther@suse.de>
2635         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2636         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2637         Do not strip INDIRECT_REFs.
2639 2014-02-14  Richard Biener  <rguenther@suse.de>
2641         PR lto/60179
2642         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2643         DECL_FUNCTION_SPECIFIC_TARGET.
2644         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2645         * tree-streamer-out.c (pack_ts_target_option): Remove.
2646         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2647         (write_ts_function_decl_tree_pointers): Do not stream
2648         DECL_FUNCTION_SPECIFIC_TARGET.
2649         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2650         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2651         (lto_input_ts_function_decl_tree_pointers): Do not stream
2652         DECL_FUNCTION_SPECIFIC_TARGET.
2654 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2656         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2657         (get_initial_def_for_induction, vectorizable_induction): Ignore
2658         debug stmts when looking for exit_phi.
2659         (vectorizable_live_operation): Fix up condition.
2661 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2663         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2664         nreverse() because it changes the content of original tree list.
2666 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2668         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2669         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2671 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2673         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2674         GNU coding standards.
2676 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2678         PR debug/60152
2679         * dwarf2out.c (gen_subprogram_die): Don't call
2680         add_calling_convention_attribute if subr_die is old_die.
2682 2014-02-13  Sharad Singhai  <singhai@google.com>
2684         * doc/optinfo.texi: Fix order of nodes.
2686 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2688         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2689         operands[2], not operands[3].
2691 2014-02-13  Richard Biener  <rguenther@suse.de>
2693         PR bootstrap/59878
2694         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2695         mention the possibility of an in-tree build.
2696         (CLooG): Update recommended version to 0.18.1, mention the
2697         possibility of an in-tree build and clarify that the ISL
2698         bundled with CLooG does not work.
2700 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2702         PR target/43546
2703         * expr.c (compress_float_constant): If x is a hard register,
2704         extend into a pseudo and then move to x.
2706 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2708         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2709         caused by bad second argument to warning_at() with -mhotpatch and
2710         nested functions (e.g. with gfortran).
2712 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2714         * opts.c (option_name): Remove "enabled by default" rider.
2716 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2718         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2720 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2721             Uros Bizjak  <ubizjak@gmail.com>
2723         PR target/60151
2724         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2725         * configure: Regenerated.
2727 2014-02-12  Richard Biener  <rguenther@suse.de>
2729         * vec.c (vec_prefix::calculate_allocation): Move as
2730         inline variant to vec.h.
2731         (vec_prefix::calculate_allocation_1): New out-of-line version.
2732         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2733         (vec_prefix::m_has_auto_buf): Rename to ...
2734         (vec_prefix::m_using_auto_storage): ... this.
2735         (vec_prefix::calculate_allocation): Inline the easy cases
2736         and dispatch to calculate_allocation_1 which doesn't need the
2737         prefix address.
2738         (va_heap::reserve): Use gcc_checking_assert.
2739         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2740         m_using_auto_storage.
2741         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2742         member and adjust.
2743         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2744         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2745         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2747 2014-02-12  Richard Biener  <rguenther@suse.de>
2749         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2750         when we found a dependence.
2752 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2754         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2755         common code...
2756         (maybe_fold_stmt): ... into this new function.
2757         * omp-low.c (lower_omp): Update comment.
2759         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2760         last use.
2762         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2763         dereference.
2765 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2767         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2768         identifiers in comments.
2769         (cortexa53_extra_costs): Likewise.
2770         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2771         (cortexa7_extra_costs): Likewise.
2772         (cortexa12_extra_costs): Likewise.
2773         (cortexa15_extra_costs): Likewise.
2774         (v7m_extra_costs): Likewise.
2776 2014-02-12  Richard Biener  <rguenther@suse.de>
2778         PR middle-end/60092
2779         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2780         of posix_memalign being successful.
2781         (lower_stmt): Restrict lowering of posix_memalign to when
2782         -ftree-bit-ccp is enabled.
2784 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2786         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2787         arg_loc.
2788         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2790 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2792         PR rtl-optimization/60116
2793         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2794         other_insn once the combination has been validated.
2796 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2798         PR lto/59468
2799         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2800         and wrapper.
2801         * ipa-devirt.c: Include demangle.h
2802         (odr_violation_reported): New static variable.
2803         (add_type_duplicate): Update odr_violations.
2804         (maybe_record_node): Add completep parameter; update it.
2805         (record_target_from_binfo): Add COMPLETEP parameter;
2806         update it as needed.
2807         (possible_polymorphic_call_targets_1): Likewise.
2808         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2809         rename FINAL to COMPLETE.
2810         (record_targets_from_bases): Sanity check we found the binfo;
2811         fix COMPLETEP updating.
2812         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2813         parameter, fix computing of COMPLETEP.
2814         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2815         at LTO time do demangling.
2816         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2817         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2818         parameter.
2819         (gimple_get_virt_method_for_binfo): Likewise.
2820         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2821         gimple_get_virt_method_for_vtable): Update prototypes.
2823 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2825         PR target/49008
2826         * genautomata.c (add_presence_absence): Fix typo with
2827         {final_}presence_list.
2829 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2831         PR target/60137
2832         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2833         for VSX/Altivec vectors that land in GPR registers.
2835 2014-02-11  Richard Henderson  <rth@redhat.com>
2836             Jakub Jelinek  <jakub@redhat.com>
2838         PR debug/59776
2839         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2840         around drhs if type conversion to lacc->type is not useless.
2842 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2844         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2845         tuning struct.
2846         (cortex-a57.cortex-a53): Likewise.
2847         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2849 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2851         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2852         arm_restrict_it.
2854 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2856         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2857         add_options_for_arm_vfp3.
2859 2014-02-11  Jeff Law  <law@redhat.com>
2861         PR middle-end/54041
2862         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2863         object with an undesirable mode.
2865 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2867         PR libgomp/60107
2868         * config/i386/sol2-9.h: New file.
2869         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2870         *-*-solaris2.9*): Use it.
2872 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2874         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2875         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2877 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2879         * config/microblaze/microblaze.c: Extend mcpu version format
2881 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
2883         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2885 2014-02-10  Richard Henderson  <rth@redhat.com>
2887         PR target/59927
2888         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2889         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2890         ms-abi vs -mno-accumulate-outgoing-args.
2891         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2892         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2893         respect to ms-abi.
2895 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2897         PR middle-end/60080
2898         * cfgexpand.c (expand_asm_operands): Attach source location to
2899         ASM_INPUT rtx objects.
2900         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2902 2014-02-10  Nick Clifton  <nickc@redhat.com>
2904         * config/mn10300/mn10300.c (popcount): New function.
2905         (mn10300_expand_prologue): Include saved registers in stack usage
2906         count.
2908 2014-02-10  Jeff Law  <law@redhat.com>
2910         PR middle-end/52306
2911         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2912         when changing the SET_DEST of a prior insn to avoid an input reload.
2914 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2916         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2917         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2918         -mcall-openbsd, or -mcall-linux.
2919         (CC1_ENDIAN_BIG_SPEC): Remove.
2920         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2921         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2922         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2923         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2924         and %cc1_endian_default.
2925         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2927 2014-02-10  Richard Biener  <rguenther@suse.de>
2929         PR tree-optimization/60115
2930         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2931         MEM_REF handling.  Properly verify that the accesses are not
2932         out of the objects bound.
2934 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2936         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2937         coretex to cortex.
2939 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2941         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2942         proper constants and fix formatting.
2943         (possible_polymorphic_call_targets): Fix formatting.
2945 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2946             Ilya Tocar  <ilya.tocar@intel.com>
2948         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2949         (_mm512_loadu_epi32): Renamed into...
2950         (_mm512_loadu_si512): This.
2951         (_mm512_storeu_epi32): Renamed into...
2952         (_mm512_storeu_si512): This.
2953         (_mm512_maskz_ceil_ps): Removed.
2954         (_mm512_maskz_ceil_pd): Ditto.
2955         (_mm512_maskz_floor_ps): Ditto.
2956         (_mm512_maskz_floor_pd): Ditto.
2957         (_mm512_floor_round_ps): Ditto.
2958         (_mm512_floor_round_pd): Ditto.
2959         (_mm512_ceil_round_ps): Ditto.
2960         (_mm512_ceil_round_pd): Ditto.
2961         (_mm512_mask_floor_round_ps): Ditto.
2962         (_mm512_mask_floor_round_pd): Ditto.
2963         (_mm512_mask_ceil_round_ps): Ditto.
2964         (_mm512_mask_ceil_round_pd): Ditto.
2965         (_mm512_maskz_floor_round_ps): Ditto.
2966         (_mm512_maskz_floor_round_pd): Ditto.
2967         (_mm512_maskz_ceil_round_ps): Ditto.
2968         (_mm512_maskz_ceil_round_pd): Ditto.
2969         (_mm512_expand_pd): Ditto.
2970         (_mm512_expand_ps): Ditto.
2971         * config/i386/i386.c (ix86_builtins): Remove
2972         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2973         (bdesc_args): Ditto.
2974         * config/i386/predicates.md (const1256_operand): New.
2975         (const_1_to_2_operand): Ditto.
2976         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2977         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2978         (*avx512pf_gatherpf<mode>sf): Ditto.
2979         (avx512pf_gatherpf<mode>df): Ditto.
2980         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2981         (*avx512pf_gatherpf<mode>df): Ditto.
2982         (avx512pf_scatterpf<mode>sf): Ditto.
2983         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2984         (*avx512pf_scatterpf<mode>sf): Ditto.
2985         (avx512pf_scatterpf<mode>df): Ditto.
2986         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2987         (*avx512pf_scatterpf<mode>df): Ditto.
2988         (avx512f_expand<mode>): Removed.
2989         (<shift_insn><mode>3<mask_name>): Change predicate type.
2991 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2993         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2994         not at the end of datarefs vector use ordered_remove to avoid
2995         reordering datarefs vector.
2997         PR c/59984
2998         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2999         mark local addressable non-static vars as GOVD_PRIVATE
3000         instead of GOVD_LOCAL.
3001         * omp-low.c (lower_omp_for): Move gimple_bind_vars
3002         and BLOCK_VARS of gimple_bind_block to new_stmt rather
3003         than copying them.
3005         PR middle-end/60092
3006         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
3007         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
3008         assume_aligned or alloc_align attributes.
3009         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3010         arguments.  Handle also assume_aligned and alloc_align attributes.
3011         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3012         calls to functions with assume_aligned or alloc_align attributes.
3013         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3015 2014-02-08  Terry Guo  <terry.guo@arm.com>
3017         * doc/invoke.texi: Document ARM -march=armv7e-m.
3019 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3021         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3022         flag on __cilkrts_rethrow builtin.
3024         PR ipa/60026
3025         * ipa-cp.c (determine_versionability): Fail at -O0
3026         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3027         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3029         Revert:
3030         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3032         PR ipa/60026
3033         * tree-inline.c (copy_forbidden): Fail for
3034         __attribute__((optimize (0))) functions.
3036 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3038         * varpool.c: Include pointer-set.h.
3039         (varpool_remove_unreferenced_decls): Variables in other partitions
3040         will not be output; be however careful to not lose information
3041         about partitioning.
3043 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3045         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3046         lookup in the vtable constructor.
3048 2014-02-07  Jeff Law  <law@redhat.com>
3050         PR target/40977
3051         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3052         define_insn_and_split.
3054         * ipa-inline.c (inline_small_functions): Fix typos.
3056 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3058         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3059         (s390_can_use_return_insn): Declare.
3060         * config/s390/s390.h (EPILOGUE_USES): Define.
3061         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3062         instructions.
3063         (s390_chunkify_start): Handle return JUMP_LABELs.
3064         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3065         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3066         (s390_can_use_return_insn): New functions.
3067         (s390_fix_long_loop_prediction): Handle conditional returns.
3068         (TARGET_SET_UP_BY_PROLOGUE): Define.
3069         * config/s390/s390.md (ANY_RETURN): New code iterator.
3070         (*creturn, *csimple_return, return, simple_return): New patterns.
