2014-02-25 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob499fb33339bf9624f128eb340b6ee85eefef4831
1 2014-02-25  Richard Biener  <rguenther@suse.de>
3         PR middle-end/60291
4         * emit-rtl.c (mem_attrs_htab): Remove.
5         (mem_attrs_htab_hash): Likewise.
6         (mem_attrs_htab_eq): Likewise.
7         (set_mem_attrs): Always allocate new mem-attrs when something
8         changed.
9         (init_emit_once): Do not allocate mem_attrs_htab.
11 2014-02-25  Richard Biener  <rguenther@suse.de>
13         PR lto/60319
14         * lto-opts.c (lto_write_options): Output non-explicit conservative
15         -fwrapv, -fno-trapv and -fno-strict-overflow.
16         * lto-wrapper.c (merge_and_complain): Handle merging those options.
17         (run_gcc): And pass them through.
19 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
20         
21         * sel-sched.c (calculate_new_fences): New parameter ptime.
22         Calculate it as a maximum over all fence cycles.
23         (sel_sched_region_2): Adjust the call to calculate_new_fences.
24         Print the final schedule timing when sched_verbose.
26 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
28         PR rtl-optimization/60292
29         * sel-sched.c (fill_vec_av_set): Do not reset target availability
30         bit fot the fence instruction.
32 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
34         * calls.h: Fix typo in comment.
36 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
38         * config/pa/pa.c (pa_output_move_double): Don't valididate when
39         adjusting offsetable addresses.
41 2014-02-24  Guozhi Wei  <carrot@google.com>
43         * sparseset.h (sparseset_pop): Fix the wrong index.
45 2014-02-24  Walter Lee  <walt@tilera.com>
47         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
48         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
49         triplet.
50         * common/config/tilegx/tilegx-common.c
51         (TARGET_DEFAULT_TARGET_FLAGS): Define.
52         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
53         (LINK_SPEC): Ditto.
54         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
55         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
56         (tilegx_gimplify_va_arg_expr): Handle big endian.
57         (tilegx_expand_unaligned_load): Ditto.
58         (tilegx_expand_unaligned_store): Ditto.
59         (TARGET_RETURN_IN_MSB): New.
60         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
61         (TARGET_ENDIAN_DEFAULT): New.
62         (TARGET_BIG_ENDIAN): Handle big endian.
63         (BYTES_BIG_ENDIAN): Ditto.
64         (WORDS_BIG_ENDIAN): Ditto.
65         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
66         (ENDIAN_SPEC): New.
67         (EXTRA_SPECS): New.
68         * config/tilegx/tilegx.md (extv): Handle big endian.
69         (extzv): Ditto.
70         (insn_st<n>): Ditto.
71         (insn_st<n>_add<bitsuffix>): Ditto.
72         (insn_stnt<n>): Ditto.
73         (insn_stnt<n>_add<bitsuffix>):Ditto.
74         (vec_interleave_highv8qi): Handle big endian.
75         (vec_interleave_highv8qi_be): New.
76         (vec_interleave_highv8qi_le): New.
77         (insn_v1int_h): Handle big endian.
78         (vec_interleave_lowv8qi): Handle big endian.
79         (vec_interleave_lowv8qi_be): New.
80         (vec_interleave_lowv8qi_le): New.
81         (insn_v1int_l): Handle big endian.
82         (vec_interleave_highv4hi): Handle big endian.
83         (vec_interleave_highv4hi_be): New.
84         (vec_interleave_highv4hi_le): New.
85         (insn_v2int_h): Handle big endian.
86         (vec_interleave_lowv4hi): Handle big endian.
87         (vec_interleave_lowv4hi_be): New.
88         (vec_interleave_lowv4hi_le): New.
89         (insn_v2int_l): Handle big endian.
90         (vec_interleave_highv2si): Handle big endian.
91         (vec_interleave_highv2si_be): New.
92         (vec_interleave_highv2si_le): New.
93         (insn_v4int_h): Handle big endian.
94         (vec_interleave_lowv2si): Handle big endian.
95         (vec_interleave_lowv2si_be): New.
96         (vec_interleave_lowv2si_le): New.
97         (insn_v4int_l): Handle big endian.
98         * config/tilegx/tilegx.opt (mbig-endian): New option.
99         (mlittle-endian): New option.
100         * doc/install.texi: Document tilegxbe-linux.
101         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
103 2014-02-24  Martin Jambor  <mjambor@suse.cz>
105         PR ipa/60266
106         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
107         there are no parameter descriptors.
109 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
111         PR rtl-optimization/60268
112         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
113         initialization to ...
114         (sched_rgn_init): ... here.
115         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
117 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
119         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
120         names.
122 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
124         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
125         definition.
127 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
129         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk
130         and define TARGET_ASM_OUTPUT_MI_THUNK and
131         TARGET_ASM_CAN_OUTPUT_MI_THUNK.
133 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
135         * config/microblaze/predicates.md: Add cmp_op predicate.
136         * config/microblaze/microblaze.md: Add branch_compare instruction 
137         which uses cmp_op predicate and emits cmp insn before branch.
138         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename 
139         to microblaze_expand_conditional_branch and consolidate logic.
140         (microblaze_expand_conditional_branch): emit branch_compare
141         insn instead of handling cmp op separate from branch insn.
143 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
145         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
146         to permit subregs.
148 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
150         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
151         define_insn with define_expand and new define_insn
152         *altivec_lve<VI_char>x_internal.
153         (altivec_stve<VI_char>x): Replace define_insn with define_expand
154         and new define_insn *altivec_stve<VI_char>x_internal.
155         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
156         prototype.
157         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
158         lve*x built-ins.
159         (altivec_expand_stvex_be): New function.
161 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
163         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
164         ARG_POINTER_REGNUM
165         to STACK_POINTER_REGNUM if !frame_pointer_needed.
166         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
167         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
169 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
171         PR target/60298
172         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
173         instead of emit_move_insn.
175 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
177         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
178         vspltw with vsldoi.
179         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
180         gen_altivec_vsumsws.
182 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
184         * config/rs6000/altivec.md (altivec_lvxl): Rename as
185         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of
186         V4SI.
187         (altivec_lvxl_<mode>): New define_expand incorporating
188         -maltivec=be semantics where needed.
189         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
190         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
191         semantics where needed.
192         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
193         (altivec_stvx_<mode>): New define_expand incorporating
194         -maltivec=be semantics where needed.
195         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
196         VM2 iterator instead of V4SI.
197         (altivec_stvxl_<mode>): New define_expand incorporating
198         -maltivec=be semantics where needed.
199         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
200         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
201         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI,
202         STVX_V2DF, STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI,
203         STVXL_V2DF, STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI,
204         STVXL_V16QI.
205         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
206         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
207         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
208         ALTIVEC_BUILTIN_STVXL.
209         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New
210         prototype.
211         (altivec_expand_stvx_be): Likewise.
212         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
213         (altivec_expand_lvx_be): Likewise.
214         (altivec_expand_stvx_be): Likewise.
215         (altivec_expand_builtin): Add cases for
216         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
217         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
218         (altivec_init_builtins): Add definitions for
219         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
220         __builtin_altivec_stvx_<mode>, and
221         __builtin_altivec_stvxl_<mode>.
223 2014-02-21  Catherine Moore  <clm@codesourcery.com>
225         * doc/invoke.texi (mvirt, mno-virt): Document.
226         * config/mips/mips.opt (mvirt): New option.
227         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
229 2014-02-21  Richard Biener  <rguenther@suse.de>
231         PR tree-optimization/60276
232         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
233         (STMT_VINFO_MIN_NEG_DIST): New macro.
234         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
235         STMT_VINFO_MIN_NEG_DIST.
236         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
237         made for negative dependence distances still hold.
239 2014-02-21  Richard Biener  <rguenther@suse.de>
241         PR middle-end/60291
242         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
243         DECL_INITIAL for globals not in the current function context.
245 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
247         PR tree-optimization/56490
248         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
249         * tree-ssa-uninit.c: Include params.h.
250         (compute_control_dep_chain): Add num_calls argument, return false
251         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
252         num_calls to recursive call.
253         (find_predicates): Change dep_chain into normal array,
254         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
255         variable and adjust compute_control_dep_chain caller.
256         (find_def_preds): Likewise.
258 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
260         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
261         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
263 2014-02-21  Nick Clifton  <nickc@redhat.com>
265         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
266         (pushhi1): Likewise.
267         (popqi1): Add mode to pre_dec.
268         (pophi1): Likewise.
270 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
272         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
273         mode for mask of V8SFmode permutation.
275 2014-02-20  Richard Henderson  <rth@redhat.com>
277         PR c++/60272
278         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
279         a new pseudo for OLDVAL.
281 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
283         PR target/57896
284         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
285         gen_reg_rtx if d->testing_p.
286         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
287         if d->testing_p and we will certainly return true.
288         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
289         if d->testing_p.
291 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
293         * emit-rtl.c (gen_reg_rtx): Assert that
294         crtl->emit.regno_pointer_align_length is non-zero.
296 2014-02-20  Richard Henderson  <rth@redhat.com>
298         PR c++/60272
299         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
300         on failure the store back into EXPECT.
302 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
303             Sandra Loosemore  <sandra@codesourcery.com>
305         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
306         * config/nios2/nios2.c (nios2_function_profiler):
307         Add -fPIC (flag_pic == 2) support.
308         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
309         (nios2_large_offset_p): New function.
310         (nios2_unspec_reloc_p): Move up position, update to use
311         nios2_large_offset_p.
312         (nios2_unspec_address): Remove function.
313         (nios2_unspec_offset): New function.
314         (nios2_large_got_address): New function.
315         (nios2_got_address): Add large offset support.
316         (nios2_legitimize_tls_address): Update usage of removed and new
317         functions.
318         (nios2_symbol_binds_local_p): New function.
319         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
320         (nios2_legitimize_address): Update to use nios2_large_offset_p.
321         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
322         (nios2_print_operand): Merge H/L processing, add hiadj/lo
323         processing for (const (unspec ...)).
324         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
326 2014-02-20  Richard Biener  <rguenther@suse.de>
328         * tree-cfg.c (replace_uses_by): Mark altered BBs before
329         doing the substitution.
330         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
332 2014-02-20  Martin Jambor  <mjambor@suse.cz>
334         PR ipa/55260
335         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
336         info when checking whether lattices are bottom.
338 2014-02-20  Richard Biener  <rguenther@suse.de>
340         PR middle-end/60221
341         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
342         regions at -O0.
344 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
346         PR ipa/58555
347         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
348         parameter specifying the scaling.
349         (inline_call): Update.
350         (want_inline_recursively): Guard division by zero.
351         (recursive_inlining): Update.
352         * ipa-inline.h (clone_inlined_nodes): Update.
354 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
356         PR target/60204
357         * config/i386/i386.c (classify_argument): Pass structures of size
358         64 bytes or less in register.
360 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
361             Kirill Yukhin  <kirill.yukhin@intel.com>
363         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
364         (_mm_rcp28_round_ss): Ditto.
365         (_mm_rsqrt28_round_sd): Ditto.
366         (_mm_rsqrt28_round_ss): Ditto.
367         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
368         (_mm_rcp14_round_ss): Ditto.
369         (_mm_rsqrt14_round_sd): Ditto.
370         (_mm_rsqrt14_round_ss): Ditto.
371         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
372         the first input operand, get rid of match_dup.
373         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
374         attribute to sse.
375         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
376         Ditto.
377         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
378         operand as the first input operand, set type attribute.
379         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
380         Set type attribute.
381         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
382         operand as the first input operand, set type attribute.
384 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
386         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
387         bit of zero.
389 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
391         PR target/60207
392         * config/i386/i386.c (construct_container): Remove TFmode check
393         for X86_64_INTEGER_CLASS.
