2014-04-07 Charles Baylis <charles.baylis@linaro.org>
[official-gcc.git] / gcc / ChangeLog
blob12fc17fdc20d08670600c954fa5a65d20e99dafd
1 2014-04-07  Charles Baylis  <charles.baylis@linaro.org>
3         PR target/60609
4         * config/arm/arm.h (ASM_OUTPUT_CASE_END): Remove.
5         (LABEL_ALIGN_AFTER_BARRIER): Align barriers which occur after
6         ADDR_DIFF_VEC.
8 2014-04-07  Richard Biener  <rguenther@suse.de>
10         PR tree-optimization/60766
11         * tree-ssa-loop-ivopts.c (cand_value_at): Compute in an
12         unsigned type.
13         (may_eliminate_iv): Convert cand_value_at result to desired
14         type.
16 2014-04-07  Jason Merrill  <jason@redhat.com>
18         PR c++/60731
19         * common.opt (-fno-gnu-unique): Add.
20         * config/elfos.h (USE_GNU_UNIQUE_OBJECT): Check it.
22 2014-04-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24         * haifa-sched.c: Fix outdated function reference and minor
25         grammar errors in introductory comment.
27 2014-04-07  Richard Biener  <rguenther@suse.de>
29         PR middle-end/60750
30         * tree-ssa-operands.c (maybe_add_call_vops): Also add VDEFs
31         for noreturn calls.
32         * tree-cfgcleanup.c (fixup_noreturn_call): Do not remove VDEFs.
34 2014-04-06  John David Anglin  <danglin@gcc.gnu.org>
36         PR debug/55794
37         * config/pa/pa.c (pa_output_function_epilogue): Skip address and code
38         size accounting for thunks.
39         (pa_asm_output_mi_thunk): Use final_start_function() and
40         final_end_function() to output function start and end directives.
42 2014-04-05  Pitchumani Sivanupandi  <Pitchumani.S@atmel.com>
44         * config/avr/avr-arch.h (avr_mcu_t): Add dev_attribute field to have device
45         specific ISA/ feature information. Remove short_sp and errata_skip ds.
46         Add avr_device_specific_features enum to have device specific info.
47         * config/avr/avr-c.c (avr_cpu_cpp_builtins): use dev_attribute to check
48         errata_skip. Add __AVR_ISA_RMW__ builtin macro if RMW ISA available.
49         * config/avr/avr-devices.c (avr_mcu_types): Update AVR_MCU macro for
50         updated device specific info.
51         * config/avr/avr-mcus.def: Merge device specific details to
52         dev_attribute field.
53         * config/avr/avr.c (avr_2word_insn_p): use dev_attribute field to check
54         errata_skip.
55         * config/avr/avr.h (AVR_HAVE_8BIT_SP): same for short sp info.
56         * config/avr/driver-avr.c (avr_device_to_as): Pass -mrmw option to
57         assembler if RMW isa supported by current device.
58         * config/avr/genmultilib.awk: Update as device info structure changed.
59         * doc/invoke.texi: Add info for __AVR_ISA_RMW__ builtin macro
61 2014-04-04  Cong Hou  <congh@google.com>
63         PR tree-optimization/60656
64         * tree-vect-stmts.c (supportable_widening_operation):
65         Fix a bug that elements in a vector with vect_used_by_reduction
66         property are incorrectly reordered when the operation on it is not
67         consistant with the one in reduction operation.
69 2014-04-04  John David Anglin  <danglin@gcc.gnu.org>
71         PR rtl-optimization/60155
72         * gcse.c (record_set_data): New function.
73         (single_set_gcse): New function.
74         (gcse_emit_move_after): Use single_set_gcse instead of single_set.
75         (hoist_code): Likewise.
76         (get_pressure_class_and_nregs): Likewise.
78 2014-04-04  Eric Botcazou  <ebotcazou@adacore.com>
80         * explow.c (probe_stack_range): Emit a final optimization blockage.
82 2014-04-04  Anthony Green  <green@moxielogic.com>
84         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2): Fix
85         typos.
87 2014-04-04  Jan Hubicka  <hubicka@ucw.cz>
89         PR ipa/59626
90         * lto-cgraph.c (input_overwrite_node): Check that partitioning
91         flags are set only during streaming.
92         * ipa.c (process_references, walk_polymorphic_call_targets,
93         symtab_remove_unreachable_nodes): Drop bodies of always inline
94         after early inlining.
95         (symtab_remove_unreachable_nodes): Remove always_inline attribute.
97 2014-04-04  Jakub Jelinek  <jakub@redhat.com>
98         Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
100         PR debug/60655
101         * dwarf2out.c (const_ok_for_output_1): Reject expressions
102         containing a NOT.
104 2014-04-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
106         PR bootstrap/60743
107         * config/arm/cortex-a53.md (cortex_a53_fdivs): Reduce reservation
108         duration.
109         (cortex_a53_fdivd): Likewise.
111 2014-04-04  Martin Jambor  <mjambor@suse.cz>
113         PR ipa/60640
114         * cgraph.h (cgraph_clone_node): New parameter added to declaration.
115         Adjust all callers.
116         * cgraph.c (clone_of_p): Also return true if thunks match.
117         (verify_edge_corresponds_to_fndecl): Removed extraneous call to
118         cgraph_function_or_thunk_node and an obsolete comment.
119         * cgraphclones.c (build_function_type_skip_args): Moved upwards in the
120         file.
121         (build_function_decl_skip_args): Likewise.
122         (set_new_clone_decl_and_node_flags): New function.
123         (duplicate_thunk_for_node): Likewise.
124         (redirect_edge_duplicating_thunks): Likewise.
125         (cgraph_clone_node): New parameter args_to_skip, pass it to
126         redirect_edge_duplicating_thunks which is called instead of
127         cgraph_redirect_edge_callee.
128         (cgraph_create_virtual_clone): Pass args_to_skip to cgraph_clone_node,
129         moved setting of a lot of flags to set_new_clone_decl_and_node_flags.
131 2014-04-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
133         PR fortran/60191
134         * fortran/trans-types.c (gfc_get_function_type): In case of recursion
135         build a variadic function type with empty argument list instead of a
136         stdarg-like function type with incomplete argument list.
138 2014-04-04  Jeff Law  <law@redhat.com>
140         PR target/60657
141         * config/arm/predicates.md (const_int_I_operand): New predicate.
142         (const_int_M_operand): Similarly.
143         * config/arm/arm.md (insv_zero): Use const_int_M_operand instead of
144         const_int_operand.
145         (insv_t2, extv_reg, extzv_t2): Likewise.
146         (load_multiple_with_writeback): Similarly for const_int_I_operand.
147         (pop_multiple_with_writeback_and_return): Likewise.
148         (vfp_pop_multiple_with_writeback): Likewise
150 2014-04-04  Richard Biener  <rguenther@suse.de>
152         PR ipa/60746
153         * tree-ssanames.c (make_ssa_name_fn): Fix assert.
154         * gimple.c (gimple_set_bb): Avoid ICEing for NULL cfun for
155         non-GIMPLE_LABELs.
156         * gimplify.h (gimple_add_tmp_var_fn): Declare.
157         * gimplify.c (gimple_add_tmp_var_fn): New function.
158         * gimple-expr.h (create_tmp_reg_fn): Declare.
159         * gimple-expr.c (create_tmp_reg_fn): New function.
160         * gimple-low.c (record_vars_into): Don't change cfun.
161         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Fix
162         code generation without cfun.
164 2014-04-04  Thomas Schwinge  <thomas@codesourcery.com>
166         PR bootstrap/60719
167         * Makefile.in (install-driver): Fix shell scripting.
169 2014-04-03  Cong Hou  <congh@google.com>
171         PR tree-optimization/60505
172         * tree-vectorizer.h (struct _stmt_vec_info): Add th field as the
173         threshold of number of iterations below which no vectorization will be
174         done.
175         * tree-vect-loop.c (new_loop_vec_info):
176         Initialize LOOP_VINFO_COST_MODEL_THRESHOLD.
177         * tree-vect-loop.c (vect_analyze_loop_operations):
178         Set LOOP_VINFO_COST_MODEL_THRESHOLD.
179         * tree-vect-loop.c (vect_transform_loop):
180         Use LOOP_VINFO_COST_MODEL_THRESHOLD.
181         * tree-vect-loop.c (vect_analyze_loop_2): Check the maximum number
182         of iterations of the loop and see if we should build the epilogue.
184 2014-04-03  Richard Biener  <rguenther@suse.de>
186         * tree-streamer.h (struct streamer_tree_cache_d): Add next_idx
187         member.
188         (streamer_tree_cache_create): Adjust.
189         * tree-streamer.c (streamer_tree_cache_add_to_node_array): Adjust
190         to allow optional nodes array.
191         (streamer_tree_cache_insert_1): Use next_idx to assign idx.
192         (streamer_tree_cache_append): Likewise.
193         (streamer_tree_cache_create): Create nodes array optionally
194         as specified by parameter.
195         * lto-streamer-out.c (create_output_block): Avoid maintaining
196         the node array in the writer cache.
197         (DFS_write_tree): Remove assertion.
198         (produce_asm_for_decls): Free the out decl state hash table
199         early.
200         * lto-streamer-in.c (lto_data_in_create): Adjust for
201         streamer_tree_cache_create prototype change.
203 2014-04-03  Richard Biener  <rguenther@suse.de>
205         * tree-streamer-out.c (streamer_write_chain): Do not temporarily
206         set TREE_CHAIN to NULL_TREE.
208 2014-04-03  Richard Biener  <rguenther@suse.de>
210         PR tree-optimization/60740
211         * graphite-scop-detection.c (stmt_simple_for_scop_p): Iterate
212         over all GIMPLE_COND operands.
214 2014-04-03  Nathan Sidwell  <nathan@codesourcery.com>
216         * doc/invoke.texi (Wnon-virtual-dtor): Adjust documentation.
217         (Weffc++): Remove Scott's numbering, merge lists and reference
218         Wnon-virtual-dtor.
220         c-family/
222         cp/
223         * class.c (accessible_nvdtor_p): New.
224         (check_bases): Don't check base destructor here ...
225         (check_bases_and_members): ... check them here.  Trigger on
226         Wnon-virtual-dtor flag.
227         (finish_struct_1): Use accessible_nvdtor_p.
229         testsuite/
230         * g++.dg/warn/Wnvdtor.C: Add non-polymorphic case.
231         * g++.dg/warn/Wnvdtor-2.C: New.
232         * g++.dg/warn/Wnvdtor-3.C: New.
233         * g++.dg/warn/Wnvdtor-4.C: New.
234         * g++.dg/warn/Weff1.C: Delete.
235         * g++.old-deja/g++.benjamin/15309-1.C: Delete.
236         * g++.old-deja/g++.benjamin/15309-2.C: Delete.
238 2014-04-03  Nick Clifton  <nickc@redhat.com>
240         * config/rl78/rl78-expand.md (movqi): Handle (SUBREG (SYMBOL_REF))
241         properly.
243 2014-04-03  Martin Jambor  <mjambor@suse.cz>
245         * ipa-cp.c (ipcp_verify_propagated_values): Also dump symtab and
246         mention gcc_unreachable before failing.
247         * ipa.c (symtab_remove_unreachable_nodes): Also print order of
248         removed symbols.
250 2014-04-02  Jan Hubicka  <hubicka@ucw.cz>
252         PR ipa/60659
253         * ipa-devirt.c (get_polymorphic_call_info): Do not ICE on type inconsistent
254         code and instead mark the context inconsistent.
255         (possible_polymorphic_call_targets): For inconsistent contexts
256         return empty complete list.
258 2014-04-02  Anthony Green  <green@moxielogic.com>
260         * config/moxie/moxie.md (zero_extendqisi2, zero_extendhisi2)
261         (extendqisi2, extendhisi2): Define.
262         * config/moxie/moxie.h (DEFAULT_SIGNED_CHAR): Change to 0.
263         (WCHAR_TYPE): Change to unsigned int.
265 2014-04-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
267         PR tree-optimization/60733
268         * gimple-ssa-strength-reduction.c (ncd_with_phi): Change required
269         insertion point for PHI candidates to be the end of the feeding
270         block for the PHI argument.
272 2014-04-02  Vladimir Makarov  <vmakarov@redhat.com>
274         PR rtl-optimization/60650
275         * lra-constraints.c (process_alt_operands): Decrease reject for
276         earlyclobber matching.
278 2014-04-02  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
280         * config/s390/s390.c (s390_expand_insv): Use GET_MODE_BITSIZE.
282 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
284         * config/spu/spu.c (pad_bb): Do not crash when the last
285         insn is CODE_FOR_blockage.
287 2014-04-02  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
289         * config/spu/spu.md ("insv"): Fail if bitoffset+bitsize
290         lies outside the target mode.
292 2014-04-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
294         PR target/60735
295         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): If we have
296         software floating point or no floating point registers, do not
297         allow any type in the FPRs.  Eliminate a test for SPE SIMD types
298         in GPRs that occurs after we tested for GPRs that would never be
299         true.
301         * config/rs6000/rs6000.md (mov<mode>_softfloat32, FMOVE64):
302         Rewrite tests to use TARGET_DOUBLE_FLOAT and TARGET_E500_DOUBLE,
303         since the FMOVE64 type is DFmode/DDmode.  If TARGET_E500_DOUBLE,
304         specifically allow DDmode, since that does not use the SPE SIMD
305         instructions.
307 2014-04-02  Richard Biener  <rguenther@suse.de>
309         PR middle-end/60729
310         * optabs.c (expand_abs_nojump): Honor flag_trapv only for
311         MODE_INTs.  Properly use negv_optab.
312         (expand_abs): Likewise.
314 2014-04-02  Richard Biener  <rguenther@suse.de>
316         PR bootstrap/60719
317         * Makefile.in (install-driver): Guard extra installs with special
318         names properly.
320 2014-04-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
322         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
323         Document vec_vgbbd.
325 2014-04-01  Richard Henderson  <rth@redhat.com>
327         PR target/60704
328         * config/i386/i386.md (*float<SWI48><MODEF>2_sse): Leave the second
329         alternative enabled before register allocation.
331 2014-04-01  Chung-Lin Tang  <cltang@codesourcery.com>
333         * config/nios2/nios2.md (unspec): Remove UNSPEC_TLS, UNSPEC_TLS_LDM.
334         * config/nios2/nios2.c (nios2_function_profiler): Fix addi operand
335         typo.
336         (nios2_large_got_address): Remove unneeded 'sym' parameter.
337         (nios2_got_address): Update nios2_large_got_address call site.
338         (nios2_delegitimize_address): New function.
339         (TARGET_DELEGITIMIZE_ADDRESS): Define to nios2_delegitimize_address.
340         * config/nios2/linux.h (GLIBC_DYNAMIC_LINKER): Define.
341         (LINK_SPEC): Specify dynamic linker using GNU_USER_DYNAMIC_LINKER.
343 2014-04-01  Martin Husemann  <martin@duskware.de>
345         * config/mips/netbsd.h (TARGET_OS_CPP_BUILTINS): Define __mips_o32
346         for -mabi=32.
348 2014-04-01  Richard Sandiford  <rdsandiford@googlemail.com>
350         PR rtl-optimization/60604
351         * recog.c (general_operand): Incorporate REG_CANNOT_CHANGE_MODE_P
352         check from register_operand.
353         (register_operand): Redefine in terms of general_operand.
354         (nonmemory_operand): Use register_operand for the non-constant cases.
356 2014-04-01  Richard Biener  <rguenther@suse.de>
358         * gimple.h (struct gimple_statement_base): Align subcode to
359         16 bits.
361 2014-04-01  Sebastian Huber  <sebastian.huber@embedded-brains.de>
363         * doc/invoke.texi (mapp-regs): Clarify.
365 2014-03-31  Ulrich Drepper  <drepper@gmail.com>
367         * config/i386/avx512fintrin.h (__v32hi): Define type.
368         (__v64qi): Likewise.
369         (_mm512_set1_epi8): Define.
370         (_mm512_set1_epi16): Define.
371         (_mm512_set4_epi32): Define.
372         (_mm512_set4_epi64): Define.
373         (_mm512_set4_pd): Define.
374         (_mm512_set4_ps): Define.
375         (_mm512_setr4_epi64): Define.
376         (_mm512_setr4_epi32): Define.
377         (_mm512_setr4_pd): Define.
378         (_mm512_setr4_ps): Define.
379         (_mm512_setzero_epi32): Define.
381 2014-03-31  Martin Jambor  <mjambor@suse.cz>
383         PR middle-end/60647
384         * tree-sra.c (callsite_has_enough_arguments_p): Renamed to
385         callsite_arguments_match_p.  Updated all callers.  Also check types of
386         corresponding formal parameters and actual arguments.
387         (not_all_callers_have_enough_arguments_p) Renamed to
388         some_callers_have_mismatched_arguments_p.
390 2014-03-31  Yuri Rumyantsev  <ysrumyan@gmail.com>
392         * tree-inline.c (copy_loops): Add missed copy of 'safelen'.
394 2014-03-31  Kugan Vivekanandarajah  <kuganv@linaro.org>
396         PR target/60034
397         * aarch64/aarch64.c (aarch64_classify_address): Fix alignment for
398         section anchor.
400 2014-03-30  Uros Bizjak  <ubizjak@gmail.com>
402         * config/i386/sse.md (FMAMODE_NOVF512): New mode iterator.
403         (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
404         Split out
405         <sd_mask_codefor>fma_fmadd_<VF_512:mode><sd_maskz_name><round_name>.
406         Use FMAMODE_NOVF512 mode iterator.
407         (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>): Ditto.
408         (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>): Ditto.
409         (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>): Ditto.
410         (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
411         Split out
412         <sd_mask_codefor>fma_fmaddsub_<VF_512:mode><sd_maskz_name><round_name>.
413         Use VF_128_256 mode iterator.
414         (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
415         Ditto.
417 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
419         * cgraph.c (cgraph_redirect_edge_call_stmt_to_callee): Clear
420         static chain if needed.
422 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
424         PR target/60697
425         * lra-constraints.c (index_part_to_reg): New.
426         (process_address): Use it.
428 2014-03-27  Jeff Law  <law@redhat.com>
429             Jakub Jelinek  <jakub@redhat.com>
431         PR target/60648
432         * expr.c (do_tablejump): Use simplify_gen_binary rather than
433         gen_rtx_{PLUS,MULT} to build up the address expression.
435         * i386/i386.c (ix86_legitimize_address): Use copy_addr_to_reg to avoid
436         creating non-canonical RTL.
438 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
440         PR ipa/60243
441         * ipa-inline.c (want_inline_small_function_p): Short circuit large
442         functions; reorganize to make cheap checks first.
443         (inline_small_functions): Do not estimate growth when dumping;
444         it is expensive.
445         * ipa-inline.h (inline_summary): Add min_size.
446         (growth_likely_positive): New function.
447         * ipa-inline-analysis.c (dump_inline_summary): Add min_size.
448         (set_cond_stmt_execution_predicate): Cleanup.
449         (estimate_edge_size_and_time): Compute min_size.
450         (estimate_calls_size_and_time): Likewise.
451         (estimate_node_size_and_time): Likewise.
452         (inline_update_overall_summary): Update min_size.
453         (do_estimate_edge_time): Likewise.
454         (do_estimate_edge_size): Update.
455         (do_estimate_edge_hints): Update.
456         (growth_likely_positive): New function.
458 2014-03-28  Jakub Jelinek  <jakub@redhat.com>
460         PR target/60693
461         * config/i386/i386.c (ix86_copy_addr_to_reg): Call copy_addr_to_reg
462         also if addr has VOIDmode.
464 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
466         * config/arm/aarch-common.c (aarch_crypto_can_dual_issue): New.
467         * config/arm/aarch-common-protos.h (aarch_crypto_can_dual_issue):
468         Declare extern.
469         * config/arm/cortex-a53.md: Add reservations and bypass for crypto
470         instructions as well as AdvancedSIMD loads.
472 2014-03-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
474         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
475         Use crypto_aese type.
476         (aarch64_crypto_aes<aesmc_op>v16qi): Use crypto_aesmc type.
477         * config/arm/arm.md (is_neon_type): Replace crypto_aes with
478         crypto_aese, crypto_aesmc.  Move to types.md.
479         * config/arm/types.md (crypto_aes): Split into crypto_aese,
480         crypto_aesmc.
481         * config/arm/iterators.md (crypto_type): Likewise.
483 2014-03-28  Jan Hubicka  <hubicka@ucw.cz>
485         * cgraph.c: Include expr.h and tree-dfa.h.
486         (cgraph_redirect_edge_call_stmt_to_callee): If call in noreturn;
487         remove LHS.
489 2014-03-28  Vladimir Makarov  <vmakarov@redhat.com>
491         PR target/60675
492         * lra-assigns.c (find_hard_regno_for): Remove unavailable hard
493         regs from checking multi-reg pseudos.
495 2014-03-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
497         * config/arm/t-aprofile (MULTILIB_MATCHES): Correct A12 rule.
499 2014-03-28  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
501         * config/rs6000/rs6000.c (fusion_gpr_load_p): Refuse optimization
502         if it would clobber the stack pointer, even temporarily.
504 2014-03-28  Eric Botcazou  <ebotcazou@adacore.com>
506         * mode-switching.c: Make small adjustments to the top comment.
508 2014-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
510         * config/rs6000/constraints.md (wD constraint): New constraint to
511         match the constant integer to get the top DImode/DFmode out of a
512         vector in a VSX register.
514         * config/rs6000/predicates.md (vsx_scalar_64bit): New predicate to
515         match the constant integer to get the top DImode/DFmode out of a
516         vector in a VSX register.
518         * config/rs6000/rs6000-builtins.def (VBPERMQ): Add vbpermq builtin
519         for ISA 2.07.
521         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
522         vbpermq builtins.
524         * config/rs6000/rs6000.c (rs6000_debug_reg_global): If
525         -mdebug=reg, print value of VECTOR_ELEMENT_SCALAR_64BIT.
527         * config/rs6000/vsx.md (vsx_extract_<mode>, V2DI/V2DF modes):
528         Optimize vec_extract of 64-bit values, where the value being
529         extracted is in the top word, where we can use scalar
530         instructions.  Add direct move and store support.  Combine the big
531         endian/little endian vector select load support into a single insn.
532         (vsx_extract_<mode>_internal1): Likewise.
533         (vsx_extract_<mode>_internal2): Likewise.
534         (vsx_extract_<mode>_load): Likewise.
535         (vsx_extract_<mode>_store): Likewise.
536         (vsx_extract_<mode>_zero): Delete, big and little endian insns are
537         combined into vsx_extract_<mode>_load.
538         (vsx_extract_<mode>_one_le): Likewise.
540         * config/rs6000/rs6000.h (VECTOR_ELEMENT_SCALAR_64BIT): Macro to
541         define the top 64-bit vector element.
543         * doc/md.texi (PowerPC and IBM RS6000 constraints): Document wD
544         constraint.
546         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
547         Document vec_vbpermq builtin.
549         PR target/60672
550         * config/rs6000/altivec.h (vec_xxsldwi): Add missing define to
551         enable use of xxsldwi and xxpermdi builtin functions.
552         (vec_xxpermdi): Likewise.