3072 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3074         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3075         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3076         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3077         REG_CFA_RESTORE list when deciding not to restore a register.
3079 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3081         * config/s390/s390.c: Include tree-pass.h and context.h.
3082         (s390_early_mach): New function, split out from...
3083         (s390_emit_prologue): ...here.
3084         (pass_data_s390_early_mach): New pass structure.
3085         (pass_s390_early_mach): New class.
3086         (s390_option_override): Create and register early_mach pass.
3087         Move to end of file.
3089 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3091         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3092         to match for the exit block.
3094 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3096         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3097         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3098         Reject misaligned operands.
3100 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3102         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3104 2014-02-07  Richard Biener  <rguenther@suse.de>
3106         PR middle-end/60092
3107         * gimple-low.c (lower_builtin_posix_memalign): New function.
3108         (lower_stmt): Call it to lower posix_memalign in a way
3109         to make alignment info accessible.
3111 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3113         PR c++/60082
3114         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3115         __builtin_setjmp_receiver.
3117 2014-02-07  Richard Biener  <rguenther@suse.de>
3119         PR middle-end/60092
3120         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3121         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3122         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3123         Handle BUILT_IN_POSIX_MEMALIGN.
3124         (find_func_clobbers): Likewise.
3125         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3126         (call_may_clobber_ref_p_1): Likewise.
3128 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3130         PR ipa/59918
3131         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3132         sanity check.
3134 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3136         PR ipa/59469
3137         * lto-cgraph.c (lto_output_node): Use
3138         symtab_get_symbol_partitioning_class.
3139         (lto_output_varpool_node): likewise.
3140         (symtab_get_symbol_partitioning_class): Move here from
3141         lto/lto-partition.c
3142         * cgraph.h (symbol_partitioning_class): Likewise.
3143         (symtab_get_symbol_partitioning_class): Declare.
3145 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3147         * ggc.h (ggc_internal_cleared_alloc): New macro.
3148         * vec.h (vec_safe_copy): Handle memory stats.
3149         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3150         * target-globals.c (save_target_globals): Likewise.
3152 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3154         PR target/60077
3155         * expr.c (emit_move_resolve_push): Export; be bit more selective
3156         on when to clear alias set.
3157         * expr.h (emit_move_resolve_push): Declare.
3158         * function.h (struct function): Add tail_call_marked.
3159         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3160         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3161         * config/i386/i386.md (TImode move expander): De not call
3162         ix86_expand_push.
3163         (FP push expanders): Preserve memory attributes.
3164         * config/i386/sse.md (push<mode>1): Remove.
3165         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3166         (ix86_expand_push): Remove.
3167         * config/i386/mmx.md (push<mode>1): Remove.
3169 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3171         PR rtl-optimization/60030
3172         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3173         lopart with paradoxical subreg before shifting it up by hprec.
3175 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3177         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3178         Remove extra newline at end of file.
3179         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3180         (arm_issue_rate): Handle cortexa57.
3181         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3182         (cortex-a57.cortex-a53): Likewise.
3184 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3186         PR target/59575
3187         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3188         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3189         bitmask.
3190         (arm_expand_prologue): Adjust all callers.
3191         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3192         info, registers also at the lowest numbered registers side.  Use
3193         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3194         XEXP.
3196         PR debug/59992
3197         * var-tracking.c (adjust_mems): Before adding a SET to
3198         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3200 2014-02-06  Alan Modra  <amodra@gmail.com>
3202         PR target/60032
3203         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3204         change SDmode to DDmode when lra_in_progress.
3206 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3208         PR middle-end/59150
3209         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3210         free_data_ref on the dr first, and before goto again also set dr
3211         to the next dr.  For simd_lane_access, free old datarefs[i] before
3212         overwriting it.  For get_vectype_for_scalar_type failure, don't
3213         free_data_ref if simd_lane_access.
3215         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3217         PR target/60062
3218         * tree.h (opts_for_fn): New inline function.
3219         (opt_for_fn): Define.
3220         * config/i386/i386.c (ix86_function_regparm): Use
3221         opt_for_fn (decl, optimize) instead of optimize.
3223 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3225         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3226         for SYMBOL_REF in large memory model.
3228 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3230         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3231         and crypto support.
3232         (cortex-a57): Likewise.
3233         (cortex-a57.cortex-a53): Likewise.
3235 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3236             Kugan Vivekanandarajah  <kuganv@linaro.org>
3238         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3239         unaligned_access.
3240         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3242 2014-02-06  Richard Biener  <rguenther@suse.de>
3244         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3245         set_loop_copy and initialize_original_copy_tables.
3247 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3249         * config/aarch64/aarch64-simd.md
3250         (aarch64_ashr_simddi): Change QI to SI.
3252 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3253             Jakub Jelinek  <jakub@redhat.com>
3255         PR middle-end/60013
3256         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3257         of the dataflow.
3259 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3261         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3262         CODE_FOR_altivec_vpku[hw]um to
3263         CODE_FOR_altivec_vpku[hw]um_direct.
3264         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3265         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3266         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3267         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3269 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3271         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3272         generation for -maltivec=be.
3273         (altivec_vsumsws): Simplify redundant test.
3275 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3277         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3278         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3279         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3280         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3281         gen_altivec_vpkuwum.
3282         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3283         BYTES_BIG_ENDIAN.
3284         (altivec_vpks<VI_char>ss): Likewise.
3285         (altivec_vpks<VI_char>us): Likewise.
3286         (altivec_vpku<VI_char>us): Likewise.
3287         (altivec_vpku<VI_char>um): Likewise.
3288         (altivec_vpku<VI_char>um_direct): New (copy of
3289         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3290         internal use).
3291         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3292         target is little endian and -maltivec=be is not specified.
3293         (*altivec_vupkhs<VU_char>_direct): New (copy of
3294         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3295         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3296         target is little endian and -maltivec=be is not specified.
3297         (*altivec_vupkls<VU_char>_direct): New (copy of
3298         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3299         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3300         little endian and -maltivec=be is not specified.
3301         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3302         little endian and -maltivec=be is not specified.
3304 2014-02-05  Richard Henderson  <rth@redhat.com>
3306         PR debug/52727
3307         * combine-stack-adj.c: Revert r206943.
3308         * sched-int.h (struct deps_desc): Add last_args_size.
3309         * sched-deps.c (init_deps): Initialize it.
3310         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3311         contain REG_ARGS_SIZE notes.
3313 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3315         * lto-cgraph.c (asm_nodes_output): Make global.
3316         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3317         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3318         (driver_handle_option): Handle OPT_fwpa.
3320 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3322         PR ipa/59947
3323         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3324         a comment typo and formatting issue.  If odr_hash hasn't been
3325         created, return vNULL and set *completep to false.
3327         PR middle-end/57499
3328         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3329         bb with no successors.
3331 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3333         PR target/59718
3334         * doc/invoke.texi (-march): Clarify documentation for ARM.
3335         (-mtune): Likewise.
3336         (-mcpu): Likewise.
3338 2014-02-05  Richard Biener  <rguenther@suse.de>
3340         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3341         when not vectorizing because of too many alias checks.
3342         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3343         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3345 2014-02-05  Nick Clifton  <nickc@redhat.com>
3347         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3348         accept extended registers in any mode when compiling for the MN10300.
3350 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3352         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3353         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3354         sanitization attributes.
3355         (can_inline_edge_p): Likewise.
3356         (sanitize_attrs_match_for_inline_p): New function.
3358 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3360         * ipa-prop.c (detect_type_change): Shor circuit testing of
3361         type changes on THIS pointer.
3363 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3365         PR target/59777
3366         * config/pa/pa.c (legitimize_tls_address): Return original address
3367         if not passed a SYMBOL_REF rtx.
3368         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3369         addresses.
3370         (pa_emit_move_sequence): Simplify TLS source operands.
3371         (pa_legitimate_constant_p): Reject all TLS constants.
3372         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3373         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3375 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3377         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3378         groups when we know they are controlled by LTO.
3379         * varasm.c (default_binds_local_p_1): If object is in other partition,
3380         it will be resolved locally.
3382 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3384         * config/host-linux.c (linux_gt_pch_use_address): Don't
3385         use SSIZE_MAX because it is not always defined.
3387 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3389         PR bootstrap/59913
3390         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3391         threshold for pseudo splitting.
3392         (update_ebb_live_info): Process call argument hard registers and
3393         hard registers from insn definition too.
3394         (max_small_class_regs_num): New constant.
3395         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3396         reloads_num only for small register classes.  Don't split for
3397         outputs of jumps.
3399 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3401         PR ipa/60058
3402         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3403         is non-null.
3405 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3407         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3408         visibility is safe.
3410 2014-02-04  Marek Polacek  <polacek@redhat.com>
3412         * gdbinit.in (pel): Define.
3414 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3416         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3417         behavior.
3419 2014-02-04  Richard Biener  <rguenther@suse.de>
3421         PR lto/59723
3422         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3423         in function context local.
3424         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3425         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3426         similar to LTO_imported_decl_ref.
3428 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3430         PR tree-optimization/60002
3431         * cgraphclones.c (build_function_decl_skip_args): Clear
3432         DECL_LANG_SPECIFIC.
3434         PR tree-optimization/60023
3435         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3436         false to gsi_replace.
3437         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3438         has been in some EH region and vec_stmt could throw, add
3439         vec_stmt into the same EH region.
3440         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3441         has no lhs, ignore it.
3442         * internal-fn.c (expand_MASK_LOAD): Likewise.
3444         PR ipa/60026
3445         * tree-inline.c (copy_forbidden): Fail for
3446         __attribute__((optimize (0))) functions.
3448         PR other/58712
3449         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3450         is set, copy one less argument.
3451         (expand_simd_clones): Don't subtract clone_info->inbranch
3452         from simd_clone_struct_alloc argument.
3454         PR rtl-optimization/57915
3455         * recog.c (simplify_while_replacing): If all unary/binary/relational
3456         operation arguments are constant, attempt to simplify those.
3458         PR middle-end/59261
3459         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3460         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3462 2014-02-04  Richard Biener  <rguenther@suse.de>
3464         PR tree-optimization/60012
3465         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3466         TBAA disambiguation to all DDRs.
3468 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3470         PR target/59788
3471         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3472         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3474 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3476         PR ipa/59882
3477         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3479 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3481         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3482         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3484 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3486         PR ipa/59831
3487         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3488         to figure out targets of polymorphic calls with known decl.
3489         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3490         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3491         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3492         (get_polymorphic_call_info): ... here.
3493         (get_polymorphic_call_info_from_invariant): New function.
3495 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3497         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3498         lookup via vtable pointer; check for type consistency
3499         and turn inconsitent facts into UNREACHABLE.
3500         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3501         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3502         type inconsistent querries; return UNREACHABLE instead.
3504 2014-02-03  Richard Henderson  <rth@twiddle.net>
3506         PR tree-opt/59924
3507         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3508         already processed this node.
3509         (normalize_one_pred_1): Pass along mark_set.
3510         (normalize_one_pred): Create and destroy a pointer_set_t.
3511         (normalize_one_pred_chain): Likewise.
3513 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3515         PR gcov-profile/58602
3516         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3518 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3520         PR ipa/59831
3521         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3522         -fno-devirtualize; try to devirtualize by the knowledge of
3523         virtual table pointer given by aggregate propagation.
3524         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3525         (ipa_print_node_jump_functions): Dump also offset that
3526         is relevant for polymorphic calls.
3527         (determine_known_aggregate_parts): Add arg_type parameter; use it
3528         instead of determining the type from pointer type.
3529         (ipa_compute_jump_functions_for_edge): Update call of
3530         determine_known_aggregate_parts.
3531         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3532         (gimple_get_virt_method_for_binfo): ... here; simplify using
3533         vtable_pointer_value_to_vtable.
3534         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3535         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3536         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3537         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3538         POINTER_PLUS_EXPR.
3539         (vtable_pointer_value_to_binfo): ... here.
3540         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3542 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3544         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3545         redef of outer loop index variable.