395 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
397         PR target/59794
398         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
399         only when -Wpsabi is enabled.
401 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
403          PR target/59799
404         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
405         passing arrays in registers are the same as for structs, so remove the
406         special case for them.
408 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
410         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
411         destination type, extract only the valid bits if the source type is not
412         integral and has a different mode.
414 2014-02-19  Richard Biener  <rguenther@suse.de>
416         PR ipa/60243
417         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
418         for all calls.
420 2014-02-19  Richard Biener  <rguenther@suse.de>
422         PR ipa/60243
423         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
424         (ipa_modify_call_arguments): Emit an argument load explicitely and
425         preserve virtual SSA form there and for the replacement call.
426         Do not update SSA form nor free dominance info.
428 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
430         * ipa.c (function_and_variable_visibility): Also clear WEAK
431         flag when disolving COMDAT_GROUP.
433 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
435         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
436         * ipa-prop.c (ipa_set_jf_known_type): Return early when
437         not devirtualizing.
438         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
439         do more sanity checks.
440         (detect_type_change): Return true when giving up early.
441         (compute_complex_assign_jump_func): Fix type parameter of
442         ipa_set_ancestor_jf.
443         (compute_complex_ancestor_jump_func): Likewise.
444         (update_jump_functions_after_inlining): Fix updating of
445         ancestor function.
446         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
448 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
450         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
451         inline clones when edge disappears.
453 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
455         PR target/60203
456         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
457         Split 64-bit moves into 2 patterns.  Do not allow the use of
458         direct move for TDmode in little endian, since the decimal value
459         has little endian bytes within a word, but the 64-bit pieces are
460         ordered in a big endian fashion, and normal subreg's of TDmode are
461         not allowed.
462         (mov<mode>_64bit_dm): Likewise.
463         (movtd_64bit_nodm): Likewise.
465 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
467         PR tree-optimization/60174
468         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
469         statement of an SSA_NAME that occurs in an abnormal PHI node.
471 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
473         PR sanitizer/60142
474         * final.c (SEEN_BB): Remove.
475         (SEEN_NOTE, SEEN_EMITTED): Renumber.
476         (final_scan_insn): Don't force_source_line on second
477         NOTE_INSN_BASIC_BLOCK.
479 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
481         PR target/60205
482         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
483         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
484         (type_natural_mode): Warn ABI change when %zmm register is not
485         available for AVX512F vector value passing.
487 2014-02-18  Kai Tietz  <ktietz@redhat.com>
489         PR target/60193
490         * config/i386/i386.c (ix86_expand_prologue): Use value in
491         rax register as displacement when restoring %r10 or %rax.
492         Fix wrong offset when restoring both registers.
494 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
496         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
497         assertion with conditional return.
499 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
500             Uros Bizjak  <ubizjak@gmail.com>
502         PR driver/60233
503         * config/i386/driver-i386.c (host_detect_local_cpu): If
504         YMM state is not saved by the OS, also clear has_f16c.  Move
505         CPUID 0x80000001 handling before YMM state saving checking.
507 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
509         PR rtl-optimization/58960
510         * haifa-sched.c (alloc_global_sched_pressure_data): New,
511         factored out from ...
512         (sched_init): ... here.
513         (free_global_sched_pressure_data): New, factored out from ...
514         (sched_finish): ... here.
515         * sched-int.h (free_global_sched_pressure_data): Declare.
516         * sched-rgn.c (nr_regions_initial): New static global.
517         (haifa_find_rgns): Initialize it.
518         (schedule_region): Disable sched-pressure for the newly
519         generated regions.
521 2014-02-17  Richard Biener  <rguenther@suse.de>
523         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
524         release SSA defs of pattern stmts.
526 2014-02-17  Richard Biener  <rguenther@suse.de>
528         * tree-inline.c (expand_call_inline): Release the virtual
529         operand defined by the call we are about to inline.
531 2014-02-17  Richard Biener  <rguenther@suse.de>
533         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
535 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
536             Ilya Tocar  <ilya.tocar@intel.com>
538         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
539         arguments order in builtin.
540         (_mm512_permutexvar_epi64): Ditto.
541         (_mm512_mask_permutexvar_epi64): Ditto
542         (_mm512_maskz_permutexvar_epi32): Ditto
543         (_mm512_permutexvar_epi32): Ditto
544         (_mm512_mask_permutexvar_epi32): Ditto
546 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
548         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
549         (p8_vmrgow): Likewise.
551 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
553         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
554         endian targets.
556 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
558         PR target/60203
559         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
560         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
561         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
562         using direct move instructions on ISA 2.07.  Also adjust
563         instruction length for 64-bit.
564         (mov<mode>_64bit, TFmode/TDmode): Likewise.
565         (mov<mode>_32bit, TFmode/TDmode): Likewise.
567 2014-02-15  Alan Modra  <amodra@gmail.com>
569         PR target/58675
570         PR target/57935
571         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
572         find_replacement on parts of insn rtl that might be reloaded.
574 2014-02-15  Richard Biener  <rguenther@suse.de>
576         PR tree-optimization/60183
577         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
578         (tree_ssa_phiprop): Calculate and free post-dominators.
580 2014-02-14  Jeff Law  <law@redhat.com>
582         PR rtl-optimization/60131
583         * ree.c (get_extended_src_reg): New function.
584         (combine_reaching_defs): Use it rather than assuming location of REG.
585         (find_and_remove_re): Verify first operand of extension is
586         a REG before adding the insns to the copy list.
588 2014-02-14  Roland McGrath  <mcgrathr@google.com>
590         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
591         * configure: Regenerated.
592         * config.in: Regenerated.
593         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
594         instead of ASM_SHORT.
596 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
597             Richard Earnshaw  <rearnsha@arm.com>
599         PR rtl-optimization/59535
600         * lra-constraints.c (process_alt_operands): Encourage alternative
601         when unassigned pseudo class is superset of the alternative class.
602         (inherit_reload_reg): Don't inherit when optimizing for code size.
603         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
604         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
605         modes not less than 4 for Thumb1.
607 2014-02-14  Kyle McMartin  <kyle@redhat.com>
609         PR pch/60010
610         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
612 2014-02-14  Richard Biener  <rguenther@suse.de>
614         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
615         (get_frame_arg): Drop the assert with langhook types_compatible_p.
616         Do not strip INDIRECT_REFs.
618 2014-02-14  Richard Biener  <rguenther@suse.de>
620         PR lto/60179
621         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
622         DECL_FUNCTION_SPECIFIC_TARGET.
623         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
624         * tree-streamer-out.c (pack_ts_target_option): Remove.
625         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
626         (write_ts_function_decl_tree_pointers): Do not stream
627         DECL_FUNCTION_SPECIFIC_TARGET.
628         * tree-streamer-in.c (unpack_ts_target_option): Remove.
629         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
630         (lto_input_ts_function_decl_tree_pointers): Do not stream
631         DECL_FUNCTION_SPECIFIC_TARGET.
633 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
635         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
636         (get_initial_def_for_induction, vectorizable_induction): Ignore
637         debug stmts when looking for exit_phi.
638         (vectorizable_live_operation): Fix up condition.
640 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
642         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
643         nreverse() because it changes the content of original tree list.
645 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
647         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
648         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
650 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
652         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
653         GNU coding standards.
655 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
657         PR debug/60152
658         * dwarf2out.c (gen_subprogram_die): Don't call
659         add_calling_convention_attribute if subr_die is old_die.
661 2014-02-13  Sharad Singhai  <singhai@google.com>
663         * doc/optinfo.texi: Fix order of nodes.
665 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
667         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
668         operands[2], not operands[3].
670 2014-02-13  Richard Biener  <rguenther@suse.de>
672         PR bootstrap/59878
673         * doc/install.texi (ISL): Update recommended version to 0.12.2,
674         mention the possibility of an in-tree build.
675         (CLooG): Update recommended version to 0.18.1, mention the
676         possibility of an in-tree build and clarify that the ISL
677         bundled with CLooG does not work.
679 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
681         PR target/43546
682         * expr.c (compress_float_constant): If x is a hard register,
683         extend into a pseudo and then move to x.
685 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
687         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
688         caused by bad second argument to warning_at() with -mhotpatch and
689         nested functions (e.g. with gfortran).
691 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
693         * opts.c (option_name): Remove "enabled by default" rider.
695 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
697         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
699 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
700             Uros Bizjak  <ubizjak@gmail.com>
702         PR target/60151
703         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
704         * configure: Regenerated.
706 2014-02-12  Richard Biener  <rguenther@suse.de>
708         * vec.c (vec_prefix::calculate_allocation): Move as
709         inline variant to vec.h.
710         (vec_prefix::calculate_allocation_1): New out-of-line version.
711         * vec.h (vec_prefix::calculate_allocation_1): Declare.
712         (vec_prefix::m_has_auto_buf): Rename to ...
713         (vec_prefix::m_using_auto_storage): ... this.
714         (vec_prefix::calculate_allocation): Inline the easy cases
715         and dispatch to calculate_allocation_1 which doesn't need the
716         prefix address.
717         (va_heap::reserve): Use gcc_checking_assert.
718         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
719         m_using_auto_storage.
720         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
721         member and adjust.
722         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
723         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
724         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
726 2014-02-12  Richard Biener  <rguenther@suse.de>
728         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
729         when we found a dependence.
731 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
733         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
734         common code...
735         (maybe_fold_stmt): ... into this new function.
736         * omp-low.c (lower_omp): Update comment.
738         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
739         last use.
741         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
742         dereference.
744 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
746         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
747         identifiers in comments.
748         (cortexa53_extra_costs): Likewise.
749         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
750         (cortexa7_extra_costs): Likewise.
751         (cortexa12_extra_costs): Likewise.
752         (cortexa15_extra_costs): Likewise.
753         (v7m_extra_costs): Likewise.
755 2014-02-12  Richard Biener  <rguenther@suse.de>
757         PR middle-end/60092
758         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
759         of posix_memalign being successful.
760         (lower_stmt): Restrict lowering of posix_memalign to when
761         -ftree-bit-ccp is enabled.
763 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
765         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
766         arg_loc.
767         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
769 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
771         PR rtl-optimization/60116
772         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
773         other_insn once the combination has been validated.
775 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
777         PR lto/59468
778         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
779         and wrapper.
780         * ipa-devirt.c: Include demangle.h
781         (odr_violation_reported): New static variable.
782         (add_type_duplicate): Update odr_violations.
783         (maybe_record_node): Add completep parameter; update it.
784         (record_target_from_binfo): Add COMPLETEP parameter;
785         update it as needed.
786         (possible_polymorphic_call_targets_1): Likewise.
787         (struct polymorphic_call_target_d): Add nonconstruction_targets;
788         rename FINAL to COMPLETE.
789         (record_targets_from_bases): Sanity check we found the binfo;
790         fix COMPLETEP updating.
791         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
792         parameter, fix computing of COMPLETEP.
793         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
794         at LTO time do demangling.
795         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
796         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
797         parameter.
798         (gimple_get_virt_method_for_binfo): Likewise.
799         * gimple-fold.h (gimple_get_virt_method_for_binfo,
800         gimple_get_virt_method_for_vtable): Update prototypes.
802 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
804         PR target/49008
805         * genautomata.c (add_presence_absence): Fix typo with
806         {final_}presence_list.
808 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
810         PR target/60137
811         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
812         for VSX/Altivec vectors that land in GPR registers.
814 2014-02-11  Richard Henderson  <rth@redhat.com>
815             Jakub Jelinek  <jakub@redhat.com>
817         PR debug/59776
818         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
819         around drhs if type conversion to lacc->type is not useless.
821 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
823         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
824         tuning struct.
825         (cortex-a57.cortex-a53): Likewise.
826         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
828 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
830         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
831         arm_restrict_it.