554         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
555         Document use of vec_xxsldwi and vec_xxpermdi builtins.
557 2014-03-27  Vladimir Makarov  <vmakarov@redhat.com>
559         PR rtl-optimization/60650
560         * lra-assign.c (find_hard_regno_for, spill_for): Add parameter
561         first_p.  Use it.
562         (find_spills_for): New.
563         (assign_by_spills): Pass the new parameter to find_hard_regno_for.
564         Spill all pseudos on the second iteration.
566 2014-03-27  Marek Polacek  <polacek@redhat.com>
568         PR c/50347
569         * doc/extend.texi (ffs Builtins): Change unsigned types to signed
570         types.
572 2014-03-27  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
574         * config/s390/s390.c (s390_can_use_return_insn): Check for
575         call-saved FPRs on 31 bit.
577 2014-03-27  Jakub Jelinek  <jakub@redhat.com>
579         PR middle-end/60682
580         * omp-low.c (lower_omp_1): For gimple_clobber_p stmts,
581         if they need regimplification, just drop them instead of
582         calling gimple_regimplify_operands on them.
584 2014-03-27  Marcus Shawcroft  <marcus.shawcroft@arm.com>
586         PR target/60580
587         * config/aarch64/aarch64.c (faked_omit_frame_pointer): Remove.
588         (aarch64_frame_pointer_required): Adjust logic.
589         (aarch64_can_eliminate): Adjust logic.
590         (aarch64_override_options_after_change): Adjust logic.
592 2014-03-27  Dehao Chen  <dehao@google.com>
594         * ipa-inline.c (early_inliner): Update node's inline info.
596 2014-03-26  Dehao Chen  <dehao@google.com>
598         * dojump.c (do_compare_rtx_and_jump): Sets correct probability for
599         compiler inserted conditional jumps for NAN float check.
601 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
603         * ubsan.h (ubsan_create_data): Change second argument's type
604         to const location_t *.
605         * ubsan.c (ubsan_source_location): If xloc.file is NULL, set it to
606         _("<unknown>").
607         (ubsan_create_data): Change second argument to const location_t *PLOC.
608         Create Loc field whenever PLOC is non-NULL.
609         (ubsan_instrument_unreachable, ubsan_expand_null_ifn,
610         ubsan_build_overflow_builtin, instrument_bool_enum_load): Adjust
611         callers.
613         PR other/59545
614         * real.c (real_to_integer2): Change type of low to UHWI.
616 2014-03-26  Tobias Burnus  <burnus@net-b.de>
618         * gcc.c (LINK_COMMAND_SPEC): Use libcilkrts.spec for -fcilkplus.
619         (CILK_SELF_SPECS): New define.
620         (driver_self_specs): Use it.
622 2014-03-26  Richard Biener  <rguenther@suse.de>
624         * tree-pretty-print.c (percent_K_format): Implement special
625         case for LTO and its stripped down BLOCK tree.
627 2014-03-26  Jakub Jelinek  <jakub@redhat.com>
629         PR sanitizer/60636
630         * ubsan.c (instrument_si_overflow): Instrument ABS_EXPR.
632         * tree-vrp.c (simplify_internal_call_using_ranges): If only
633         one range is range_int_cst_p, but not both, at least optimize
634         addition/subtraction of 0 and multiplication by 0 or 1.
635         * gimple-fold.c (gimple_fold_call): Fold
636         IFN_UBSAN_CHECK_{ADD,SUB,MUL}.
637         (gimple_fold_stmt_to_constant_1): If both op0 and op1 aren't
638         INTEGER_CSTs, try to fold at least x * 0 and y - y.
640 2014-03-26  Eric Botcazou  <ebotcazou@adacore.com>
642         PR rtl-optimization/60452
643         * rtlanal.c (rtx_addr_can_trap_p_1): Fix head comment.
644         <case REG>: Return 1 for invalid offsets from the frame pointer.
646 2014-03-26  Marek Polacek  <polacek@redhat.com>
648         PR c/37428
649         * doc/extend.texi (C Extensions): Mention variable-length arrays in
650         a structure/union.
652 2014-03-26  Marek Polacek  <polacek@redhat.com>
654         PR c/39525
655         * doc/extend.texi (Designated Inits): Describe what happens to omitted
656         field members.
658 2014-03-26  Marek Polacek  <polacek@redhat.com>
660         PR other/59545
661         * ira-color.c (update_conflict_hard_regno_costs): Perform the
662         multiplication in unsigned type.
664 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
666         * doc/install.texi: Document nds32le-*-elf and nds32be-*-elf.
668 2014-03-26  Chung-Ju Wu  <jasonwucj@gmail.com>
670         * doc/contrib.texi: Add myself as Andes nds32 port contributor.
672 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
674         PR ipa/60315
675         * cif-code.def (UNREACHABLE) New code.
676         * ipa-inline.c (inline_small_functions): Skip edges to
677         __builtlin_unreachable.
678         (estimate_edge_growth): Allow edges to __builtlin_unreachable.
679         * ipa-inline-analysis.c (edge_set_predicate): Redirect edges with false
680         predicate to __bulitin_unreachable.
681         (set_cond_stmt_execution_predicate): Fix issue when
682         invert_tree_comparison returns ERROR_MARK.
683         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow): Do not
684         propagate to inline clones.
685         * cgraph.c (verify_edge_corresponds_to_fndecl): Allow redirection
686         to unreachable.
687         * ipa-cp.c (create_specialized_node): Be ready for new node to appear.
688         * cgraphclones.c (cgraph_clone_node): If call destination is already
689         ureachable, do not redirect it back.
690         * tree-inline.c (fold_marked_statements): Hanlde calls becoming
691         unreachable.
693 2014-03-25  Jan Hubicka  <hubicka@ucw.cz>
695         * ipa-pure-const.c (propagate_pure_const, propagate_nothrow):
696         Do not modify inline clones.
698 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
700         * config/i386/i386.md (general_sext_operand): New mode attr.
701         (addv<mode>4, subv<mode>4, mulv<mode>4): If operands[2] is CONST_INT,
702         don't generate (sign_extend (const_int)).
703         (*addv<mode>4, *subv<mode>4, *mulv<mode>4): Disallow CONST_INT_P
704         operands[2].  Use We constraint instead of <i> and
705         <general_sext_operand> predicate instead of <general_operand>.
706         (*addv<mode>4_1, *subv<mode>4_1, *mulv<mode>4_1): New insns.
707         * config/i386/constraints.md (We): New constraint.
708         * config/i386/predicates.md (x86_64_sext_operand,
709         sext_operand): New predicates.
711 2014-03-25  Martin Jambor  <mjambor@suse.cz>
713         PR ipa/60600
714         * ipa-cp.c (ipa_get_indirect_edge_target_1): Redirect type
715         inconsistent devirtualizations to __builtin_unreachable.
717 2014-03-25  Marek Polacek  <polacek@redhat.com>
719         PR c/35449
720         * doc/extend.texi (Example of asm with clobbered asm reg): Fix typo.
722 2014-03-25  Alan Lawrence  <alan.lawrence@arm.com>
724         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Reverse
725         order of elements for big-endian.
727 2014-03-25  Richard Biener  <rguenther@suse.de>
729         PR middle-end/60635
730         * gimplify-me.c (gimple_regimplify_operands): Update the
731         re-gimplifed stmt.
733 2014-03-25  Martin Jambor  <mjambor@suse.cz>
735         PR ipa/59176
736         * lto-cgraph.c (lto_output_node): Stream body_removed flag.
737         (lto_output_varpool_node): Likewise.
738         (input_overwrite_node): Likewise.
739         (input_varpool_node): Likewise.
741 2014-03-25  Richard Biener  <rguenther@suse.de>
743         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like OPT_fpie.
744         (run_gcc): Likewise.
746 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
748         * combine.c (simplify_compare_const): Add MODE argument.
749         Handle mode_width 0 as very large mode_width.
750         (try_combine, simplify_comparison): Adjust callers.
752         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
753         type to avoid signed integer overflow.
754         * explow.c (plus_constant): Likewise.
756 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
758         * doc/generic.texi: Correct typos.
760 2014-03-24  Tobias Burnus  <burnus@net-b.de>
762         * doc/invoke.texi (-flto): Expand section about
763         using static libraries with LTO.
765 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
767         PR rtl-optimization/60501
768         * optabs.def (addptr3_optab): New optab.
769         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
770         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
771         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
773         * lra.c (emit_add3_insn): Use the addptr pattern if available.
775         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
777 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
779         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
780         _mm512_set1_pd.
782         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
783         (_mm256_undefined_ps): Define.
784         (_mm256_undefined_pd): Define.
785         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
786         (_mm_undefined_pd): Define.
787         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
788         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
789         (_mm512_undefined_ps): Define.
790         (_mm512_undefined_pd): Define.
791         Use _mm*_undefined_*.
792         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
794 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
796         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
797         (lshr_simd): DI mode added.
798         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
799         (aarch64_ushr_simddi): Likewise.
800         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
801         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
802         (vshrd_n_u64): Likewise.
804 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
806         * Makefile.in (s-macro_list): Depend on cc1.
808 2014-03-23  Teresa Johnson  <tejohnson@google.com>
810         * ipa-utils.c (ipa_print_order): Use specified dump file.
812 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
814         PR rtl-optimization/60601
815         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
817         * gcc.c (eval_spec_function): Initialize save_growing_value.
819 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
821         PR sanitizer/60613
822         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
823         code == MINUS_EXPR, never swap op0 with op1.
825         * toplev.c (init_local_tick): Avoid signed integer multiplication
826         overflow.
827         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
828         shift by first operand's bitsize.
830 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
832         PR target/60610
833         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
834         redefine to 1 or 0.
835         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
836         TARGET_ISA_64BIT_P(x).
838 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
840         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
841         pattern for vector nor instead of subtract from splat(-1).
842         (altivec_expand_vec_perm_const_le): Likewise.
844 2014-03-21  Richard Henderson  <rth@twiddle.net>
846         PR target/60598
847         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
848         related insns after epilogue_completed.
850 2014-03-21  Martin Jambor  <mjambor@suse.cz>
852         PR ipa/59176
853         * cgraph.h (symtab_node): New flag body_removed.
854         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
855         when removing bodies.
856         * symtab.c (dump_symtab_base): Dump body_removed flag.
857         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
858         had their bodies removed.
860 2014-03-21  Martin Jambor  <mjambor@suse.cz>
862         PR ipa/60419
863         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
864         in the border.
866 2014-03-21  Richard Biener  <rguenther@suse.de>
868         PR tree-optimization/60577
869         * tree-core.h (struct tree_base): Document nothrow_flag use
870         in VAR_DECL_NONALIASED.
871         * tree.h (VAR_DECL_NONALIASED): New.
872         (may_be_aliased): Adjust.
873         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
875 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
877         * expr.c (expand_expr_real_1): Remove outdated comment.
879 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
881         PR middle-end/60597
882         * ira.c (adjust_cleared_regs): Call copy_rtx on
883         *reg_equiv[REGNO (loc)].src_p before passing it to
884         simplify_replace_fn_rtx.
886         PR target/60568
887         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
888         into CONST, put pic register as first operand of PLUS.  Use
889         gen_const_mem for both 32-bit and 64-bit PIC got loads.
891 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
893         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
895 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
897         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
898         around for store forwarding issue in the FPU on the UT699.
899         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
900         loads and operations if -mfix-ut699 is specified.
901         (divtf3_hq): Tweak attribute.
902         (sqrttf2_hq): Likewise.
904 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
906         * calls.c (store_one_arg): Remove incorrect const qualification on the
907         type of the temporary.
908         * cfgexpand.c (expand_return): Likewise.
909         * expr.c (expand_constructor): Likewise.
910         (expand_expr_real_1): Likewise.
912 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
914         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
915         of parts.
917 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
919         PR target/60039
920         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
922 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
924         * config/arm/aarch-common-protos.h
925         (alu_cost_table): Fix spelling of "extend".
926         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
928 2014-03-19  Richard Biener  <rguenther@suse.de>
930         PR middle-end/60553
931         * tree-core.h (tree_type_common): Re-order pointer members
932         to reduce recursion depth during GC walks.
934 2014-03-19  Marek Polacek  <polacek@redhat.com>
936         PR sanitizer/60569
937         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
938         before accessing it.
940 2014-03-19  Richard Biener  <rguenther@suse.de>
942         PR lto/59543
943         * lto-streamer-in.c (input_function): In WPA stage do not drop
944         debug stmts.
946 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
948         PR tree-optimization/60559
949         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
950         with build_zero_cst assignment.
952 2014-03-18  Kai Tietz  <ktietz@redhat.com>
954         PR rtl-optimization/56356
955         * sdbout.c (sdbout_parms): Verify that parms'
956         incoming argument is valid.
957         (sdbout_reg_parms): Likewise.
959 2014-03-18  Richard Henderson  <rth@redhat.com>
961         PR target/60562
962         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
963         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
964         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
966 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
968         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
969         * doc/plugins.texi (Plugin callbacks): Mention PLUGIN_INCLUDE_FILE.
970         Italicize plugin event names in description.  Explain that
971         PLUGIN_PRAGMAS has no sense for lto1.  Explain PLUGIN_INCLUDE_FILE.
972         Remind that no GCC functions should be called after PLUGIN_FINISH.
973         Explain what pragmas with expansion are.
975 2014-03-18  Martin Liska  <mliska@suse.cz>
977         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
978         gimple call statement is update.
979         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
980         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
982 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
984         PR sanitizer/60557
985         * ubsan.c (ubsan_instrument_unreachable): Call
986         initialize_sanitizer_builtins.
987         (ubsan_pass): Likewise.
989         PR sanitizer/60535
990         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
991         varpool_finalize_decl instead of rest_of_decl_compilation.
993 2014-03-18  Richard Biener  <rguenther@suse.de>
995         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
996         by using bitmap_and_compl instead of bitmap_and_compl_into.
997         (df_rd_transfer_function): Likewise.
999 2014-03-18  Richard Biener  <rguenther@suse.de>
1001         * doc/lto.texi (fresolution): Fix typo.
1003 2014-03-18  Richard Biener  <rguenther@suse.de>
1005         * doc/invoke.texi (flto): Update for changes in 4.9.
1007 2014-03-18  Richard Biener  <rguenther@suse.de>
1009         * doc/loop.texi: Remove section on the removed lambda framework.
1010         Update loop docs with recent changes in preserving loop structure.
1012 2014-03-18  Richard Biener  <rguenther@suse.de>
1014         * doc/lto.texi (-fresolution): Document.
1016 2014-03-18  Richard Biener  <rguenther@suse.de>
1018         * doc/contrib.texi: Adjust my name.
1020 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
1022         PR ipa/58721
1023         * internal-fn.c: Include diagnostic-core.h.
1024         (expand_BUILTIN_EXPECT): New function.
1025         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
1026         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
1027         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
1028         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
1029         IFN_BUILTIN_EXPECT.
1030         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
1031         Revert 3 argument __builtin_expect code.
1032         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
1033         * gimple-fold.c (gimple_fold_call): Likewise.
1034         * tree.h (fold_builtin_expect): New prototype.
1035         * builtins.c (build_builtin_expect_predicate): Add predictor
1036         argument, if non-NULL, create 3 argument __builtin_expect.
1037         (fold_builtin_expect): No longer static.  Add ARG2 argument,
1038         pass it through to build_builtin_expect_predicate.
1039         (fold_builtin_2): Adjust caller.
1040         (fold_builtin_3): Handle BUILT_IN_EXPECT.
1041         * internal-fn.def (BUILTIN_EXPECT): New.
1043 2014-03-18  Tobias Burnus  <burnus@net-b.de>
1045         PR ipa/58721
1046         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
1047         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
1048         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
1050 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
1052         PR ipa/58721
1053         * predict.c (combine_predictions_for_bb): Fix up formatting.
1054         (expr_expected_value_1, expr_expected_value): Add predictor argument,
1055         fill what it points to if non-NULL.
1056         (tree_predict_by_opcode): Adjust caller, use the predictor.
1057         * predict.def (PRED_COMPARE_AND_SWAP): Add.
1059 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
1061         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
1062         proper constant for the store mode.
1064 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
1066         * symtab.c (change_decl_assembler_name): Fix transparent alias
1067         chain construction.
1069 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
1071         * config/aarch64/aarch64.c: Correct the comments about the
1072         aarch64 stack layout.
1074 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
1076         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
1077         check for GF_OMP_FOR_KIND_FOR.
1079 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
1081         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
1082         ymm and zmm register names.
1084 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
1086         PR target/60516
1087         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
1088         note creation for the 2010-08-31 changes.
1090 2014-03-17  Marek Polacek  <polacek@redhat.com>
1092         PR middle-end/60534
1093         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
1094         as -fno-tree-loop-vectorize.
1095         (expand_omp_simd): Likewise.
1097 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
1099         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
1100         (eligible_for_call_delay): New prototype.
1101         * config/sparc/sparc.c (tls_call_delay): Rename into...
1102         (eligible_for_call_delay): ...this.  Return false if the instruction
1103         cannot be put in the delay slot of a branch.
1104         (eligible_for_restore_insn): Simplify.
1105         (eligible_for_return_delay): Return false if the instruction cannot be
1106         put in the delay slot of a branch and simplify.
1107         (eligible_for_sibcall_delay): Return false if the instruction cannot be
1108         put in the delay slot of a branch.
1109         * config/sparc/sparc.md (fix_ut699): New attribute.
1110         (tls_call_delay): Delete.
1111         (in_call_delay): Reimplement.
1112         (eligible_for_sibcall_delay): Rename into...
1113         (in_sibcall_delay): ...this.
1114         (eligible_for_return_delay): Rename into...
1115         (in_return_delay): ...this.
1116         (in_branch_delay): Reimplement.
1117         (in_uncond_branch_delay): Delete.
1118         (in_annul_branch_delay): Delete.
1120 2014-03-14  Richard Henderson  <rth@redhat.com>
1122         PR target/60525
1123         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
1124         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
1125         (*floathi<X87MODEF>2_i387_with_temp): Remove.
1126         (floathi splitters): Remove.
1127         (float<SWI48x>xf2): New pattern.
1128         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
1129         code that tried to handle DImode for 32-bit, but which was excluded
1130         by the pattern's condition.  Drop allocation of stack temporary.
1131         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
1132         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
1133         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
1134         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
1135         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
1136         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
1137         (*float<SWI48><MODEF>2_sse_interunit): Remove.
1138         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
1139         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
1140         (*float<SWI48x><X87MODEF>2_i387): Remove.
1141         (all float _with_temp splitters): Remove.
1142         (*float<SWI48x><MODEF>2_i387): New pattern.
1143         (*float<SWI48><MODEF>2_sse): New pattern.
1144         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
1145         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
1147 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
1148             Marek Polacek  <polacek@redhat.com>
1150         PR middle-end/60484
1151         * common.opt (dump_base_name_prefixed): New Variable.
1152         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
1153         if x_dump_base_name_prefixed is already set, set it at the end.
1155 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
1157         PR rtl-optimization/60508
1158         * lra-constraints.c (get_reload_reg): Add new parameter
1159         in_subreg_p.
1160         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
1161         Pass the new parameter values.
1163 2014-03-14  Richard Biener  <rguenther@suse.de>
1165         * common.opt: Revert unintented changes from r205065.
1166         * opts.c: Likewise.
1168 2014-03-14  Richard Biener  <rguenther@suse.de>
1170         PR middle-end/60518
1171         * cfghooks.c (split_block): Properly adjust all loops the
1172         block was a latch of.
1174 2014-03-14  Martin Jambor  <mjambor@suse.cz>
1176         PR lto/60461
1177         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
1178         and simplify it.
1180 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
1182         PR target/59396
1183         * config/avr/avr.c (avr_set_current_function): Pass function name
1184         through default_strip_name_encoding before sanity checking instead
1185         of skipping the first char of the assembler name.
1187 2014-03-13  Richard Henderson  <rth@redhat.com>
1189         PR debug/60438
1190         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
1191         (ix86_force_to_memory, ix86_free_from_memory): Remove.
1192         * config/i386/i386-protos.h: Likewise.
1193         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
1194         in the expander instead of a splitter.
1195         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
1196         any possibility of requiring a memory.
1197         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
1198         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
1199         (fp branch splitters): Update for ix86_split_fp_branch.
1200         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
1201         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
1202         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
1203         (*fop_<MODEF>_2_i387): Remove f/r alternative.
1204         (*fop_<MODEF>_3_i387): Likewise.
1205         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
1206         (splitters for the fop_* register patterns): Remove.
1207         (fscalexf4_i387): Rename from *fscalexf4_i387.
1208         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
1210 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1212         PR tree-optimization/59779
1213         * tree-dfa.c (get_ref_base_and_extent): Use double_int
1214         type for bitsize and maxsize instead of HOST_WIDE_INT.
1216 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
1218         PR rtl-optimization/57320
1219         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
1220         the CFG after thread_prologue_and_epilogue_insns.
1222 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
1224         PR rtl-optimization/57189
1225         * lra-constraints.c (process_alt_operands): Disfavor spilling
1226         vector pseudos.
1228 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
1230         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic messages.
1232 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
1234         PR tree-optimization/59025
1235         PR middle-end/60418
1236         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
1237         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
1239 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
1241         PR target/60486
1242         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
1243         calls of avr_out_plus_1.
1245 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
1247         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
1248         BB's single pred and update the father loop's latch info later.
1250 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
1252         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
1253         (VEC_M): Likewise.
1254         (VEC_N): Likewise.
1255         (VEC_R): Likewise.
1256         (VEC_base): Likewise.
1257         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
1258         registers, we need to swap double words in little endian mode.
1260         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
1261         to be a container mode for 128-bit integer operations added in ISA
1262         2.07.  Unlike TImode and PTImode, the preferred register set is
1263         the Altivec/VMX registers for the 128-bit operations.
1265         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
1266         declarations.
1267         (rs6000_split_128bit_ok_p): Likewise.
1269         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
1270         macros for creating ISA 2.07 normal and overloaded builtin
1271         functions with 3 arguments.
1272         (BU_P8V_OVERLOAD_3): Likewise.
1273         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
1274         for use as overloaded functions.
1275         (VPERM_1TI_UNS): Likewise.
1276         (VSEL_1TI): Likewise.
1277         (VSEL_1TI_UNS): Likewise.
1278         (ST_INTERNAL_1ti): Likewise.
1279         (LD_INTERNAL_1ti): Likewise.
1280         (XXSEL_1TI): Likewise.
1281         (XXSEL_1TI_UNS): Likewise.
1282         (VPERM_1TI): Likewise.
1283         (VPERM_1TI_UNS): Likewise.
1284         (XXPERMDI_1TI): Likewise.
1285         (SET_1TI): Likewise.
1286         (LXVD2X_V1TI): Likewise.
1287         (STXVD2X_V1TI): Likewise.
1288         (VEC_INIT_V1TI): Likewise.
1289         (VEC_SET_V1TI): Likewise.
1290         (VEC_EXT_V1TI): Likewise.
1291         (EQV_V1TI): Likewise.
1292         (NAND_V1TI): Likewise.
1293         (ORC_V1TI): Likewise.
1294         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
1295         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
1296         overloaded builtin.
1297         (VADDUQM): Likewise.
1298         (VSUBCUQ): Likewise.
1299         (VADDEUQM): Likewise.