3547 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3549         PR c++/53017
3550         PR c++/59211
3551         * doc/extend.texi (Function Attributes): Typo.
3553 2014-02-03  Cong Hou  <congh@google.com>
3555         PR tree-optimization/60000
3556         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3557         if the vectorized statement is a store.  A store statement can only
3558         appear at the end of pattern statements.
3560 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3562         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3563         (ix86_option_override_internal): Default long double to 64-bit for
3564         32-bit Bionic and to 128-bit for 64-bit Bionic.
3566         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3567         TARGET_LONG_DOUBLE_128 is true.
3568         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3570         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3571         (mlong-double-64): Negate -mlong-double-128.
3572         (mlong-double-128): New option.
3574         * config/i386/i386-c.c (ix86_target_macros): Define
3575         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3577         * doc/invoke.texi: Document -mlong-double-128.
3579 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3581         PR rtl-optimization/60024
3582         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3584 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3586         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3588 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3590         PR rtl-optimization/57662
3591         * sel-sched.c (code_motion_path_driver): Do not mark already not
3592         existing blocks in the visiting bitmap.
3594 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3596         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3597         on the insn being emitted.
3599 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3600             Will Deacon  <will.deacon@arm.com>
3602         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3604 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3606         * config/arm/arm-tables.opt: Regenerate.
3608 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3610         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3611         for vector types other than V16QImode.
3612         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3613         define_expand, and call altivec_expand_vec_perm_le when producing
3614         code with little endian element order.
3615         (*altivec_vperm_<mode>_internal): New insn having previous
3616         behavior of altivec_vperm_<mode>.
3617         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3618         altivec_expand_vec_perm_le when producing code with little endian
3619         element order.
3620         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3621         behavior of altivec_vperm_<mode>_uns.
3623 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3625         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3626         (altivec_vsumsws): Add handling for -maltivec=be with a little
3627         endian target.
3628         (altivec_vsumsws_direct): New.
3629         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3630         gen_altivec_vsumsws.
3632 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3634         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3635         vtable_pointer_value_to_binfo): New functions.
3636         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3637         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3639 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3641         * config/nios2/nios2.md (load_got_register): Initialize GOT
3642         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3643         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3645 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3647         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3648         preserverd by passthrough, do not propagate the type.
3650 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3652         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3653         (mips_atomic_assign_expand_fenv): New function.
3654         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3656 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3658         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3659         (__builtin_mips_set_fcsr): Likewise.
3660         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3661         MIPS_USI_FTYPE_VOID.
3662         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3663         (mips16_expand_set_fcsr): Likewise.
3664         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3665         (mips16_set_fcsr_stub): Likewise.
3666         (mips16_get_fcsr_one_only_stub): New class.
3667         (mips16_set_fcsr_one_only_stub): Likewise.
3668         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3669         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3670         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3671         (hard_float): New availability predicate.
3672         (mips_builtins): Add get_fcsr and set_fcsr.
3673         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3674         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3675         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3676         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3677         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3678         patterns.
3680 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3682         * config/mips/mips.c (mips_one_only_stub): New class.
3683         (mips_need_mips16_rdhwr_p): Replace with...
3684         (mips16_rdhwr_stub): ...this new variable.
3685         (mips16_stub_call_address): New function.
3686         (mips16_rdhwr_one_only_stub): New class.
3687         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3688         (mips_output_mips16_rdhwr): Delete.
3689         (mips_finish_stub): New function.
3690         (mips_code_end): Use it to handle rdhwr stubs.
3692 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3694         PR target/60017
3695         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3696         when calculating size of integer atomic types.
3698 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3700         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3702 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3704         PR tree-optimization/60003
3705         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3706         * profile.c (branch_prob): Use gimple_call_builtin_p
3707         to check for BUILT_IN_SETJMP_RECEIVER.
3708         * tree-inline.c (copy_bb): Call notice_special_calls.
3710 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3712         PR bootstrap/59985
3713         * lra-constraints.c (process_alt_operands): Update reload_sum only
3714         on the first pass.
3716 2014-01-31  Richard Henderson  <rth@redhat.com>
3718         PR middle-end/60004
3719         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3720         until after else_eh is processed.
3722 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3724         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3725         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3726         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3727         in smmintrin.h, remove them.
3728         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3729         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3730         * config/i386/i386.md (ROUND_SAE): Fix value.
3731         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3732         (const48_operand): New.
3733         * config/i386/subst.md (round), (round_expand): Use
3734         const_4_or_8_to_11_operand.
3735         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3737 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3739         * config/i386/constraints.md (Yk): Swap meaning with k.
3740         * config/i386/i386.md (movhi_internal): Change Yk to k.
3741         (movqi_internal): Ditto.
3742         (*k<logic><mode>): Ditto.
3743         (*andhi_1): Ditto.
3744         (*andqi_1): Ditto.
3745         (kandn<mode>): Ditto.
3746         (*<code>hi_1): Ditto.
3747         (*<code>qi_1): Ditto.
3748         (kxnor<mode>): Ditto.
3749         (kortestzhi): Ditto.
3750         (kortestchi): Ditto.
3751         (kunpckhi): Ditto.
3752         (*one_cmplhi2_1): Ditto.
3753         (*one_cmplqi2_1): Ditto.
3754         * config/i386/sse.md (): Change k to Yk.
3755         (avx512f_load<mode>_mask): Ditto.
3756         (avx512f_blendm<mode>): Ditto.
3757         (avx512f_store<mode>_mask): Ditto.
3758         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3759         (avx512f_storedqu<mode>_mask): Ditto.
3760         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3761         Ditto.
3762         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3763         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3764         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3765         (avx512f_maskcmp<mode>3): Ditto.
3766         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3767         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3768         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3769         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3770         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3771         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3772         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3773         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3774         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3775         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3776         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3777         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3778         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3779         (vec_extract_lo_<mode>_maskm): Ditto.
3780         (vec_extract_hi_<mode>_maskm): Ditto.
3781         (avx512f_vternlog<mode>_mask): Ditto.
3782         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3783         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3784         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3785         (avx512f_<code>v8div16qi2_mask): Ditto.
3786         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3787         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3788         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3789         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3790         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3791         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3792         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3793         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3794         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3795         (avx512cd_maskb_vec_dupv8di): Ditto.
3796         (avx512cd_maskw_vec_dupv16si): Ditto.
3797         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3798         (avx512f_vpermi2var<mode>3_mask): Ditto.
3799         (avx512f_vpermi2var<mode>3_mask): Ditto.
3800         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3801         (*avx512f_gathersi<mode>): Ditto.
3802         (*avx512f_gathersi<mode>_2): Ditto.
3803         (*avx512f_gatherdi<mode>): Ditto.
3804         (*avx512f_gatherdi<mode>_2): Ditto.
3805         (*avx512f_scattersi<mode>): Ditto.
3806         (*avx512f_scatterdi<mode>): Ditto.
3807         (avx512f_compress<mode>_mask): Ditto.
3808         (avx512f_compressstore<mode>_mask): Ditto.
3809         (avx512f_expand<mode>_mask): Ditto.
3810         * config/i386/subst.md (mask): Change k to Yk.
3811         (mask_scalar_merge): Ditto.
3812         (sd): Ditto.
3814 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3816         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3818 2014-01-31  Richard Biener  <rguenther@suse.de>
3820         PR middle-end/59990
3821         * builtins.c (fold_builtin_memory_op): Make sure to not
3822         use a floating-point mode or a boolean or enumeral type for
3823         the copy operation.
3825 2014-01-30  DJ Delorie  <dj@redhat.com>
3827         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3828         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3829         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3830         whenever main() has an epilogue.
3832 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3834         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3835         unused variable "field".
3836         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3837         (vsx_mergeh_<mode>): Likewise.
3838         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3839         (altivec_vmrghh): Likewise.
3840         (altivec_vmrghw): Likewise.
3841         (altivec_vmrglb): Likewise.
3842         (altivec_vmrglh): Likewise.
3843         (altivec_vmrglw): Likewise.
3844         (altivec_vspltb): Add missing uses.
3845         (altivec_vsplth): Likewise.
3846         (altivec_vspltw): Likewise.
3847         (altivec_vspltsf): Likewise.
3849 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3851         PR target/59923
3852         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3853         frame related instructions.
3855 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3857         PR rtl-optimization/59959
3858         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3859         any reload of register whose subreg is invalid.
3861 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3863         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3864         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3865         Add missing return type - void.
3867 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3869         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3870         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3871         remove element index adjustment for endian (now handled in vsx.md
3872         and altivec.md).
3873         (altivec_expand_vec_perm_const): Use
3874         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3875         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3876         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3877         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3878         define_expand and a new define_insn *altivec_vspltb_internal;
3879         adjust for -maltivec=be on a little endian target.
3880         (altivec_vspltb_direct): New.
3881         (altivec_vsplth): Divide into a define_expand and a new
3882         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3883         little endian target.
3884         (altivec_vsplth_direct): New.
3885         (altivec_vspltw): Divide into a define_expand and a new
3886         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3887         little endian target.
3888         (altivec_vspltw_direct): New.
3889         (altivec_vspltsf): Divide into a define_expand and a new
3890         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3891         a little endian target.
3893 2014-01-30  Richard Biener  <rguenther@suse.de>
3895         PR tree-optimization/59993
3896         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3897         can propagate form the earlier stmt and avoid the transform
3898         when the intermediate result is needed.
3900 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3902         * README.Portability: Fix typo.
3904 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
3906         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3907         comparison_operator with ordered_comparison_operator.
3909 2014-01-30  Nick Clifton  <nickc@redhat.com>
3911         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3912         Rename to mn10300_store_multiple_regs.
3913         * config/mn10300/mn10300.c: Likewise.
3914         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3915         store_multiple_regs.
3916         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3917         Call mn10300_store_multiple_regs.
3919 2014-01-30  Nick Clifton  <nickc@redhat.com>
3920             DJ Delorie  <dj@redhat.com>
3922         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3923         %fp 2 to keep registers after it properly word-aligned.
3924         (rl78_alloc_physical_registers_umul): Handle the case where both
3925         input operands are the same.
3927 2014-01-30  Richard Biener  <rguenther@suse.de>
3929         PR tree-optimization/59903
3930         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3931         check properly.
3933 2014-01-30  Jason Merrill  <jason@redhat.com>
3935         PR c++/59633
3936         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3938         PR c++/59645
3939         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3941 2014-01-30  Richard Biener  <rguenther@suse.de>
3943         PR tree-optimization/59951
3944         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3946 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3948         PR target/59784
3949         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3950         SFmode to DFmode case.
3952 2014-01-29  DJ Delorie  <dj@redhat.com>
3954         * config/msp430/msp430.opt (-minrt): New.
3955         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3956         if -minrt given.
3957         (ENDFILE_SPEC): Likewise.
3959 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3961         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3962         (estimate_function_body_sizes): Use it.
3964 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3966         PR c++/58561
3967         * dwarf2out.c (is_cxx_auto): New.
3968         (is_base_type): Use it.
3969         (gen_type_die_with_usage): Likewise.
3971 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3973         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3974         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3975         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3976         -maltivec=be with LE targets.
3977         (vsx_mergeh_<mode>): Likewise.
3978         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3979         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3980         (altivec_vmrghb): Replace with define_expand and new
3981         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3982         (altivec_vmrghb_direct): New define_insn.
3983         (altivec_vmrghh): Replace with define_expand and new
3984         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3985         (altivec_vmrghh_direct): New define_insn.
3986         (altivec_vmrghw): Replace with define_expand and new
3987         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3988         (altivec_vmrghw_direct): New define_insn.
3989         (*altivec_vmrghsf): Adjust for endianness.
3990         (altivec_vmrglb): Replace with define_expand and new
3991         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3992         (altivec_vmrglb_direct): New define_insn.
3993         (altivec_vmrglh): Replace with define_expand and new
3994         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3995         (altivec_vmrglh_direct): New define_insn.