833 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
835         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
836         add_options_for_arm_vfp3.
838 2014-02-11  Jeff Law  <law@redhat.com>
840         PR middle-end/54041
841         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
842         object with an undesirable mode.
844 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
846         PR libgomp/60107
847         * config/i386/sol2-9.h: New file.
848         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
849         *-*-solaris2.9*): Use it.
851 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
853         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
854         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
856 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
858         * config/microblaze/microblaze.c: Extend mcpu version format
860 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
862         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
864 2014-02-10  Richard Henderson  <rth@redhat.com>
866         PR target/59927
867         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
868         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
869         ms-abi vs -mno-accumulate-outgoing-args.
870         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
871         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
872         respect to ms-abi.
874 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
876         PR middle-end/60080
877         * cfgexpand.c (expand_asm_operands): Attach source location to
878         ASM_INPUT rtx objects.
879         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
881 2014-02-10  Nick Clifton  <nickc@redhat.com>
883         * config/mn10300/mn10300.c (popcount): New function.
884         (mn10300_expand_prologue): Include saved registers in stack usage
885         count.
887 2014-02-10  Jeff Law  <law@redhat.com>
889         PR middle-end/52306
890         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
891         when changing the SET_DEST of a prior insn to avoid an input reload.
893 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
895         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
896         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
897         -mcall-openbsd, or -mcall-linux.
898         (CC1_ENDIAN_BIG_SPEC): Remove.
899         (CC1_ENDIAN_LITTLE_SPEC): Remove.
900         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
901         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
902         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
903         and %cc1_endian_default.
904         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
906 2014-02-10  Richard Biener  <rguenther@suse.de>
908         PR tree-optimization/60115
909         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
910         MEM_REF handling.  Properly verify that the accesses are not
911         out of the objects bound.
913 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
915         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
916         coretex to cortex.
918 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
920         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
921         proper constants and fix formatting.
922         (possible_polymorphic_call_targets): Fix formatting.
924 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
925             Ilya Tocar  <ilya.tocar@intel.com>
927         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
928         (_mm512_loadu_epi32): Renamed into...
929         (_mm512_loadu_si512): This.
930         (_mm512_storeu_epi32): Renamed into...
931         (_mm512_storeu_si512): This.
932         (_mm512_maskz_ceil_ps): Removed.
933         (_mm512_maskz_ceil_pd): Ditto.
934         (_mm512_maskz_floor_ps): Ditto.
935         (_mm512_maskz_floor_pd): Ditto.
936         (_mm512_floor_round_ps): Ditto.
937         (_mm512_floor_round_pd): Ditto.
938         (_mm512_ceil_round_ps): Ditto.
939         (_mm512_ceil_round_pd): Ditto.
940         (_mm512_mask_floor_round_ps): Ditto.
941         (_mm512_mask_floor_round_pd): Ditto.
942         (_mm512_mask_ceil_round_ps): Ditto.
943         (_mm512_mask_ceil_round_pd): Ditto.
944         (_mm512_maskz_floor_round_ps): Ditto.
945         (_mm512_maskz_floor_round_pd): Ditto.
946         (_mm512_maskz_ceil_round_ps): Ditto.
947         (_mm512_maskz_ceil_round_pd): Ditto.
948         (_mm512_expand_pd): Ditto.
949         (_mm512_expand_ps): Ditto.
950         * config/i386/i386.c (ix86_builtins): Remove
951         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
952         (bdesc_args): Ditto.
953         * config/i386/predicates.md (const1256_operand): New.
954         (const_1_to_2_operand): Ditto.
955         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
956         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
957         (*avx512pf_gatherpf<mode>sf): Ditto.
958         (avx512pf_gatherpf<mode>df): Ditto.
959         (*avx512pf_gatherpf<mode>df_mask): Ditto.
960         (*avx512pf_gatherpf<mode>df): Ditto.
961         (avx512pf_scatterpf<mode>sf): Ditto.
962         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
963         (*avx512pf_scatterpf<mode>sf): Ditto.
964         (avx512pf_scatterpf<mode>df): Ditto.
965         (*avx512pf_scatterpf<mode>df_mask): Ditto.
966         (*avx512pf_scatterpf<mode>df): Ditto.
967         (avx512f_expand<mode>): Removed.
968         (<shift_insn><mode>3<mask_name>): Change predicate type.
970 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
972         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
973         not at the end of datarefs vector use ordered_remove to avoid
974         reordering datarefs vector.
976         PR c/59984
977         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
978         mark local addressable non-static vars as GOVD_PRIVATE
979         instead of GOVD_LOCAL.
980         * omp-low.c (lower_omp_for): Move gimple_bind_vars
981         and BLOCK_VARS of gimple_bind_block to new_stmt rather
982         than copying them.
984         PR middle-end/60092
985         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
986         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
987         assume_aligned or alloc_align attributes.
988         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
989         arguments.  Handle also assume_aligned and alloc_align attributes.
990         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
991         calls to functions with assume_aligned or alloc_align attributes.
992         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
994 2014-02-08  Terry Guo  <terry.guo@arm.com>
996         * doc/invoke.texi: Document ARM -march=armv7e-m.
998 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1000         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
1001         flag on __cilkrts_rethrow builtin.
1003         PR ipa/60026
1004         * ipa-cp.c (determine_versionability): Fail at -O0
1005         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
1006         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
1008         Revert:
1009         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1011         PR ipa/60026
1012         * tree-inline.c (copy_forbidden): Fail for
1013         __attribute__((optimize (0))) functions.
1015 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
1017         * varpool.c: Include pointer-set.h.
1018         (varpool_remove_unreferenced_decls): Variables in other partitions
1019         will not be output; be however careful to not lose information
1020         about partitioning.
1022 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
1024         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
1025         lookup in the vtable constructor.
1027 2014-02-07  Jeff Law  <law@redhat.com>
1029         PR target/40977
1030         * config/m68k/m68k.md (ashldi_extsi): Turn into a
1031         define_insn_and_split.
1033         * ipa-inline.c (inline_small_functions): Fix typos.
1035 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1037         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
1038         (s390_can_use_return_insn): Declare.
1039         * config/s390/s390.h (EPILOGUE_USES): Define.
1040         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
1041         instructions.
1042         (s390_chunkify_start): Handle return JUMP_LABELs.
1043         (s390_early_mach): Emit a main_pool instruction on the entry edge.
1044         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
1045         (s390_can_use_return_insn): New functions.
1046         (s390_fix_long_loop_prediction): Handle conditional returns.
1047         (TARGET_SET_UP_BY_PROLOGUE): Define.
1048         * config/s390/s390.md (ANY_RETURN): New code iterator.
1049         (*creturn, *csimple_return, return, simple_return): New patterns.
1051 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1053         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
1054         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
1055         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
1056         REG_CFA_RESTORE list when deciding not to restore a register.
1058 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1060         * config/s390/s390.c: Include tree-pass.h and context.h.
1061         (s390_early_mach): New function, split out from...
1062         (s390_emit_prologue): ...here.
1063         (pass_data_s390_early_mach): New pass structure.
1064         (pass_s390_early_mach): New class.
1065         (s390_option_override): Create and register early_mach pass.
1066         Move to end of file.
1068 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1070         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
1071         to match for the exit block.
1073 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1075         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
1076         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
1077         Reject misaligned operands.
1079 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1081         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
1083 2014-02-07  Richard Biener  <rguenther@suse.de>
1085         PR middle-end/60092
1086         * gimple-low.c (lower_builtin_posix_memalign): New function.
1087         (lower_stmt): Call it to lower posix_memalign in a way
1088         to make alignment info accessible.
1090 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
1092         PR c++/60082
1093         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
1094         __builtin_setjmp_receiver.
1096 2014-02-07  Richard Biener  <rguenther@suse.de>
1098         PR middle-end/60092
1099         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
1100         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
1101         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1102         Handle BUILT_IN_POSIX_MEMALIGN.
1103         (find_func_clobbers): Likewise.
1104         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
1105         (call_may_clobber_ref_p_1): Likewise.
1107 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1109         PR ipa/59918
1110         * ipa-devirt.c (record_target_from_binfo): Remove overactive
1111         sanity check.
1113 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1115         PR ipa/59469
1116         * lto-cgraph.c (lto_output_node): Use
1117         symtab_get_symbol_partitioning_class.
1118         (lto_output_varpool_node): likewise.
1119         (symtab_get_symbol_partitioning_class): Move here from
1120         lto/lto-partition.c
1121         * cgraph.h (symbol_partitioning_class): Likewise.
1122         (symtab_get_symbol_partitioning_class): Declare.
1124 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1126         * ggc.h (ggc_internal_cleared_alloc): New macro.
1127         * vec.h (vec_safe_copy): Handle memory stats.
1128         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
1129         * target-globals.c (save_target_globals): Likewise.
1131 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1133         PR target/60077
1134         * expr.c (emit_move_resolve_push): Export; be bit more selective
1135         on when to clear alias set.
1136         * expr.h (emit_move_resolve_push): Declare.
1137         * function.h (struct function): Add tail_call_marked.
1138         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
1139         * config/i386/i386-protos.h (ix86_expand_push): Remove.
1140         * config/i386/i386.md (TImode move expander): De not call
1141         ix86_expand_push.
1142         (FP push expanders): Preserve memory attributes.
1143         * config/i386/sse.md (push<mode>1): Remove.
1144         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
1145         (ix86_expand_push): Remove.
1146         * config/i386/mmx.md (push<mode>1): Remove.
1148 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1150         PR rtl-optimization/60030
1151         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
1152         lopart with paradoxical subreg before shifting it up by hprec.
1154 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1156         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
1157         Remove extra newline at end of file.
1158         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
1159         (arm_issue_rate): Handle cortexa57.
1160         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
1161         (cortex-a57.cortex-a53): Likewise.
1163 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1165         PR target/59575
1166         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
1167         don't record in REG_FRAME_RELATED_EXPR registers not set in that
1168         bitmask.
1169         (arm_expand_prologue): Adjust all callers.
1170         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
1171         info, registers also at the lowest numbered registers side.  Use
1172         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
1173         XEXP.
1175         PR debug/59992
1176         * var-tracking.c (adjust_mems): Before adding a SET to
1177         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
1179 2014-02-06  Alan Modra  <amodra@gmail.com>
1181         PR target/60032
1182         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
1183         change SDmode to DDmode when lra_in_progress.
1185 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1187         PR middle-end/59150
1188         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
1189         free_data_ref on the dr first, and before goto again also set dr
1190         to the next dr.  For simd_lane_access, free old datarefs[i] before
1191         overwriting it.  For get_vectype_for_scalar_type failure, don't
1192         free_data_ref if simd_lane_access.
1194         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
1196         PR target/60062
1197         * tree.h (opts_for_fn): New inline function.
1198         (opt_for_fn): Define.
1199         * config/i386/i386.c (ix86_function_regparm): Use
1200         opt_for_fn (decl, optimize) instead of optimize.
1202 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1204         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
1205         for SYMBOL_REF in large memory model.
1207 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1209         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
1210         and crypto support.
1211         (cortex-a57): Likewise.
1212         (cortex-a57.cortex-a53): Likewise.
1214 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
1215             Kugan Vivekanandarajah  <kuganv@linaro.org>
1217         * config/arm/arm.c (arm_vector_alignment_reachable): Check
1218         unaligned_access.
1219         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
1221 2014-02-06  Richard Biener  <rguenther@suse.de>
1223         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
1224         set_loop_copy and initialize_original_copy_tables.
1226 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
1228         * config/aarch64/aarch64-simd.md
1229         (aarch64_ashr_simddi): Change QI to SI.
1231 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
1232             Jakub Jelinek  <jakub@redhat.com>
1234         PR middle-end/60013
1235         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
1236         of the dataflow.