1300         (VADDECUQ): Likewise.
1301         (VSUBEUQM): Likewise.
1302         (VSUBECUQ): Likewise.
1304         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
1305         __int128_t and __uint128_t types.
1306         (__uint128_type): Likewise.
1307         (altivec_categorize_keyword): Add support for vector __int128_t,
1308         vector __uint128_t, vector __int128, and vector unsigned __int128
1309         as a container type for TImode operations that need to be done in
1310         VSX/Altivec registers.
1311         (rs6000_macro_to_expand): Likewise.
1312         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
1313         to support 128-bit integer instructions vaddcuq, vadduqm,
1314         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
1315         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
1317         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
1318         for V1TImode, and set up preferences to use VSX/Altivec registers.
1319         Setup VSX reload handlers.
1320         (rs6000_debug_reg_global): Likewise.
1321         (rs6000_init_hard_regno_mode_ok): Likewise.
1322         (rs6000_preferred_simd_mode): Likewise.
1323         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
1324         (easy_altivec_constant): Likewise.
1325         (output_vec_const_move): Likewise.
1326         (rs6000_expand_vector_set): Convert V1TImode set and extract to
1327         simple move.
1328         (rs6000_expand_vector_extract): Likewise.
1329         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
1330         addressing.
1331         (rs6000_const_vec): Add support for V1TImode.
1332         (rs6000_emit_le_vsx_load): Swap double words when loading or
1333         storing TImode/V1TImode.
1334         (rs6000_emit_le_vsx_store): Likewise.
1335         (rs6000_emit_le_vsx_move): Likewise.
1336         (rs6000_emit_move): Add support for V1TImode.
1337         (altivec_expand_ld_builtin): Likewise.
1338         (altivec_expand_st_builtin): Likewise.
1339         (altivec_expand_vec_init_builtin): Likewise.
1340         (altivec_expand_builtin): Likewise.
1341         (rs6000_init_builtins): Add support for V1TImode type.  Add
1342         support for ISA 2.07 128-bit integer builtins.  Define type names
1343         for the VSX/Altivec vector types.
1344         (altivec_init_builtins): Add support for overloaded vector
1345         functions with V1TImode type.
1346         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
1347         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
1348         external function.
1349         (rs6000_split_128bit_ok_p): Likewise.
1350         (rs6000_handle_altivec_attribute): Create V1TImode from vector
1351         __int128_t and vector __uint128_t.
1353         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
1354         and mode attributes.
1355         (VSX_M): Likewise.
1356         (VSX_M2): Likewise.
1357         (VSm): Likewise.
1358         (VSs): Likewise.
1359         (VSr): Likewise.
1360         (VSv): Likewise.
1361         (VS_scalar): Likewise.
1362         (VS_double): Likewise.
1363         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
1365         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
1366         we support the ISA 2.07 128-bit integer arithmetic instructions.
1367         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
1368         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
1369         and TImode types for use with the builtin functions.
1370         (V1TI_type_node): Likewise.
1371         (unsigned_V1TI_type_node): Likewise.
1372         (intTI_type_internal_node): Likewise.
1373         (uintTI_type_internal_node): Likewise.
1375         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
1376         128-bit builtin functions.
1377         (UNSPEC_VADDEUQM): Likewise.
1378         (UNSPEC_VADDECUQ): Likewise.
1379         (UNSPEC_VSUBCUQ): Likewise.
1380         (UNSPEC_VSUBEUQM): Likewise.
1381         (UNSPEC_VSUBECUQ): Likewise.
1382         (VM): Add V1TImode to vector mode iterators.
1383         (VM2): Likewise.
1384         (VI_unit): Likewise.
1385         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
1386         (altivec_vaddcuq): Likewise.
1387         (altivec_vsubuqm): Likewise.
1388         (altivec_vsubcuq): Likewise.
1389         (altivec_vaddeuqm): Likewise.
1390         (altivec_vaddecuq): Likewise.
1391         (altivec_vsubeuqm): Likewise.
1392         (altivec_vsubecuq): Likewise.
1394         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
1395         mode iterators.
1396         (BOOL_128): Likewise.
1397         (BOOL_REGS_OUTPUT): Likewise.
1398         (BOOL_REGS_OP1): Likewise.
1399         (BOOL_REGS_OP2): Likewise.
1400         (BOOL_REGS_UNARY): Likewise.
1401         (BOOL_REGS_AND_CR0): Likewise.
1403         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
1404         128-bit integer builtin support.
1405         (vec_vadduqm): Likewise.
1406         (vec_vaddecuq): Likewise.
1407         (vec_vaddeuqm): Likewise.
1408         (vec_vsubecuq): Likewise.
1409         (vec_vsubeuqm): Likewise.
1410         (vec_vsubcuq): Likewise.
1411         (vec_vsubuqm): Likewise.
1413         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
1414         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
1415         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
1416         128-bit integer add/subtract to ISA 2.07.
1418 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
1420         * config/arc/arc.c (arc_predicate_delay_insns):
1421         Fix third argument passed to conditionalize_nonjump.
1423 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
1425         * config/aarch64/aarch64-builtins.c
1426         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
1427         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
1428         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
1429         instead of __builtin_lfloor.
1430         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
1432 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1434         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
1435         (tree_ssa_ifcombine_bb_1): New function.
1436         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
1437         is an empty forwarder block to then_bb or vice versa and then_bb
1438         and else_bb are effectively swapped.
1440 2014-03-12  Christian Bruel  <christian.bruel@st.com>
1442         PR target/60264
1443         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
1444         REG_CFA_DEF_CFA note.
1445         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
1446         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
1448 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
1450         PR tree-optimization/60454
1451         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
1453 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1455         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
1456         Do not define target_cpu_default2 to generic.
1457         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
1458         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
1459         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
1461 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1462             Marc Glisse  <marc.glisse@inria.fr>
1464         PR tree-optimization/60502
1465         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
1466         instead of build_low_bits_mask.
1468 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
1470         PR middle-end/60482
1471         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
1472         if there are multiple uses, but op doesn't live on E edge.
1473         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
1474         clobber stmts before __builtin_unreachable.
1476 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
1478         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
1479         hard_frame_pointer_rtx.
1480         * cse.c (cse_insn): Remove volatile check.
1481         * cselib.c (cselib_process_insn): Likewise.
1482         * dse.c (scan_insn): Likewise.
1484 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1486         * config/arc/arc.c (conditionalize_nonjump): New function,
1487         broken out of ...
1488         (arc_ifcvt): ... this.
1489         (arc_predicate_delay_insns): Use it.
1491 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
1493         * config/arc/predicates.md (extend_operand): During/after reload,
1494         allow const_int_operand.
1495         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
1496         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
1497         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
1498         to "i".
1499         (umulsi3_highpart_i): Likewise.
1501 2014-03-11  Richard Biener  <rguenther@suse.de>
1503         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
1504         Add asserts to guard possible wrong-code bugs.
1506 2014-03-11  Richard Biener  <rguenther@suse.de>
1508         PR tree-optimization/60429
1509         PR tree-optimization/60485
1510         * tree-ssa-structalias.c (set_union_with_increment): Properly
1511         take into account all fields that overlap the shifted vars.
1512         (do_sd_constraint): Likewise.
1513         (do_ds_constraint): Likewise.
1514         (get_constraint_for_ptr_offset): Likewise.
1516 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
1518         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
1519         (nios2_compute_frame_layout):
1520         Add calculation of cfun->machine->fp_save_offset.
1521         (nios2_expand_prologue): Correct setting of frame pointer register
1522         in prologue.
1523         (nios2_expand_epilogue): Update recovery of stack pointer from
1524         frame pointer accordingly.
1525         (nios2_initial_elimination_offset): Update calculation of offset
1526         for eliminating to HARD_FRAME_POINTER_REGNUM.
1528 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
1530         PR ipa/60457
1531         * ipa.c (symtab_remove_unreachable_nodes): Don't call
1532         cgraph_get_create_node on VAR_DECLs.
1534 2014-03-10  Richard Biener  <rguenther@suse.de>
1536         PR middle-end/60474
1537         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
1539 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
1541         * config/vms/vms.opt (vms_float_format): New variable.
1543 2014-03-08  Tobias Burnus  <burnus@net-b.de>
1545         * doc/invoke.texi (-fcilkplus): Update implementation status.
1547 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
1548             Richard Biener  <rguenther@suse.de>
1550         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
1551         consistently accross all TUs.
1552         (run_gcc): Enable -fshort-double automatically at link at link-time
1553         and disallow override.
1555 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
1557         PR target/58271
1558         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
1559         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
1560         if they can't be used.
1562 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1564         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
1565         for Solaris 11/x86 ld.
1566         * configure: Regenerate.
1568 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1570         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
1571         (LIB_TLS_SPEC): Save as ld_tls_libs.
1572         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
1573         (HAVE_AS_IX86_TLSLDM): New test.
1574         * configure, config.in: Regenerate.
1575         * config/i386/i386.c (legitimize_tls_address): Fall back to
1576         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
1577         cannot support TLS_MODEL_LOCAL_DYNAMIC.
1578         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
1579         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
1581 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
1583         * common.opt (fira-loop-pressure): Mark as optimization.
1585 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
1587         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
1588         an OpenMP mappable type.
1590 2014-03-06  Matthias Klose  <doko@ubuntu.com>
1592         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
1593         MULTILIB_OSDIRNAMES is not defined.
1595 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
1596             Meador Inge  <meadori@codesourcery.com>
1598         PR target/58595
1599         * config/arm/arm.c (arm_tls_symbol_p): Remove.
1600         (arm_legitimize_address): Call legitimize_tls_address for any
1601         arm_tls_referenced_p expression, handle constant addend.  Call it
1602         before testing for !TARGET_ARM.
1603         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
1605 2014-03-06  Richard Biener  <rguenther@suse.de>
1607         PR middle-end/60445
1608         PR lto/60424
1609         PR lto/60427
1610         Revert
1611         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1613         * tree-streamer.c (record_common_node): Assert we don't record
1614         nodes with type double.
1615         (preload_common_node): Skip type double, complex double and double
1616         pointer since it is now frontend dependent due to fshort-double option.
1618 2014-03-06  Richard Biener  <rguenther@suse.de>
1620         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
1621         or -fno-lto is specified and the linker has full plugin support.
1622         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
1623         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
1624         * lto-wrapper.c (merge_and_complain): Merge compile-time
1625         optimization levels.
1626         (run_gcc): And pass it through to the link options.
1628 2014-03-06  Alexandre Oliva  <aoliva@redhat.com>
1630         PR debug/60381
1631         Revert:
1632         2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1633         PR debug/59992
1634         * cselib.c (remove_useless_values): Skip to avoid quadratic
1635         behavior if the condition moved from...
1636         (cselib_process_insn): ... here holds.
1638 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1640         PR plugins/59335
1641         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
1642         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
1644         PR plugins/59335
1645         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
1646         (TM_H): Add x86-tune.def.
1648 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1650         * config/aarch64/aarch64.c (generic_tunings):
1651         Use cortexa57_extra_costs.
1653 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
1655         PR lto/60404
1656         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
1657         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
1658         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
1659         cost for in_lto_p.
1661 2014-03-04  Heiher  <r@hev.cc>
1663         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
1664         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
1666 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
1668         * config/i386/predicates.md (const2356_operand): Change to ...
1669         (const2367_operand): ... this.
1670         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
1671         const2367_operand.
1672         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1673         (*avx512pf_scatterpf<mode>sf): Ditto.
1674         (avx512pf_scatterpf<mode>df): Ditto.
1675         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1676         (*avx512pf_scatterpf<mode>df): Ditto.
1677         * config/i386/i386.c (ix86_expand_builtin): Update
1678         incorrect hint operand error message.
1680 2014-03-04  Richard Biener  <rguenther@suse.de>
1682         * lto-section-in.c (lto_get_section_data): Fix const cast.
1684 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
1686         * tree-streamer.c (record_common_node): Assert we don't record
1687         nodes with type double.
1688         (preload_common_node): Skip type double, complex double and double
1689         pointer since it is now frontend dependent due to fshort-double option.
1691 2014-03-04  Richard Biener  <rguenther@suse.de>
1693         PR lto/60405
1694         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
1695         (lto_input_toplevel_asms): Likewise.
1696         * lto-section-in.c (lto_get_section_data): Instead do it here
1697         for every section.
1699 2014-03-04  Richard Biener  <rguenther@suse.de>
1701         PR tree-optimization/60382
1702         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
1703         dead PHIs a reduction.
1705 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
1707         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
1708         hint value.
1709         (_mm_prefetch): Move out of GCC target("sse") pragma.
1710         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
1711         GCC target("prfchw") pragma.
1712         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
1713         for locality <= 2.
1714         * config/i386/i386.c (ix86_option_override_internal): Enable
1715         -mprfchw with -mprefetchwt1.
1717 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1719         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
1720         Mark as varying.
1722 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
1724         * opts.h (CL_PCH_IGNORE): Define.
1725         * targhooks.c (option_affects_pch_p):
1726         Return false for options that have CL_PCH_IGNORE set.
1727         * opt-functions.awk: Process PchIgnore.
1728         * doc/options.texi: Document PchIgnore.
1730         * config/arc/arc.opt (misize): Add PchIgnore property.
1732 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1734         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1735         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1736         constraint on constants to permit them being loaded into
1737         GENERAL_REGS or BASE_REGS.
1739 2014-03-03  Nick Clifton  <nickc@redhat.com>
1741         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1742         anti-cacnonical alternatives.
1743         (negandhi3_real): New pattern.
1744         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1746 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1748         * config/avr/avr-mcus.def: Remove atxmega16x1.
1749         * config/avr/avr-tables.opt: Regenerate.
1750         * config/avr/t-multilib: Regenerate.
1751         * doc/avr-mmcu.texi: Regenerate.
1753 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1754             Mircea Namolaru  <mircea.namolaru@inria.fr>
1756         PR tree-optimization/58028
1757         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1758         scalar dimensions.
1760 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1762         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1763         not handled by recognizers.
1765 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1767         PR middle-end/60175
1768         * function.c (expand_function_end): Don't emit
1769         clobber_return_register sequence if clobber_after is a BARRIER.
1770         * cfgexpand.c (construct_exit_block): Append instructions before
1771         return_label to prev_bb.
1773 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1775         * config/rs6000/constraints.md: Document reserved use of "wc".
1777 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1779         PR ipa/60150
1780         * ipa.c (function_and_variable_visibility): When dissolving comdat
1781         group, also set all symbols to local.
1783 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1785         PR ipa/60306
1787         Revert:
1788         2013-12-14   Jan Hubicka  <jh@suse.cz>
1789         PR middle-end/58477
1790         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1792 2014-03-02  Jon Beniston  <jon@beniston.com>
1794         PR bootstrap/48230
1795         PR bootstrap/50927
1796         PR bootstrap/52466
1797         PR target/46898
1798         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1799           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1800         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1801         (simple_return, *simple_return): New patterns
1802         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1803         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1805 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1807         * dwarf2out.c (gen_subprogram_die): Tidy.
1809 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1811         PR target/60071
1812         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1813         (*mov_t_msb_neg_negc): ... this new insn.
1815 2014-02-28  Jason Merrill  <jason@redhat.com>
1817         PR c++/58678
1818         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1819         function.
1821 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1823         PR c++/60314
1824         * dwarf2out.c (decltype_auto_die): New static.
1825         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1826         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1827         (is_cxx_auto): Likewise.
1829 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1831         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1832         we are not using general regs only.
1834 2014-02-28  Richard Biener  <rguenther@suse.de>
1836         PR target/60280
1837         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1838         previous fix and only allow to remove trivial pre-headers
1839         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1840         (remove_forwarder_block): Properly update the latch of a loop.
1842 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1844         PR debug/59992
1845         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1846         (cselib_preserved_hash_table): New.
1847         (preserve_constants_and_equivs): Move preserved vals to it.
1848         (cselib_find_slot): Look it up first.
1849         (cselib_init): Initialize it.
1850         (cselib_finish): Release it.
1851         (dump_cselib_table): Dump it.
1853 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1855         PR debug/59992
1856         * cselib.c (remove_useless_values): Skip to avoid quadratic
1857         behavior if the condition moved from...
1858         (cselib_process_insn): ... here holds.
1860 2014-02-28  Alexandre Oliva  <aoliva@redhat.com>
1862         PR debug/57232
1863         * var-tracking.c (vt_initialize): Apply the same condition to
1864         preserve the CFA base value.
1866 2014-02-28  Joey Ye  <joey.ye@arm.com>
1868         PR target/PR60169
1869         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1870         if reload in progress or completed.
1872 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1874         PR middle-end/60147
1875         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1876         NAMELIST_DECL.
1878 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1880         * doc/tm.texi.in (Condition Code Status): Update documention for
1881         relative locations of cc0-setter and cc0-user.
1883 2014-02-27  Jeff Law  <law@redhat.com>
1885         PR rtl-optimization/52714
1886         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1887         into two independent simple sets, if I3 is a jump, ensure the
1888         pattern we place into I3 is a (set (pc) ...).
1890 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1891             Jeff Law  <law@redhat.com>
1893         PR rtl-optimization/49847
1894         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1895         are in different blocks.
1896         * doc/tm.texi (Condition Code Status): Update documention for
1897         relative locations of cc0-setter and cc0-user.
1899 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1901         PR target/59222
1902         * lra.c (lra_emit_add): Check SUBREG too.
1904 2014-02-27  Andreas Schwab  <schwab@suse.de>
1906         * config/m68k/m68k.c (m68k_option_override): Disable
1907         -flive-range-shrinkage for classic m68k.
1908         (m68k_override_options_after_change): Likewise.
1910 2014-02-27  Marek Polacek  <polacek@redhat.com>
1912         PR middle-end/59223
1913         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1914         -Wmaybe-uninitialized.
1916 2014-02-27  Alan Modra  <amodra@gmail.com>
1918         PR target/57936
1919         * reload1.c (emit_input_reload_insns): When reload_override_in,
1920         set old to rl->in_reg when rl->in_reg is a subreg.
1922 2014-02-26  Richard Biener  <rguenther@suse.de>
1924         PR bootstrap/60343
1925         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1927 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1929         * common/config/i386/predicates.md (const1256_operand): Remove.
1930         (const2356_operand): New.
1931         (const_1_to_2_operand): Remove.
1932         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1933         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1934         (*avx512pf_gatherpf<mode>sf): Ditto.
1935         (avx512pf_gatherpf<mode>df): Ditto.
1936         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1937         (*avx512pf_gatherpf<mode>df): Ditto.
1938         (avx512pf_scatterpf<mode>sf): Ditto.
1939         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1940         (*avx512pf_scatterpf<mode>sf): Ditto.
1941         (avx512pf_scatterpf<mode>df): Ditto.
1942         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1943         (*avx512pf_scatterpf<mode>df): Ditto.
1944         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1946 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1948         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1949         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1950         (_mm512_mask_testn_epi64_mask): Move to ...
1951         * config/i386/avx512cdintrin.h: Here.
1952         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1953         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1954         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1955         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1956         TARGET_AVX512F from TARGET_AVX512CD.
1958 2014-02-26  Richard Biener  <rguenther@suse.de>
1960         PR ipa/60327
1961         * ipa.c (walk_polymorphic_call_targets): Properly guard
1962         call to inline_update_overall_summary.
1964 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1966         PR target/60280
1967         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1968         and latches only if requested.  Fix latch if it is removed.
1969         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1970         LOOPS_HAVE_PREHEADERS.
1972 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1974         * builtins.c (expand_builtin_thread_pointer): Create a new target
1975         when the target is NULL.
1977 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1979         PR rtl-optimization/60317
1980         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1981         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1982         * lra-assigns.c: Include params.h.
1983         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1984         other reload pseudos considerations.
1986 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1988         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1989         to use canonical form for nor<mode>3.
1991 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1993         PR target/55426
1994         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1995         conversions.
1997 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1999         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
2000         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
2001         (ix86_handle_option): Handle OPT_mprefetchwt1.
2002         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
2003         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2004         PREFETCHWT1 CPUID.
2005         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2006         OPTION_MASK_ISA_PREFETCHWT1.
2007         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
2008         (PTA_PREFETCHWT1): New.
2009         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
2010         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
2011         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
2012         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
2013         (*prefetch_avx512pf_<mode>_: Change into ...
2014         (*prefetch_prefetchwt1_<mode>: This.
2015         * config/i386/i386.opt (mprefetchwt1): New.
2016         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
2017         (_mm_prefetch): Handle intent to write.
2018         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
2020 2014-02-25  Richard Biener  <rguenther@suse.de>
2022         PR middle-end/60291
2023         * emit-rtl.c (mem_attrs_htab): Remove.
2024         (mem_attrs_htab_hash): Likewise.
2025         (mem_attrs_htab_eq): Likewise.
2026         (set_mem_attrs): Always allocate new mem-attrs when something changed.
2027         (init_emit_once): Do not allocate mem_attrs_htab.
2029 2014-02-25  Richard Biener  <rguenther@suse.de>
2031         PR lto/60319
2032         * lto-opts.c (lto_write_options): Output non-explicit conservative
2033         -fwrapv, -fno-trapv and -fno-strict-overflow.
2034         * lto-wrapper.c (merge_and_complain): Handle merging those options.
2035         (run_gcc): And pass them through.
2037 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2039         * sel-sched.c (calculate_new_fences): New parameter ptime.
2040         Calculate it as a maximum over all fence cycles.
2041         (sel_sched_region_2): Adjust the call to calculate_new_fences.
2042         Print the final schedule timing when sched_verbose.
2044 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
2046         PR rtl-optimization/60292
2047         * sel-sched.c (fill_vec_av_set): Do not reset target availability
2048         bit fot the fence instruction.
2050 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
2052         * calls.h: Fix typo in comment.
2054 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
2056         * config/pa/pa.c (pa_output_move_double): Don't valididate when
2057         adjusting offsetable addresses.
2059 2014-02-24  Guozhi Wei  <carrot@google.com>
2061         * sparseset.h (sparseset_pop): Fix the wrong index.
2063 2014-02-24  Walter Lee  <walt@tilera.com>
2065         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
2066         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
2067         triplet.
2068         * common/config/tilegx/tilegx-common.c
2069         (TARGET_DEFAULT_TARGET_FLAGS): Define.
2070         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
2071         (LINK_SPEC): Ditto.
2072         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
2073         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
2074         (tilegx_gimplify_va_arg_expr): Handle big endian.
2075         (tilegx_expand_unaligned_load): Ditto.
2076         (tilegx_expand_unaligned_store): Ditto.
2077         (TARGET_RETURN_IN_MSB): New.
2078         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
2079         (TARGET_ENDIAN_DEFAULT): New.
2080         (TARGET_BIG_ENDIAN): Handle big endian.
2081         (BYTES_BIG_ENDIAN): Ditto.
2082         (WORDS_BIG_ENDIAN): Ditto.
2083         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
2084         (ENDIAN_SPEC): New.
2085         (EXTRA_SPECS): New.
2086         * config/tilegx/tilegx.md (extv): Handle big endian.
2087         (extzv): Ditto.
2088         (insn_st<n>): Ditto.
2089         (insn_st<n>_add<bitsuffix>): Ditto.
2090         (insn_stnt<n>): Ditto.