3996         (altivec_vmrglw): Replace with define_expand and new
3997         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3998         (altivec_vmrglw_direct): New define_insn.
3999         (*altivec_vmrglsf): Adjust for endianness.
4000         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4001         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4002         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
4003         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
4004         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4005         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4006         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
4007         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4009 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4011         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4012         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4013         whitespace.
4015 2014-01-29  Richard Biener  <rguenther@suse.de>
4017         PR tree-optimization/58742
4018         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4019         associate_pointerplus_align.
4020         (associate_pointerplus_diff): New function.
4021         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4022         and associate_pointerplus_diff.
4024 2014-01-29  Richard Biener  <rguenther@suse.de>
4026         * lto-streamer.h (LTO_major_version): Bump to 3.
4027         (LTO_minor_version): Reset to 0.
4029 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4031         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4032         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4033         (arm_file_start): Generate correct asm header for armv7ve.
4034         * config/arm/bpabi.h: Add multilib support for armv7ve.
4035         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4036         and cortex-a15 to armv7ve.
4037         * config/arm/t-aprofile: Add multilib support for armv7ve.
4038         * doc/invoke.texi: Document -march=armv7ve.
4040 2014-01-29  Richard Biener  <rguenther@suse.de>
4042         PR tree-optimization/58742
4043         * tree-ssa-forwprop.c (associate_plusminus): Return true
4044         if we changed sth, defer EH cleanup to ...
4045         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4046         (simplify_mult): New function.
4048 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4050         PR middle-end/59917
4051         PR tree-optimization/59920
4052         * tree.c (build_common_builtin_nodes): Remove
4053         __builtin_setjmp_dispatcher initialization.
4054         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4055         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4056         instead of gsi_after_labels + manually skipping debug stmts.
4057         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4058         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4059         * tree-inline.c (copy_edges_for_bb): Remove
4060         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4061         argument.  Ignore computed_goto_p stmts.  Don't call
4062         make_abnormal_goto_edges.  If a call might need abnormal edges
4063         for non-local gotos, see if it already has an edge to
4064         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4065         with true argument, don't do anything then, otherwise add
4066         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4067         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4068         caller.
4069         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4070         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4071         (lower_stmt): Don't set data->calls_builtin_setjmp.
4072         (lower_builtin_setjmp): Adjust comment.
4073         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4074         * tree-cfg.c (found_computed_goto): Remove.
4075         (factor_computed_gotos): Remove.
4076         (make_goto_expr_edges): Return bool, true for computed gotos.
4077         Don't call make_abnormal_goto_edges.
4078         (build_gimple_cfg): Don't set found_computed_goto, don't call
4079         factor_computed_gotos.
4080         (computed_goto_p): No longer static.
4081         (make_blocks): Don't set found_computed_goto.
4082         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4083         (make_edges): If make_goto_expr_edges returns true, push bb
4084         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4085         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4086         vector.  Record mapping between bbs and OpenMP regions if there
4087         are any, adjust make_gimple_omp_edges caller.  Call
4088         handle_abnormal_edges.
4089         (make_abnormal_goto_edges): Remove.
4090         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4091         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4092         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4093         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4094         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4095         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4096         filling *region also set *region_idx to (*region)->entry->index.
4098         PR other/58712
4099         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4100         For REGs set ORIGINAL_REGNO.
4102 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4104         * doc/md.texi: Mention that a target shouldn't implement
4105         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4106         than hi/lo pair.
4108 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4110         PR tree-optimization/59594
4111         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4112         a copy of the datarefs vector rather than the vector itself.
4114 2014-01-28  Jason Merrill  <jason@redhat.com>
4116         PR c++/53756
4117         * dwarf2out.c (auto_die): New static.
4118         (gen_type_die_with_usage): Handle C++1y 'auto'.
4119         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4120         on definition.
4122 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4124         PR target/59672
4125         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4126         (SPEC_X32): Likewise.
4127         (SPEC_64): Likewise.
4128         * config/i386/i386.c (ix86_option_override_internal): Turn off
4129         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4130         for TARGET_16BIT.
4131         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4132         * config/i386/i386.h (TARGET_16BIT): New macro.
4133         (TARGET_16BIT_P): Likewise.
4134         * config/i386/i386.opt: Add m16.
4135         * doc/invoke.texi: Document -m16.
4137 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4139         PR preprocessor/59935
4140         * input.c (location_get_source_line): Bail out on when line number
4141         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4143 2014-01-28  Richard Biener  <rguenther@suse.de>
4145         PR tree-optimization/58742
4146         * tree-ssa-forwprop.c (associate_plusminus): Handle
4147         pointer subtraction of the form (T)(P + A) - (T)P.
4149 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4151         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4152         at const_int_cost.
4154 2014-01-28  Richard Biener  <rguenther@suse.de>
4156         Revert
4157         2014-01-28  Richard Biener  <rguenther@suse.de>
4159         PR rtl-optimization/45364
4160         PR rtl-optimization/59890
4161         * var-tracking.c (local_get_addr_clear_given_value): Handle
4162         already cleared slot.
4163         (val_reset): Handle not allocated local_get_addr_cache.
4164         (vt_find_locations): Use post-order on the inverted CFG.
4166 2014-01-28  Richard Biener  <rguenther@suse.de>
4168         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4170 2014-01-28  Richard Biener  <rguenther@suse.de>
4172         PR rtl-optimization/45364
4173         PR rtl-optimization/59890
4174         * var-tracking.c (local_get_addr_clear_given_value): Handle
4175         already cleared slot.
4176         (val_reset): Handle not allocated local_get_addr_cache.
4177         (vt_find_locations): Use post-order on the inverted CFG.
4179 2014-01-28  Alan Modra  <amodra@gmail.com>
4181         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4182         * configure.ac <recursive call for build != host>: Define
4183         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4184         and LD_FOR_BUILD too.
4185         * configure: Regenerate.
4187 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4189         * config/i386/i386.c (get_builtin_code_for_version): Separate
4190         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4191         Broadwell from Haswell.
4193 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4195         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4196         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4197         * config/mips/mips.c (mips_option_override): Change setting
4198         of TARGET_DSP.
4199         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4200         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4201         Change from Mask to Var.
4203 2014-01-27  Jeff Law  <law@redhat.com>
4205         * ipa-inline.c (inline_small_functions): Fix typo.
4207 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4209         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4210         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4211         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4212         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4213         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4214         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4215         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4216         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4217         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4218         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4219         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4220         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4221         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4222         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4223         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4224         (_mm512_storeu_epi64): Ditto.
4225         (_mm512_cmpge_epi32_mask): Ditto.
4226         (_mm512_cmpge_epu32_mask): Ditto.
4227         (_mm512_cmpge_epi64_mask): Ditto.
4228         (_mm512_cmpge_epu64_mask): Ditto.
4229         (_mm512_cmple_epi32_mask): Ditto.
4230         (_mm512_cmple_epu32_mask): Ditto.
4231         (_mm512_cmple_epi64_mask): Ditto.
4232         (_mm512_cmple_epu64_mask): Ditto.
4233         (_mm512_cmplt_epi32_mask): Ditto.
4234         (_mm512_cmplt_epu32_mask): Ditto.
4235         (_mm512_cmplt_epi64_mask): Ditto.
4236         (_mm512_cmplt_epu64_mask): Ditto.
4237         (_mm512_cmpneq_epi32_mask): Ditto.
4238         (_mm512_cmpneq_epu32_mask): Ditto.
4239         (_mm512_cmpneq_epi64_mask): Ditto.
4240         (_mm512_cmpneq_epu64_mask): Ditto.
4241         (_mm512_expand_pd): Ditto.
4242         (_mm512_expand_ps): Ditto.
4243         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4244         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4245         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4246         * config/i386/i386.c (ix86_builtins): Add
4247         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4248         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4249         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4250         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4251         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4252         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4253         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4254         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4255         IX86_BUILTIN_PMOVUSQW512_MEM.
4256         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4257         __builtin_ia32_pmovsqd512mem_mask,
4258         __builtin_ia32_pmovqd512mem_mask,
4259         __builtin_ia32_pmovusqw512mem_mask,
4260         __builtin_ia32_pmovsqw512mem_mask,
4261         __builtin_ia32_pmovqw512mem_mask,
4262         __builtin_ia32_pmovusdw512mem_mask,
4263         __builtin_ia32_pmovsdw512mem_mask,
4264         __builtin_ia32_pmovdw512mem_mask,
4265         __builtin_ia32_pmovqb512mem_mask,
4266         __builtin_ia32_pmovusqb512mem_mask,
4267         __builtin_ia32_pmovsqb512mem_mask,
4268         __builtin_ia32_pmovusdb512mem_mask,
4269         __builtin_ia32_pmovsdb512mem_mask,
4270         __builtin_ia32_pmovdb512mem_mask.
4271         (bdesc_args): Add __builtin_ia32_expanddf512,
4272         __builtin_ia32_expandsf512.
4273         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4274         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4275         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4276         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4277         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4278         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4279         (avx512f_<code>v8div16qi2_mask_store): This.
4280         (avx512f_expand<mode>): New.
4282 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4284         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4285         New.
4286         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4287         (_mm512_prefetch_i32scatter_pd): Ditto.
4288         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4289         (_mm512_prefetch_i64scatter_pd): Ditto.
4290         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4291         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4292         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4293         (_mm512_prefetch_i32scatter_ps): Ditto.
4294         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4295         (_mm512_prefetch_i64scatter_ps): Ditto.
4296         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4297         * config/i386/i386-builtin-types.def: Define
4298         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4299         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4300         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4301         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4302         IX86_BUILTIN_SCATTERPFQPD.
4303         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4304         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4305         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4306         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4307         __builtin_ia32_scatterpfqps.
4308         (ix86_expand_builtin): Expand new built-ins.
4309         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4310         fix memory access data type.
4311         (*avx512pf_gatherpf<mode>_mask): Ditto.
4312         (*avx512pf_gatherpf<mode>): Ditto.
4313         (avx512pf_scatterpf<mode>): Ditto.
4314         (*avx512pf_scatterpf<mode>_mask): Ditto.
4315         (*avx512pf_scatterpf<mode>): Ditto.
4316         (GATHER_SCATTER_SF_MEM_MODE): New.
4317         (avx512pf_gatherpf<mode>df): Ditto.
4318         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4319         (*avx512pf_scatterpf<mode>df): Ditto.
4321 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4323         PR bootstrap/59934
4324         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4325         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4326         reached.
4328 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4330         * common/config/arm/arm-common.c
4331         (arm_rewrite_mcpu): Handle multiple names.
4332         * config/arm/arm.h
4333         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4335 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4337         * gimple-builder.h (create_gimple_tmp): Delete.
4339 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4341         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4342         words comparisons.
4344 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4346         * config/pa/pa.md (call): Generate indirect long calls to non-local
4347         functions when outputing 32-bit code.
4348         (call_value): Likewise except for special call to buggy powf function.
4350         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4351         portable runtime and PIC indirect calls.
4352         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4353         and PIC call sequences.  Use ldo instead of blr to set return register
4354         in PIC call sequence.
4356 2014-01-25  Walter Lee  <walt@tilera.com>
4358         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4359         avoid clobbering a live register.
4361 2014-01-25  Walter Lee  <walt@tilera.com>
4363         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4364         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4365         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4366         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4368 2014-01-25  Walter Lee  <walt@tilera.com>
4370         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4371         arguments on even registers.
4372         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4373         STACK_BOUNDARY.
4374         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4375         (BIGGEST_ALIGNMENT): Ditto.
4376         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4378 2014-01-25  Walter Lee  <walt@tilera.com>
4380         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4381         insns before bundling.
4382         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4384 2014-01-25  Walter Lee  <walt@tilera.com>
4386         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4387         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4388         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4390 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4392         * config/mips/constraints.md (kl): Delete.
4393         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4394         define expands, using...
4395         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4396         instructions for MIPS16.
4397         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4398         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4400 2014-01-25  Walter Lee  <walt@tilera.com>
4402         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4403         (clzdi2): Ditto.