1238 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1240         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1241         CODE_FOR_altivec_vpku[hw]um to
1242         CODE_FOR_altivec_vpku[hw]um_direct.
1243         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1244         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1245         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1246         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1248 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1250         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1251         generation for -maltivec=be.
1252         (altivec_vsumsws): Simplify redundant test.
1254 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1256         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1257         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1258         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1259         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1260         gen_altivec_vpkuwum.
1261         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1262         BYTES_BIG_ENDIAN.
1263         (altivec_vpks<VI_char>ss): Likewise.
1264         (altivec_vpks<VI_char>us): Likewise.
1265         (altivec_vpku<VI_char>us): Likewise.
1266         (altivec_vpku<VI_char>um): Likewise.
1267         (altivec_vpku<VI_char>um_direct): New (copy of
1268         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1269         internal use).
1270         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1271         target is little endian and -maltivec=be is not specified.
1272         (*altivec_vupkhs<VU_char>_direct): New (copy of
1273         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1274         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1275         target is little endian and -maltivec=be is not specified.
1276         (*altivec_vupkls<VU_char>_direct): New (copy of
1277         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1278         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1279         little endian and -maltivec=be is not specified.
1280         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1281         little endian and -maltivec=be is not specified.
1283 2014-02-05  Richard Henderson  <rth@redhat.com>
1285         PR debug/52727
1286         * combine-stack-adj.c: Revert r206943.
1287         * sched-int.h (struct deps_desc): Add last_args_size.
1288         * sched-deps.c (init_deps): Initialize it.
1289         (sched_analyze_insn): Add OUTPUT dependencies between insns that
1290         contain REG_ARGS_SIZE notes.
1292 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
1294         * lto-cgraph.c (asm_nodes_output): Make global.
1295         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1296         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1297         (driver_handle_option): Handle OPT_fwpa.
1299 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
1301         PR ipa/59947
1302         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1303         a comment typo and formatting issue.  If odr_hash hasn't been
1304         created, return vNULL and set *completep to false.
1306         PR middle-end/57499
1307         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1308         bb with no successors.
1310 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
1312         PR target/59718
1313         * doc/invoke.texi (-march): Clarify documentation for ARM.
1314         (-mtune): Likewise.
1315         (-mcpu): Likewise.
1317 2014-02-05  Richard Biener  <rguenther@suse.de>
1319         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1320         when not vectorizing because of too many alias checks.
1321         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1322         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1324 2014-02-05  Nick Clifton  <nickc@redhat.com>
1326         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1327         accept extended registers in any mode when compiling for the MN10300.
1329 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
1331         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1332         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1333         sanitization attributes.
1334         (can_inline_edge_p): Likewise.
1335         (sanitize_attrs_match_for_inline_p): New function.
1337 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1339         * ipa-prop.c (detect_type_change): Shor circuit testing of
1340         type changes on THIS pointer.
1342 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
1344         PR target/59777
1345         * config/pa/pa.c (legitimize_tls_address): Return original address
1346         if not passed a SYMBOL_REF rtx.
1347         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1348         addresses.
1349         (pa_emit_move_sequence): Simplify TLS source operands.
1350         (pa_legitimate_constant_p): Reject all TLS constants.
1351         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1352         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1354 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1356         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1357         groups when we know they are controlled by LTO.
1358         * varasm.c (default_binds_local_p_1): If object is in other partition,
1359         it will be resolved locally.
1361 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1363         * config/host-linux.c (linux_gt_pch_use_address): Don't
1364         use SSIZE_MAX because it is not always defined.
1366 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
1368         PR bootstrap/59913
1369         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1370         threshold for pseudo splitting.
1371         (update_ebb_live_info): Process call argument hard registers and
1372         hard registers from insn definition too.
1373         (max_small_class_regs_num): New constant.
1374         (inherit_in_ebb): Update live hard regs through EBBs.  Update
1375         reloads_num only for small register classes.  Don't split for
1376         outputs of jumps.
1378 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
1380         PR ipa/60058
1381         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1382         is non-null.
1384 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1386         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1387         visibility is safe.
1389 2014-02-04  Marek Polacek  <polacek@redhat.com>
1391         * gdbinit.in (pel): Define.
1393 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1395         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1396         behavior.
1398 2014-02-04  Richard Biener  <rguenther@suse.de>
1400         PR lto/59723
1401         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1402         in function context local.
1403         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1404         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1405         similar to LTO_imported_decl_ref.
1407 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1409         PR tree-optimization/60002
1410         * cgraphclones.c (build_function_decl_skip_args): Clear
1411         DECL_LANG_SPECIFIC.
1413         PR tree-optimization/60023
1414         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1415         false to gsi_replace.
1416         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1417         has been in some EH region and vec_stmt could throw, add
1418         vec_stmt into the same EH region.
1419         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1420         has no lhs, ignore it.
1421         * internal-fn.c (expand_MASK_LOAD): Likewise.
1423         PR ipa/60026
1424         * tree-inline.c (copy_forbidden): Fail for
1425         __attribute__((optimize (0))) functions.
1427         PR other/58712
1428         * omp-low.c (simd_clone_struct_copy): If from->inbranch
1429         is set, copy one less argument.
1430         (expand_simd_clones): Don't subtract clone_info->inbranch
1431         from simd_clone_struct_alloc argument.
1433         PR rtl-optimization/57915
1434         * recog.c (simplify_while_replacing): If all unary/binary/relational
1435         operation arguments are constant, attempt to simplify those.
1437         PR middle-end/59261
1438         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1439         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1441 2014-02-04  Richard Biener  <rguenther@suse.de>
1443         PR tree-optimization/60012
1444         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1445         TBAA disambiguation to all DDRs.
1447 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1449         PR target/59788
1450         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1451         (LINK_SPEC): Use it for -shared, -shared-libgcc.
1453 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1455         PR ipa/59882
1456         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1458 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1460         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1461         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1463 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1465         PR ipa/59831
1466         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1467         to figure out targets of polymorphic calls with known decl.
1468         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1469         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1470         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1471         (get_polymorphic_call_info): ... here.
1472         (get_polymorphic_call_info_from_invariant): New function.
1474 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1476         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1477         lookup via vtable pointer; check for type consistency
1478         and turn inconsitent facts into UNREACHABLE.
1479         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1480         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1481         type inconsistent querries; return UNREACHABLE instead.
1483 2014-02-03  Richard Henderson  <rth@twiddle.net>
1485         PR tree-opt/59924
1486         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1487         already processed this node.
1488         (normalize_one_pred_1): Pass along mark_set.
1489         (normalize_one_pred): Create and destroy a pointer_set_t.
1490         (normalize_one_pred_chain): Likewise.
1492 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
1494         PR gcov-profile/58602
1495         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1497 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1499         PR ipa/59831
1500         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1501         -fno-devirtualize; try to devirtualize by the knowledge of
1502         virtual table pointer given by aggregate propagation.
1503         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1504         (ipa_print_node_jump_functions): Dump also offset that
1505         is relevant for polymorphic calls.
1506         (determine_known_aggregate_parts): Add arg_type parameter; use it
1507         instead of determining the type from pointer type.
1508         (ipa_compute_jump_functions_for_edge): Update call of
1509         determine_known_aggregate_parts.
1510         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1511         (gimple_get_virt_method_for_binfo): ... here; simplify using
1512         vtable_pointer_value_to_vtable.
1513         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1514         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1515         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1516         (vtable_pointer_value_to_vtable): Break out from ...; handle also
1517         POINTER_PLUS_EXPR.
1518         (vtable_pointer_value_to_binfo): ... here.
1519         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1521 2014-02-03  Teresa Johnson  <tejohnson@google.com>
1523         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1524         redef of outer loop index variable.
1526 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
1528         PR c++/53017
1529         PR c++/59211
1530         * doc/extend.texi (Function Attributes): Typo.
1532 2014-02-03  Cong Hou  <congh@google.com>
1534         PR tree-optimization/60000
1535         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1536         if the vectorized statement is a store.  A store statement can only
1537         appear at the end of pattern statements.
1539 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1541         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1542         (ix86_option_override_internal): Default long double to 64-bit for
1543         32-bit Bionic and to 128-bit for 64-bit Bionic.
1545         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1546         TARGET_LONG_DOUBLE_128 is true.
1547         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1549         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1550         (mlong-double-64): Negate -mlong-double-128.
1551         (mlong-double-128): New option.
1553         * config/i386/i386-c.c (ix86_target_macros): Define
1554         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1556         * doc/invoke.texi: Document -mlong-double-128.
1558 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1560         PR rtl-optimization/60024
1561         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1563 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
1565         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1567 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1569         PR rtl-optimization/57662
1570         * sel-sched.c (code_motion_path_driver): Do not mark already not
1571         existing blocks in the visiting bitmap.
1573 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1575         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1576         on the insn being emitted.
1578 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
1579             Will Deacon  <will.deacon@arm.com>
1581         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1583 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1585         * config/arm/arm-tables.opt: Regenerate.
1587 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1589         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1590         for vector types other than V16QImode.
1591         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1592         define_expand, and call altivec_expand_vec_perm_le when producing
1593         code with little endian element order.
1594         (*altivec_vperm_<mode>_internal): New insn having previous
1595         behavior of altivec_vperm_<mode>.
1596         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1597         altivec_expand_vec_perm_le when producing code with little endian
1598         element order.
1599         (*altivec_vperm_<mode>_uns_internal): New insn having previous
1600         behavior of altivec_vperm_<mode>_uns.
1602 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1604         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1605         (altivec_vsumsws): Add handling for -maltivec=be with a little
1606         endian target.
1607         (altivec_vsumsws_direct): New.
1608         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1609         gen_altivec_vsumsws.
1611 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1613         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1614         vtable_pointer_value_to_binfo): New functions.
1615         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1616         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1618 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
1620         * config/nios2/nios2.md (load_got_register): Initialize GOT
1621         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1622         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1624 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1626         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1627         preserverd by passthrough, do not propagate the type.
1629 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1631         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1632         (mips_atomic_assign_expand_fenv): New function.
1633         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1635 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1637         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1638         (__builtin_mips_set_fcsr): Likewise.
1639         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1640         MIPS_USI_FTYPE_VOID.
1641         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1642         (mips16_expand_set_fcsr): Likewise.
1643         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1644         (mips16_set_fcsr_stub): Likewise.
1645         (mips16_get_fcsr_one_only_stub): New class.
1646         (mips16_set_fcsr_one_only_stub): Likewise.
1647         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1648         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1649         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1650         (hard_float): New availability predicate.
1651         (mips_builtins): Add get_fcsr and set_fcsr.
1652         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1653         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1654         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1655         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1656         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1657         patterns.
1659 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1661         * config/mips/mips.c (mips_one_only_stub): New class.
1662         (mips_need_mips16_rdhwr_p): Replace with...
1663         (mips16_rdhwr_stub): ...this new variable.
1664         (mips16_stub_call_address): New function.
1665         (mips16_rdhwr_one_only_stub): New class.
1666         (mips_expand_thread_pointer): Use mips16_stub_call_address.
1667         (mips_output_mips16_rdhwr): Delete.
1668         (mips_finish_stub): New function.
1669         (mips_code_end): Use it to handle rdhwr stubs.
1671 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
1673         PR target/60017
1674         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1675         when calculating size of integer atomic types.
1677 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
1679         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1681 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
1683         PR tree-optimization/60003
1684         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1685         * profile.c (branch_prob): Use gimple_call_builtin_p
1686         to check for BUILT_IN_SETJMP_RECEIVER.
1687         * tree-inline.c (copy_bb): Call notice_special_calls.
1689 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1691         PR bootstrap/59985
1692         * lra-constraints.c (process_alt_operands): Update reload_sum only
1693         on the first pass.