2091         (insn_stnt<n>_add<bitsuffix>):Ditto.
2092         (vec_interleave_highv8qi): Handle big endian.
2093         (vec_interleave_highv8qi_be): New.
2094         (vec_interleave_highv8qi_le): New.
2095         (insn_v1int_h): Handle big endian.
2096         (vec_interleave_lowv8qi): Handle big endian.
2097         (vec_interleave_lowv8qi_be): New.
2098         (vec_interleave_lowv8qi_le): New.
2099         (insn_v1int_l): Handle big endian.
2100         (vec_interleave_highv4hi): Handle big endian.
2101         (vec_interleave_highv4hi_be): New.
2102         (vec_interleave_highv4hi_le): New.
2103         (insn_v2int_h): Handle big endian.
2104         (vec_interleave_lowv4hi): Handle big endian.
2105         (vec_interleave_lowv4hi_be): New.
2106         (vec_interleave_lowv4hi_le): New.
2107         (insn_v2int_l): Handle big endian.
2108         (vec_interleave_highv2si): Handle big endian.
2109         (vec_interleave_highv2si_be): New.
2110         (vec_interleave_highv2si_le): New.
2111         (insn_v4int_h): Handle big endian.
2112         (vec_interleave_lowv2si): Handle big endian.
2113         (vec_interleave_lowv2si_be): New.
2114         (vec_interleave_lowv2si_le): New.
2115         (insn_v4int_l): Handle big endian.
2116         * config/tilegx/tilegx.opt (mbig-endian): New option.
2117         (mlittle-endian): New option.
2118         * doc/install.texi: Document tilegxbe-linux.
2119         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
2121 2014-02-24  Martin Jambor  <mjambor@suse.cz>
2123         PR ipa/60266
2124         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
2125         there are no parameter descriptors.
2127 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
2129         PR rtl-optimization/60268
2130         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
2131         initialization to ...
2132         (sched_rgn_init): ... here.
2133         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
2135 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2137         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
2138         names.
2140 2014-02-23  Edgar E. Iglesias  <edgar.iglesias@xilinx.com>
2142         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
2143         definition.
2145 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2147         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
2148         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
2150 2014-02-23  David Holsgrove  <david.holsgrove@xilinx.com>
2152         * config/microblaze/predicates.md: Add cmp_op predicate.
2153         * config/microblaze/microblaze.md: Add branch_compare instruction
2154         which uses cmp_op predicate and emits cmp insn before branch.
2155         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
2156         to microblaze_expand_conditional_branch and consolidate logic.
2157         (microblaze_expand_conditional_branch): emit branch_compare
2158         insn instead of handling cmp op separate from branch insn.
2160 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2162         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
2163         to permit subregs.
2165 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2167         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
2168         define_insn with define_expand and new define_insn
2169         *altivec_lve<VI_char>x_internal.
2170         (altivec_stve<VI_char>x): Replace define_insn with define_expand
2171         and new define_insn *altivec_stve<VI_char>x_internal.
2172         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
2173         prototype.
2174         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
2175         lve*x built-ins.
2176         (altivec_expand_stvex_be): New function.
2178 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
2180         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
2181         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
2182         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
2183         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
2185 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
2187         PR target/60298
2188         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
2189         instead of emit_move_insn.
2191 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2193         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
2194         vspltw with vsldoi.
2195         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
2196         gen_altivec_vsumsws.
2198 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2200         * config/rs6000/altivec.md (altivec_lvxl): Rename as
2201         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
2202         (altivec_lvxl_<mode>): New define_expand incorporating
2203         -maltivec=be semantics where needed.
2204         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
2205         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
2206         semantics where needed.
2207         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
2208         (altivec_stvx_<mode>): New define_expand incorporating
2209         -maltivec=be semantics where needed.
2210         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
2211         VM2 iterator instead of V4SI.
2212         (altivec_stvxl_<mode>): New define_expand incorporating
2213         -maltivec=be semantics where needed.
2214         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
2215         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
2216         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
2217         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
2218         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
2219         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
2220         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
2221         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
2222         ALTIVEC_BUILTIN_STVXL.
2223         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
2224         (altivec_expand_stvx_be): Likewise.
2225         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
2226         (altivec_expand_lvx_be): Likewise.
2227         (altivec_expand_stvx_be): Likewise.
2228         (altivec_expand_builtin): Add cases for
2229         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
2230         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
2231         (altivec_init_builtins): Add definitions for
2232         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
2233         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
2235 2014-02-21  Catherine Moore  <clm@codesourcery.com>
2237         * doc/invoke.texi (mvirt, mno-virt): Document.
2238         * config/mips/mips.opt (mvirt): New option.
2239         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
2241 2014-02-21  Richard Biener  <rguenther@suse.de>
2243         PR tree-optimization/60276
2244         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
2245         (STMT_VINFO_MIN_NEG_DIST): New macro.
2246         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
2247         STMT_VINFO_MIN_NEG_DIST.
2248         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
2249         made for negative dependence distances still hold.
2251 2014-02-21  Richard Biener  <rguenther@suse.de>
2253         PR middle-end/60291
2254         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
2255         DECL_INITIAL for globals not in the current function context.
2257 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2259         PR tree-optimization/56490
2260         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
2261         * tree-ssa-uninit.c: Include params.h.
2262         (compute_control_dep_chain): Add num_calls argument, return false
2263         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
2264         num_calls to recursive call.
2265         (find_predicates): Change dep_chain into normal array,
2266         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
2267         variable and adjust compute_control_dep_chain caller.
2268         (find_def_preds): Likewise.
2270 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
2272         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
2273         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
2275 2014-02-21  Nick Clifton  <nickc@redhat.com>
2277         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
2278         (pushhi1): Likewise.
2279         (popqi1): Add mode to pre_dec.
2280         (pophi1): Likewise.
2282 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
2284         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
2285         mode for mask of V8SFmode permutation.
2287 2014-02-20  Richard Henderson  <rth@redhat.com>
2289         PR c++/60272
2290         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
2291         a new pseudo for OLDVAL.
2293 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
2295         PR target/57896
2296         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
2297         gen_reg_rtx if d->testing_p.
2298         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
2299         if d->testing_p and we will certainly return true.
2300         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
2301         if d->testing_p.
2303 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
2305         * emit-rtl.c (gen_reg_rtx): Assert that
2306         crtl->emit.regno_pointer_align_length is non-zero.
2308 2014-02-20  Richard Henderson  <rth@redhat.com>
2310         PR c++/60272
2311         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
2312         on failure the store back into EXPECT.
2314 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
2315             Sandra Loosemore  <sandra@codesourcery.com>
2317         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
2318         * config/nios2/nios2.c (nios2_function_profiler): Add
2319         -fPIC (flag_pic == 2) support.
2320         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
2321         (nios2_large_offset_p): New function.
2322         (nios2_unspec_reloc_p): Move up position, update to use
2323         nios2_large_offset_p.
2324         (nios2_unspec_address): Remove function.
2325         (nios2_unspec_offset): New function.
2326         (nios2_large_got_address): New function.
2327         (nios2_got_address): Add large offset support.
2328         (nios2_legitimize_tls_address): Update usage of removed and new
2329         functions.
2330         (nios2_symbol_binds_local_p): New function.
2331         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
2332         (nios2_legitimize_address): Update to use nios2_large_offset_p.
2333         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
2334         (nios2_print_operand): Merge H/L processing, add hiadj/lo
2335         processing for (const (unspec ...)).
2336         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
2338 2014-02-20  Richard Biener  <rguenther@suse.de>
2340         * tree-cfg.c (replace_uses_by): Mark altered BBs before
2341         doing the substitution.
2342         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
2344 2014-02-20  Martin Jambor  <mjambor@suse.cz>
2346         PR ipa/55260
2347         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
2348         info when checking whether lattices are bottom.
2350 2014-02-20  Richard Biener  <rguenther@suse.de>
2352         PR middle-end/60221
2353         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
2354         regions at -O0.
2356 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
2358         PR ipa/58555
2359         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
2360         parameter specifying the scaling.
2361         (inline_call): Update.
2362         (want_inline_recursively): Guard division by zero.
2363         (recursive_inlining): Update.
2364         * ipa-inline.h (clone_inlined_nodes): Update.
2366 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2368         PR target/60204
2369         * config/i386/i386.c (classify_argument): Pass structures of size
2370         64 bytes or less in register.
2372 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
2373             Kirill Yukhin  <kirill.yukhin@intel.com>
2375         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
2376         (_mm_rcp28_round_ss): Ditto.
2377         (_mm_rsqrt28_round_sd): Ditto.
2378         (_mm_rsqrt28_round_ss): Ditto.
2379         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
2380         (_mm_rcp14_round_ss): Ditto.
2381         (_mm_rsqrt14_round_sd): Ditto.
2382         (_mm_rsqrt14_round_ss): Ditto.
2383         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
2384         the first input operand, get rid of match_dup.
2385         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
2386         attribute to sse.
2387         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2388         Ditto.
2389         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
2390         operand as the first input operand, set type attribute.
2391         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2392         Set type attribute.
2393         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
2394         operand as the first input operand, set type attribute.
2396 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2398         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
2399         bit of zero.
2401 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
2403         PR target/60207
2404         * config/i386/i386.c (construct_container): Remove TFmode check
2405         for X86_64_INTEGER_CLASS.
2407 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
2409         PR target/59794
2410         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
2411         only when -Wpsabi is enabled.
2413 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
2415          PR target/59799
2416         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
2417         passing arrays in registers are the same as for structs, so remove the
2418         special case for them.
2420 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
2422         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
2423         destination type, extract only the valid bits if the source type is not
2424         integral and has a different mode.
2426 2014-02-19  Richard Biener  <rguenther@suse.de>
2428         PR ipa/60243
2429         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
2430         for all calls.
2432 2014-02-19  Richard Biener  <rguenther@suse.de>
2434         PR ipa/60243
2435         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
2436         (ipa_modify_call_arguments): Emit an argument load explicitely and
2437         preserve virtual SSA form there and for the replacement call.
2438         Do not update SSA form nor free dominance info.
2440 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2442         * ipa.c (function_and_variable_visibility): Also clear WEAK
2443         flag when disolving COMDAT_GROUP.
2445 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2447         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
2448         * ipa-prop.c (ipa_set_jf_known_type): Return early when
2449         not devirtualizing.
2450         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
2451         do more sanity checks.
2452         (detect_type_change): Return true when giving up early.
2453         (compute_complex_assign_jump_func): Fix type parameter of
2454         ipa_set_ancestor_jf.
2455         (compute_complex_ancestor_jump_func): Likewise.
2456         (update_jump_functions_after_inlining): Fix updating of
2457         ancestor function.
2458         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
2460 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
2462         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
2463         inline clones when edge disappears.
2465 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
2467         PR target/60203
2468         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
2469         Split 64-bit moves into 2 patterns.  Do not allow the use of
2470         direct move for TDmode in little endian, since the decimal value
2471         has little endian bytes within a word, but the 64-bit pieces are
2472         ordered in a big endian fashion, and normal subreg's of TDmode are
2473         not allowed.
2474         (mov<mode>_64bit_dm): Likewise.
2475         (movtd_64bit_nodm): Likewise.
2477 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2479         PR tree-optimization/60174
2480         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
2481         statement of an SSA_NAME that occurs in an abnormal PHI node.
2483 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2485         PR sanitizer/60142
2486         * final.c (SEEN_BB): Remove.
2487         (SEEN_NOTE, SEEN_EMITTED): Renumber.
2488         (final_scan_insn): Don't force_source_line on second
2489         NOTE_INSN_BASIC_BLOCK.
2491 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
2493         PR target/60205
2494         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
2495         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
2496         (type_natural_mode): Warn ABI change when %zmm register is not
2497         available for AVX512F vector value passing.
2499 2014-02-18  Kai Tietz  <ktietz@redhat.com>
2501         PR target/60193
2502         * config/i386/i386.c (ix86_expand_prologue): Use value in
2503         rax register as displacement when restoring %r10 or %rax.
2504         Fix wrong offset when restoring both registers.
2506 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
2508         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
2509         assertion with conditional return.
2511 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
2512             Uros Bizjak  <ubizjak@gmail.com>
2514         PR driver/60233
2515         * config/i386/driver-i386.c (host_detect_local_cpu): If
2516         YMM state is not saved by the OS, also clear has_f16c.  Move
2517         CPUID 0x80000001 handling before YMM state saving checking.
2519 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
2521         PR rtl-optimization/58960
2522         * haifa-sched.c (alloc_global_sched_pressure_data): New,
2523         factored out from ...
2524         (sched_init): ... here.
2525         (free_global_sched_pressure_data): New, factored out from ...
2526         (sched_finish): ... here.
2527         * sched-int.h (free_global_sched_pressure_data): Declare.
2528         * sched-rgn.c (nr_regions_initial): New static global.
2529         (haifa_find_rgns): Initialize it.
2530         (schedule_region): Disable sched-pressure for the newly
2531         generated regions.
2533 2014-02-17  Richard Biener  <rguenther@suse.de>
2535         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
2536         release SSA defs of pattern stmts.
2538 2014-02-17  Richard Biener  <rguenther@suse.de>
2540         * tree-inline.c (expand_call_inline): Release the virtual
2541         operand defined by the call we are about to inline.
2543 2014-02-17  Richard Biener  <rguenther@suse.de>
2545         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
2547 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
2548             Ilya Tocar  <ilya.tocar@intel.com>
2550         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
2551         arguments order in builtin.
2552         (_mm512_permutexvar_epi64): Ditto.
2553         (_mm512_mask_permutexvar_epi64): Ditto
2554         (_mm512_maskz_permutexvar_epi32): Ditto
2555         (_mm512_permutexvar_epi32): Ditto
2556         (_mm512_mask_permutexvar_epi32): Ditto
2558 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2560         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
2561         (p8_vmrgow): Likewise.
2563 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2565         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
2566         endian targets.
2568 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
2570         PR target/60203
2571         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
2572         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
2573         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
2574         using direct move instructions on ISA 2.07.  Also adjust
2575         instruction length for 64-bit.
2576         (mov<mode>_64bit, TFmode/TDmode): Likewise.
2577         (mov<mode>_32bit, TFmode/TDmode): Likewise.
2579 2014-02-15  Alan Modra  <amodra@gmail.com>
2581         PR target/58675
2582         PR target/57935
2583         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
2584         find_replacement on parts of insn rtl that might be reloaded.
2586 2014-02-15  Richard Biener  <rguenther@suse.de>
2588         PR tree-optimization/60183
2589         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
2590         (tree_ssa_phiprop): Calculate and free post-dominators.
2592 2014-02-14  Jeff Law  <law@redhat.com>
2594         PR rtl-optimization/60131
2595         * ree.c (get_extended_src_reg): New function.
2596         (combine_reaching_defs): Use it rather than assuming location of REG.
2597         (find_and_remove_re): Verify first operand of extension is
2598         a REG before adding the insns to the copy list.
2600 2014-02-14  Roland McGrath  <mcgrathr@google.com>
2602         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
2603         * configure: Regenerated.
2604         * config.in: Regenerated.
2605         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
2606         instead of ASM_SHORT.
2608 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
2609             Richard Earnshaw  <rearnsha@arm.com>
2611         PR rtl-optimization/59535
2612         * lra-constraints.c (process_alt_operands): Encourage alternative
2613         when unassigned pseudo class is superset of the alternative class.
2614         (inherit_reload_reg): Don't inherit when optimizing for code size.
2615         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
2616         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
2617         modes not less than 4 for Thumb1.
2619 2014-02-14  Kyle McMartin  <kyle@redhat.com>
2621         PR pch/60010
2622         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
2624 2014-02-14  Richard Biener  <rguenther@suse.de>
2626         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
2627         (get_frame_arg): Drop the assert with langhook types_compatible_p.
2628         Do not strip INDIRECT_REFs.
2630 2014-02-14  Richard Biener  <rguenther@suse.de>
2632         PR lto/60179
2633         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
2634         DECL_FUNCTION_SPECIFIC_TARGET.
2635         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
2636         * tree-streamer-out.c (pack_ts_target_option): Remove.
2637         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
2638         (write_ts_function_decl_tree_pointers): Do not stream
2639         DECL_FUNCTION_SPECIFIC_TARGET.
2640         * tree-streamer-in.c (unpack_ts_target_option): Remove.
2641         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
2642         (lto_input_ts_function_decl_tree_pointers): Do not stream
2643         DECL_FUNCTION_SPECIFIC_TARGET.
2645 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
2647         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
2648         (get_initial_def_for_induction, vectorizable_induction): Ignore
2649         debug stmts when looking for exit_phi.
2650         (vectorizable_live_operation): Fix up condition.
2652 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2654         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
2655         nreverse() because it changes the content of original tree list.
2657 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2659         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
2660         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
2662 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
2664         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
2665         GNU coding standards.
2667 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2669         PR debug/60152
2670         * dwarf2out.c (gen_subprogram_die): Don't call
2671         add_calling_convention_attribute if subr_die is old_die.
2673 2014-02-13  Sharad Singhai  <singhai@google.com>
2675         * doc/optinfo.texi: Fix order of nodes.
2677 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
2679         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
2680         operands[2], not operands[3].
2682 2014-02-13  Richard Biener  <rguenther@suse.de>
2684         PR bootstrap/59878
2685         * doc/install.texi (ISL): Update recommended version to 0.12.2,
2686         mention the possibility of an in-tree build.
2687         (CLooG): Update recommended version to 0.18.1, mention the
2688         possibility of an in-tree build and clarify that the ISL
2689         bundled with CLooG does not work.
2691 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
2693         PR target/43546
2694         * expr.c (compress_float_constant): If x is a hard register,
2695         extend into a pseudo and then move to x.
2697 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
2699         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
2700         caused by bad second argument to warning_at() with -mhotpatch and
2701         nested functions (e.g. with gfortran).
2703 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
2705         * opts.c (option_name): Remove "enabled by default" rider.
2707 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
2709         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
2711 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
2712             Uros Bizjak  <ubizjak@gmail.com>
2714         PR target/60151
2715         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
2716         * configure: Regenerated.
2718 2014-02-12  Richard Biener  <rguenther@suse.de>
2720         * vec.c (vec_prefix::calculate_allocation): Move as
2721         inline variant to vec.h.
2722         (vec_prefix::calculate_allocation_1): New out-of-line version.
2723         * vec.h (vec_prefix::calculate_allocation_1): Declare.
2724         (vec_prefix::m_has_auto_buf): Rename to ...
2725         (vec_prefix::m_using_auto_storage): ... this.
2726         (vec_prefix::calculate_allocation): Inline the easy cases
2727         and dispatch to calculate_allocation_1 which doesn't need the
2728         prefix address.
2729         (va_heap::reserve): Use gcc_checking_assert.
2730         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
2731         m_using_auto_storage.
2732         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
2733         member and adjust.
2734         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2735         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2736         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2738 2014-02-12  Richard Biener  <rguenther@suse.de>
2740         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2741         when we found a dependence.
2743 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2745         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2746         common code...
2747         (maybe_fold_stmt): ... into this new function.
2748         * omp-low.c (lower_omp): Update comment.
2750         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2751         last use.
2753         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2754         dereference.
2756 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2758         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2759         identifiers in comments.
2760         (cortexa53_extra_costs): Likewise.
2761         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2762         (cortexa7_extra_costs): Likewise.
2763         (cortexa12_extra_costs): Likewise.
2764         (cortexa15_extra_costs): Likewise.
2765         (v7m_extra_costs): Likewise.
2767 2014-02-12  Richard Biener  <rguenther@suse.de>
2769         PR middle-end/60092
2770         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2771         of posix_memalign being successful.
2772         (lower_stmt): Restrict lowering of posix_memalign to when
2773         -ftree-bit-ccp is enabled.
2775 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2777         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2778         arg_loc.
2779         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2781 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2783         PR rtl-optimization/60116
2784         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2785         other_insn once the combination has been validated.
2787 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2789         PR lto/59468
2790         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2791         and wrapper.
2792         * ipa-devirt.c: Include demangle.h
2793         (odr_violation_reported): New static variable.
2794         (add_type_duplicate): Update odr_violations.
2795         (maybe_record_node): Add completep parameter; update it.
2796         (record_target_from_binfo): Add COMPLETEP parameter;
2797         update it as needed.
2798         (possible_polymorphic_call_targets_1): Likewise.
2799         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2800         rename FINAL to COMPLETE.
2801         (record_targets_from_bases): Sanity check we found the binfo;
2802         fix COMPLETEP updating.
2803         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2804         parameter, fix computing of COMPLETEP.
2805         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2806         at LTO time do demangling.
2807         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2808         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2809         parameter.
2810         (gimple_get_virt_method_for_binfo): Likewise.
2811         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2812         gimple_get_virt_method_for_vtable): Update prototypes.
2814 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2816         PR target/49008
2817         * genautomata.c (add_presence_absence): Fix typo with
2818         {final_}presence_list.
2820 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2822         PR target/60137
2823         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2824         for VSX/Altivec vectors that land in GPR registers.
2826 2014-02-11  Richard Henderson  <rth@redhat.com>
2827             Jakub Jelinek  <jakub@redhat.com>
2829         PR debug/59776
2830         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2831         around drhs if type conversion to lacc->type is not useless.
2833 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2835         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2836         tuning struct.
2837         (cortex-a57.cortex-a53): Likewise.
2838         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2840 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2842         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2843         arm_restrict_it.
2845 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2847         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2848         add_options_for_arm_vfp3.
2850 2014-02-11  Jeff Law  <law@redhat.com>
2852         PR middle-end/54041
2853         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2854         object with an undesirable mode.
2856 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2858         PR libgomp/60107
2859         * config/i386/sol2-9.h: New file.
2860         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2861         *-*-solaris2.9*): Use it.
2863 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2865         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2866         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2868 2014-02-10  Nagaraju Mekala  <nagaraju.mekala@xilinx.com>
2870         * config/microblaze/microblaze.c: Extend mcpu version format
2872 2014-02-10  David Holsgrove  <david.holsgrove@xilinx.com>
2874         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2876 2014-02-10  Richard Henderson  <rth@redhat.com>
2878         PR target/59927
2879         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2880         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2881         ms-abi vs -mno-accumulate-outgoing-args.
2882         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2883         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2884         respect to ms-abi.
2886 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2888         PR middle-end/60080
2889         * cfgexpand.c (expand_asm_operands): Attach source location to
2890         ASM_INPUT rtx objects.
2891         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2893 2014-02-10  Nick Clifton  <nickc@redhat.com>
2895         * config/mn10300/mn10300.c (popcount): New function.
2896         (mn10300_expand_prologue): Include saved registers in stack usage
2897         count.
2899 2014-02-10  Jeff Law  <law@redhat.com>
2901         PR middle-end/52306
2902         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2903         when changing the SET_DEST of a prior insn to avoid an input reload.
2905 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2907         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2908         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2909         -mcall-openbsd, or -mcall-linux.
2910         (CC1_ENDIAN_BIG_SPEC): Remove.
2911         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2912         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2913         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2914         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2915         and %cc1_endian_default.
2916         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2918 2014-02-10  Richard Biener  <rguenther@suse.de>
2920         PR tree-optimization/60115
2921         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2922         MEM_REF handling.  Properly verify that the accesses are not
2923         out of the objects bound.
2925 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2927         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2928         coretex to cortex.