4404         (ffsdi2): Ditto.
4406 2014-01-25  Walter Lee  <walt@tilera.com>
4408         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4409         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4411 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4413         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4414         Handle XOR.
4416 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4418         * print-rtl.c (in_call_function_usage): New var.
4419         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4420         EXPR_LIST mode as mode and not as reg note name.
4422         PR middle-end/59561
4423         * cfgloopmanip.c (copy_loop_info): If
4424         loop->warned_aggressive_loop_optimizations, make sure
4425         the flag is set in target loop too.
4427 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4429         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4430         flag_cilkplus.
4431         * builtins.def: Likewise.
4432         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4433         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4434         * ira.c (ira_setup_eliminable_regset): Likewise.
4435         * omp-low.c (gate_expand_omp): Likewise.
4436         (execute_lower_omp): Likewise.
4437         (diagnose_sb_0): Likewise.
4438         (gate_diagnose_omp_blocks): Likewise.
4439         (simd_clone_clauses_extract): Likewise.
4440         (gate): Likewise.
4442 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4444         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4445         correction for little endian...
4446         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4447         here.
4449 2014-01-24  Jeff Law  <law@redhat.com>
4451         PR tree-optimization/59919
4452         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4453         for non-returning calls.
4455 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4457         * common/config/aarch64/aarch64-common.c
4458         (aarch64_rewrite_mcpu): Handle multiple names.
4459         * config/aarch64/aarch64.h
4460         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4462 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4464         * input.c (add_file_to_cache_tab): Handle the case where fopen
4465         returns NULL.
4467 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4469         PR target/59929
4470         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4471         from push operand if code of push isn't PRE_DEC.
4473 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4475         PR target/59909
4476         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4477         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4478         it is only used for non-atomic loads/stores.
4480         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4481         -mquad-memory or -mquad-memory-atomic switches.
4483         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4484         -mquad-memory-atomic to ISA 2.07 support.
4486         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4487         to separate support of normal quad word memory operations (ldq, stq)
4488         from the atomic quad word memory operations.
4490         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4491         support to separate non-atomic quad word operations from atomic
4492         quad word operations.  Disable non-atomic quad word operations in
4493         little endian mode so that we don't have to swap words after the
4494         load and before the store.
4495         (quad_load_store_p): Add comment about atomic quad word support.
4496         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4497         options printed with -mdebug=reg.
4499         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4500         -mquad-memory-atomic as the test for whether we have quad word
4501         atomic instructions.
4502         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4503         or -mp8-vector are used, allow byte/half-word atomic operations.
4505         * config/rs6000/sync.md (load_lockedti): Insure that the address
4506         is a proper indexed or indirect address for the lqarx instruction.
4507         On little endian systems, swap the hi/lo registers after the lqarx
4508         instruction.
4509         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4510         insure the address is valid for the lqarx instruction.
4511         (store_conditionalti): Insure that the address is a proper indexed
4512         or indirect address for the stqcrx. instruction.  On little endian
4513         systems, swap the hi/lo registers before doing the stqcrx.
4514         instruction.
4515         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4516         insure the address is valid for the stqcrx. instruction.
4518         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4519         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4520         type of quad memory support is available.
4522 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4524         PR regression/59915
4525         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4526         there is a danger of looping.
4528 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4530         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4531         force flag_ira_loop_pressure if set via command line.
4533 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4535         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4536         (ashr_simd): New builtin handling DI mode.
4537         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4538         (aarch64_sshr_simddi): New match pattern.
4539         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4540         (vshrd_n_s64): Likewise.
4541         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4543 2014-01-23  Nick Clifton  <nickc@redhat.com>
4545         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4546         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4547         favour of mcu specific scripts.
4548         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4549         430x multilibs.
4551 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4552             Alex Velenko  <Alex.Velenko@arm.com>
4554         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4555         (vaddv_s16): Likewise.
4556         (vaddv_s32): Likewise.
4557         (vaddv_u8): Likewise.
4558         (vaddv_u16): Likewise.
4559         (vaddv_u32): Likewise.
4560         (vaddvq_s8): Likewise.
4561         (vaddvq_s16): Likewise.
4562         (vaddvq_s32): Likewise.
4563         (vaddvq_s64): Likewise.
4564         (vaddvq_u8): Likewise.
4565         (vaddvq_u16): Likewise.
4566         (vaddvq_u32): Likewise.
4567         (vaddvq_u64): Likewise.
4568         (vaddv_f32): Likewise.
4569         (vaddvq_f32): Likewise.
4570         (vaddvq_f64): Likewise.
4571         (vmaxv_f32): Likewise.
4572         (vmaxv_s8): Likewise.
4573         (vmaxv_s16): Likewise.
4574         (vmaxv_s32): Likewise.
4575         (vmaxv_u8): Likewise.
4576         (vmaxv_u16): Likewise.
4577         (vmaxv_u32): Likewise.
4578         (vmaxvq_f32): Likewise.
4579         (vmaxvq_f64): Likewise.
4580         (vmaxvq_s8): Likewise.
4581         (vmaxvq_s16): Likewise.
4582         (vmaxvq_s32): Likewise.
4583         (vmaxvq_u8): Likewise.
4584         (vmaxvq_u16): Likewise.
4585         (vmaxvq_u32): Likewise.
4586         (vmaxnmv_f32): Likewise.
4587         (vmaxnmvq_f32): Likewise.
4588         (vmaxnmvq_f64): Likewise.
4589         (vminv_f32): Likewise.
4590         (vminv_s8): Likewise.
4591         (vminv_s16): Likewise.
4592         (vminv_s32): Likewise.
4593         (vminv_u8): Likewise.
4594         (vminv_u16): Likewise.
4595         (vminv_u32): Likewise.
4596         (vminvq_f32): Likewise.
4597         (vminvq_f64): Likewise.
4598         (vminvq_s8): Likewise.
4599         (vminvq_s16): Likewise.
4600         (vminvq_s32): Likewise.
4601         (vminvq_u8): Likewise.
4602         (vminvq_u16): Likewise.
4603         (vminvq_u32): Likewise.
4604         (vminnmv_f32): Likewise.
4605         (vminnmvq_f32): Likewise.
4606         (vminnmvq_f64): Likewise.
4608 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4610         * config/aarch64/aarch64-simd.md
4611         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4612         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4613         (*aarch64_mul3_elt<mode>): Likewise.
4614         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4615         (*aarch64_mul3_elt_to_64v2df): Likewise.
4616         (*aarch64_mla_elt<mode>): Likewise.
4617         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4618         (*aarch64_mls_elt<mode>): Likewise.
4619         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4620         (*aarch64_fma4_elt<mode>): Likewise.
4621         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4622         (*aarch64_fma4_elt_to_64v2df): Likewise.
4623         (*aarch64_fnma4_elt<mode>): Likewise.
4624         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4625         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4626         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4627         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4628         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4629         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4630         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4631         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4632         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4634 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4636         * config/aarch64/aarch64-simd.md
4637         (aarch64_be_checked_get_lane<mode>): New define_expand.
4638         * config/aarch64/aarch64-simd-builtins.def
4639         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4640         New builtin definition.
4641         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4642         Use new safe be builtin.
4644 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4646         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4647         New define_insn.
4648         (aarch64_be_st1<mode>): Likewise.
4649         (aarch_ld1<VALL:mode>): Define_expand modified.
4650         (aarch_st1<VALL:mode>): Likewise.
4651         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4652         (UNSPEC_ST1): Likewise.
4654 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
4656         * config/microblaze/microblaze.md: Add trap insn and attribute
4658 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4660         PR preprocessor/58580
4661         * input.h (location_get_source_line): Take an additional line_size
4662         parameter.
4663         (void diagnostics_file_cache_fini): Declare new function.
4664         * input.c (struct fcache): New type.
4665         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4666         New static constants.
4667         (diagnostic_file_cache_init, total_lines_num)
4668         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4669         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4670         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4671         (get_next_line, read_next_line, goto_next_line, read_line_num):
4672         New static function definitions.
4673         (diagnostic_file_cache_fini): New function.
4674         (location_get_source_line): Take an additional output line_len
4675         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4676         read_line_num.
4677         * diagnostic.c (diagnostic_finish): Call
4678         diagnostic_file_cache_fini.
4679         (adjust_line): Take an additional input parameter for the length
4680         of the line, rather than calculating it with strlen.
4681         (diagnostic_show_locus): Adjust the use of
4682         location_get_source_line and adjust_line with respect to their new
4683         signature.  While displaying a line now, do not stop at the first
4684         null byte.  Rather, display the zero byte as a space and keep
4685         going until we reach the size of the line.
4686         * Makefile.in: Add vec.o to OBJS-libcommon
4688 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4689             Ilya Tocar     <ilya.tocar@intel.com>
4691         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4692         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4693         (__builtin_ia32_kmov16): Ditto.
4694         * config/i386/i386.md (UNSPEC_KMOV): New.
4695         (kmovw): Ditto.
4697 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4699         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4700         (_mm512_storeu_si512): Ditto.
4702 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4704         PR target/52125
4705         * rtl.h (get_referenced_operands): Declare.
4706         * recog.c (get_referenced_operands): New function.
4707         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4708         operands have been referenced when recording LO_SUM references.
4710 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
4712         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4714 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4716         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4717         Enable for generic and recent AMD targets.
4719 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4721         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4722         ARG_SIZE note when adjustment was eliminated.
4724 2014-01-22  Jeff Law  <law@redhat.com>
4726         PR tree-optimization/59597
4727         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4728         in file.  Accept new argument REGISTERING and use it to modify
4729         dump output appropriately.
4730         (register_jump_thread): Corresponding changes.
4731         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4732         thread paths involving joiner blocks.  Add code to dump cancelled
4733         jump threading paths.
4735 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4737         PR rtl-optimization/59477
4738         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4739         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4741 2014-01-22  Tom Tromey  <tromey@redhat.com>
4743         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4744         PARAMS.
4745         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4747 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4749         PR rtl-optimization/59896
4750         * lra-constraints.c (process_alt_operands): Check unused note for
4751         matched operands of insn with no output reloads.
4753 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4755         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4756         (mips_move_from_gpr_cost): Likewise.
4758 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4760         PR rtl-optimization/59858
4761         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4762         ira_class_hard_regs_num.
4763         (process_alt_operands): Increase reject for dying matched operand.
4765 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4767         PR target/59003
4768         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4769         smaller than size, perform several stores or loads and stores
4770         at dst + count - size to store or copy all of size bytes, rather
4771         than just last modesize bytes.
4773 2014-01-20  DJ Delorie  <dj@redhat.com>
4775         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4776         that CLOBBERs are REGs before propogating their values.
4778 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4780         PR middle-end/59789
4781         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4782         (cgraph_inline_failed_type): New function.
4783         * cgraph.h (DEFCIFCODE): Add type.
4784         (cgraph_inline_failed_type_t): New enum.
4785         (cgraph_inline_failed_type): New prototype.
4786         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4787         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4788         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4789         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4790         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4791         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4792         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4793         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4794         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4795         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4796         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4797         OPTIMIZATION_MISMATCH.
4798         * tree-inline.c (expand_call_inline): Emit errors during
4799         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4801 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4803         PR target/59685
4804         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4805         mode attribute in insn output.
4807 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4809         * output.h (output_constant): Delete.
4810         * varasm.c (output_constant): Make private.
4812 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4814         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4816 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4818         PR middle-end/59860
4819         * tree.h (fold_builtin_strcat): New prototype.
4820         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4821         argument, if non-NULL, don't call c_strlen.  Optimize
4822         directly into __builtin_memcpy instead of __builtin_strcpy.
4823         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4824         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4826 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4828         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4829         for SImode_address_operand operands, having only a REG argument.
4831 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4833         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4834         loader name using mbig-endian.
4835         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4837 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4839         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4840         (-mtune): Likewise.
4841         (-mcpu): Likewise.
4843 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4845         * config/aarch64/aarch64-protos.h
4846         (aarch64_cannot_change_mode_class_ptr): Declare.