1695 2014-01-31  Richard Henderson  <rth@redhat.com>
1697         PR middle-end/60004
1698         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1699         until after else_eh is processed.
1701 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1703         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1704         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1705         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1706         in smmintrin.h, remove them.
1707         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1708         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1709         * config/i386/i386.md (ROUND_SAE): Fix value.
1710         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1711         (const48_operand): New.
1712         * config/i386/subst.md (round), (round_expand): Use
1713         const_4_or_8_to_11_operand.
1714         (round_saeonly), (round_saeonly_expand): Use const48_operand.
1716 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1718         * config/i386/constraints.md (Yk): Swap meaning with k.
1719         * config/i386/i386.md (movhi_internal): Change Yk to k.
1720         (movqi_internal): Ditto.
1721         (*k<logic><mode>): Ditto.
1722         (*andhi_1): Ditto.
1723         (*andqi_1): Ditto.
1724         (kandn<mode>): Ditto.
1725         (*<code>hi_1): Ditto.
1726         (*<code>qi_1): Ditto.
1727         (kxnor<mode>): Ditto.
1728         (kortestzhi): Ditto.
1729         (kortestchi): Ditto.
1730         (kunpckhi): Ditto.
1731         (*one_cmplhi2_1): Ditto.
1732         (*one_cmplqi2_1): Ditto.
1733         * config/i386/sse.md (): Change k to Yk.
1734         (avx512f_load<mode>_mask): Ditto.
1735         (avx512f_blendm<mode>): Ditto.
1736         (avx512f_store<mode>_mask): Ditto.
1737         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1738         (avx512f_storedqu<mode>_mask): Ditto.
1739         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1740         Ditto.
1741         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1742         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1743         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1744         (avx512f_maskcmp<mode>3): Ditto.
1745         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1746         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1747         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1748         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1749         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1750         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1751         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1752         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1753         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1754         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1755         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1756         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1757         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1758         (vec_extract_lo_<mode>_maskm): Ditto.
1759         (vec_extract_hi_<mode>_maskm): Ditto.
1760         (avx512f_vternlog<mode>_mask): Ditto.
1761         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1762         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1763         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1764         (avx512f_<code>v8div16qi2_mask): Ditto.
1765         (avx512f_<code>v8div16qi2_mask_store): Ditto.
1766         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1767         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1768         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1769         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1770         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1771         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1772         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1773         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1774         (avx512cd_maskb_vec_dupv8di): Ditto.
1775         (avx512cd_maskw_vec_dupv16si): Ditto.
1776         (avx512f_vpermi2var<mode>3_maskz): Ditto.
1777         (avx512f_vpermi2var<mode>3_mask): Ditto.
1778         (avx512f_vpermi2var<mode>3_mask): Ditto.
1779         (avx512f_vpermt2var<mode>3_maskz): Ditto.
1780         (*avx512f_gathersi<mode>): Ditto.
1781         (*avx512f_gathersi<mode>_2): Ditto.
1782         (*avx512f_gatherdi<mode>): Ditto.
1783         (*avx512f_gatherdi<mode>_2): Ditto.
1784         (*avx512f_scattersi<mode>): Ditto.
1785         (*avx512f_scatterdi<mode>): Ditto.
1786         (avx512f_compress<mode>_mask): Ditto.
1787         (avx512f_compressstore<mode>_mask): Ditto.
1788         (avx512f_expand<mode>_mask): Ditto.
1789         * config/i386/subst.md (mask): Change k to Yk.
1790         (mask_scalar_merge): Ditto.
1791         (sd): Ditto.
1793 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
1795         * doc/extend.texi (Vector Extensions): Document ?: in C++.
1797 2014-01-31  Richard Biener  <rguenther@suse.de>
1799         PR middle-end/59990
1800         * builtins.c (fold_builtin_memory_op): Make sure to not
1801         use a floating-point mode or a boolean or enumeral type for
1802         the copy operation.
1804 2014-01-30  DJ Delorie  <dj@redhat.com>
1806         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1807         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1808         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1809         whenever main() has an epilogue.
1811 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1813         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1814         unused variable "field".
1815         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1816         (vsx_mergeh_<mode>): Likewise.
1817         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1818         (altivec_vmrghh): Likewise.
1819         (altivec_vmrghw): Likewise.
1820         (altivec_vmrglb): Likewise.
1821         (altivec_vmrglh): Likewise.
1822         (altivec_vmrglw): Likewise.
1823         (altivec_vspltb): Add missing uses.
1824         (altivec_vsplth): Likewise.
1825         (altivec_vspltw): Likewise.
1826         (altivec_vspltsf): Likewise.
1828 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1830         PR target/59923
1831         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1832         frame related instructions.
1834 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1836         PR rtl-optimization/59959
1837         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1838         any reload of register whose subreg is invalid.
1840 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1842         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1843         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1844         Add missing return type - void.
1846 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1848         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1849         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1850         remove element index adjustment for endian (now handled in vsx.md
1851         and altivec.md).
1852         (altivec_expand_vec_perm_const): Use
1853         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1854         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1855         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1856         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1857         define_expand and a new define_insn *altivec_vspltb_internal;
1858         adjust for -maltivec=be on a little endian target.
1859         (altivec_vspltb_direct): New.
1860         (altivec_vsplth): Divide into a define_expand and a new
1861         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1862         little endian target.
1863         (altivec_vsplth_direct): New.
1864         (altivec_vspltw): Divide into a define_expand and a new
1865         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1866         little endian target.
1867         (altivec_vspltw_direct): New.
1868         (altivec_vspltsf): Divide into a define_expand and a new
1869         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1870         a little endian target.
1872 2014-01-30  Richard Biener  <rguenther@suse.de>
1874         PR tree-optimization/59993
1875         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1876         can propagate form the earlier stmt and avoid the transform
1877         when the intermediate result is needed.
1879 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1881         * README.Portability: Fix typo.
1883 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1885         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1886         comparison_operator with ordered_comparison_operator.
1888 2014-01-30  Nick Clifton  <nickc@redhat.com>
1890         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1891         Rename to mn10300_store_multiple_regs.
1892         * config/mn10300/mn10300.c: Likewise.
1893         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1894         store_multiple_regs.
1895         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1896         Call mn10300_store_multiple_regs.
1898 2014-01-30  Nick Clifton  <nickc@redhat.com>
1899             DJ Delorie  <dj@redhat.com>
1901         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1902         %fp 2 to keep registers after it properly word-aligned.
1903         (rl78_alloc_physical_registers_umul): Handle the case where both
1904         input operands are the same.
1906 2014-01-30  Richard Biener  <rguenther@suse.de>
1908         PR tree-optimization/59903
1909         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1910         check properly.
1912 2014-01-30  Jason Merrill  <jason@redhat.com>
1914         PR c++/59633
1915         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1917         PR c++/59645
1918         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1920 2014-01-30  Richard Biener  <rguenther@suse.de>
1922         PR tree-optimization/59951
1923         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1925 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1927         PR target/59784
1928         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1929         SFmode to DFmode case.
1931 2014-01-29  DJ Delorie  <dj@redhat.com>
1933         * config/msp430/msp430.opt (-minrt): New.
1934         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1935         if -minrt given.
1936         (ENDFILE_SPEC): Likewise.
1938 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1940         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1941         (estimate_function_body_sizes): Use it.
1943 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1945         PR c++/58561
1946         * dwarf2out.c (is_cxx_auto): New.
1947         (is_base_type): Use it.
1948         (gen_type_die_with_usage): Likewise.
1950 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1952         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1953         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1954         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1955         -maltivec=be with LE targets.
1956         (vsx_mergeh_<mode>): Likewise.
1957         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1958         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1959         (altivec_vmrghb): Replace with define_expand and new
1960         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1961         (altivec_vmrghb_direct): New define_insn.
1962         (altivec_vmrghh): Replace with define_expand and new
1963         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1964         (altivec_vmrghh_direct): New define_insn.
1965         (altivec_vmrghw): Replace with define_expand and new
1966         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1967         (altivec_vmrghw_direct): New define_insn.
1968         (*altivec_vmrghsf): Adjust for endianness.
1969         (altivec_vmrglb): Replace with define_expand and new
1970         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1971         (altivec_vmrglb_direct): New define_insn.
1972         (altivec_vmrglh): Replace with define_expand and new
1973         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1974         (altivec_vmrglh_direct): New define_insn.
1975         (altivec_vmrglw): Replace with define_expand and new
1976         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1977         (altivec_vmrglw_direct): New define_insn.
1978         (*altivec_vmrglsf): Adjust for endianness.
1979         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1980         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1981         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1982         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1983         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1984         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1985         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1986         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1988 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1990         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1991         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1992         whitespace.
1994 2014-01-29  Richard Biener  <rguenther@suse.de>
1996         PR tree-optimization/58742
1997         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1998         associate_pointerplus_align.
1999         (associate_pointerplus_diff): New function.
2000         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
2001         and associate_pointerplus_diff.
2003 2014-01-29  Richard Biener  <rguenther@suse.de>
2005         * lto-streamer.h (LTO_major_version): Bump to 3.
2006         (LTO_minor_version): Reset to 0.
2008 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
2010         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
2011         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
2012         (arm_file_start): Generate correct asm header for armv7ve.
2013         * config/arm/bpabi.h: Add multilib support for armv7ve.
2014         * config/arm/driver-arm.c: Change the architectures of cortex-a7
2015         and cortex-a15 to armv7ve.
2016         * config/arm/t-aprofile: Add multilib support for armv7ve.
2017         * doc/invoke.texi: Document -march=armv7ve.
2019 2014-01-29  Richard Biener  <rguenther@suse.de>
2021         PR tree-optimization/58742
2022         * tree-ssa-forwprop.c (associate_plusminus): Return true
2023         if we changed sth, defer EH cleanup to ...
2024         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
2025         (simplify_mult): New function.
2027 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
2029         PR middle-end/59917
2030         PR tree-optimization/59920
2031         * tree.c (build_common_builtin_nodes): Remove
2032         __builtin_setjmp_dispatcher initialization.
2033         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
2034         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
2035         instead of gsi_after_labels + manually skipping debug stmts.
2036         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
2037         ignore bbs with IFN_ABNORMAL_DISPATCHER.
2038         * tree-inline.c (copy_edges_for_bb): Remove
2039         can_make_abnormal_goto argument, instead add abnormal_goto_dest
2040         argument.  Ignore computed_goto_p stmts.  Don't call
2041         make_abnormal_goto_edges.  If a call might need abnormal edges
2042         for non-local gotos, see if it already has an edge to
2043         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
2044         with true argument, don't do anything then, otherwise add
2045         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
2046         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
2047         caller.
2048         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
2049         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
2050         (lower_stmt): Don't set data->calls_builtin_setjmp.
2051         (lower_builtin_setjmp): Adjust comment.
2052         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
2053         * tree-cfg.c (found_computed_goto): Remove.
2054         (factor_computed_gotos): Remove.
2055         (make_goto_expr_edges): Return bool, true for computed gotos.
2056         Don't call make_abnormal_goto_edges.
2057         (build_gimple_cfg): Don't set found_computed_goto, don't call
2058         factor_computed_gotos.
2059         (computed_goto_p): No longer static.
2060         (make_blocks): Don't set found_computed_goto.
2061         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
2062         (make_edges): If make_goto_expr_edges returns true, push bb
2063         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
2064         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
2065         vector.  Record mapping between bbs and OpenMP regions if there
2066         are any, adjust make_gimple_omp_edges caller.  Call
2067         handle_abnormal_edges.
2068         (make_abnormal_goto_edges): Remove.
2069         * tree-cfg.h (make_abnormal_goto_edges): Remove.
2070         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
2071         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
2072         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
2073         * internal-fn.def (ABNORMAL_DISPATCHER): New.
2074         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
2075         filling *region also set *region_idx to (*region)->entry->index.