2930 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2932         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2933         proper constants and fix formatting.
2934         (possible_polymorphic_call_targets): Fix formatting.
2936 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2937             Ilya Tocar  <ilya.tocar@intel.com>
2939         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2940         (_mm512_loadu_epi32): Renamed into...
2941         (_mm512_loadu_si512): This.
2942         (_mm512_storeu_epi32): Renamed into...
2943         (_mm512_storeu_si512): This.
2944         (_mm512_maskz_ceil_ps): Removed.
2945         (_mm512_maskz_ceil_pd): Ditto.
2946         (_mm512_maskz_floor_ps): Ditto.
2947         (_mm512_maskz_floor_pd): Ditto.
2948         (_mm512_floor_round_ps): Ditto.
2949         (_mm512_floor_round_pd): Ditto.
2950         (_mm512_ceil_round_ps): Ditto.
2951         (_mm512_ceil_round_pd): Ditto.
2952         (_mm512_mask_floor_round_ps): Ditto.
2953         (_mm512_mask_floor_round_pd): Ditto.
2954         (_mm512_mask_ceil_round_ps): Ditto.
2955         (_mm512_mask_ceil_round_pd): Ditto.
2956         (_mm512_maskz_floor_round_ps): Ditto.
2957         (_mm512_maskz_floor_round_pd): Ditto.
2958         (_mm512_maskz_ceil_round_ps): Ditto.
2959         (_mm512_maskz_ceil_round_pd): Ditto.
2960         (_mm512_expand_pd): Ditto.
2961         (_mm512_expand_ps): Ditto.
2962         * config/i386/i386.c (ix86_builtins): Remove
2963         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2964         (bdesc_args): Ditto.
2965         * config/i386/predicates.md (const1256_operand): New.
2966         (const_1_to_2_operand): Ditto.
2967         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2968         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2969         (*avx512pf_gatherpf<mode>sf): Ditto.
2970         (avx512pf_gatherpf<mode>df): Ditto.
2971         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2972         (*avx512pf_gatherpf<mode>df): Ditto.
2973         (avx512pf_scatterpf<mode>sf): Ditto.
2974         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2975         (*avx512pf_scatterpf<mode>sf): Ditto.
2976         (avx512pf_scatterpf<mode>df): Ditto.
2977         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2978         (*avx512pf_scatterpf<mode>df): Ditto.
2979         (avx512f_expand<mode>): Removed.
2980         (<shift_insn><mode>3<mask_name>): Change predicate type.
2982 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2984         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2985         not at the end of datarefs vector use ordered_remove to avoid
2986         reordering datarefs vector.
2988         PR c/59984
2989         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2990         mark local addressable non-static vars as GOVD_PRIVATE
2991         instead of GOVD_LOCAL.
2992         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2993         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2994         than copying them.
2996         PR middle-end/60092
2997         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2998         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2999         assume_aligned or alloc_align attributes.
3000         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
3001         arguments.  Handle also assume_aligned and alloc_align attributes.
3002         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
3003         calls to functions with assume_aligned or alloc_align attributes.
3004         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
3006 2014-02-08  Terry Guo  <terry.guo@arm.com>
3008         * doc/invoke.texi: Document ARM -march=armv7e-m.
3010 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
3012         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
3013         flag on __cilkrts_rethrow builtin.
3015         PR ipa/60026
3016         * ipa-cp.c (determine_versionability): Fail at -O0
3017         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
3018         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
3020         Revert:
3021         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3023         PR ipa/60026
3024         * tree-inline.c (copy_forbidden): Fail for
3025         __attribute__((optimize (0))) functions.
3027 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3029         * varpool.c: Include pointer-set.h.
3030         (varpool_remove_unreferenced_decls): Variables in other partitions
3031         will not be output; be however careful to not lose information
3032         about partitioning.
3034 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
3036         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
3037         lookup in the vtable constructor.
3039 2014-02-07  Jeff Law  <law@redhat.com>
3041         PR target/40977
3042         * config/m68k/m68k.md (ashldi_extsi): Turn into a
3043         define_insn_and_split.
3045         * ipa-inline.c (inline_small_functions): Fix typos.
3047 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3049         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
3050         (s390_can_use_return_insn): Declare.
3051         * config/s390/s390.h (EPILOGUE_USES): Define.
3052         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
3053         instructions.
3054         (s390_chunkify_start): Handle return JUMP_LABELs.
3055         (s390_early_mach): Emit a main_pool instruction on the entry edge.
3056         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
3057         (s390_can_use_return_insn): New functions.
3058         (s390_fix_long_loop_prediction): Handle conditional returns.
3059         (TARGET_SET_UP_BY_PROLOGUE): Define.
3060         * config/s390/s390.md (ANY_RETURN): New code iterator.
3061         (*creturn, *csimple_return, return, simple_return): New patterns.
3063 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3065         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
3066         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
3067         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
3068         REG_CFA_RESTORE list when deciding not to restore a register.
3070 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3072         * config/s390/s390.c: Include tree-pass.h and context.h.
3073         (s390_early_mach): New function, split out from...
3074         (s390_emit_prologue): ...here.
3075         (pass_data_s390_early_mach): New pass structure.
3076         (pass_s390_early_mach): New class.
3077         (s390_option_override): Create and register early_mach pass.
3078         Move to end of file.
3080 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3082         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
3083         to match for the exit block.
3085 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3087         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
3088         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
3089         Reject misaligned operands.
3091 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3093         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
3095 2014-02-07  Richard Biener  <rguenther@suse.de>
3097         PR middle-end/60092
3098         * gimple-low.c (lower_builtin_posix_memalign): New function.
3099         (lower_stmt): Call it to lower posix_memalign in a way
3100         to make alignment info accessible.
3102 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
3104         PR c++/60082
3105         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
3106         __builtin_setjmp_receiver.
3108 2014-02-07  Richard Biener  <rguenther@suse.de>
3110         PR middle-end/60092
3111         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
3112         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
3113         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
3114         Handle BUILT_IN_POSIX_MEMALIGN.
3115         (find_func_clobbers): Likewise.
3116         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
3117         (call_may_clobber_ref_p_1): Likewise.
3119 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3121         PR ipa/59918
3122         * ipa-devirt.c (record_target_from_binfo): Remove overactive
3123         sanity check.
3125 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3127         PR ipa/59469
3128         * lto-cgraph.c (lto_output_node): Use
3129         symtab_get_symbol_partitioning_class.
3130         (lto_output_varpool_node): likewise.
3131         (symtab_get_symbol_partitioning_class): Move here from
3132         lto/lto-partition.c
3133         * cgraph.h (symbol_partitioning_class): Likewise.
3134         (symtab_get_symbol_partitioning_class): Declare.
3136 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3138         * ggc.h (ggc_internal_cleared_alloc): New macro.
3139         * vec.h (vec_safe_copy): Handle memory stats.
3140         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
3141         * target-globals.c (save_target_globals): Likewise.
3143 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
3145         PR target/60077
3146         * expr.c (emit_move_resolve_push): Export; be bit more selective
3147         on when to clear alias set.
3148         * expr.h (emit_move_resolve_push): Declare.
3149         * function.h (struct function): Add tail_call_marked.
3150         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
3151         * config/i386/i386-protos.h (ix86_expand_push): Remove.
3152         * config/i386/i386.md (TImode move expander): De not call
3153         ix86_expand_push.
3154         (FP push expanders): Preserve memory attributes.
3155         * config/i386/sse.md (push<mode>1): Remove.
3156         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
3157         (ix86_expand_push): Remove.
3158         * config/i386/mmx.md (push<mode>1): Remove.
3160 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3162         PR rtl-optimization/60030
3163         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
3164         lopart with paradoxical subreg before shifting it up by hprec.
3166 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3168         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
3169         Remove extra newline at end of file.
3170         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
3171         (arm_issue_rate): Handle cortexa57.
3172         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
3173         (cortex-a57.cortex-a53): Likewise.
3175 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3177         PR target/59575
3178         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
3179         don't record in REG_FRAME_RELATED_EXPR registers not set in that
3180         bitmask.
3181         (arm_expand_prologue): Adjust all callers.
3182         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
3183         info, registers also at the lowest numbered registers side.  Use
3184         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
3185         XEXP.
3187         PR debug/59992
3188         * var-tracking.c (adjust_mems): Before adding a SET to
3189         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
3191 2014-02-06  Alan Modra  <amodra@gmail.com>
3193         PR target/60032
3194         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
3195         change SDmode to DDmode when lra_in_progress.
3197 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
3199         PR middle-end/59150
3200         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
3201         free_data_ref on the dr first, and before goto again also set dr
3202         to the next dr.  For simd_lane_access, free old datarefs[i] before
3203         overwriting it.  For get_vectype_for_scalar_type failure, don't
3204         free_data_ref if simd_lane_access.
3206         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
3208         PR target/60062
3209         * tree.h (opts_for_fn): New inline function.
3210         (opt_for_fn): Define.
3211         * config/i386/i386.c (ix86_function_regparm): Use
3212         opt_for_fn (decl, optimize) instead of optimize.
3214 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3216         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
3217         for SYMBOL_REF in large memory model.
3219 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3221         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
3222         and crypto support.
3223         (cortex-a57): Likewise.
3224         (cortex-a57.cortex-a53): Likewise.
3226 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
3227             Kugan Vivekanandarajah  <kuganv@linaro.org>
3229         * config/arm/arm.c (arm_vector_alignment_reachable): Check
3230         unaligned_access.
3231         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
3233 2014-02-06  Richard Biener  <rguenther@suse.de>
3235         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
3236         set_loop_copy and initialize_original_copy_tables.
3238 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
3240         * config/aarch64/aarch64-simd.md
3241         (aarch64_ashr_simddi): Change QI to SI.
3243 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3244             Jakub Jelinek  <jakub@redhat.com>
3246         PR middle-end/60013
3247         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
3248         of the dataflow.
3250 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3252         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
3253         CODE_FOR_altivec_vpku[hw]um to
3254         CODE_FOR_altivec_vpku[hw]um_direct.
3255         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
3256         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
3257         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
3258         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
3260 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3262         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
3263         generation for -maltivec=be.
3264         (altivec_vsumsws): Simplify redundant test.
3266 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3268         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
3269         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
3270         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
3271         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
3272         gen_altivec_vpkuwum.
3273         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
3274         BYTES_BIG_ENDIAN.
3275         (altivec_vpks<VI_char>ss): Likewise.
3276         (altivec_vpks<VI_char>us): Likewise.
3277         (altivec_vpku<VI_char>us): Likewise.
3278         (altivec_vpku<VI_char>um): Likewise.
3279         (altivec_vpku<VI_char>um_direct): New (copy of
3280         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
3281         internal use).
3282         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
3283         target is little endian and -maltivec=be is not specified.
3284         (*altivec_vupkhs<VU_char>_direct): New (copy of
3285         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
3286         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
3287         target is little endian and -maltivec=be is not specified.
3288         (*altivec_vupkls<VU_char>_direct): New (copy of
3289         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
3290         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
3291         little endian and -maltivec=be is not specified.
3292         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
3293         little endian and -maltivec=be is not specified.
3295 2014-02-05  Richard Henderson  <rth@redhat.com>
3297         PR debug/52727
3298         * combine-stack-adj.c: Revert r206943.
3299         * sched-int.h (struct deps_desc): Add last_args_size.
3300         * sched-deps.c (init_deps): Initialize it.
3301         (sched_analyze_insn): Add OUTPUT dependencies between insns that
3302         contain REG_ARGS_SIZE notes.
3304 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
3306         * lto-cgraph.c (asm_nodes_output): Make global.
3307         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
3308         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
3309         (driver_handle_option): Handle OPT_fwpa.
3311 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
3313         PR ipa/59947
3314         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
3315         a comment typo and formatting issue.  If odr_hash hasn't been
3316         created, return vNULL and set *completep to false.
3318         PR middle-end/57499
3319         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
3320         bb with no successors.
3322 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
3324         PR target/59718
3325         * doc/invoke.texi (-march): Clarify documentation for ARM.
3326         (-mtune): Likewise.
3327         (-mcpu): Likewise.
3329 2014-02-05  Richard Biener  <rguenther@suse.de>
3331         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
3332         when not vectorizing because of too many alias checks.
3333         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
3334         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
3336 2014-02-05  Nick Clifton  <nickc@redhat.com>
3338         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
3339         accept extended registers in any mode when compiling for the MN10300.
3341 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
3343         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
3344         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
3345         sanitization attributes.
3346         (can_inline_edge_p): Likewise.
3347         (sanitize_attrs_match_for_inline_p): New function.
3349 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3351         * ipa-prop.c (detect_type_change): Shor circuit testing of
3352         type changes on THIS pointer.
3354 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
3356         PR target/59777
3357         * config/pa/pa.c (legitimize_tls_address): Return original address
3358         if not passed a SYMBOL_REF rtx.
3359         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
3360         addresses.
3361         (pa_emit_move_sequence): Simplify TLS source operands.
3362         (pa_legitimate_constant_p): Reject all TLS constants.
3363         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
3364         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
3366 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3368         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
3369         groups when we know they are controlled by LTO.
3370         * varasm.c (default_binds_local_p_1): If object is in other partition,
3371         it will be resolved locally.
3373 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3375         * config/host-linux.c (linux_gt_pch_use_address): Don't
3376         use SSIZE_MAX because it is not always defined.
3378 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
3380         PR bootstrap/59913
3381         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
3382         threshold for pseudo splitting.
3383         (update_ebb_live_info): Process call argument hard registers and
3384         hard registers from insn definition too.
3385         (max_small_class_regs_num): New constant.
3386         (inherit_in_ebb): Update live hard regs through EBBs.  Update
3387         reloads_num only for small register classes.  Don't split for
3388         outputs of jumps.
3390 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
3392         PR ipa/60058
3393         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
3394         is non-null.
3396 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
3398         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
3399         visibility is safe.
3401 2014-02-04  Marek Polacek  <polacek@redhat.com>
3403         * gdbinit.in (pel): Define.
3405 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3407         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
3408         behavior.
3410 2014-02-04  Richard Biener  <rguenther@suse.de>
3412         PR lto/59723
3413         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
3414         in function context local.
3415         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
3416         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
3417         similar to LTO_imported_decl_ref.
3419 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
3421         PR tree-optimization/60002
3422         * cgraphclones.c (build_function_decl_skip_args): Clear
3423         DECL_LANG_SPECIFIC.
3425         PR tree-optimization/60023
3426         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
3427         false to gsi_replace.
3428         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
3429         has been in some EH region and vec_stmt could throw, add
3430         vec_stmt into the same EH region.
3431         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
3432         has no lhs, ignore it.
3433         * internal-fn.c (expand_MASK_LOAD): Likewise.
3435         PR ipa/60026
3436         * tree-inline.c (copy_forbidden): Fail for
3437         __attribute__((optimize (0))) functions.
3439         PR other/58712
3440         * omp-low.c (simd_clone_struct_copy): If from->inbranch
3441         is set, copy one less argument.
3442         (expand_simd_clones): Don't subtract clone_info->inbranch
3443         from simd_clone_struct_alloc argument.
3445         PR rtl-optimization/57915
3446         * recog.c (simplify_while_replacing): If all unary/binary/relational
3447         operation arguments are constant, attempt to simplify those.
3449         PR middle-end/59261
3450         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
3451         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
3453 2014-02-04  Richard Biener  <rguenther@suse.de>
3455         PR tree-optimization/60012
3456         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
3457         TBAA disambiguation to all DDRs.
3459 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3461         PR target/59788
3462         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
3463         (LINK_SPEC): Use it for -shared, -shared-libgcc.
3465 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3467         PR ipa/59882
3468         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
3470 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3472         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
3473         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
3475 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3477         PR ipa/59831
3478         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
3479         to figure out targets of polymorphic calls with known decl.
3480         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3481         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
3482         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
3483         (get_polymorphic_call_info): ... here.
3484         (get_polymorphic_call_info_from_invariant): New function.
3486 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3488         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
3489         lookup via vtable pointer; check for type consistency
3490         and turn inconsitent facts into UNREACHABLE.
3491         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3492         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
3493         type inconsistent querries; return UNREACHABLE instead.
3495 2014-02-03  Richard Henderson  <rth@twiddle.net>
3497         PR tree-opt/59924
3498         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
3499         already processed this node.
3500         (normalize_one_pred_1): Pass along mark_set.
3501         (normalize_one_pred): Create and destroy a pointer_set_t.
3502         (normalize_one_pred_chain): Likewise.
3504 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
3506         PR gcov-profile/58602
3507         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
3509 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
3511         PR ipa/59831
3512         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
3513         -fno-devirtualize; try to devirtualize by the knowledge of
3514         virtual table pointer given by aggregate propagation.
3515         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
3516         (ipa_print_node_jump_functions): Dump also offset that
3517         is relevant for polymorphic calls.
3518         (determine_known_aggregate_parts): Add arg_type parameter; use it
3519         instead of determining the type from pointer type.
3520         (ipa_compute_jump_functions_for_edge): Update call of
3521         determine_known_aggregate_parts.
3522         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
3523         (gimple_get_virt_method_for_binfo): ... here; simplify using
3524         vtable_pointer_value_to_vtable.
3525         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
3526         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
3527         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
3528         (vtable_pointer_value_to_vtable): Break out from ...; handle also
3529         POINTER_PLUS_EXPR.
3530         (vtable_pointer_value_to_binfo): ... here.
3531         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
3533 2014-02-03  Teresa Johnson  <tejohnson@google.com>
3535         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
3536         redef of outer loop index variable.
3538 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
3540         PR c++/53017
3541         PR c++/59211
3542         * doc/extend.texi (Function Attributes): Typo.
3544 2014-02-03  Cong Hou  <congh@google.com>
3546         PR tree-optimization/60000
3547         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
3548         if the vectorized statement is a store.  A store statement can only
3549         appear at the end of pattern statements.
3551 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3553         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
3554         (ix86_option_override_internal): Default long double to 64-bit for
3555         32-bit Bionic and to 128-bit for 64-bit Bionic.
3557         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
3558         TARGET_LONG_DOUBLE_128 is true.
3559         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
3561         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
3562         (mlong-double-64): Negate -mlong-double-128.
3563         (mlong-double-128): New option.
3565         * config/i386/i386-c.c (ix86_target_macros): Define
3566         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
3568         * doc/invoke.texi: Document -mlong-double-128.
3570 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
3572         PR rtl-optimization/60024
3573         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
3575 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
3577         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
3579 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3581         PR rtl-optimization/57662
3582         * sel-sched.c (code_motion_path_driver): Do not mark already not
3583         existing blocks in the visiting bitmap.
3585 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
3587         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
3588         on the insn being emitted.
3590 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
3591             Will Deacon  <will.deacon@arm.com>
3593         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
3595 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3597         * config/arm/arm-tables.opt: Regenerate.
3599 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3601         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
3602         for vector types other than V16QImode.
3603         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
3604         define_expand, and call altivec_expand_vec_perm_le when producing
3605         code with little endian element order.
3606         (*altivec_vperm_<mode>_internal): New insn having previous
3607         behavior of altivec_vperm_<mode>.
3608         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
3609         altivec_expand_vec_perm_le when producing code with little endian
3610         element order.
3611         (*altivec_vperm_<mode>_uns_internal): New insn having previous
3612         behavior of altivec_vperm_<mode>_uns.
3614 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3616         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
3617         (altivec_vsumsws): Add handling for -maltivec=be with a little
3618         endian target.
3619         (altivec_vsumsws_direct): New.
3620         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
3621         gen_altivec_vsumsws.
3623 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3625         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
3626         vtable_pointer_value_to_binfo): New functions.
3627         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
3628         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
3630 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
3632         * config/nios2/nios2.md (load_got_register): Initialize GOT
3633         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
3634         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
3636 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
3638         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
3639         preserverd by passthrough, do not propagate the type.
3641 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3643         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
3644         (mips_atomic_assign_expand_fenv): New function.
3645         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
3647 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3649         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
3650         (__builtin_mips_set_fcsr): Likewise.
3651         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
3652         MIPS_USI_FTYPE_VOID.
3653         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
3654         (mips16_expand_set_fcsr): Likewise.
3655         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
3656         (mips16_set_fcsr_stub): Likewise.
3657         (mips16_get_fcsr_one_only_stub): New class.
3658         (mips16_set_fcsr_one_only_stub): Likewise.
3659         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
3660         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
3661         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
3662         (hard_float): New availability predicate.
3663         (mips_builtins): Add get_fcsr and set_fcsr.
3664         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
3665         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
3666         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
3667         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
3668         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
3669         patterns.
3671 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
3673         * config/mips/mips.c (mips_one_only_stub): New class.
3674         (mips_need_mips16_rdhwr_p): Replace with...
3675         (mips16_rdhwr_stub): ...this new variable.
3676         (mips16_stub_call_address): New function.
3677         (mips16_rdhwr_one_only_stub): New class.
3678         (mips_expand_thread_pointer): Use mips16_stub_call_address.
3679         (mips_output_mips16_rdhwr): Delete.
3680         (mips_finish_stub): New function.
3681         (mips_code_end): Use it to handle rdhwr stubs.
3683 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
3685         PR target/60017
3686         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
3687         when calculating size of integer atomic types.
3689 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
3691         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
3693 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
3695         PR tree-optimization/60003
3696         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
3697         * profile.c (branch_prob): Use gimple_call_builtin_p
3698         to check for BUILT_IN_SETJMP_RECEIVER.
3699         * tree-inline.c (copy_bb): Call notice_special_calls.
3701 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
3703         PR bootstrap/59985
3704         * lra-constraints.c (process_alt_operands): Update reload_sum only
3705         on the first pass.
3707 2014-01-31  Richard Henderson  <rth@redhat.com>
3709         PR middle-end/60004
3710         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
3711         until after else_eh is processed.
3713 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3715         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
3716         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
3717         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
3718         in smmintrin.h, remove them.
3719         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
3720         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
3721         * config/i386/i386.md (ROUND_SAE): Fix value.
3722         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
3723         (const48_operand): New.
3724         * config/i386/subst.md (round), (round_expand): Use
3725         const_4_or_8_to_11_operand.
3726         (round_saeonly), (round_saeonly_expand): Use const48_operand.
3728 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
3730         * config/i386/constraints.md (Yk): Swap meaning with k.
3731         * config/i386/i386.md (movhi_internal): Change Yk to k.
3732         (movqi_internal): Ditto.
3733         (*k<logic><mode>): Ditto.
3734         (*andhi_1): Ditto.
3735         (*andqi_1): Ditto.
3736         (kandn<mode>): Ditto.
3737         (*<code>hi_1): Ditto.
3738         (*<code>qi_1): Ditto.
3739         (kxnor<mode>): Ditto.
3740         (kortestzhi): Ditto.
3741         (kortestchi): Ditto.
3742         (kunpckhi): Ditto.
3743         (*one_cmplhi2_1): Ditto.
3744         (*one_cmplqi2_1): Ditto.
3745         * config/i386/sse.md (): Change k to Yk.
3746         (avx512f_load<mode>_mask): Ditto.
3747         (avx512f_blendm<mode>): Ditto.
3748         (avx512f_store<mode>_mask): Ditto.
3749         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3750         (avx512f_storedqu<mode>_mask): Ditto.