4847         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4848         aarch64_cannot_change_mode_class_ptr): New.
4849         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4850         backend hook aarch64_cannot_change_mode_class.
4852 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4854         * common/config/aarch64/aarch64-common.c
4855         (aarch64_handle_option): Don't handle any option order logic here.
4856         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4857         selected_cpu, warn on architecture version mismatch.
4858         (aarch64_override_options): Fix parsing order for option strings.
4860 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4861             Iain Sandoe  <iain@codesourcery.com>
4863         PR bootstrap/59496
4864         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4865         warning.  Amend comment to reflect current functionality.
4867 2014-01-20  Richard Biener  <rguenther@suse.de>
4869         PR middle-end/59860
4870         * builtins.c (fold_builtin_strcat): Remove case better handled
4871         by tree-ssa-strlen.c.
4873 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4875         * config/aarch64/aarch64.opt
4876         (mcpu, march, mtune): Make case-insensitive.
4878 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4880         PR target/59880
4881         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4882         if operands[1] is a REG or ZERO_EXTEND of a REG.
4884 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4886         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4888 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4890         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4891         long non-pic millicode calls.
4893 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4895         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4897 2014-01-19  Kito Cheng  <kito@0xlab.org>
4899         * builtins.c (expand_movstr): Check movstr expand done or fail.
4901 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4902             H.J. Lu  <hongjiu.lu@intel.com>
4904         PR target/59379
4905         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4906         to DImode for zero-extended addresses.
4908 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4910         PR rtl-optimization/57763
4911         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4912         on the new indirect jump_insn and increment LABEL_NUSES (label).
4914 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4916         PR bootstrap/59580
4917         PR bootstrap/59583
4918         * config.gcc (x86_archs): New variable.
4919         (x86_64_archs): Likewise.
4920         (x86_cpus): Likewise.
4921         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4922         --with-arch/--with-cpu= options.
4923         Support --with-arch=/--with-cpu={nehalem,westmere,
4924         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4926 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4928         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4929         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4931 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4933         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4935 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4937         PR target/58944
4938         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4939         clear cpp_get_options (parse_in)->warn_unused_macros for
4940         ix86_target_macros_internal with cpp_define.
4942 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4944         * jump.c (delete_related_insns): Keep (use (insn))s.
4945         * reorg.c (redundant_insn): Check for barriers too.
4947 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4949         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4951 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4953         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4954         call to $$dyncall when TARGET_LONG_CALLS is true.
4956 2014-01-17  Jeff Law  <law@redhat.com>
4958         * ree.c (combine_set_extension): Temporarily disable test for
4959         changing number of hard registers.
4961 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4963         PR middle-end/58125
4964         * ipa-inline-analysis.c (inline_free_summary):
4965         Do not free summary of aliases.
4967 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4969         PR middle-end/59706
4970         * gimplify.c (gimplify_expr): Use create_tmp_var
4971         instead of create_tmp_var_raw.  If cond doesn't have
4972         integral type, don't add the IFN_ANNOTATE builtin at all.
4974 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4976         PR ipa/59736
4977         * ipa-cp.c (prev_edge_clone): New variable.
4978         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4979         Also resize prev_edge_clone vector.
4980         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4981         (ipcp_edge_removal_hook): New function.
4982         (ipcp_driver): Register ipcp_edge_removal_hook.
4984 2014-01-17  Andrew Pinski  <apinski@cavium.com>
4985             Steve Ellcey  <sellcey@mips.com>
4987         PR target/59462
4988         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4989         of operator mode.
4991 2014-01-17  Jeff Law  <law@redhat.com>
4993         PR middle-end/57904
4994         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4995         so that pass_ccp runs first.
4997 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4999         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
5000         (ix86_adjust_cost): Use !TARGET_XXX.
5001         (do_reorder_for_imul): Likewise.
5002         (swap_top_of_ready_list): Likewise.
5003         (ix86_sched_reorder): Likewise.
5005 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
5007         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5008         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5009         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5010         (intel_memset): New.  Duplicate slm_memset.
5011         (intel_cost): New.  Duplicate slm_cost.
5012         (m_INTEL): New macro.
5013         (processor_target_table): Add "intel".
5014         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5015         with PROCESSOR_INTEL for "intel".
5016         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5017         PROCESSOR_SILVERMONT.
5018         (ix86_issue_rate): Likewise.
5019         (ix86_adjust_cost): Likewise.
5020         (ia32_multipass_dfa_lookahead): Likewise.
5021         (swap_top_of_ready_list): Likewise.
5022         (ix86_sched_reorder): Likewise.
5023         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5024         instead of TARGET_OPT_AGU.
5025         * config/i386/i386.h (TARGET_INTEL): New.
5026         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5027         (processor_type): Add PROCESSOR_INTEL.
5028         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5029         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5031 2014-01-17  Marek Polacek  <polacek@redhat.com>
5033         PR c/58346
5034         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5035         size is zero.
5037 2014-01-17  Richard Biener  <rguenther@suse.de>
5039         PR tree-optimization/46590
5040         * opts.c (default_options_table): Add entries for
5041         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5042         all enabled at -O1 but not for -Og.
5043         * common.opt (fbranch-count-reg): Remove Init(1).
5044         (fmove-loop-invariants): Likewise.
5045         (ftree-pta): Likewise.
5047 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5049         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5050         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5051         decls to at least the GCC 4.8 used alignments.
5053         PR fortran/59440
5054         * tree-nested.c (convert_nonlocal_reference_stmt,
5055         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5056         of GIMPLE_BIND stmts, adjust associated decls.
5058 2014-01-17  Richard Biener  <rguenther@suse.de>
5060         PR tree-optimization/46590
5061         * vec.h (vec<>::bseach): New member function implementing
5062         binary search according to C89 bsearch.
5063         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5064         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5065         bitmap pointer again.  Make accesses_in_loop a flat array.
5066         (mem_ref_obstack): New global.
5067         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5068         (mark_ref_stored): Likewise.
5069         (ref_indep_loop_p_2): Likewise.
5070         (set_ref_stored_in_loop): New helper function.
5071         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5072         (memref_free): Adjust.
5073         (record_mem_ref_loc): Simplify.
5074         (gather_mem_refs_stmt): Adjust.
5075         (sort_locs_in_loop_postorder_cmp): New function.
5076         (analyze_memory_references): Sort accesses_in_loop after
5077         loop postorder number.
5078         (find_ref_loc_in_loop_cmp): New function.
5079         (for_all_locs_in_loop): Find relevant cluster of locs in
5080         accesses_in_loop and iterate without recursion.
5081         (execute_sm): Avoid uninit warning.
5082         (struct ref_always_accessed): Simplify.
5083         (ref_always_accessed::operator ()): Likewise.
5084         (ref_always_accessed_p): Likewise.
5085         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5086         loop postorder numbers here.
5087         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5088         numbers.
5090 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5092         PR c++/57945
5093         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5094         on decls for which assemble_alias has been called.
5096 2014-01-17  Nick Clifton  <nickc@redhat.com>
5098         * config/msp430/msp430.opt: (mcpu): New option.
5099         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5100         (msp430_option_override): Parse target_cpu.  If the MCU name
5101         matches a generic string, clear target_mcu.
5102         (msp430_attr): Allow numeric interrupt values up to 63.
5103         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5104         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5105         option.
5106         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5107         Add mcpu matches.
5108         * config/msp430/msp430.md (popm): Use %J rather than %I.
5109         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5110         (addhi_cy_i): Use immediate_operand for operand 2.
5111         * doc/invoke.texi: Document -mcpu option.
5113 2014-01-17  Richard Biener  <rguenther@suse.de>
5115         PR rtl-optimization/38518
5116         * df.h (df_analyze_loop): Declare.
5117         * df-core.c: Include cfgloop.h.
5118         (df_analyze_1): Split out main part of df_analyze.
5119         (df_analyze): Adjust.
5120         (loop_inverted_post_order_compute): New function.
5121         (loop_post_order_compute): Likewise.
5122         (df_analyze_loop): New function avoiding whole-function
5123         postorder computes.
5124         * loop-invariant.c (find_defs): Use df_analyze_loop.
5125         (find_invariants): Adjust.
5126         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5128 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5130         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5131         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5133 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5135         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5136         traversal when removing references.
5138 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5140         PR ipa/59775
5141         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5143 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5145         PR middle-end/56791
5146         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5147         pushing a reload for an autoinc when we had previously reloaded an
5148         inner part of the address.
5150 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5152         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5153         field.
5154         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5155         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5156         when not giving up or versioning for alias only because of
5157         loop->safelen.
5158         (vect_analyze_data_ref_dependences): Set to true.
5159         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5160         is a GIMPLE_PHI.
5161         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5162         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5163         to the condition.
5165         PR middle-end/58344
5166         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5168         PR target/59839
5169         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5170         operand 0 predicate for gathers, use a new pseudo as subtarget.
5172 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5174         PR middle-end/59609
5175         * lra-constraints.c (process_alt_operands): Add printing debug info.
5176         Check absence of input/output reloads for matched operands too.
5178 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5180         PR rtl-optimization/59835
5181         * ira.c (ira_init_register_move_cost): Increase cost for
5182         impossible modes.
5184 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5186         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5188 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5190         PR target/59780
5191         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5192         non-register objects.  Use gen_(high/low)part more consistently.
5193         Fix assertions.
5195 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5197         PR target/59844
5198         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5199         endian support, remove tests for WORDS_BIG_ENDIAN.
5200         (p8_mfvsrd_3_<mode>): Likewise.
5201         (reload_gpr_from_vsx<mode>): Likewise.
5202         (reload_gpr_from_vsxsf): Likewise.
5203         (p8_mfvsrd_4_disf): Likewise.
5205 2014-01-16  Richard Biener  <rguenther@suse.de>
5207         PR rtl-optimization/46590
5208         * lcm.c (compute_antinout_edge): Use postorder iteration.
5209         (compute_laterin): Use inverted postorder iteration.
5211 2014-01-16  Nick Clifton  <nickc@redhat.com>
5213         PR middle-end/28865
5214         * varasm.c (output_constant): Return the number of bytes actually
5215         emitted.
5216         (output_constructor_array_range): Update the field size with the
5217         number of bytes emitted by output_constant.
5218         (output_constructor_regular_field): Likewise.  Also do not
5219         complain if the total number of bytes emitted is now greater
5220         than the expected fieldpos.
5221         * output.h (output_constant): Update prototype and descriptive comment.
5223 2014-01-16  Marek Polacek  <polacek@redhat.com>
5225         PR middle-end/59827
5226         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5227         it is error_mark_node.
5229 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5231         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5232         VALID_AVX256_REG_OR_OI_MODE.
5234 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5236         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5237         current procedure should be profiled.
5239 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5241         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5242         of moving from/to the STACK_REG register class.
5244 2014-01-15  Richard Henderson  <rth@redhat.com>
5246         PR debug/54694
5247         * reginfo.c (global_regs_decl): Globalize.
5248         * rtl.h (global_regs_decl): Declare.
5249         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5250         reserved via global_regs.
5252 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5254         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5256 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5258         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5259         and vmulosh rather than call gen_vec_widen_smult_*.
5260         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5261         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5262         (vec_widen_smult_even_v16qi): Likewise.
5263         (vec_widen_umult_even_v8hi): Likewise.
5264         (vec_widen_smult_even_v8hi): Likewise.
5265         (vec_widen_umult_odd_v16qi): Likewise.
5266         (vec_widen_smult_odd_v16qi): Likewise.
5267         (vec_widen_umult_odd_v8hi): Likewise.
5268         (vec_widen_smult_odd_v8hi): Likewise.
5269         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5270         vmuloub rather than call gen_vec_widen_umult_*.
5271         (vec_widen_umult_lo_v16qi): Likewise.
5272         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5273         vmulosb rather than call gen_vec_widen_smult_*.
5274         (vec_widen_smult_lo_v16qi): Likewise.
5275         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5276         rather than call gen_vec_widen_umult_*.