2077         PR other/58712
2078         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
2079         For REGs set ORIGINAL_REGNO.
2081 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
2083         * doc/md.texi: Mention that a target shouldn't implement
2084         vec_widen_(s|u)mul_even/odd pair if it is less efficient
2085         than hi/lo pair.
2087 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
2089         PR tree-optimization/59594
2090         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
2091         a copy of the datarefs vector rather than the vector itself.
2093 2014-01-28  Jason Merrill  <jason@redhat.com>
2095         PR c++/53756
2096         * dwarf2out.c (auto_die): New static.
2097         (gen_type_die_with_usage): Handle C++1y 'auto'.
2098         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
2099         on definition.
2101 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2103         PR target/59672
2104         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
2105         (SPEC_X32): Likewise.
2106         (SPEC_64): Likewise.
2107         * config/i386/i386.c (ix86_option_override_internal): Turn off
2108         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
2109         for TARGET_16BIT.
2110         (x86_file_start): Output .code16gcc for TARGET_16BIT.
2111         * config/i386/i386.h (TARGET_16BIT): New macro.
2112         (TARGET_16BIT_P): Likewise.
2113         * config/i386/i386.opt: Add m16.
2114         * doc/invoke.texi: Document -m16.
2116 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
2118         PR preprocessor/59935
2119         * input.c (location_get_source_line): Bail out on when line number
2120         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
2122 2014-01-28  Richard Biener  <rguenther@suse.de>
2124         PR tree-optimization/58742
2125         * tree-ssa-forwprop.c (associate_plusminus): Handle
2126         pointer subtraction of the form (T)(P + A) - (T)P.
2128 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2130         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
2131         at const_int_cost.
2133 2014-01-28  Richard Biener  <rguenther@suse.de>
2135         Revert
2136         2014-01-28  Richard Biener  <rguenther@suse.de>
2138         PR rtl-optimization/45364
2139         PR rtl-optimization/59890
2140         * var-tracking.c (local_get_addr_clear_given_value): Handle
2141         already cleared slot.
2142         (val_reset): Handle not allocated local_get_addr_cache.
2143         (vt_find_locations): Use post-order on the inverted CFG.
2145 2014-01-28  Richard Biener  <rguenther@suse.de>
2147         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
2149 2014-01-28  Richard Biener  <rguenther@suse.de>
2151         PR rtl-optimization/45364
2152         PR rtl-optimization/59890
2153         * var-tracking.c (local_get_addr_clear_given_value): Handle
2154         already cleared slot.
2155         (val_reset): Handle not allocated local_get_addr_cache.
2156         (vt_find_locations): Use post-order on the inverted CFG.
2158 2014-01-28  Alan Modra  <amodra@gmail.com>
2160         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
2161         * configure.ac <recursive call for build != host>: Define
2162         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
2163         and LD_FOR_BUILD too.
2164         * configure: Regenerate.
2166 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
2168         * config/i386/i386.c (get_builtin_code_for_version): Separate
2169         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
2170         Broadwell from Haswell.
2172 2014-01-27  Steve Ellcey  <sellcey@mips.com>
2174         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2175         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
2176         * config/mips/mips.c (mips_option_override): Change setting
2177         of TARGET_DSP.
2178         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
2179         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
2180         Change from Mask to Var.
2182 2014-01-27  Jeff Law  <law@redhat.com>
2184         * ipa-inline.c (inline_small_functions): Fix typo.
2186 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
2188         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
2189         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
2190         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
2191         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
2192         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
2193         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
2194         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
2195         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
2196         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
2197         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
2198         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
2199         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
2200         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
2201         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
2202         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
2203         (_mm512_storeu_epi64): Ditto.
2204         (_mm512_cmpge_epi32_mask): Ditto.
2205         (_mm512_cmpge_epu32_mask): Ditto.
2206         (_mm512_cmpge_epi64_mask): Ditto.
2207         (_mm512_cmpge_epu64_mask): Ditto.
2208         (_mm512_cmple_epi32_mask): Ditto.
2209         (_mm512_cmple_epu32_mask): Ditto.
2210         (_mm512_cmple_epi64_mask): Ditto.
2211         (_mm512_cmple_epu64_mask): Ditto.
2212         (_mm512_cmplt_epi32_mask): Ditto.
2213         (_mm512_cmplt_epu32_mask): Ditto.
2214         (_mm512_cmplt_epi64_mask): Ditto.
2215         (_mm512_cmplt_epu64_mask): Ditto.
2216         (_mm512_cmpneq_epi32_mask): Ditto.
2217         (_mm512_cmpneq_epu32_mask): Ditto.
2218         (_mm512_cmpneq_epi64_mask): Ditto.
2219         (_mm512_cmpneq_epu64_mask): Ditto.
2220         (_mm512_expand_pd): Ditto.
2221         (_mm512_expand_ps): Ditto.
2222         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
2223         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
2224         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
2225         * config/i386/i386.c (ix86_builtins): Add
2226         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
2227         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
2228         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
2229         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
2230         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
2231         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
2232         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
2233         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
2234         IX86_BUILTIN_PMOVUSQW512_MEM.
2235         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
2236         __builtin_ia32_pmovsqd512mem_mask,
2237         __builtin_ia32_pmovqd512mem_mask,
2238         __builtin_ia32_pmovusqw512mem_mask,
2239         __builtin_ia32_pmovsqw512mem_mask,
2240         __builtin_ia32_pmovqw512mem_mask,
2241         __builtin_ia32_pmovusdw512mem_mask,
2242         __builtin_ia32_pmovsdw512mem_mask,
2243         __builtin_ia32_pmovdw512mem_mask,
2244         __builtin_ia32_pmovqb512mem_mask,
2245         __builtin_ia32_pmovusqb512mem_mask,
2246         __builtin_ia32_pmovsqb512mem_mask,
2247         __builtin_ia32_pmovusdb512mem_mask,
2248         __builtin_ia32_pmovsdb512mem_mask,
2249         __builtin_ia32_pmovdb512mem_mask.
2250         (bdesc_args): Add __builtin_ia32_expanddf512,
2251         __builtin_ia32_expandsf512.
2252         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
2253         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
2254         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
2255         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
2256         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
2257         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
2258         (avx512f_<code>v8div16qi2_mask_store): This.
2259         (avx512f_expand<mode>): New.
2261 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
2263         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2264         New.
2265         (_mm512_mask_prefetch_i64gather_pd): Ditto.
2266         (_mm512_prefetch_i32scatter_pd): Ditto.
2267         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2268         (_mm512_prefetch_i64scatter_pd): Ditto.
2269         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2270         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2271         (_mm512_mask_prefetch_i64gather_ps): Ditto.
2272         (_mm512_prefetch_i32scatter_ps): Ditto.
2273         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2274         (_mm512_prefetch_i64scatter_ps): Ditto.
2275         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2276         * config/i386/i386-builtin-types.def: Define
2277         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2278         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2279         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2280         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2281         IX86_BUILTIN_SCATTERPFQPD.
2282         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2283         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2284         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2285         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2286         __builtin_ia32_scatterpfqps.
2287         (ix86_expand_builtin): Expand new built-ins.
2288         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2289         fix memory access data type.
2290         (*avx512pf_gatherpf<mode>_mask): Ditto.
2291         (*avx512pf_gatherpf<mode>): Ditto.
2292         (avx512pf_scatterpf<mode>): Ditto.
2293         (*avx512pf_scatterpf<mode>_mask): Ditto.
2294         (*avx512pf_scatterpf<mode>): Ditto.
2295         (GATHER_SCATTER_SF_MEM_MODE): New.
2296         (avx512pf_gatherpf<mode>df): Ditto.
2297         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2298         (*avx512pf_scatterpf<mode>df): Ditto.
2300 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
2302         PR bootstrap/59934
2303         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2304         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2305         reached.
2307 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2309         * common/config/arm/arm-common.c
2310         (arm_rewrite_mcpu): Handle multiple names.
2311         * config/arm/arm.h
2312         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2314 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2316         * gimple-builder.h (create_gimple_tmp): Delete.
2318 2014-01-27  Christian Bruel  <christian.bruel@st.com>
2320         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2321         words comparisons.
2323 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
2325         * config/pa/pa.md (call): Generate indirect long calls to non-local
2326         functions when outputing 32-bit code.
2327         (call_value): Likewise except for special call to buggy powf function.
2329         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2330         portable runtime and PIC indirect calls.
2331         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2332         and PIC call sequences.  Use ldo instead of blr to set return register
2333         in PIC call sequence.
2335 2014-01-25  Walter Lee  <walt@tilera.com>
2337         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2338         avoid clobbering a live register.
2340 2014-01-25  Walter Lee  <walt@tilera.com>
2342         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2343         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2344         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2345         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2347 2014-01-25  Walter Lee  <walt@tilera.com>
2349         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2350         arguments on even registers.
2351         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2352         STACK_BOUNDARY.
2353         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2354         (BIGGEST_ALIGNMENT): Ditto.
2355         (BIGGEST_FIELD_ALIGNMENT): Ditto.
2357 2014-01-25  Walter Lee  <walt@tilera.com>
2359         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2360         insns before bundling.
2361         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2363 2014-01-25  Walter Lee  <walt@tilera.com>
2365         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2366         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2367         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2369 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2371         * config/mips/constraints.md (kl): Delete.
2372         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2373         define expands, using...
2374         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2375         instructions for MIPS16.
2376         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2377         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2379 2014-01-25  Walter Lee  <walt@tilera.com>
2381         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2382         (clzdi2): Ditto.
2383         (ffsdi2): Ditto.
2385 2014-01-25  Walter Lee  <walt@tilera.com>
2387         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2388         (TARGET_EXPAND_TO_RTL_HOOK): Define.
2390 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2392         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2393         Handle XOR.
2395 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
2397         * print-rtl.c (in_call_function_usage): New var.
2398         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2399         EXPR_LIST mode as mode and not as reg note name.
2401         PR middle-end/59561
2402         * cfgloopmanip.c (copy_loop_info): If
2403         loop->warned_aggressive_loop_optimizations, make sure
2404         the flag is set in target loop too.
2406 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
2408         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2409         flag_cilkplus.
2410         * builtins.def: Likewise.
2411         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2412         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2413         * ira.c (ira_setup_eliminable_regset): Likewise.
2414         * omp-low.c (gate_expand_omp): Likewise.
2415         (execute_lower_omp): Likewise.
2416         (diagnose_sb_0): Likewise.
2417         (gate_diagnose_omp_blocks): Likewise.
2418         (simd_clone_clauses_extract): Likewise.
2419         (gate): Likewise.
2421 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2423         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2424         correction for little endian...
2425         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2426         here.
2428 2014-01-24  Jeff Law  <law@redhat.com>
2430         PR tree-optimization/59919
2431         * tree-vrp.c (find_assert_locations_1): Do not register asserts
2432         for non-returning calls.
2434 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
2436         * common/config/aarch64/aarch64-common.c
2437         (aarch64_rewrite_mcpu): Handle multiple names.
2438         * config/aarch64/aarch64.h
2439         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2441 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
2443         * input.c (add_file_to_cache_tab): Handle the case where fopen
2444         returns NULL.
2446 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
2448         PR target/59929
2449         * config/i386/i386.md (pushsf splitter): Get stack adjustment
2450         from push operand if code of push isn't PRE_DEC.
2452 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
2454         PR target/59909
2455         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2456         -mquad-memory-atomic.  Update -mquad-memory documentation to say
2457         it is only used for non-atomic loads/stores.
2459         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2460         -mquad-memory or -mquad-memory-atomic switches.
2462         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2463         -mquad-memory-atomic to ISA 2.07 support.
2465         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2466         to separate support of normal quad word memory operations (ldq, stq)
2467         from the atomic quad word memory operations.