3751         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3752         Ditto.
3753         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3754         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3755         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3756         (avx512f_maskcmp<mode>3): Ditto.
3757         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3758         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3759         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3760         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3761         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3762         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3763         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3764         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3765         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3766         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3767         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3768         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3769         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3770         (vec_extract_lo_<mode>_maskm): Ditto.
3771         (vec_extract_hi_<mode>_maskm): Ditto.
3772         (avx512f_vternlog<mode>_mask): Ditto.
3773         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3774         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3775         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3776         (avx512f_<code>v8div16qi2_mask): Ditto.
3777         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3778         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3779         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3780         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3781         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3782         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3783         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3784         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3785         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3786         (avx512cd_maskb_vec_dupv8di): Ditto.
3787         (avx512cd_maskw_vec_dupv16si): Ditto.
3788         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3789         (avx512f_vpermi2var<mode>3_mask): Ditto.
3790         (avx512f_vpermi2var<mode>3_mask): Ditto.
3791         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3792         (*avx512f_gathersi<mode>): Ditto.
3793         (*avx512f_gathersi<mode>_2): Ditto.
3794         (*avx512f_gatherdi<mode>): Ditto.
3795         (*avx512f_gatherdi<mode>_2): Ditto.
3796         (*avx512f_scattersi<mode>): Ditto.
3797         (*avx512f_scatterdi<mode>): Ditto.
3798         (avx512f_compress<mode>_mask): Ditto.
3799         (avx512f_compressstore<mode>_mask): Ditto.
3800         (avx512f_expand<mode>_mask): Ditto.
3801         * config/i386/subst.md (mask): Change k to Yk.
3802         (mask_scalar_merge): Ditto.
3803         (sd): Ditto.
3805 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3807         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3809 2014-01-31  Richard Biener  <rguenther@suse.de>
3811         PR middle-end/59990
3812         * builtins.c (fold_builtin_memory_op): Make sure to not
3813         use a floating-point mode or a boolean or enumeral type for
3814         the copy operation.
3816 2014-01-30  DJ Delorie  <dj@redhat.com>
3818         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3819         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3820         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3821         whenever main() has an epilogue.
3823 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3825         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3826         unused variable "field".
3827         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3828         (vsx_mergeh_<mode>): Likewise.
3829         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3830         (altivec_vmrghh): Likewise.
3831         (altivec_vmrghw): Likewise.
3832         (altivec_vmrglb): Likewise.
3833         (altivec_vmrglh): Likewise.
3834         (altivec_vmrglw): Likewise.
3835         (altivec_vspltb): Add missing uses.
3836         (altivec_vsplth): Likewise.
3837         (altivec_vspltw): Likewise.
3838         (altivec_vspltsf): Likewise.
3840 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3842         PR target/59923
3843         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3844         frame related instructions.
3846 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3848         PR rtl-optimization/59959
3849         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3850         any reload of register whose subreg is invalid.
3852 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3854         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3855         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3856         Add missing return type - void.
3858 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3860         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3861         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3862         remove element index adjustment for endian (now handled in vsx.md
3863         and altivec.md).
3864         (altivec_expand_vec_perm_const): Use
3865         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3866         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3867         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3868         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3869         define_expand and a new define_insn *altivec_vspltb_internal;
3870         adjust for -maltivec=be on a little endian target.
3871         (altivec_vspltb_direct): New.
3872         (altivec_vsplth): Divide into a define_expand and a new
3873         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3874         little endian target.
3875         (altivec_vsplth_direct): New.
3876         (altivec_vspltw): Divide into a define_expand and a new
3877         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3878         little endian target.
3879         (altivec_vspltw_direct): New.
3880         (altivec_vspltsf): Divide into a define_expand and a new
3881         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3882         a little endian target.
3884 2014-01-30  Richard Biener  <rguenther@suse.de>
3886         PR tree-optimization/59993
3887         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3888         can propagate form the earlier stmt and avoid the transform
3889         when the intermediate result is needed.
3891 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3893         * README.Portability: Fix typo.
3895 2014-01-30  David Holsgrove  <david.holsgrove@xilinx.com>
3897         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3898         comparison_operator with ordered_comparison_operator.
3900 2014-01-30  Nick Clifton  <nickc@redhat.com>
3902         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3903         Rename to mn10300_store_multiple_regs.
3904         * config/mn10300/mn10300.c: Likewise.
3905         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3906         store_multiple_regs.
3907         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3908         Call mn10300_store_multiple_regs.
3910 2014-01-30  Nick Clifton  <nickc@redhat.com>
3911             DJ Delorie  <dj@redhat.com>
3913         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3914         %fp 2 to keep registers after it properly word-aligned.
3915         (rl78_alloc_physical_registers_umul): Handle the case where both
3916         input operands are the same.
3918 2014-01-30  Richard Biener  <rguenther@suse.de>
3920         PR tree-optimization/59903
3921         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3922         check properly.
3924 2014-01-30  Jason Merrill  <jason@redhat.com>
3926         PR c++/59633
3927         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3929         PR c++/59645
3930         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3932 2014-01-30  Richard Biener  <rguenther@suse.de>
3934         PR tree-optimization/59951
3935         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3937 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3939         PR target/59784
3940         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3941         SFmode to DFmode case.
3943 2014-01-29  DJ Delorie  <dj@redhat.com>
3945         * config/msp430/msp430.opt (-minrt): New.
3946         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3947         if -minrt given.
3948         (ENDFILE_SPEC): Likewise.
3950 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3952         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3953         (estimate_function_body_sizes): Use it.
3955 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3957         PR c++/58561
3958         * dwarf2out.c (is_cxx_auto): New.
3959         (is_base_type): Use it.
3960         (gen_type_die_with_usage): Likewise.
3962 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3964         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3965         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3966         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3967         -maltivec=be with LE targets.
3968         (vsx_mergeh_<mode>): Likewise.
3969         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3970         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3971         (altivec_vmrghb): Replace with define_expand and new
3972         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3973         (altivec_vmrghb_direct): New define_insn.
3974         (altivec_vmrghh): Replace with define_expand and new
3975         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3976         (altivec_vmrghh_direct): New define_insn.
3977         (altivec_vmrghw): Replace with define_expand and new
3978         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3979         (altivec_vmrghw_direct): New define_insn.
3980         (*altivec_vmrghsf): Adjust for endianness.
3981         (altivec_vmrglb): Replace with define_expand and new
3982         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3983         (altivec_vmrglb_direct): New define_insn.
3984         (altivec_vmrglh): Replace with define_expand and new
3985         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3986         (altivec_vmrglh_direct): New define_insn.
3987         (altivec_vmrglw): Replace with define_expand and new
3988         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3989         (altivec_vmrglw_direct): New define_insn.
3990         (*altivec_vmrglsf): Adjust for endianness.
3991         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3992         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3993         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3994         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3995         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3996         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3997         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3998         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
4000 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4002         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
4003         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
4004         whitespace.
4006 2014-01-29  Richard Biener  <rguenther@suse.de>
4008         PR tree-optimization/58742
4009         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
4010         associate_pointerplus_align.
4011         (associate_pointerplus_diff): New function.
4012         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
4013         and associate_pointerplus_diff.
4015 2014-01-29  Richard Biener  <rguenther@suse.de>
4017         * lto-streamer.h (LTO_major_version): Bump to 3.
4018         (LTO_minor_version): Reset to 0.
4020 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
4022         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
4023         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
4024         (arm_file_start): Generate correct asm header for armv7ve.
4025         * config/arm/bpabi.h: Add multilib support for armv7ve.
4026         * config/arm/driver-arm.c: Change the architectures of cortex-a7
4027         and cortex-a15 to armv7ve.
4028         * config/arm/t-aprofile: Add multilib support for armv7ve.
4029         * doc/invoke.texi: Document -march=armv7ve.
4031 2014-01-29  Richard Biener  <rguenther@suse.de>
4033         PR tree-optimization/58742
4034         * tree-ssa-forwprop.c (associate_plusminus): Return true
4035         if we changed sth, defer EH cleanup to ...
4036         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
4037         (simplify_mult): New function.
4039 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4041         PR middle-end/59917
4042         PR tree-optimization/59920
4043         * tree.c (build_common_builtin_nodes): Remove
4044         __builtin_setjmp_dispatcher initialization.
4045         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
4046         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
4047         instead of gsi_after_labels + manually skipping debug stmts.
4048         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
4049         ignore bbs with IFN_ABNORMAL_DISPATCHER.
4050         * tree-inline.c (copy_edges_for_bb): Remove
4051         can_make_abnormal_goto argument, instead add abnormal_goto_dest
4052         argument.  Ignore computed_goto_p stmts.  Don't call
4053         make_abnormal_goto_edges.  If a call might need abnormal edges
4054         for non-local gotos, see if it already has an edge to
4055         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
4056         with true argument, don't do anything then, otherwise add
4057         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
4058         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
4059         caller.
4060         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
4061         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
4062         (lower_stmt): Don't set data->calls_builtin_setjmp.
4063         (lower_builtin_setjmp): Adjust comment.
4064         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
4065         * tree-cfg.c (found_computed_goto): Remove.
4066         (factor_computed_gotos): Remove.
4067         (make_goto_expr_edges): Return bool, true for computed gotos.
4068         Don't call make_abnormal_goto_edges.
4069         (build_gimple_cfg): Don't set found_computed_goto, don't call
4070         factor_computed_gotos.
4071         (computed_goto_p): No longer static.
4072         (make_blocks): Don't set found_computed_goto.
4073         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
4074         (make_edges): If make_goto_expr_edges returns true, push bb
4075         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
4076         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
4077         vector.  Record mapping between bbs and OpenMP regions if there
4078         are any, adjust make_gimple_omp_edges caller.  Call
4079         handle_abnormal_edges.
4080         (make_abnormal_goto_edges): Remove.
4081         * tree-cfg.h (make_abnormal_goto_edges): Remove.
4082         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
4083         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
4084         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
4085         * internal-fn.def (ABNORMAL_DISPATCHER): New.
4086         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
4087         filling *region also set *region_idx to (*region)->entry->index.
4089         PR other/58712
4090         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
4091         For REGs set ORIGINAL_REGNO.
4093 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
4095         * doc/md.texi: Mention that a target shouldn't implement
4096         vec_widen_(s|u)mul_even/odd pair if it is less efficient
4097         than hi/lo pair.
4099 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
4101         PR tree-optimization/59594
4102         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
4103         a copy of the datarefs vector rather than the vector itself.
4105 2014-01-28  Jason Merrill  <jason@redhat.com>
4107         PR c++/53756
4108         * dwarf2out.c (auto_die): New static.
4109         (gen_type_die_with_usage): Handle C++1y 'auto'.
4110         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
4111         on definition.
4113 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
4115         PR target/59672
4116         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
4117         (SPEC_X32): Likewise.
4118         (SPEC_64): Likewise.
4119         * config/i386/i386.c (ix86_option_override_internal): Turn off
4120         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
4121         for TARGET_16BIT.
4122         (x86_file_start): Output .code16gcc for TARGET_16BIT.
4123         * config/i386/i386.h (TARGET_16BIT): New macro.
4124         (TARGET_16BIT_P): Likewise.
4125         * config/i386/i386.opt: Add m16.
4126         * doc/invoke.texi: Document -m16.
4128 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
4130         PR preprocessor/59935
4131         * input.c (location_get_source_line): Bail out on when line number
4132         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
4134 2014-01-28  Richard Biener  <rguenther@suse.de>
4136         PR tree-optimization/58742
4137         * tree-ssa-forwprop.c (associate_plusminus): Handle
4138         pointer subtraction of the form (T)(P + A) - (T)P.
4140 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4142         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
4143         at const_int_cost.
4145 2014-01-28  Richard Biener  <rguenther@suse.de>
4147         Revert
4148         2014-01-28  Richard Biener  <rguenther@suse.de>
4150         PR rtl-optimization/45364
4151         PR rtl-optimization/59890
4152         * var-tracking.c (local_get_addr_clear_given_value): Handle
4153         already cleared slot.
4154         (val_reset): Handle not allocated local_get_addr_cache.
4155         (vt_find_locations): Use post-order on the inverted CFG.
4157 2014-01-28  Richard Biener  <rguenther@suse.de>
4159         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
4161 2014-01-28  Richard Biener  <rguenther@suse.de>
4163         PR rtl-optimization/45364
4164         PR rtl-optimization/59890
4165         * var-tracking.c (local_get_addr_clear_given_value): Handle
4166         already cleared slot.
4167         (val_reset): Handle not allocated local_get_addr_cache.
4168         (vt_find_locations): Use post-order on the inverted CFG.
4170 2014-01-28  Alan Modra  <amodra@gmail.com>
4172         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
4173         * configure.ac <recursive call for build != host>: Define
4174         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
4175         and LD_FOR_BUILD too.
4176         * configure: Regenerate.
4178 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
4180         * config/i386/i386.c (get_builtin_code_for_version): Separate
4181         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
4182         Broadwell from Haswell.
4184 2014-01-27  Steve Ellcey  <sellcey@mips.com>
4186         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4187         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
4188         * config/mips/mips.c (mips_option_override): Change setting
4189         of TARGET_DSP.
4190         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
4191         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
4192         Change from Mask to Var.
4194 2014-01-27  Jeff Law  <law@redhat.com>
4196         * ipa-inline.c (inline_small_functions): Fix typo.
4198 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
4200         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
4201         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
4202         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
4203         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
4204         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
4205         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
4206         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
4207         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
4208         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
4209         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
4210         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
4211         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
4212         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
4213         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
4214         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
4215         (_mm512_storeu_epi64): Ditto.
4216         (_mm512_cmpge_epi32_mask): Ditto.
4217         (_mm512_cmpge_epu32_mask): Ditto.
4218         (_mm512_cmpge_epi64_mask): Ditto.
4219         (_mm512_cmpge_epu64_mask): Ditto.
4220         (_mm512_cmple_epi32_mask): Ditto.
4221         (_mm512_cmple_epu32_mask): Ditto.
4222         (_mm512_cmple_epi64_mask): Ditto.
4223         (_mm512_cmple_epu64_mask): Ditto.
4224         (_mm512_cmplt_epi32_mask): Ditto.
4225         (_mm512_cmplt_epu32_mask): Ditto.
4226         (_mm512_cmplt_epi64_mask): Ditto.
4227         (_mm512_cmplt_epu64_mask): Ditto.
4228         (_mm512_cmpneq_epi32_mask): Ditto.
4229         (_mm512_cmpneq_epu32_mask): Ditto.
4230         (_mm512_cmpneq_epi64_mask): Ditto.
4231         (_mm512_cmpneq_epu64_mask): Ditto.
4232         (_mm512_expand_pd): Ditto.
4233         (_mm512_expand_ps): Ditto.
4234         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
4235         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
4236         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
4237         * config/i386/i386.c (ix86_builtins): Add
4238         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
4239         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
4240         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
4241         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
4242         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
4243         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
4244         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
4245         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
4246         IX86_BUILTIN_PMOVUSQW512_MEM.
4247         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
4248         __builtin_ia32_pmovsqd512mem_mask,
4249         __builtin_ia32_pmovqd512mem_mask,
4250         __builtin_ia32_pmovusqw512mem_mask,
4251         __builtin_ia32_pmovsqw512mem_mask,
4252         __builtin_ia32_pmovqw512mem_mask,
4253         __builtin_ia32_pmovusdw512mem_mask,
4254         __builtin_ia32_pmovsdw512mem_mask,
4255         __builtin_ia32_pmovdw512mem_mask,
4256         __builtin_ia32_pmovqb512mem_mask,
4257         __builtin_ia32_pmovusqb512mem_mask,
4258         __builtin_ia32_pmovsqb512mem_mask,
4259         __builtin_ia32_pmovusdb512mem_mask,
4260         __builtin_ia32_pmovsdb512mem_mask,
4261         __builtin_ia32_pmovdb512mem_mask.
4262         (bdesc_args): Add __builtin_ia32_expanddf512,
4263         __builtin_ia32_expandsf512.
4264         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
4265         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
4266         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
4267         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
4268         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
4269         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
4270         (avx512f_<code>v8div16qi2_mask_store): This.
4271         (avx512f_expand<mode>): New.
4273 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
4275         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
4276         New.
4277         (_mm512_mask_prefetch_i64gather_pd): Ditto.
4278         (_mm512_prefetch_i32scatter_pd): Ditto.
4279         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
4280         (_mm512_prefetch_i64scatter_pd): Ditto.
4281         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
4282         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
4283         (_mm512_mask_prefetch_i64gather_ps): Ditto.
4284         (_mm512_prefetch_i32scatter_ps): Ditto.
4285         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
4286         (_mm512_prefetch_i64scatter_ps): Ditto.
4287         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
4288         * config/i386/i386-builtin-types.def: Define
4289         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
4290         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
4291         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
4292         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
4293         IX86_BUILTIN_SCATTERPFQPD.
4294         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
4295         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
4296         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
4297         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
4298         __builtin_ia32_scatterpfqps.
4299         (ix86_expand_builtin): Expand new built-ins.
4300         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
4301         fix memory access data type.
4302         (*avx512pf_gatherpf<mode>_mask): Ditto.
4303         (*avx512pf_gatherpf<mode>): Ditto.
4304         (avx512pf_scatterpf<mode>): Ditto.
4305         (*avx512pf_scatterpf<mode>_mask): Ditto.
4306         (*avx512pf_scatterpf<mode>): Ditto.
4307         (GATHER_SCATTER_SF_MEM_MODE): New.
4308         (avx512pf_gatherpf<mode>df): Ditto.
4309         (*avx512pf_gatherpf<mode>df_mask): Ditto.
4310         (*avx512pf_scatterpf<mode>df): Ditto.
4312 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
4314         PR bootstrap/59934
4315         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
4316         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
4317         reached.
4319 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4321         * common/config/arm/arm-common.c
4322         (arm_rewrite_mcpu): Handle multiple names.
4323         * config/arm/arm.h
4324         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4326 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
4328         * gimple-builder.h (create_gimple_tmp): Delete.
4330 2014-01-27  Christian Bruel  <christian.bruel@st.com>
4332         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
4333         words comparisons.
4335 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
4337         * config/pa/pa.md (call): Generate indirect long calls to non-local
4338         functions when outputing 32-bit code.
4339         (call_value): Likewise except for special call to buggy powf function.
4341         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
4342         portable runtime and PIC indirect calls.
4343         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
4344         and PIC call sequences.  Use ldo instead of blr to set return register
4345         in PIC call sequence.
4347 2014-01-25  Walter Lee  <walt@tilera.com>
4349         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
4350         avoid clobbering a live register.
4352 2014-01-25  Walter Lee  <walt@tilera.com>
4354         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
4355         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
4356         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
4357         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
4359 2014-01-25  Walter Lee  <walt@tilera.com>
4361         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
4362         arguments on even registers.
4363         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
4364         STACK_BOUNDARY.
4365         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
4366         (BIGGEST_ALIGNMENT): Ditto.
4367         (BIGGEST_FIELD_ALIGNMENT): Ditto.
4369 2014-01-25  Walter Lee  <walt@tilera.com>
4371         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
4372         insns before bundling.
4373         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
4375 2014-01-25  Walter Lee  <walt@tilera.com>
4377         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
4378         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
4379         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
4381 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4383         * config/mips/constraints.md (kl): Delete.
4384         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
4385         define expands, using...
4386         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
4387         instructions for MIPS16.
4388         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
4389         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
4391 2014-01-25  Walter Lee  <walt@tilera.com>
4393         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
4394         (clzdi2): Ditto.
4395         (ffsdi2): Ditto.
4397 2014-01-25  Walter Lee  <walt@tilera.com>
4399         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
4400         (TARGET_EXPAND_TO_RTL_HOOK): Define.
4402 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
4404         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
4405         Handle XOR.
4407 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
4409         * print-rtl.c (in_call_function_usage): New var.
4410         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
4411         EXPR_LIST mode as mode and not as reg note name.
4413         PR middle-end/59561
4414         * cfgloopmanip.c (copy_loop_info): If
4415         loop->warned_aggressive_loop_optimizations, make sure
4416         the flag is set in target loop too.
4418 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
4420         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
4421         flag_cilkplus.
4422         * builtins.def: Likewise.
4423         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
4424         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
4425         * ira.c (ira_setup_eliminable_regset): Likewise.
4426         * omp-low.c (gate_expand_omp): Likewise.
4427         (execute_lower_omp): Likewise.
4428         (diagnose_sb_0): Likewise.
4429         (gate_diagnose_omp_blocks): Likewise.
4430         (simd_clone_clauses_extract): Likewise.
4431         (gate): Likewise.
4433 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4435         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
4436         correction for little endian...
4437         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
4438         here.
4440 2014-01-24  Jeff Law  <law@redhat.com>
4442         PR tree-optimization/59919
4443         * tree-vrp.c (find_assert_locations_1): Do not register asserts
4444         for non-returning calls.
4446 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
4448         * common/config/aarch64/aarch64-common.c
4449         (aarch64_rewrite_mcpu): Handle multiple names.
4450         * config/aarch64/aarch64.h
4451         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
4453 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
4455         * input.c (add_file_to_cache_tab): Handle the case where fopen
4456         returns NULL.
4458 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
4460         PR target/59929
4461         * config/i386/i386.md (pushsf splitter): Get stack adjustment
4462         from push operand if code of push isn't PRE_DEC.
4464 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
4466         PR target/59909
4467         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4468         -mquad-memory-atomic.  Update -mquad-memory documentation to say
4469         it is only used for non-atomic loads/stores.
4471         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
4472         -mquad-memory or -mquad-memory-atomic switches.
4474         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
4475         -mquad-memory-atomic to ISA 2.07 support.
4477         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
4478         to separate support of normal quad word memory operations (ldq, stq)
4479         from the atomic quad word memory operations.
4481         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
4482         support to separate non-atomic quad word operations from atomic
4483         quad word operations.  Disable non-atomic quad word operations in
4484         little endian mode so that we don't have to swap words after the
4485         load and before the store.
4486         (quad_load_store_p): Add comment about atomic quad word support.
4487         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
4488         options printed with -mdebug=reg.
4490         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
4491         -mquad-memory-atomic as the test for whether we have quad word
4492         atomic instructions.
4493         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
4494         or -mp8-vector are used, allow byte/half-word atomic operations.
4496         * config/rs6000/sync.md (load_lockedti): Insure that the address
4497         is a proper indexed or indirect address for the lqarx instruction.
4498         On little endian systems, swap the hi/lo registers after the lqarx
4499         instruction.
4500         (load_lockedpti): Use indexed_or_indirect_operand predicate to
4501         insure the address is valid for the lqarx instruction.
4502         (store_conditionalti): Insure that the address is a proper indexed
4503         or indirect address for the stqcrx. instruction.  On little endian
4504         systems, swap the hi/lo registers before doing the stqcrx.
4505         instruction.
4506         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
4507         insure the address is valid for the stqcrx. instruction.
4509         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
4510         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
4511         type of quad memory support is available.
4513 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
4515         PR regression/59915
4516         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
4517         there is a danger of looping.
4519 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
4521         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4522         force flag_ira_loop_pressure if set via command line.
4524 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4526         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
4527         (ashr_simd): New builtin handling DI mode.
4528         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
4529         (aarch64_sshr_simddi): New match pattern.