5277         (vec_widen_umult_lo_v8hi): Likewise.
5278         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5279         rather than call gen_vec_widen_smult_*.
5280         (vec_widen_smult_lo_v8hi): Likewise.
5282 2014-01-15  Jeff Law  <law@redhat.com>
5284         PR tree-optimization/59747
5285         * ree.c (find_and_remove_re): Properly handle case where a second
5286         eliminated extension requires widening a copy created for elimination
5287         of a prior extension.
5288         (combine_set_extension): Ensure that the number of hard regs needed
5289         for a destination register does not change when we widen it.
5291 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5293         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5294         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5295         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5296         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5297         (avr-*-rtems*): Likewise.
5298         (bfin*-rtems*): Likewise.
5299         (moxie-*-rtems*): Likewise.
5300         (h8300-*-rtems*): Likewise.
5301         (i[34567]86-*-rtems*): Likewise.
5302         (lm32-*-rtems*): Likewise.
5303         (m32r-*-rtems*): Likewise.
5304         (m68k-*-rtems*): Likewise.
5305         (microblaze*-*-rtems*): Likewise.
5306         (mips*-*-rtems*): Likewise.
5307         (powerpc-*-rtems*): Likewise.
5308         (sh-*-rtems*): Likewise.
5309         (sparc-*-rtems*): Likewise.
5310         (sparc64-*-rtems*): Likewise.
5311         (v850-*-rtems*): Likewise.
5312         (m32c-*-rtems*): Likewise.
5314 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5316         PR rtl-optimization/59511
5317         * ira.c (ira_init_register_move_cost): Use memory costs for some
5318         cases of register move cost calculations.
5319         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5320         instead of BB frequency.
5321         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5322         * lra-assigns.c (find_hard_regno_for): Ditto.
5324 2014-01-15  Richard Biener  <rguenther@suse.de>
5326         PR tree-optimization/59822
5327         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5328         (vectorizable_load): Use it to hoist defs of uses of invariant
5329         loads out of the loop.
5331 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5332             Kugan Vivekanandarajah  <kuganv@linaro.org>
5334         PR target/59695
5335         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5336         truncation.
5338 2014-01-15  Richard Biener  <rguenther@suse.de>
5340         PR rtl-optimization/59802
5341         * lcm.c (compute_available): Use inverted postorder to seed
5342         the initial worklist.
5344 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5346         PR target/59803
5347         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5348         ADDR_REGS for invalid symrefs in non-PIC code.
5350 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5352         PR other/58712
5353         * builtins.c (determine_block_size): Initialize *probable_max_size
5354         even if len_rtx is CONST_INT.
5356 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5358         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5359         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5360         (cortexa53_tunings): Likewise.
5361         (aarch64_sched_issue_rate): New function.
5362         (TARGET_SCHED_ISSUE_RATE): Define.
5364 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5366         * ira-costs.c (find_costs_and_classes): Add missed
5367         ira_init_register_move_cost_if_necessary.
5369 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5371         PR target/59787
5372         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5374 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5376         PR target/59794
5377         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5378         to indicate if type is used for function return value.  Warn ABI
5379         change if the vector mode isn't available for function return value.
5380         (ix86_function_arg_advance): Pass false to type_natural_mode.
5381         (ix86_function_arg): Likewise.
5382         (ix86_gimplify_va_arg): Likewise.
5383         (function_arg_32): Don't warn ABI change.
5384         (ix86_function_value): Pass true to type_natural_mode.
5385         (ix86_return_in_memory): Likewise.
5386         (ix86_struct_value_rtx): Removed.
5387         (TARGET_STRUCT_VALUE_RTX): Likewise.
5389 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5391         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5392         converting a conditional jump into a conditional return.
5394 2014-01-14  Richard Biener  <rguenther@suse.de>
5396         PR tree-optimization/58921
5397         PR tree-optimization/59006
5398         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5399         hoisting invariant stmts.
5400         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5401         invariant loads on the preheader edge if possible.
5403 2014-01-14  Joey Ye  <joey.ye@arm.com>
5405         * doc/plugin.texi (Building GCC plugins): Update to C++.
5407 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5409         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5410         (_mm_rcp28_round_ss): Ditto.
5411         (_mm_rsqrt28_round_sd): Ditto.
5412         (_mm_rsqrt28_round_ss): Ditto.
5413         (_mm_rcp28_sd): Ditto.
5414         (_mm_rcp28_ss): Ditto.
5415         (_mm_rsqrt28_sd): Ditto.
5416         (_mm_rsqrt28_ss): Ditto.
5417         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5418         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5419         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5420         (IX86_BUILTIN_RCP28SD): Ditto.
5421         (IX86_BUILTIN_RCP28SS): Ditto.
5422         (IX86_BUILTIN_RSQRT28SD): Ditto.
5423         (IX86_BUILTIN_RSQRT28SS): Ditto.
5424         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5425         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5426         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5427         (ix86_expand_special_args_builtin): Expand new FTYPE.
5428         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5429         (srcp14<mode>): Make insn unary.
5430         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5431         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5432         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5433         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5434         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5435         Fix rounding: make it SAE only.
5436         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5437         Ditto.
5438         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5439         Ditto.
5440         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5441         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5442         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5443         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5444         (round_saeonly_mask_scalar_operand4): Ditto.
5445         (round_saeonly_mask_scalar_op3): Ditto.
5446         (round_saeonly_mask_scalar_op4): Ditto.
5448 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5450         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5451         Implement -maltivec=be for vec_insert and vec_extract.
5453 2014-01-10  DJ Delorie  <dj@redhat.com>
5455         * config/msp430/msp430.md (call_internal): Don't allow memory
5456         references with SP as the base register.
5457         (call_value_internal): Likewise.
5458         * config/msp430/constraints.md (Yc): New.  For memory references
5459         that don't use SP as a base register.
5461         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5462         "an integer without a # prefix"
5463         * config/msp430/msp430.md (epilogue_helper): Use it.
5465 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5467         PR target/59617
5468         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5469         AVX512F gather builtins.
5470         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5471         on gather decls with INTEGER_TYPE masktype.
5472         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5473         directly into the builtin rather than hoisting it before loop.
5475         PR tree-optimization/59387
5476         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5477         (scev_const_prop): If folded_casts and type has undefined overflow,
5478         use force_gimple_operand instead of force_gimple_operand_gsi and
5479         for each added stmt if it is assign with
5480         arith_code_with_undefined_signed_overflow, call
5481         rewrite_to_defined_overflow.
5482         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5483         gimple-fold.h instead.
5484         (arith_code_with_undefined_signed_overflow,
5485         rewrite_to_defined_overflow): Moved to ...
5486         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5487         rewrite_to_defined_overflow): ... here.  No longer static.
5488         Include gimplify-me.h.
5489         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5490         rewrite_to_defined_overflow): New prototypes.
5492 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5494         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5496 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5498         * builtins.c (get_object_alignment_2): Minor tweak.
5499         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5501 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5503         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5504         optimized non constant lengths.
5506 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5508         PR libgomp/59194
5509         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5510         load as __atomic_load_N if possible.
5512 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5514         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5515         target parameter.
5516         (rs6000_expand_builtin): Adjust call.
5518 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5520         PR target/58115
5521         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5522         * config/rs6000/rs6000.c: Include target-globals.h.
5523         (rs6000_set_current_function): Instead of doing target_reinit
5524         unconditionally, use save_target_globals_default_opts and
5525         restore_target_globals.
5527         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5528         FPSCR.
5529         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5530         (rs6000_expand_builtin): Handle mffs and mtfsf.
5531         (rs6000_init_builtins): Define mffs and mtfsf.
5532         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5533         (rs6000_mffs): New pattern.
5534         (rs6000_mtfsf): New pattern.
5536 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5538         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5539         Start narrowing with START.  Apply candidate-use pair
5540         and check overall cost in narrowing.
5541         (iv_ca_prune): Pass new argument.
5543 2014-01-10  Jeff Law  <law@redhat.com>
5545         PR middle-end/59743
5546         * ree.c (combine_reaching_defs): Ensure the defining statement
5547         occurs before the extension when optimizing extensions with
5548         different source and destination hard registers.
5550 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5552         PR ipa/58585
5553         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5554         vtables into the type inheritance graph.
5556 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5558         PR rtl-optimization/59754
5559         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5560         modes in the REGNO != REGNO case.
5562 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5564         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5566 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5568         PR tree-optimization/59745
5569         * tree-predcom.c (tree_predictive_commoning_loop): Call
5570         free_affine_expand_cache if giving up because components is NULL.
5572         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5573         GC in payload of target_globals struct instead of allocating them on
5574         the heap and the larger structs separately using GC.
5575         * target-globals.h (struct target_globals): Make regs, hard_regs,
5576         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5577         of GTY((skip)) and change type to void *.
5578         (reset_target_globals): Cast loads from those fields to corresponding
5579         types.
5581 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5583         PR plugins/59335
5584         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5585         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5586         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5588 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5590         PR target/59744
5591         * aarch64-modes.def (CC_Zmode): New flags mode.
5592         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5593         represents an equality.
5594         (aarch64_get_condition_code): Handle CC_Zmode.
5595         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5597 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5599         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5600         extraction in good case.
5602 2014-01-10  Richard Biener  <rguenther@suse.de>
5604         PR tree-optimization/59374
5605         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5606         checking after SLP discovery.  Mark stmts not participating
5607         in any SLP instance properly.
5609 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5611         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5612         when handling a SET rtx.
5614 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5616         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5617         (cortex-a57): Likewise.
5618         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5620 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5622         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5623         non-iwmmxt builtins.
5625 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5627         PR ipa/58252
5628         PR ipa/59226
5629         * ipa-devirt.c record_target_from_binfo): Take as argument
5630         stack of binfos and lookup matching one for virtual inheritance.
5631         (possible_polymorphic_call_targets_1): Update.
5633 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5635         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5636         kernel strings for Loongson-2E/2F/3A.
5638 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5640         PR middle-end/59670
5641         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5642         is_gimple_call before calling gimple_call_internal_p.
5644 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5646         * Makefile.in (TREE_FLOW_H): Remove.
5647         (TREE_SSA_H): Add file names from tree-flow.h.
5648         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5649         * tree.h: Remove tree-flow.h reference.
5650         * hash-table.h: Remove tree-flow.h reference.
5651         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5652         reference with tree-ssa-loop.h.
5654 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5656         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5657         default element-order behavior for -maltivec.
5658         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5659         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5660         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5661         when targeting big endian, at least for now.
5662         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5664 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5666         PR middle-end/47735
5667         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5668         var satisfies use_register_for_decl, just take into account type
5669         alignment, rather than decl alignment.
5671         PR tree-optimization/59622
5672         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5673         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5674         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5675         Don't devirtualize for inplace at all.  For targets.length () == 1,
5676         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5678 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5680         * config/i386/i386.md (cpu): Remove the unused btver1.
5682 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5684         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5686 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5688         PR target/58115
5689         * tree-core.h (struct target_globals): New forward declaration.
5690         (struct tree_target_option): Add globals field.
5691         * tree.h (TREE_TARGET_GLOBALS): Define.
5692         (prepare_target_option_nodes_for_pch): New prototype.
5693         * target-globals.h (struct target_globals): Define even if
5694         !SWITCHABLE_TARGET.
5695         * tree.c (prepare_target_option_node_for_pch,
5696         prepare_target_option_nodes_for_pch): New functions.
5697         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5698         * config/i386/i386.c: Include target-globals.h.
5699         (ix86_set_current_function): Instead of doing target_reinit
5700         unconditionally, use save_target_globals_default_opts and
5701         restore_target_globals.
5703 2014-01-09  Richard Biener  <rguenther@suse.de>
5705         PR tree-optimization/59715
5706         * tree-cfg.h (split_critical_edges): Declare.
5707         * tree-cfg.c (split_critical_edges): Export.
5708         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5710 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5712         * cfgexpand.c (expand_stack_vars): Optionally disable
5713         asan stack protection.