2469         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2470         support to separate non-atomic quad word operations from atomic
2471         quad word operations.  Disable non-atomic quad word operations in
2472         little endian mode so that we don't have to swap words after the
2473         load and before the store.
2474         (quad_load_store_p): Add comment about atomic quad word support.
2475         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2476         options printed with -mdebug=reg.
2478         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2479         -mquad-memory-atomic as the test for whether we have quad word
2480         atomic instructions.
2481         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2482         or -mp8-vector are used, allow byte/half-word atomic operations.
2484         * config/rs6000/sync.md (load_lockedti): Insure that the address
2485         is a proper indexed or indirect address for the lqarx instruction.
2486         On little endian systems, swap the hi/lo registers after the lqarx
2487         instruction.
2488         (load_lockedpti): Use indexed_or_indirect_operand predicate to
2489         insure the address is valid for the lqarx instruction.
2490         (store_conditionalti): Insure that the address is a proper indexed
2491         or indirect address for the stqcrx. instruction.  On little endian
2492         systems, swap the hi/lo registers before doing the stqcrx.
2493         instruction.
2494         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2495         insure the address is valid for the stqcrx. instruction.
2497         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2498         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2499         type of quad memory support is available.
2501 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
2503         PR regression/59915
2504         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2505         there is a danger of looping.
2507 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
2509         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2510         force flag_ira_loop_pressure if set via command line.
2512 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2514         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2515         (ashr_simd): New builtin handling DI mode.
2516         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2517         (aarch64_sshr_simddi): New match pattern.
2518         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2519         (vshrd_n_s64): Likewise.
2520         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2522 2014-01-23  Nick Clifton  <nickc@redhat.com>
2524         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2525         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2526         favour of mcu specific scripts.
2527         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2528         430x multilibs.
2530 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2531             Alex Velenko  <Alex.Velenko@arm.com>
2533         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2534         (vaddv_s16): Likewise.
2535         (vaddv_s32): Likewise.
2536         (vaddv_u8): Likewise.
2537         (vaddv_u16): Likewise.
2538         (vaddv_u32): Likewise.
2539         (vaddvq_s8): Likewise.
2540         (vaddvq_s16): Likewise.
2541         (vaddvq_s32): Likewise.
2542         (vaddvq_s64): Likewise.
2543         (vaddvq_u8): Likewise.
2544         (vaddvq_u16): Likewise.
2545         (vaddvq_u32): Likewise.
2546         (vaddvq_u64): Likewise.
2547         (vaddv_f32): Likewise.
2548         (vaddvq_f32): Likewise.
2549         (vaddvq_f64): Likewise.
2550         (vmaxv_f32): Likewise.
2551         (vmaxv_s8): Likewise.
2552         (vmaxv_s16): Likewise.
2553         (vmaxv_s32): Likewise.
2554         (vmaxv_u8): Likewise.
2555         (vmaxv_u16): Likewise.
2556         (vmaxv_u32): Likewise.
2557         (vmaxvq_f32): Likewise.
2558         (vmaxvq_f64): Likewise.
2559         (vmaxvq_s8): Likewise.
2560         (vmaxvq_s16): Likewise.
2561         (vmaxvq_s32): Likewise.
2562         (vmaxvq_u8): Likewise.
2563         (vmaxvq_u16): Likewise.
2564         (vmaxvq_u32): Likewise.
2565         (vmaxnmv_f32): Likewise.
2566         (vmaxnmvq_f32): Likewise.
2567         (vmaxnmvq_f64): Likewise.
2568         (vminv_f32): Likewise.
2569         (vminv_s8): Likewise.
2570         (vminv_s16): Likewise.
2571         (vminv_s32): Likewise.
2572         (vminv_u8): Likewise.
2573         (vminv_u16): Likewise.
2574         (vminv_u32): Likewise.
2575         (vminvq_f32): Likewise.
2576         (vminvq_f64): Likewise.
2577         (vminvq_s8): Likewise.
2578         (vminvq_s16): Likewise.
2579         (vminvq_s32): Likewise.
2580         (vminvq_u8): Likewise.
2581         (vminvq_u16): Likewise.
2582         (vminvq_u32): Likewise.
2583         (vminnmv_f32): Likewise.
2584         (vminnmvq_f32): Likewise.
2585         (vminnmvq_f64): Likewise.
2587 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2589         * config/aarch64/aarch64-simd.md
2590         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2591         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2592         (*aarch64_mul3_elt<mode>): Likewise.
2593         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2594         (*aarch64_mul3_elt_to_64v2df): Likewise.
2595         (*aarch64_mla_elt<mode>): Likewise.
2596         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2597         (*aarch64_mls_elt<mode>): Likewise.
2598         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2599         (*aarch64_fma4_elt<mode>): Likewise.
2600         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2601         (*aarch64_fma4_elt_to_64v2df): Likewise.
2602         (*aarch64_fnma4_elt<mode>): Likewise.
2603         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2604         (*aarch64_fnma4_elt_to_64v2df): Likewise.
2605         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2606         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2607         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2608         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2609         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2610         (aarch64_sqdmull_lane<mode>_internal): Likewise.
2611         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2613 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2615         * config/aarch64/aarch64-simd.md
2616         (aarch64_be_checked_get_lane<mode>): New define_expand.
2617         * config/aarch64/aarch64-simd-builtins.def
2618         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2619         New builtin definition.
2620         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2621         Use new safe be builtin.
2623 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2625         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2626         New define_insn.
2627         (aarch64_be_st1<mode>): Likewise.
2628         (aarch_ld1<VALL:mode>): Define_expand modified.
2629         (aarch_st1<VALL:mode>): Likewise.
2630         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2631         (UNSPEC_ST1): Likewise.
2633 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
2635         * config/microblaze/microblaze.md: Add trap insn and attribute
2637 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
2639         PR preprocessor/58580
2640         * input.h (location_get_source_line): Take an additional line_size
2641         parameter.
2642         (void diagnostics_file_cache_fini): Declare new function.
2643         * input.c (struct fcache): New type.
2644         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2645         New static constants.
2646         (diagnostic_file_cache_init, total_lines_num)
2647         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2648         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2649         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2650         (get_next_line, read_next_line, goto_next_line, read_line_num):
2651         New static function definitions.
2652         (diagnostic_file_cache_fini): New function.
2653         (location_get_source_line): Take an additional output line_len
2654         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
2655         read_line_num.
2656         * diagnostic.c (diagnostic_finish): Call
2657         diagnostic_file_cache_fini.
2658         (adjust_line): Take an additional input parameter for the length
2659         of the line, rather than calculating it with strlen.
2660         (diagnostic_show_locus): Adjust the use of
2661         location_get_source_line and adjust_line with respect to their new
2662         signature.  While displaying a line now, do not stop at the first
2663         null byte.  Rather, display the zero byte as a space and keep
2664         going until we reach the size of the line.
2665         * Makefile.in: Add vec.o to OBJS-libcommon
2667 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2668             Ilya Tocar     <ilya.tocar@intel.com>
2670         * config/i386/avx512fintrin.h (_mm512_kmov): New.
2671         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2672         (__builtin_ia32_kmov16): Ditto.
2673         * config/i386/i386.md (UNSPEC_KMOV): New.
2674         (kmovw): Ditto.
2676 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2678         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2679         (_mm512_storeu_si512): Ditto.
2681 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
2683         PR target/52125
2684         * rtl.h (get_referenced_operands): Declare.
2685         * recog.c (get_referenced_operands): New function.
2686         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2687         operands have been referenced when recording LO_SUM references.
2689 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
2691         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2693 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2695         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2696         Enable for generic and recent AMD targets.
2698 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2700         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2701         ARG_SIZE note when adjustment was eliminated.
2703 2014-01-22  Jeff Law  <law@redhat.com>
2705         PR tree-optimization/59597
2706         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2707         in file.  Accept new argument REGISTERING and use it to modify
2708         dump output appropriately.
2709         (register_jump_thread): Corresponding changes.
2710         (mark_threaded_blocks): Reinstate code to cancel unprofitable
2711         thread paths involving joiner blocks.  Add code to dump cancelled
2712         jump threading paths.
2714 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
2716         PR rtl-optimization/59477
2717         * lra-constraints.c (inherit_in_ebb): Process call for living hard
2718         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
2720 2014-01-22  Tom Tromey  <tromey@redhat.com>
2722         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2723         PARAMS.
2724         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2726 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2728         PR rtl-optimization/59896
2729         * lra-constraints.c (process_alt_operands): Check unused note for
2730         matched operands of insn with no output reloads.
2732 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
2734         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2735         (mips_move_from_gpr_cost): Likewise.
2737 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2739         PR rtl-optimization/59858
2740         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2741         ira_class_hard_regs_num.
2742         (process_alt_operands): Increase reject for dying matched operand.
2744 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
2746         PR target/59003
2747         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2748         smaller than size, perform several stores or loads and stores
2749         at dst + count - size to store or copy all of size bytes, rather
2750         than just last modesize bytes.
2752 2014-01-20  DJ Delorie  <dj@redhat.com>
2754         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2755         that CLOBBERs are REGs before propogating their values.
2757 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2759         PR middle-end/59789
2760         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2761         (cgraph_inline_failed_type): New function.
2762         * cgraph.h (DEFCIFCODE): Add type.
2763         (cgraph_inline_failed_type_t): New enum.
2764         (cgraph_inline_failed_type): New prototype.
2765         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2766         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2767         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2768         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2769         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2770         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2771         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2772         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2773         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2774         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2775         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2776         OPTIMIZATION_MISMATCH.
2777         * tree-inline.c (expand_call_inline): Emit errors during
2778         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2780 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2782         PR target/59685
2783         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2784         mode attribute in insn output.
2786 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
2788         * output.h (output_constant): Delete.
2789         * varasm.c (output_constant): Make private.
2791 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
2793         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2795 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2797         PR middle-end/59860
2798         * tree.h (fold_builtin_strcat): New prototype.
2799         * builtins.c (fold_builtin_strcat): No longer static.  Add len
2800         argument, if non-NULL, don't call c_strlen.  Optimize
2801         directly into __builtin_memcpy instead of __builtin_strcpy.
2802         (fold_builtin_2): Adjust fold_builtin_strcat caller.
2803         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2805 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2807         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2808         for SImode_address_operand operands, having only a REG argument.
2810 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2812         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2813         loader name using mbig-endian.
2814         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2816 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2818         * doc/invoke.texi (-march): Clarify documentation for AArch64.
2819         (-mtune): Likewise.
2820         (-mcpu): Likewise.
2822 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
2824         * config/aarch64/aarch64-protos.h
2825         (aarch64_cannot_change_mode_class_ptr): Declare.
2826         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2827         aarch64_cannot_change_mode_class_ptr): New.
2828         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2829         backend hook aarch64_cannot_change_mode_class.
2831 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2833         * common/config/aarch64/aarch64-common.c
2834         (aarch64_handle_option): Don't handle any option order logic here.
2835         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2836         selected_cpu, warn on architecture version mismatch.
2837         (aarch64_override_options): Fix parsing order for option strings.
2839 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2840             Iain Sandoe  <iain@codesourcery.com>
2842         PR bootstrap/59496
2843         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2844         warning.  Amend comment to reflect current functionality.
2846 2014-01-20  Richard Biener  <rguenther@suse.de>
2848         PR middle-end/59860
2849         * builtins.c (fold_builtin_strcat): Remove case better handled
2850         by tree-ssa-strlen.c.
2852 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2854         * config/aarch64/aarch64.opt
2855         (mcpu, march, mtune): Make case-insensitive.
2857 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2859         PR target/59880
2860         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2861         if operands[1] is a REG or ZERO_EXTEND of a REG.
2863 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2865         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2867 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2869         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2870         long non-pic millicode calls.