4530         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
4531         (vshrd_n_s64): Likewise.
4532         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
4534 2014-01-23  Nick Clifton  <nickc@redhat.com>
4536         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
4537         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
4538         favour of mcu specific scripts.
4539         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
4540         430x multilibs.
4542 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4543             Alex Velenko  <Alex.Velenko@arm.com>
4545         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
4546         (vaddv_s16): Likewise.
4547         (vaddv_s32): Likewise.
4548         (vaddv_u8): Likewise.
4549         (vaddv_u16): Likewise.
4550         (vaddv_u32): Likewise.
4551         (vaddvq_s8): Likewise.
4552         (vaddvq_s16): Likewise.
4553         (vaddvq_s32): Likewise.
4554         (vaddvq_s64): Likewise.
4555         (vaddvq_u8): Likewise.
4556         (vaddvq_u16): Likewise.
4557         (vaddvq_u32): Likewise.
4558         (vaddvq_u64): Likewise.
4559         (vaddv_f32): Likewise.
4560         (vaddvq_f32): Likewise.
4561         (vaddvq_f64): Likewise.
4562         (vmaxv_f32): Likewise.
4563         (vmaxv_s8): Likewise.
4564         (vmaxv_s16): Likewise.
4565         (vmaxv_s32): Likewise.
4566         (vmaxv_u8): Likewise.
4567         (vmaxv_u16): Likewise.
4568         (vmaxv_u32): Likewise.
4569         (vmaxvq_f32): Likewise.
4570         (vmaxvq_f64): Likewise.
4571         (vmaxvq_s8): Likewise.
4572         (vmaxvq_s16): Likewise.
4573         (vmaxvq_s32): Likewise.
4574         (vmaxvq_u8): Likewise.
4575         (vmaxvq_u16): Likewise.
4576         (vmaxvq_u32): Likewise.
4577         (vmaxnmv_f32): Likewise.
4578         (vmaxnmvq_f32): Likewise.
4579         (vmaxnmvq_f64): Likewise.
4580         (vminv_f32): Likewise.
4581         (vminv_s8): Likewise.
4582         (vminv_s16): Likewise.
4583         (vminv_s32): Likewise.
4584         (vminv_u8): Likewise.
4585         (vminv_u16): Likewise.
4586         (vminv_u32): Likewise.
4587         (vminvq_f32): Likewise.
4588         (vminvq_f64): Likewise.
4589         (vminvq_s8): Likewise.
4590         (vminvq_s16): Likewise.
4591         (vminvq_s32): Likewise.
4592         (vminvq_u8): Likewise.
4593         (vminvq_u16): Likewise.
4594         (vminvq_u32): Likewise.
4595         (vminnmv_f32): Likewise.
4596         (vminnmvq_f32): Likewise.
4597         (vminnmvq_f64): Likewise.
4599 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
4601         * config/aarch64/aarch64-simd.md
4602         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
4603         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
4604         (*aarch64_mul3_elt<mode>): Likewise.
4605         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
4606         (*aarch64_mul3_elt_to_64v2df): Likewise.
4607         (*aarch64_mla_elt<mode>): Likewise.
4608         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
4609         (*aarch64_mls_elt<mode>): Likewise.
4610         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
4611         (*aarch64_fma4_elt<mode>): Likewise.
4612         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
4613         (*aarch64_fma4_elt_to_64v2df): Likewise.
4614         (*aarch64_fnma4_elt<mode>): Likewise.
4615         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
4616         (*aarch64_fnma4_elt_to_64v2df): Likewise.
4617         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
4618         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
4619         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4620         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
4621         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
4622         (aarch64_sqdmull_lane<mode>_internal): Likewise.
4623         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
4625 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4627         * config/aarch64/aarch64-simd.md
4628         (aarch64_be_checked_get_lane<mode>): New define_expand.
4629         * config/aarch64/aarch64-simd-builtins.def
4630         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
4631         New builtin definition.
4632         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
4633         Use new safe be builtin.
4635 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
4637         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
4638         New define_insn.
4639         (aarch64_be_st1<mode>): Likewise.
4640         (aarch_ld1<VALL:mode>): Define_expand modified.
4641         (aarch_st1<VALL:mode>): Likewise.
4642         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
4643         (UNSPEC_ST1): Likewise.
4645 2014-01-23  David Holsgrove  <david.holsgrove@xilinx.com>
4647         * config/microblaze/microblaze.md: Add trap insn and attribute
4649 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
4651         PR preprocessor/58580
4652         * input.h (location_get_source_line): Take an additional line_size
4653         parameter.
4654         (void diagnostics_file_cache_fini): Declare new function.
4655         * input.c (struct fcache): New type.
4656         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
4657         New static constants.
4658         (diagnostic_file_cache_init, total_lines_num)
4659         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
4660         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
4661         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
4662         (get_next_line, read_next_line, goto_next_line, read_line_num):
4663         New static function definitions.
4664         (diagnostic_file_cache_fini): New function.
4665         (location_get_source_line): Take an additional output line_len
4666         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
4667         read_line_num.
4668         * diagnostic.c (diagnostic_finish): Call
4669         diagnostic_file_cache_fini.
4670         (adjust_line): Take an additional input parameter for the length
4671         of the line, rather than calculating it with strlen.
4672         (diagnostic_show_locus): Adjust the use of
4673         location_get_source_line and adjust_line with respect to their new
4674         signature.  While displaying a line now, do not stop at the first
4675         null byte.  Rather, display the zero byte as a space and keep
4676         going until we reach the size of the line.
4677         * Makefile.in: Add vec.o to OBJS-libcommon
4679 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4680             Ilya Tocar     <ilya.tocar@intel.com>
4682         * config/i386/avx512fintrin.h (_mm512_kmov): New.
4683         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
4684         (__builtin_ia32_kmov16): Ditto.
4685         * config/i386/i386.md (UNSPEC_KMOV): New.
4686         (kmovw): Ditto.
4688 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
4690         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
4691         (_mm512_storeu_si512): Ditto.
4693 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
4695         PR target/52125
4696         * rtl.h (get_referenced_operands): Declare.
4697         * recog.c (get_referenced_operands): New function.
4698         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
4699         operands have been referenced when recording LO_SUM references.
4701 2014-01-22  David Holsgrove  <david.holsgrove@xilinx.com>
4703         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
4705 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4707         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
4708         Enable for generic and recent AMD targets.
4710 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
4712         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
4713         ARG_SIZE note when adjustment was eliminated.
4715 2014-01-22  Jeff Law  <law@redhat.com>
4717         PR tree-optimization/59597
4718         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
4719         in file.  Accept new argument REGISTERING and use it to modify
4720         dump output appropriately.
4721         (register_jump_thread): Corresponding changes.
4722         (mark_threaded_blocks): Reinstate code to cancel unprofitable
4723         thread paths involving joiner blocks.  Add code to dump cancelled
4724         jump threading paths.
4726 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
4728         PR rtl-optimization/59477
4729         * lra-constraints.c (inherit_in_ebb): Process call for living hard
4730         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
4732 2014-01-22  Tom Tromey  <tromey@redhat.com>
4734         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4735         PARAMS.
4736         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4738 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4740         PR rtl-optimization/59896
4741         * lra-constraints.c (process_alt_operands): Check unused note for
4742         matched operands of insn with no output reloads.
4744 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4746         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4747         (mips_move_from_gpr_cost): Likewise.
4749 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4751         PR rtl-optimization/59858
4752         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4753         ira_class_hard_regs_num.
4754         (process_alt_operands): Increase reject for dying matched operand.
4756 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4758         PR target/59003
4759         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4760         smaller than size, perform several stores or loads and stores
4761         at dst + count - size to store or copy all of size bytes, rather
4762         than just last modesize bytes.
4764 2014-01-20  DJ Delorie  <dj@redhat.com>
4766         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4767         that CLOBBERs are REGs before propogating their values.
4769 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4771         PR middle-end/59789
4772         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4773         (cgraph_inline_failed_type): New function.
4774         * cgraph.h (DEFCIFCODE): Add type.
4775         (cgraph_inline_failed_type_t): New enum.
4776         (cgraph_inline_failed_type): New prototype.
4777         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4778         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4779         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4780         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4781         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4782         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4783         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4784         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4785         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4786         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4787         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4788         OPTIMIZATION_MISMATCH.
4789         * tree-inline.c (expand_call_inline): Emit errors during
4790         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4792 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4794         PR target/59685
4795         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4796         mode attribute in insn output.
4798 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4800         * output.h (output_constant): Delete.
4801         * varasm.c (output_constant): Make private.
4803 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4805         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4807 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4809         PR middle-end/59860
4810         * tree.h (fold_builtin_strcat): New prototype.
4811         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4812         argument, if non-NULL, don't call c_strlen.  Optimize
4813         directly into __builtin_memcpy instead of __builtin_strcpy.
4814         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4815         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4817 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4819         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4820         for SImode_address_operand operands, having only a REG argument.
4822 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4824         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4825         loader name using mbig-endian.
4826         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4828 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4830         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4831         (-mtune): Likewise.
4832         (-mcpu): Likewise.
4834 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4836         * config/aarch64/aarch64-protos.h
4837         (aarch64_cannot_change_mode_class_ptr): Declare.
4838         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4839         aarch64_cannot_change_mode_class_ptr): New.
4840         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4841         backend hook aarch64_cannot_change_mode_class.
4843 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4845         * common/config/aarch64/aarch64-common.c
4846         (aarch64_handle_option): Don't handle any option order logic here.
4847         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4848         selected_cpu, warn on architecture version mismatch.
4849         (aarch64_override_options): Fix parsing order for option strings.
4851 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4852             Iain Sandoe  <iain@codesourcery.com>
4854         PR bootstrap/59496
4855         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4856         warning.  Amend comment to reflect current functionality.
4858 2014-01-20  Richard Biener  <rguenther@suse.de>
4860         PR middle-end/59860
4861         * builtins.c (fold_builtin_strcat): Remove case better handled
4862         by tree-ssa-strlen.c.
4864 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4866         * config/aarch64/aarch64.opt
4867         (mcpu, march, mtune): Make case-insensitive.
4869 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4871         PR target/59880
4872         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4873         if operands[1] is a REG or ZERO_EXTEND of a REG.
4875 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4877         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4879 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4881         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4882         long non-pic millicode calls.
4884 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4886         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4888 2014-01-19  Kito Cheng  <kito@0xlab.org>
4890         * builtins.c (expand_movstr): Check movstr expand done or fail.
4892 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4893             H.J. Lu  <hongjiu.lu@intel.com>
4895         PR target/59379
4896         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4897         to DImode for zero-extended addresses.
4899 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4901         PR rtl-optimization/57763
4902         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4903         on the new indirect jump_insn and increment LABEL_NUSES (label).
4905 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4907         PR bootstrap/59580
4908         PR bootstrap/59583
4909         * config.gcc (x86_archs): New variable.
4910         (x86_64_archs): Likewise.
4911         (x86_cpus): Likewise.
4912         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4913         --with-arch/--with-cpu= options.
4914         Support --with-arch=/--with-cpu={nehalem,westmere,
4915         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4917 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4919         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4920         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4922 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4924         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4926 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4928         PR target/58944
4929         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4930         clear cpp_get_options (parse_in)->warn_unused_macros for
4931         ix86_target_macros_internal with cpp_define.
4933 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4935         * jump.c (delete_related_insns): Keep (use (insn))s.
4936         * reorg.c (redundant_insn): Check for barriers too.
4938 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4940         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4942 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4944         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4945         call to $$dyncall when TARGET_LONG_CALLS is true.
4947 2014-01-17  Jeff Law  <law@redhat.com>
4949         * ree.c (combine_set_extension): Temporarily disable test for
4950         changing number of hard registers.
4952 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4954         PR middle-end/58125
4955         * ipa-inline-analysis.c (inline_free_summary):
4956         Do not free summary of aliases.
4958 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4960         PR middle-end/59706
4961         * gimplify.c (gimplify_expr): Use create_tmp_var
4962         instead of create_tmp_var_raw.  If cond doesn't have
4963         integral type, don't add the IFN_ANNOTATE builtin at all.
4965 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4967         PR ipa/59736
4968         * ipa-cp.c (prev_edge_clone): New variable.
4969         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4970         Also resize prev_edge_clone vector.
4971         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4972         (ipcp_edge_removal_hook): New function.
4973         (ipcp_driver): Register ipcp_edge_removal_hook.
4975 2014-01-17  Andrew Pinski  <apinski@cavium.com>
4976             Steve Ellcey  <sellcey@mips.com>
4978         PR target/59462
4979         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4980         of operator mode.
4982 2014-01-17  Jeff Law  <law@redhat.com>
4984         PR middle-end/57904
4985         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4986         so that pass_ccp runs first.
4988 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4990         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4991         (ix86_adjust_cost): Use !TARGET_XXX.
4992         (do_reorder_for_imul): Likewise.
4993         (swap_top_of_ready_list): Likewise.
4994         (ix86_sched_reorder): Likewise.
4996 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4998         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4999         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
5000         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
5001         (intel_memset): New.  Duplicate slm_memset.
5002         (intel_cost): New.  Duplicate slm_cost.
5003         (m_INTEL): New macro.
5004         (processor_target_table): Add "intel".
5005         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
5006         with PROCESSOR_INTEL for "intel".
5007         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
5008         PROCESSOR_SILVERMONT.
5009         (ix86_issue_rate): Likewise.
5010         (ix86_adjust_cost): Likewise.
5011         (ia32_multipass_dfa_lookahead): Likewise.
5012         (swap_top_of_ready_list): Likewise.
5013         (ix86_sched_reorder): Likewise.
5014         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
5015         instead of TARGET_OPT_AGU.
5016         * config/i386/i386.h (TARGET_INTEL): New.
5017         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
5018         (processor_type): Add PROCESSOR_INTEL.
5019         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
5020         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
5022 2014-01-17  Marek Polacek  <polacek@redhat.com>
5024         PR c/58346
5025         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
5026         size is zero.
5028 2014-01-17  Richard Biener  <rguenther@suse.de>
5030         PR tree-optimization/46590
5031         * opts.c (default_options_table): Add entries for
5032         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
5033         all enabled at -O1 but not for -Og.
5034         * common.opt (fbranch-count-reg): Remove Init(1).
5035         (fmove-loop-invariants): Likewise.
5036         (ftree-pta): Likewise.
5038 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
5040         * config/i386/i386.c (ix86_data_alignment): For compatibility with
5041         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
5042         decls to at least the GCC 4.8 used alignments.
5044         PR fortran/59440
5045         * tree-nested.c (convert_nonlocal_reference_stmt,
5046         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
5047         of GIMPLE_BIND stmts, adjust associated decls.
5049 2014-01-17  Richard Biener  <rguenther@suse.de>
5051         PR tree-optimization/46590
5052         * vec.h (vec<>::bseach): New member function implementing
5053         binary search according to C89 bsearch.
5054         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
5055         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
5056         bitmap pointer again.  Make accesses_in_loop a flat array.
5057         (mem_ref_obstack): New global.
5058         (outermost_indep_loop): Adjust for mem_ref->stored changes.
5059         (mark_ref_stored): Likewise.
5060         (ref_indep_loop_p_2): Likewise.
5061         (set_ref_stored_in_loop): New helper function.
5062         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
5063         (memref_free): Adjust.
5064         (record_mem_ref_loc): Simplify.
5065         (gather_mem_refs_stmt): Adjust.
5066         (sort_locs_in_loop_postorder_cmp): New function.
5067         (analyze_memory_references): Sort accesses_in_loop after
5068         loop postorder number.
5069         (find_ref_loc_in_loop_cmp): New function.
5070         (for_all_locs_in_loop): Find relevant cluster of locs in
5071         accesses_in_loop and iterate without recursion.
5072         (execute_sm): Avoid uninit warning.
5073         (struct ref_always_accessed): Simplify.
5074         (ref_always_accessed::operator ()): Likewise.
5075         (ref_always_accessed_p): Likewise.
5076         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
5077         loop postorder numbers here.
5078         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
5079         numbers.
5081 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
5083         PR c++/57945
5084         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
5085         on decls for which assemble_alias has been called.
5087 2014-01-17  Nick Clifton  <nickc@redhat.com>
5089         * config/msp430/msp430.opt: (mcpu): New option.
5090         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
5091         (msp430_option_override): Parse target_cpu.  If the MCU name
5092         matches a generic string, clear target_mcu.
5093         (msp430_attr): Allow numeric interrupt values up to 63.
5094         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
5095         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
5096         option.
5097         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
5098         Add mcpu matches.
5099         * config/msp430/msp430.md (popm): Use %J rather than %I.
5100         (addsi3): Use msp430_nonimmediate_operand for operand 2.
5101         (addhi_cy_i): Use immediate_operand for operand 2.
5102         * doc/invoke.texi: Document -mcpu option.
5104 2014-01-17  Richard Biener  <rguenther@suse.de>
5106         PR rtl-optimization/38518
5107         * df.h (df_analyze_loop): Declare.
5108         * df-core.c: Include cfgloop.h.
5109         (df_analyze_1): Split out main part of df_analyze.
5110         (df_analyze): Adjust.
5111         (loop_inverted_post_order_compute): New function.
5112         (loop_post_order_compute): Likewise.
5113         (df_analyze_loop): New function avoiding whole-function
5114         postorder computes.
5115         * loop-invariant.c (find_defs): Use df_analyze_loop.
5116         (find_invariants): Adjust.
5117         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
5119 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
5121         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
5122         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
5124 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
5126         * ipa-ref.c (ipa_remove_stmt_references): Fix references
5127         traversal when removing references.
5129 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
5131         PR ipa/59775
5132         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
5134 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
5136         PR middle-end/56791
5137         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
5138         pushing a reload for an autoinc when we had previously reloaded an
5139         inner part of the address.
5141 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
5143         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
5144         field.
5145         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
5146         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
5147         when not giving up or versioning for alias only because of
5148         loop->safelen.
5149         (vect_analyze_data_ref_dependences): Set to true.
5150         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
5151         is a GIMPLE_PHI.
5152         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
5153         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
5154         to the condition.
5156         PR middle-end/58344
5157         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
5159         PR target/59839
5160         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
5161         operand 0 predicate for gathers, use a new pseudo as subtarget.
5163 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5165         PR middle-end/59609
5166         * lra-constraints.c (process_alt_operands): Add printing debug info.
5167         Check absence of input/output reloads for matched operands too.
5169 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
5171         PR rtl-optimization/59835
5172         * ira.c (ira_init_register_move_cost): Increase cost for
5173         impossible modes.
5175 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
5177         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
5179 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
5181         PR target/59780
5182         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
5183         non-register objects.  Use gen_(high/low)part more consistently.
5184         Fix assertions.
5186 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
5188         PR target/59844
5189         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
5190         endian support, remove tests for WORDS_BIG_ENDIAN.
5191         (p8_mfvsrd_3_<mode>): Likewise.
5192         (reload_gpr_from_vsx<mode>): Likewise.
5193         (reload_gpr_from_vsxsf): Likewise.
5194         (p8_mfvsrd_4_disf): Likewise.
5196 2014-01-16  Richard Biener  <rguenther@suse.de>
5198         PR rtl-optimization/46590
5199         * lcm.c (compute_antinout_edge): Use postorder iteration.
5200         (compute_laterin): Use inverted postorder iteration.
5202 2014-01-16  Nick Clifton  <nickc@redhat.com>
5204         PR middle-end/28865
5205         * varasm.c (output_constant): Return the number of bytes actually
5206         emitted.
5207         (output_constructor_array_range): Update the field size with the
5208         number of bytes emitted by output_constant.
5209         (output_constructor_regular_field): Likewise.  Also do not
5210         complain if the total number of bytes emitted is now greater
5211         than the expected fieldpos.
5212         * output.h (output_constant): Update prototype and descriptive comment.
5214 2014-01-16  Marek Polacek  <polacek@redhat.com>
5216         PR middle-end/59827
5217         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
5218         it is error_mark_node.
5220 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
5222         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
5223         VALID_AVX256_REG_OR_OI_MODE.
5225 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
5227         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
5228         current procedure should be profiled.
5230 2014-01-15  Andrew Pinski  <apinski@cavium.com>
5232         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
5233         of moving from/to the STACK_REG register class.
5235 2014-01-15  Richard Henderson  <rth@redhat.com>
5237         PR debug/54694
5238         * reginfo.c (global_regs_decl): Globalize.
5239         * rtl.h (global_regs_decl): Declare.
5240         * ira.c (do_reload): Diagnose frame_pointer_needed and it
5241         reserved via global_regs.
5243 2014-01-15  Teresa Johnson  <tejohnson@google.com>
5245         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
5247 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
5249         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
5250         and vmulosh rather than call gen_vec_widen_smult_*.
5251         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
5252         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
5253         (vec_widen_smult_even_v16qi): Likewise.
5254         (vec_widen_umult_even_v8hi): Likewise.
5255         (vec_widen_smult_even_v8hi): Likewise.
5256         (vec_widen_umult_odd_v16qi): Likewise.
5257         (vec_widen_smult_odd_v16qi): Likewise.
5258         (vec_widen_umult_odd_v8hi): Likewise.
5259         (vec_widen_smult_odd_v8hi): Likewise.
5260         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
5261         vmuloub rather than call gen_vec_widen_umult_*.
5262         (vec_widen_umult_lo_v16qi): Likewise.
5263         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
5264         vmulosb rather than call gen_vec_widen_smult_*.
5265         (vec_widen_smult_lo_v16qi): Likewise.
5266         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
5267         rather than call gen_vec_widen_umult_*.
5268         (vec_widen_umult_lo_v8hi): Likewise.
5269         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
5270         rather than call gen_vec_widen_smult_*.
5271         (vec_widen_smult_lo_v8hi): Likewise.
5273 2014-01-15  Jeff Law  <law@redhat.com>
5275         PR tree-optimization/59747
5276         * ree.c (find_and_remove_re): Properly handle case where a second
5277         eliminated extension requires widening a copy created for elimination
5278         of a prior extension.
5279         (combine_set_extension): Ensure that the number of hard regs needed
5280         for a destination register does not change when we widen it.
5282 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
5284         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
5285         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
5286         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
5287         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
5288         (avr-*-rtems*): Likewise.
5289         (bfin*-rtems*): Likewise.
5290         (moxie-*-rtems*): Likewise.
5291         (h8300-*-rtems*): Likewise.
5292         (i[34567]86-*-rtems*): Likewise.
5293         (lm32-*-rtems*): Likewise.
5294         (m32r-*-rtems*): Likewise.
5295         (m68k-*-rtems*): Likewise.
5296         (microblaze*-*-rtems*): Likewise.
5297         (mips*-*-rtems*): Likewise.
5298         (powerpc-*-rtems*): Likewise.
5299         (sh-*-rtems*): Likewise.
5300         (sparc-*-rtems*): Likewise.
5301         (sparc64-*-rtems*): Likewise.
5302         (v850-*-rtems*): Likewise.
5303         (m32c-*-rtems*): Likewise.
5305 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
5307         PR rtl-optimization/59511
5308         * ira.c (ira_init_register_move_cost): Use memory costs for some
5309         cases of register move cost calculations.
5310         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
5311         instead of BB frequency.
5312         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
5313         * lra-assigns.c (find_hard_regno_for): Ditto.