5714         (expand_used_vars): Likewise.
5715         (partition_stack_vars): Likewise.
5716         * asan.c (asan_emit_stack_protection): Optionally disable
5717         after return stack usage.
5718         (instrument_derefs): Optionally disable memory access instrumentation.
5719         (instrument_builtin_call): Likewise.
5720         (instrument_strlen_call): Likewise.
5721         (asan_protect_global): Optionally disable global variables protection.
5722         * doc/invoke.texi: Added doc for new options.
5723         * params.def: Added new options.
5724         * params.h: Likewise.
5726 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5728         PR rtl-optimization/59724
5729         * ifcvt.c (cond_exec_process_if_block): Don't call
5730         flow_find_head_matching_sequence with 0 longest_match.
5731         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5732         non-active insns if !stop_after.
5733         (try_head_merge_bb): Revert 2014-01-07 changes.
5735 2014-01-08  Jeff Law  <law@redhat.com>
5737         * ree.c (get_sub_rtx): New function, extracted from...
5738         (merge_def_and_ext): Here.
5739         (combine_reaching_defs): Use get_sub_rtx.
5741 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5743         * cgraph.h (varpool_variable_node): Do not choke on null node.
5745 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5747         * config/mips/mips.md (simple_return): Attempt to use JRC
5748         for microMIPS.
5749         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5751 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5753         PR rtl-optimization/59137
5754         * reorg.c (steal_delay_list_from_target): Call update_block for
5755         elided insns.
5756         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5758 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5760         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5761         two duplicate entries.
5763 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5765         Revert:
5766         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5768         * config/mips/mips.c (mips_truncated_op_cost): New function.
5769         (mips_rtx_costs): Adjust test for BADDU.
5770         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5772         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5774         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5775         (*baddu_si): ...this new pattern.
5777 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5779         PR ipa/59722
5780         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5782 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5784         PR middle-end/57748
5785         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5786         inner_reference_p.
5787         (expand_expr, expand_normal): Adjust.
5788         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5789         inner_reference_p. Use inner_reference_p to expand inner references.
5790         (store_expr): Adjust.
5791         * cfgexpand.c (expand_call_stmt): Adjust.
5793 2014-01-08  Rong Xu  <xur@google.com>
5795         * gcov-io.c (gcov_var): Move from gcov-io.h.
5796         (gcov_position): Ditto.
5797         (gcov_is_error): Ditto.
5798         (gcov_rewrite): Ditto.
5799         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5800         only part to libgcc/libgcov.h.
5802 2014-01-08  Marek Polacek  <polacek@redhat.com>
5804         PR middle-end/59669
5805         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5807 2014-01-08  Marek Polacek  <polacek@redhat.com>
5809         PR sanitizer/59667
5810         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5812 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5814         PR rtl-optimization/59649
5815         * stor-layout.c (get_mode_bounds): For BImode return
5816         0 and STORE_FLAG_VALUE.
5818 2014-01-08  Richard Biener  <rguenther@suse.de>
5820         PR middle-end/59630
5821         * gimple.h (is_gimple_builtin_call): Remove.
5822         (gimple_builtin_call_types_compatible_p): New.
5823         (gimple_call_builtin_p): New overload.
5824         * gimple.c (is_gimple_builtin_call): Remove.
5825         (validate_call): Rename to ...
5826         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5827         check return types.
5828         (validate_type): New static function.
5829         (gimple_call_builtin_p): New overload and adjust.
5830         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5831         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5832         (gimple_fold_stmt_to_constant_1): Likewise.
5833         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5835 2014-01-08  Richard Biener  <rguenther@suse.de>
5837         PR middle-end/59471
5838         * gimplify.c (gimplify_expr): Gimplify register-register type
5839         VIEW_CONVERT_EXPRs to separate stmts.
5841 2014-01-07  Jeff Law  <law@redhat.com>
5843         PR middle-end/53623
5844         * ree.c (combine_set_extension): Handle case where source
5845         and destination registers in an extension insn are different.
5846         (combine_reaching_defs): Allow source and destination registers
5847         in extension to be different under limited circumstances.
5848         (add_removable_extension): Remove restriction that the
5849         source and destination registers in the extension are the same.
5850         (find_and_remove_re): Emit a copy from the extension's
5851         destination to its source after the defining insn if
5852         the source and destination registers are different.
5854         PR middle-end/59285
5855         * ifcvt.c (merge_if_block): If we are merging a block with more than
5856         one successor with a block with no successors, remove any BARRIER
5857         after the second block.
5859 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
5861         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5863 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5865         PR target/59652
5866         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5867         for 14-bit register offsets when INT14_OK_STRICT is false.
5869 2014-01-07  Roland Stigge  <stigge@antcom.de>
5870             Michael Meissner  <meissner@linux.vnet.ibm.com>
5872         PR 57386/target
5873         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5874         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5876 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5878         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5879         -mcpu.
5881 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5883         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5884         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5885         rtx is const0_rtx or not.
5887 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5889         PR target/58115
5890         * target-globals.c (save_target_globals): Remove this_fn_optab
5891         handling.
5892         * toplev.c: Include optabs.h.
5893         (target_reinit): Temporarily restore the global options if another
5894         set of options are in force.
5896 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5898         PR rtl-optimization/58668
5899         * cfgcleanup.c (flow_find_cross_jump): Don't count
5900         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5901         to determine what is counted.
5902         (flow_find_head_matching_sequence): Use active_insn_p to determine
5903         what is counted.
5904         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5905         counting change.
5906         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5907         determine what is counted.
5909         PR tree-optimization/59643
5910         * tree-predcom.c (split_data_refs_to_components): If one dr is
5911         read and one write, determine_offset fails and the write isn't
5912         in the bad component, just put the read into the bad component.
5914 2014-01-07  Mike Stump  <mikestump@comcast.net>
5915             Jakub Jelinek  <jakub@redhat.com>
5917         PR pch/59436
5918         * tree-core.h (struct tree_optimization_option): Change optabs
5919         type from unsigned char * to void *.
5920         * optabs.c (init_tree_optimization_optabs): Adjust
5921         TREE_OPTIMIZATION_OPTABS initialization.
5923 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5925         PR target/59644
5926         * config/i386/i386.h (struct machine_function): Add
5927         no_drap_save_restore field.
5928         * config/i386/i386.c (ix86_save_reg): Use
5929         !cfun->machine->no_drap_save_restore instead of
5930         crtl->stack_realign_needed.
5931         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5932         this function clears frame_pointer_needed.  Set
5933         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5934         and DRAP reg is needed.
5936 2014-01-06  Marek Polacek  <polacek@redhat.com>
5938         PR c/57773
5939         * doc/implement-c.texi: Mention that other integer types are
5940         permitted as bit-field types in strictly conforming mode.
5942 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5944         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5945         is newly allocated.
5947 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5949         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5951 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5953         PR ipa/59008
5954         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5955         to int.
5956         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5958 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5960         PR debug/59350
5961         PR debug/59510
5962         * var-tracking.c (add_stores): Preserve the value of the source even if
5963         we don't record the store.
5965 2014-01-06  Terry Guo  <terry.guo@arm.com>
5967         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5969 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5971         PR bootstrap/59541
5972         * config/darwin.c (darwin_function_section): Adjust return values to
5973         correspond to optimisation changes made in r206070.
5975 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5977         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5978         from prefetch_block tune setting.
5979         (nocona_cost): Correct size of prefetch block to 64.
5981 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5983         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5984         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5985         used to save the static chain register in the computation of the offset
5986         from which the FP registers need to be restored.
5988 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5990         PR tree-optimization/59519
5991         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5992         ICE if get_current_def (current_new_name) is already non-NULL, as long
5993         as it is a phi result of some other phi in *new_exit_bb that has
5994         the same argument.
5996         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5997         or vmovdqu* for misaligned_operand.
5998         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5999         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
6000         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
6001         aligned_mem for AVX512F masked aligned load and store builtins and for
6002         non-temporal moves.
6004 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
6006         PR tree-optimization/59651
6007         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
6008         Address range for negative step should be added by TYPE_SIZE_UNIT.
6010 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6012         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6013         overlapping registers also for registers other than the stack pointer.
6015 2014-01-03  Marek Polacek  <polacek@redhat.com>
6017         PR other/59661
6018         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6019         __builtin_FILE.
6021 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6023         PR target/59625
6024         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6025         asm goto as jump.
6027         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6028         (push splitter): Use <P:MODE_SIZE> instead of
6029         GET_MODE_SIZE (<P:MODE>mode).
6030         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6031         (mov -1, reg peephole2): Likewise.
6032         * config/i386/sse.md (*mov<mode>_internal,
6033         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6034         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6035         *<code><mode>3, *andnot<mode>3<mask_name>,
6036         <mask_codefor><code><mode>3<mask_name>): Likewise.
6037         * config/i386/subst.md (mask_mode512bit_condition,
6038         sd_mask_mode512bit_condition): Likewise.
6040 2014-01-02  Xinliang David Li  <davidxl@google.com>
6042         PR tree-optimization/59303
6043         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6044         (dump_predicates): Better output format.
6045         (pred_equal_p): New function.
6046         (is_neq_relop_p): Ditto.
6047         (is_neq_zero_form_p): Ditto.
6048         (pred_expr_equal_p): Ditto.
6049         (pred_neg_p): Ditto.
6050         (simplify_pred): Ditto.
6051         (simplify_preds_2): Ditto.
6052         (simplify_preds_3): Ditto.
6053         (simplify_preds_4): Ditto.
6054         (simplify_preds): Ditto.
6055         (push_pred): Ditto.
6056         (push_to_worklist): Ditto.
6057         (get_pred_info_from_cmp): Ditto.
6058         (is_degenerated_phi): Ditto.
6059         (normalize_one_pred_1): Ditto.
6060         (normalize_one_pred): Ditto.
6061         (normalize_one_pred_chain): Ditto.
6062         (normalize_preds): Ditto.
6063         (normalize_cond_1): Remove function.
6064         (normalize_cond): Ditto.
6065         (is_gcond_subset_of): Ditto.
6066         (is_subset_of_any): Ditto.
6067         (is_or_set_subset_of): Ditto.
6068         (is_and_set_subset_of): Ditto.
6069         (is_norm_cond_subset_of): Ditto.
6070         (pred_chain_length_cmp): Ditto.
6071         (convert_control_dep_chain_into_preds): Type change.
6072         (find_predicates): Ditto.
6073         (find_def_preds): Ditto.
6074         (destroy_predicates_vecs): Ditto.
6075         (find_matching_predicates_in_rest_chains): Ditto.
6076         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6077         (is_pred_expr_subset): Ditto.
6078         (is_pred_chain_subset_of): Ditto.
6079         (is_included_in): Ditto.
6080         (is_superset_of): Ditto.
6082 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6084         Update copyright years.
6086 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6088         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6089         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6090         config/arc/arc.md, config/arc/arc.opt,
6091         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6092         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6093         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6094         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6095         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6096         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6097         vtable-verify.c, vtable-verify.h: Use the standard form for the
6098         copyright notice.
6100 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6102         * gcc.c (process_command): Update copyright notice dates.
6103         * gcov-dump.c: Ditto.
6104         * gcov.c: Ditto.
6105         * doc/cpp.texi: Bump @copying's copyright year.
6106         * doc/cppinternals.texi: Ditto.
6107         * doc/gcc.texi: Ditto.
6108         * doc/gccint.texi: Ditto.
6109         * doc/gcov.texi: Ditto.
6110         * doc/install.texi: Ditto.
6111         * doc/invoke.texi: Ditto.
6113 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6115         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6117 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6119         * config/i386/sse.md (*mov<mode>_internal): Guard
6120         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6122         PR rtl-optimization/59647
6123         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6124         new_rtx into UNSIGNED_FLOAT rtxes.
6126 Copyright (C) 2014 Free Software Foundation, Inc.
6128 Copying and distribution of this file, with or without modification,
6129 are permitted in any medium without royalty provided the copyright
6130 notice and this notice are preserved.