5315 2014-01-15  Richard Biener  <rguenther@suse.de>
5317         PR tree-optimization/59822
5318         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
5319         (vectorizable_load): Use it to hoist defs of uses of invariant
5320         loads out of the loop.
5322 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
5323             Kugan Vivekanandarajah  <kuganv@linaro.org>
5325         PR target/59695
5326         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
5327         truncation.
5329 2014-01-15  Richard Biener  <rguenther@suse.de>
5331         PR rtl-optimization/59802
5332         * lcm.c (compute_available): Use inverted postorder to seed
5333         the initial worklist.
5335 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5337         PR target/59803
5338         * config/s390/s390.c (s390_preferred_reload_class): Don't return
5339         ADDR_REGS for invalid symrefs in non-PIC code.
5341 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
5343         PR other/58712
5344         * builtins.c (determine_block_size): Initialize *probable_max_size
5345         even if len_rtx is CONST_INT.
5347 2014-01-14  Andrew Pinski  <apinski@cavium.com>
5349         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
5350         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
5351         (cortexa53_tunings): Likewise.
5352         (aarch64_sched_issue_rate): New function.
5353         (TARGET_SCHED_ISSUE_RATE): Define.
5355 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5357         * ira-costs.c (find_costs_and_classes): Add missed
5358         ira_init_register_move_cost_if_necessary.
5360 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
5362         PR target/59787
5363         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
5365 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
5367         PR target/59794
5368         * config/i386/i386.c (type_natural_mode): Add a bool parameter
5369         to indicate if type is used for function return value.  Warn ABI
5370         change if the vector mode isn't available for function return value.
5371         (ix86_function_arg_advance): Pass false to type_natural_mode.
5372         (ix86_function_arg): Likewise.
5373         (ix86_gimplify_va_arg): Likewise.
5374         (function_arg_32): Don't warn ABI change.
5375         (ix86_function_value): Pass true to type_natural_mode.
5376         (ix86_return_in_memory): Likewise.
5377         (ix86_struct_value_rtx): Removed.
5378         (TARGET_STRUCT_VALUE_RTX): Likewise.
5380 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
5382         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
5383         converting a conditional jump into a conditional return.
5385 2014-01-14  Richard Biener  <rguenther@suse.de>
5387         PR tree-optimization/58921
5388         PR tree-optimization/59006
5389         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
5390         hoisting invariant stmts.
5391         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
5392         invariant loads on the preheader edge if possible.
5394 2014-01-14  Joey Ye  <joey.ye@arm.com>
5396         * doc/plugin.texi (Building GCC plugins): Update to C++.
5398 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
5400         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
5401         (_mm_rcp28_round_ss): Ditto.
5402         (_mm_rsqrt28_round_sd): Ditto.
5403         (_mm_rsqrt28_round_ss): Ditto.
5404         (_mm_rcp28_sd): Ditto.
5405         (_mm_rcp28_ss): Ditto.
5406         (_mm_rsqrt28_sd): Ditto.
5407         (_mm_rsqrt28_ss): Ditto.
5408         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
5409         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
5410         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
5411         (IX86_BUILTIN_RCP28SD): Ditto.
5412         (IX86_BUILTIN_RCP28SS): Ditto.
5413         (IX86_BUILTIN_RSQRT28SD): Ditto.
5414         (IX86_BUILTIN_RSQRT28SS): Ditto.
5415         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
5416         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
5417         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
5418         (ix86_expand_special_args_builtin): Expand new FTYPE.
5419         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
5420         (srcp14<mode>): Make insn unary.
5421         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
5422         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
5423         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
5424         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
5425         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
5426         Fix rounding: make it SAE only.
5427         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
5428         Ditto.
5429         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
5430         Ditto.
5431         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
5432         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
5433         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
5434         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
5435         (round_saeonly_mask_scalar_operand4): Ditto.
5436         (round_saeonly_mask_scalar_op3): Ditto.
5437         (round_saeonly_mask_scalar_op4): Ditto.
5439 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5441         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5442         Implement -maltivec=be for vec_insert and vec_extract.
5444 2014-01-10  DJ Delorie  <dj@redhat.com>
5446         * config/msp430/msp430.md (call_internal): Don't allow memory
5447         references with SP as the base register.
5448         (call_value_internal): Likewise.
5449         * config/msp430/constraints.md (Yc): New.  For memory references
5450         that don't use SP as a base register.
5452         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
5453         "an integer without a # prefix"
5454         * config/msp430/msp430.md (epilogue_helper): Use it.
5456 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5458         PR target/59617
5459         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
5460         AVX512F gather builtins.
5461         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
5462         on gather decls with INTEGER_TYPE masktype.
5463         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
5464         directly into the builtin rather than hoisting it before loop.
5466         PR tree-optimization/59387
5467         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
5468         (scev_const_prop): If folded_casts and type has undefined overflow,
5469         use force_gimple_operand instead of force_gimple_operand_gsi and
5470         for each added stmt if it is assign with
5471         arith_code_with_undefined_signed_overflow, call
5472         rewrite_to_defined_overflow.
5473         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
5474         gimple-fold.h instead.
5475         (arith_code_with_undefined_signed_overflow,
5476         rewrite_to_defined_overflow): Moved to ...
5477         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
5478         rewrite_to_defined_overflow): ... here.  No longer static.
5479         Include gimplify-me.h.
5480         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
5481         rewrite_to_defined_overflow): New prototypes.
5483 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5485         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
5487 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
5489         * builtins.c (get_object_alignment_2): Minor tweak.
5490         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
5492 2014-01-13  Christian Bruel  <christian.bruel@st.com>
5494         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
5495         optimized non constant lengths.
5497 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
5499         PR libgomp/59194
5500         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
5501         load as __atomic_load_N if possible.
5503 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5505         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
5506         target parameter.
5507         (rs6000_expand_builtin): Adjust call.
5509 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
5511         PR target/58115
5512         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
5513         * config/rs6000/rs6000.c: Include target-globals.h.
5514         (rs6000_set_current_function): Instead of doing target_reinit
5515         unconditionally, use save_target_globals_default_opts and
5516         restore_target_globals.
5518         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
5519         FPSCR.
5520         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
5521         (rs6000_expand_builtin): Handle mffs and mtfsf.
5522         (rs6000_init_builtins): Define mffs and mtfsf.
5523         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
5524         (rs6000_mffs): New pattern.
5525         (rs6000_mtfsf): New pattern.
5527 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
5529         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
5530         Start narrowing with START.  Apply candidate-use pair
5531         and check overall cost in narrowing.
5532         (iv_ca_prune): Pass new argument.
5534 2014-01-10  Jeff Law  <law@redhat.com>
5536         PR middle-end/59743
5537         * ree.c (combine_reaching_defs): Ensure the defining statement
5538         occurs before the extension when optimizing extensions with
5539         different source and destination hard registers.
5541 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5543         PR ipa/58585
5544         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
5545         vtables into the type inheritance graph.
5547 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5549         PR rtl-optimization/59754
5550         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
5551         modes in the REGNO != REGNO case.
5553 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5555         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
5557 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5559         PR tree-optimization/59745
5560         * tree-predcom.c (tree_predictive_commoning_loop): Call
5561         free_affine_expand_cache if giving up because components is NULL.
5563         * target-globals.c (save_target_globals): Allocate < 4KB structs using
5564         GC in payload of target_globals struct instead of allocating them on
5565         the heap and the larger structs separately using GC.
5566         * target-globals.h (struct target_globals): Make regs, hard_regs,
5567         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
5568         of GTY((skip)) and change type to void *.
5569         (reset_target_globals): Cast loads from those fields to corresponding
5570         types.
5572 2014-01-10  Steve Ellcey  <sellcey@mips.com>
5574         PR plugins/59335
5575         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
5576         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
5577         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
5579 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
5581         PR target/59744
5582         * aarch64-modes.def (CC_Zmode): New flags mode.
5583         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
5584         represents an equality.
5585         (aarch64_get_condition_code): Handle CC_Zmode.
5586         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
5588 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
5590         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
5591         extraction in good case.
5593 2014-01-10  Richard Biener  <rguenther@suse.de>
5595         PR tree-optimization/59374
5596         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
5597         checking after SLP discovery.  Mark stmts not participating
5598         in any SLP instance properly.
5600 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5602         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
5603         when handling a SET rtx.
5605 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5607         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
5608         (cortex-a57): Likewise.
5609         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
5611 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5613         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
5614         non-iwmmxt builtins.
5616 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
5618         PR ipa/58252
5619         PR ipa/59226
5620         * ipa-devirt.c record_target_from_binfo): Take as argument
5621         stack of binfos and lookup matching one for virtual inheritance.
5622         (possible_polymorphic_call_targets_1): Update.
5624 2014-01-10  Huacai Chen  <chenhc@lemote.com>
5626         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
5627         kernel strings for Loongson-2E/2F/3A.
5629 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
5631         PR middle-end/59670
5632         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
5633         is_gimple_call before calling gimple_call_internal_p.
5635 2014-01-09  Steve Ellcey  <sellcey@mips.com>
5637         * Makefile.in (TREE_FLOW_H): Remove.
5638         (TREE_SSA_H): Add file names from tree-flow.h.
5639         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
5640         * tree.h: Remove tree-flow.h reference.
5641         * hash-table.h: Remove tree-flow.h reference.
5642         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
5643         reference with tree-ssa-loop.h.
5645 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5647         * doc/invoke.texi: Add -maltivec={be,le} options, and document
5648         default element-order behavior for -maltivec.
5649         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
5650         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
5651         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
5652         when targeting big endian, at least for now.
5653         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
5655 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5657         PR middle-end/47735
5658         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
5659         var satisfies use_register_for_decl, just take into account type
5660         alignment, rather than decl alignment.
5662         PR tree-optimization/59622
5663         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
5664         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
5665         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
5666         Don't devirtualize for inplace at all.  For targets.length () == 1,
5667         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
5669 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5671         * config/i386/i386.md (cpu): Remove the unused btver1.
5673 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
5675         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
5677 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5679         PR target/58115
5680         * tree-core.h (struct target_globals): New forward declaration.
5681         (struct tree_target_option): Add globals field.
5682         * tree.h (TREE_TARGET_GLOBALS): Define.
5683         (prepare_target_option_nodes_for_pch): New prototype.
5684         * target-globals.h (struct target_globals): Define even if
5685         !SWITCHABLE_TARGET.
5686         * tree.c (prepare_target_option_node_for_pch,
5687         prepare_target_option_nodes_for_pch): New functions.
5688         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
5689         * config/i386/i386.c: Include target-globals.h.
5690         (ix86_set_current_function): Instead of doing target_reinit
5691         unconditionally, use save_target_globals_default_opts and
5692         restore_target_globals.
5694 2014-01-09  Richard Biener  <rguenther@suse.de>
5696         PR tree-optimization/59715
5697         * tree-cfg.h (split_critical_edges): Declare.
5698         * tree-cfg.c (split_critical_edges): Export.
5699         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
5701 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
5703         * cfgexpand.c (expand_stack_vars): Optionally disable
5704         asan stack protection.
5705         (expand_used_vars): Likewise.
5706         (partition_stack_vars): Likewise.
5707         * asan.c (asan_emit_stack_protection): Optionally disable
5708         after return stack usage.
5709         (instrument_derefs): Optionally disable memory access instrumentation.
5710         (instrument_builtin_call): Likewise.
5711         (instrument_strlen_call): Likewise.
5712         (asan_protect_global): Optionally disable global variables protection.
5713         * doc/invoke.texi: Added doc for new options.
5714         * params.def: Added new options.
5715         * params.h: Likewise.
5717 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
5719         PR rtl-optimization/59724
5720         * ifcvt.c (cond_exec_process_if_block): Don't call
5721         flow_find_head_matching_sequence with 0 longest_match.
5722         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
5723         non-active insns if !stop_after.
5724         (try_head_merge_bb): Revert 2014-01-07 changes.
5726 2014-01-08  Jeff Law  <law@redhat.com>
5728         * ree.c (get_sub_rtx): New function, extracted from...
5729         (merge_def_and_ext): Here.
5730         (combine_reaching_defs): Use get_sub_rtx.
5732 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
5734         * cgraph.h (varpool_variable_node): Do not choke on null node.
5736 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5738         * config/mips/mips.md (simple_return): Attempt to use JRC
5739         for microMIPS.
5740         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5742 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5744         PR rtl-optimization/59137
5745         * reorg.c (steal_delay_list_from_target): Call update_block for
5746         elided insns.
5747         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5749 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5751         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5752         two duplicate entries.
5754 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5756         Revert:
5757         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5759         * config/mips/mips.c (mips_truncated_op_cost): New function.
5760         (mips_rtx_costs): Adjust test for BADDU.
5761         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5763         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5765         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5766         (*baddu_si): ...this new pattern.
5768 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5770         PR ipa/59722
5771         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5773 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5775         PR middle-end/57748
5776         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5777         inner_reference_p.
5778         (expand_expr, expand_normal): Adjust.
5779         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5780         inner_reference_p. Use inner_reference_p to expand inner references.
5781         (store_expr): Adjust.
5782         * cfgexpand.c (expand_call_stmt): Adjust.
5784 2014-01-08  Rong Xu  <xur@google.com>
5786         * gcov-io.c (gcov_var): Move from gcov-io.h.
5787         (gcov_position): Ditto.
5788         (gcov_is_error): Ditto.
5789         (gcov_rewrite): Ditto.
5790         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5791         only part to libgcc/libgcov.h.
5793 2014-01-08  Marek Polacek  <polacek@redhat.com>
5795         PR middle-end/59669
5796         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5798 2014-01-08  Marek Polacek  <polacek@redhat.com>
5800         PR sanitizer/59667
5801         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5803 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5805         PR rtl-optimization/59649
5806         * stor-layout.c (get_mode_bounds): For BImode return
5807         0 and STORE_FLAG_VALUE.
5809 2014-01-08  Richard Biener  <rguenther@suse.de>
5811         PR middle-end/59630
5812         * gimple.h (is_gimple_builtin_call): Remove.
5813         (gimple_builtin_call_types_compatible_p): New.
5814         (gimple_call_builtin_p): New overload.
5815         * gimple.c (is_gimple_builtin_call): Remove.
5816         (validate_call): Rename to ...
5817         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5818         check return types.
5819         (validate_type): New static function.
5820         (gimple_call_builtin_p): New overload and adjust.
5821         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5822         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5823         (gimple_fold_stmt_to_constant_1): Likewise.
5824         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5826 2014-01-08  Richard Biener  <rguenther@suse.de>
5828         PR middle-end/59471
5829         * gimplify.c (gimplify_expr): Gimplify register-register type
5830         VIEW_CONVERT_EXPRs to separate stmts.
5832 2014-01-07  Jeff Law  <law@redhat.com>
5834         PR middle-end/53623
5835         * ree.c (combine_set_extension): Handle case where source
5836         and destination registers in an extension insn are different.
5837         (combine_reaching_defs): Allow source and destination registers
5838         in extension to be different under limited circumstances.
5839         (add_removable_extension): Remove restriction that the
5840         source and destination registers in the extension are the same.
5841         (find_and_remove_re): Emit a copy from the extension's
5842         destination to its source after the defining insn if
5843         the source and destination registers are different.
5845         PR middle-end/59285
5846         * ifcvt.c (merge_if_block): If we are merging a block with more than
5847         one successor with a block with no successors, remove any BARRIER
5848         after the second block.
5850 2014-01-07  Dan Xio Qiang  <ziyan01@163.com>
5852         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5854 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5856         PR target/59652
5857         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5858         for 14-bit register offsets when INT14_OK_STRICT is false.
5860 2014-01-07  Roland Stigge  <stigge@antcom.de>
5861             Michael Meissner  <meissner@linux.vnet.ibm.com>
5863         PR 57386/target
5864         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5865         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5867 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5869         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5870         -mcpu.
5872 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5874         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5875         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5876         rtx is const0_rtx or not.
5878 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5880         PR target/58115
5881         * target-globals.c (save_target_globals): Remove this_fn_optab
5882         handling.
5883         * toplev.c: Include optabs.h.
5884         (target_reinit): Temporarily restore the global options if another
5885         set of options are in force.
5887 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5889         PR rtl-optimization/58668
5890         * cfgcleanup.c (flow_find_cross_jump): Don't count
5891         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5892         to determine what is counted.
5893         (flow_find_head_matching_sequence): Use active_insn_p to determine
5894         what is counted.
5895         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5896         counting change.
5897         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5898         determine what is counted.
5900         PR tree-optimization/59643
5901         * tree-predcom.c (split_data_refs_to_components): If one dr is
5902         read and one write, determine_offset fails and the write isn't
5903         in the bad component, just put the read into the bad component.
5905 2014-01-07  Mike Stump  <mikestump@comcast.net>
5906             Jakub Jelinek  <jakub@redhat.com>
5908         PR pch/59436
5909         * tree-core.h (struct tree_optimization_option): Change optabs
5910         type from unsigned char * to void *.
5911         * optabs.c (init_tree_optimization_optabs): Adjust
5912         TREE_OPTIMIZATION_OPTABS initialization.
5914 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5916         PR target/59644
5917         * config/i386/i386.h (struct machine_function): Add
5918         no_drap_save_restore field.
5919         * config/i386/i386.c (ix86_save_reg): Use
5920         !cfun->machine->no_drap_save_restore instead of
5921         crtl->stack_realign_needed.
5922         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5923         this function clears frame_pointer_needed.  Set
5924         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5925         and DRAP reg is needed.
5927 2014-01-06  Marek Polacek  <polacek@redhat.com>
5929         PR c/57773
5930         * doc/implement-c.texi: Mention that other integer types are
5931         permitted as bit-field types in strictly conforming mode.
5933 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5935         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5936         is newly allocated.
5938 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5940         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5942 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5944         PR ipa/59008
5945         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5946         to int.
5947         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5949 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5951         PR debug/59350
5952         PR debug/59510
5953         * var-tracking.c (add_stores): Preserve the value of the source even if
5954         we don't record the store.
5956 2014-01-06  Terry Guo  <terry.guo@arm.com>
5958         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5960 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5962         PR bootstrap/59541
5963         * config/darwin.c (darwin_function_section): Adjust return values to
5964         correspond to optimisation changes made in r206070.
5966 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5968         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5969         from prefetch_block tune setting.
5970         (nocona_cost): Correct size of prefetch block to 64.
5972 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5974         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5975         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5976         used to save the static chain register in the computation of the offset
5977         from which the FP registers need to be restored.
5979 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5981         PR tree-optimization/59519
5982         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5983         ICE if get_current_def (current_new_name) is already non-NULL, as long
5984         as it is a phi result of some other phi in *new_exit_bb that has
5985         the same argument.
5987         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5988         or vmovdqu* for misaligned_operand.
5989         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5990         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5991         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5992         aligned_mem for AVX512F masked aligned load and store builtins and for
5993         non-temporal moves.
5995 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5997         PR tree-optimization/59651
5998         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5999         Address range for negative step should be added by TYPE_SIZE_UNIT.
6001 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
6003         * config/m68k/m68k.c (handle_move_double): Handle pushes with
6004         overlapping registers also for registers other than the stack pointer.
6006 2014-01-03  Marek Polacek  <polacek@redhat.com>
6008         PR other/59661
6009         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
6010         __builtin_FILE.
6012 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
6014         PR target/59625
6015         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
6016         asm goto as jump.
6018         * config/i386/i386.md (MODE_SIZE): New mode attribute.
6019         (push splitter): Use <P:MODE_SIZE> instead of
6020         GET_MODE_SIZE (<P:MODE>mode).
6021         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
6022         (mov -1, reg peephole2): Likewise.
6023         * config/i386/sse.md (*mov<mode>_internal,
6024         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
6025         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
6026         *<code><mode>3, *andnot<mode>3<mask_name>,
6027         <mask_codefor><code><mode>3<mask_name>): Likewise.
6028         * config/i386/subst.md (mask_mode512bit_condition,
6029         sd_mask_mode512bit_condition): Likewise.
6031 2014-01-02  Xinliang David Li  <davidxl@google.com>
6033         PR tree-optimization/59303
6034         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
6035         (dump_predicates): Better output format.
6036         (pred_equal_p): New function.
6037         (is_neq_relop_p): Ditto.
6038         (is_neq_zero_form_p): Ditto.
6039         (pred_expr_equal_p): Ditto.
6040         (pred_neg_p): Ditto.
6041         (simplify_pred): Ditto.
6042         (simplify_preds_2): Ditto.
6043         (simplify_preds_3): Ditto.
6044         (simplify_preds_4): Ditto.
6045         (simplify_preds): Ditto.
6046         (push_pred): Ditto.
6047         (push_to_worklist): Ditto.
6048         (get_pred_info_from_cmp): Ditto.
6049         (is_degenerated_phi): Ditto.
6050         (normalize_one_pred_1): Ditto.
6051         (normalize_one_pred): Ditto.
6052         (normalize_one_pred_chain): Ditto.
6053         (normalize_preds): Ditto.
6054         (normalize_cond_1): Remove function.
6055         (normalize_cond): Ditto.
6056         (is_gcond_subset_of): Ditto.
6057         (is_subset_of_any): Ditto.
6058         (is_or_set_subset_of): Ditto.
6059         (is_and_set_subset_of): Ditto.
6060         (is_norm_cond_subset_of): Ditto.
6061         (pred_chain_length_cmp): Ditto.
6062         (convert_control_dep_chain_into_preds): Type change.
6063         (find_predicates): Ditto.
6064         (find_def_preds): Ditto.
6065         (destroy_predicates_vecs): Ditto.
6066         (find_matching_predicates_in_rest_chains): Ditto.
6067         (use_pred_not_overlap_with_undef_path_pred): Ditto.
6068         (is_pred_expr_subset): Ditto.
6069         (is_pred_chain_subset_of): Ditto.
6070         (is_included_in): Ditto.
6071         (is_superset_of): Ditto.
6073 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6075         Update copyright years.
6077 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
6079         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
6080         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
6081         config/arc/arc.md, config/arc/arc.opt,
6082         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
6083         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
6084         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
6085         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
6086         config/linux-protos.h, config/linux.c, config/winnt-c.c,
6087         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
6088         vtable-verify.c, vtable-verify.h: Use the standard form for the
6089         copyright notice.
6091 2014-01-02  Tobias Burnus  <burnus@net-b.de>
6093         * gcc.c (process_command): Update copyright notice dates.
6094         * gcov-dump.c: Ditto.
6095         * gcov.c: Ditto.
6096         * doc/cpp.texi: Bump @copying's copyright year.
6097         * doc/cppinternals.texi: Ditto.
6098         * doc/gcc.texi: Ditto.
6099         * doc/gccint.texi: Ditto.
6100         * doc/gcov.texi: Ditto.
6101         * doc/install.texi: Ditto.
6102         * doc/invoke.texi: Ditto.
6104 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
6106         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
6108 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
6110         * config/i386/sse.md (*mov<mode>_internal): Guard
6111         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
6113         PR rtl-optimization/59647
6114         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
6115         new_rtx into UNSIGNED_FLOAT rtxes.
6117 Copyright (C) 2014 Free Software Foundation, Inc.
6119 Copying and distribution of this file, with or without modification,
6120 are permitted in any medium without royalty provided the copyright
6121 notice and this notice are preserved.