2872 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2874         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2876 2014-01-19  Kito Cheng  <kito@0xlab.org>
2878         * builtins.c (expand_movstr): Check movstr expand done or fail.
2880 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2881             H.J. Lu  <hongjiu.lu@intel.com>
2883         PR target/59379
2884         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2885         to DImode for zero-extended addresses.
2887 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2889         PR rtl-optimization/57763
2890         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2891         on the new indirect jump_insn and increment LABEL_NUSES (label).
2893 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2895         PR bootstrap/59580
2896         PR bootstrap/59583
2897         * config.gcc (x86_archs): New variable.
2898         (x86_64_archs): Likewise.
2899         (x86_cpus): Likewise.
2900         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2901         --with-arch/--with-cpu= options.
2902         Support --with-arch=/--with-cpu={nehalem,westmere,
2903         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2905 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2907         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2908         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2910 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2912         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2914 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2916         PR target/58944
2917         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2918         clear cpp_get_options (parse_in)->warn_unused_macros for
2919         ix86_target_macros_internal with cpp_define.
2921 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2923         * jump.c (delete_related_insns): Keep (use (insn))s.
2924         * reorg.c (redundant_insn): Check for barriers too.
2926 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2928         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2930 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2932         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2933         call to $$dyncall when TARGET_LONG_CALLS is true.
2935 2014-01-17  Jeff Law  <law@redhat.com>
2937         * ree.c (combine_set_extension): Temporarily disable test for
2938         changing number of hard registers.
2940 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2942         PR middle-end/58125
2943         * ipa-inline-analysis.c (inline_free_summary):
2944         Do not free summary of aliases.
2946 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2948         PR middle-end/59706
2949         * gimplify.c (gimplify_expr): Use create_tmp_var
2950         instead of create_tmp_var_raw.  If cond doesn't have
2951         integral type, don't add the IFN_ANNOTATE builtin at all.
2953 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2955         PR ipa/59736
2956         * ipa-cp.c (prev_edge_clone): New variable.
2957         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2958         Also resize prev_edge_clone vector.
2959         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2960         (ipcp_edge_removal_hook): New function.
2961         (ipcp_driver): Register ipcp_edge_removal_hook.
2963 2014-01-17  Andrew Pinski <apinski@cavium.com>
2964             Steve Ellcey  <sellcey@mips.com>
2966         PR target/59462
2967         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2968         of operator mode.
2970 2014-01-17  Jeff Law  <law@redhat.com>
2972         PR middle-end/57904
2973         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2974         so that pass_ccp runs first.
2976 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2978         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2979         (ix86_adjust_cost): Use !TARGET_XXX.
2980         (do_reorder_for_imul): Likewise.
2981         (swap_top_of_ready_list): Likewise.
2982         (ix86_sched_reorder): Likewise.
2984 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2986         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2987         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2988         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2989         (intel_memset): New.  Duplicate slm_memset.
2990         (intel_cost): New.  Duplicate slm_cost.
2991         (m_INTEL): New macro.
2992         (processor_target_table): Add "intel".
2993         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2994         with PROCESSOR_INTEL for "intel".
2995         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2996         PROCESSOR_SILVERMONT.
2997         (ix86_issue_rate): Likewise.
2998         (ix86_adjust_cost): Likewise.
2999         (ia32_multipass_dfa_lookahead): Likewise.
3000         (swap_top_of_ready_list): Likewise.
3001         (ix86_sched_reorder): Likewise.
3002         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
3003         instead of TARGET_OPT_AGU.
3004         * config/i386/i386.h (TARGET_INTEL): New.
3005         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
3006         (processor_type): Add PROCESSOR_INTEL.
3007         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
3008         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
3010 2014-01-17  Marek Polacek  <polacek@redhat.com>
3012         PR c/58346
3013         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
3014         size is zero.
3016 2014-01-17  Richard Biener  <rguenther@suse.de>
3018         PR tree-optimization/46590
3019         * opts.c (default_options_table): Add entries for
3020         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
3021         all enabled at -O1 but not for -Og.
3022         * common.opt (fbranch-count-reg): Remove Init(1).
3023         (fmove-loop-invariants): Likewise.
3024         (ftree-pta): Likewise.
3026 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3028         * config/i386/i386.c (ix86_data_alignment): For compatibility with
3029         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
3030         decls to at least the GCC 4.8 used alignments.
3032         PR fortran/59440
3033         * tree-nested.c (convert_nonlocal_reference_stmt,
3034         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
3035         of GIMPLE_BIND stmts, adjust associated decls.
3037 2014-01-17  Richard Biener  <rguenther@suse.de>
3039         PR tree-optimization/46590
3040         * vec.h (vec<>::bseach): New member function implementing
3041         binary search according to C89 bsearch.
3042         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
3043         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
3044         bitmap pointer again.  Make accesses_in_loop a flat array.
3045         (mem_ref_obstack): New global.
3046         (outermost_indep_loop): Adjust for mem_ref->stored changes.
3047         (mark_ref_stored): Likewise.
3048         (ref_indep_loop_p_2): Likewise.
3049         (set_ref_stored_in_loop): New helper function.
3050         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
3051         (memref_free): Adjust.
3052         (record_mem_ref_loc): Simplify.
3053         (gather_mem_refs_stmt): Adjust.
3054         (sort_locs_in_loop_postorder_cmp): New function.
3055         (analyze_memory_references): Sort accesses_in_loop after
3056         loop postorder number.
3057         (find_ref_loc_in_loop_cmp): New function.
3058         (for_all_locs_in_loop): Find relevant cluster of locs in
3059         accesses_in_loop and iterate without recursion.
3060         (execute_sm): Avoid uninit warning.
3061         (struct ref_always_accessed): Simplify.
3062         (ref_always_accessed::operator ()): Likewise.
3063         (ref_always_accessed_p): Likewise.
3064         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
3065         loop postorder numbers here.
3066         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
3067         numbers.
3069 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3071         PR c++/57945
3072         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
3073         on decls for which assemble_alias has been called.
3075 2014-01-17  Nick Clifton  <nickc@redhat.com>
3077         * config/msp430/msp430.opt: (mcpu): New option.
3078         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
3079         (msp430_option_override): Parse target_cpu.  If the MCU name
3080         matches a generic string, clear target_mcu.
3081         (msp430_attr): Allow numeric interrupt values up to 63.
3082         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
3083         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
3084         option.
3085         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
3086         Add mcpu matches.
3087         * config/msp430/msp430.md (popm): Use %J rather than %I.
3088         (addsi3): Use msp430_nonimmediate_operand for operand 2.
3089         (addhi_cy_i): Use immediate_operand for operand 2.
3090         * doc/invoke.texi: Document -mcpu option.
3092 2014-01-17  Richard Biener  <rguenther@suse.de>
3094         PR rtl-optimization/38518
3095         * df.h (df_analyze_loop): Declare.
3096         * df-core.c: Include cfgloop.h.
3097         (df_analyze_1): Split out main part of df_analyze.
3098         (df_analyze): Adjust.
3099         (loop_inverted_post_order_compute): New function.
3100         (loop_post_order_compute): Likewise.
3101         (df_analyze_loop): New function avoiding whole-function
3102         postorder computes.
3103         * loop-invariant.c (find_defs): Use df_analyze_loop.
3104         (find_invariants): Adjust.
3105         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
3107 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
3109         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
3110         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
3112 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
3114         * ipa-ref.c (ipa_remove_stmt_references): Fix references
3115         traversal when removing references.
3117 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
3119         PR ipa/59775
3120         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
3122 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
3124         PR middle-end/56791
3125         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
3126         pushing a reload for an autoinc when we had previously reloaded an
3127         inner part of the address.
3129 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
3131         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
3132         field.
3133         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
3134         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
3135         when not giving up or versioning for alias only because of
3136         loop->safelen.
3137         (vect_analyze_data_ref_dependences): Set to true.
3138         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
3139         is a GIMPLE_PHI.
3140         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
3141         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
3142         to the condition.
3144         PR middle-end/58344
3145         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
3147         PR target/59839
3148         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
3149         operand 0 predicate for gathers, use a new pseudo as subtarget.
3151 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
3153         PR middle-end/59609
3154         * lra-constraints.c (process_alt_operands): Add printing debug info.
3155         Check absence of input/output reloads for matched operands too.
3157 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
3159         PR rtl-optimization/59835
3160         * ira.c (ira_init_register_move_cost): Increase cost for
3161         impossible modes.
3163 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
3165         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
3167 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
3169         PR target/59780
3170         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
3171         non-register objects.  Use gen_(high/low)part more consistently.
3172         Fix assertions.
3174 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
3176         PR target/59844
3177         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
3178         endian support, remove tests for WORDS_BIG_ENDIAN.
3179         (p8_mfvsrd_3_<mode>): Likewise.
3180         (reload_gpr_from_vsx<mode>): Likewise.
3181         (reload_gpr_from_vsxsf): Likewise.
3182         (p8_mfvsrd_4_disf): Likewise.
3184 2014-01-16  Richard Biener  <rguenther@suse.de>
3186         PR rtl-optimization/46590
3187         * lcm.c (compute_antinout_edge): Use postorder iteration.
3188         (compute_laterin): Use inverted postorder iteration.
3190 2014-01-16  Nick Clifton  <nickc@redhat.com>
3192         PR middle-end/28865
3193         * varasm.c (output_constant): Return the number of bytes actually
3194         emitted.
3195         (output_constructor_array_range): Update the field size with the
3196         number of bytes emitted by output_constant.
3197         (output_constructor_regular_field): Likewise.  Also do not
3198         complain if the total number of bytes emitted is now greater
3199         than the expected fieldpos.
3200         * output.h (output_constant): Update prototype and descriptive comment.
3202 2014-01-16  Marek Polacek  <polacek@redhat.com>
3204         PR middle-end/59827
3205         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3206         it is error_mark_node.
3208 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
3210         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3211         VALID_AVX256_REG_OR_OI_MODE.
3213 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
3215         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
3216         current procedure should be profiled.
3218 2014-01-15  Andrew Pinski  <apinski@cavium.com>
3220         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
3221         of moving from/to the STACK_REG register class.
3223 2014-01-15  Richard Henderson  <rth@redhat.com>
3225         PR debug/54694
3226         * reginfo.c (global_regs_decl): Globalize.
3227         * rtl.h (global_regs_decl): Declare.
3228         * ira.c (do_reload): Diagnose frame_pointer_needed and it
3229         reserved via global_regs.
3231 2014-01-15  Teresa Johnson  <tejohnson@google.com>
3233         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
3235 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
3237         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
3238         and vmulosh rather than call gen_vec_widen_smult_*.
3239         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
3240         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
3241         (vec_widen_smult_even_v16qi): Likewise.
3242         (vec_widen_umult_even_v8hi): Likewise.
3243         (vec_widen_smult_even_v8hi): Likewise.
3244         (vec_widen_umult_odd_v16qi): Likewise.
3245         (vec_widen_smult_odd_v16qi): Likewise.
3246         (vec_widen_umult_odd_v8hi): Likewise.
3247         (vec_widen_smult_odd_v8hi): Likewise.
3248         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
3249         vmuloub rather than call gen_vec_widen_umult_*.
3250         (vec_widen_umult_lo_v16qi): Likewise.
3251         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
3252         vmulosb rather than call gen_vec_widen_smult_*.
3253         (vec_widen_smult_lo_v16qi): Likewise.
3254         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
3255         rather than call gen_vec_widen_umult_*.
3256         (vec_widen_umult_lo_v8hi): Likewise.
3257         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
3258         rather than call gen_vec_widen_smult_*.
3259         (vec_widen_smult_lo_v8hi): Likewise.
3261 2014-01-15  Jeff Law  <law@redhat.com>
3263         PR tree-optimization/59747
3264         * ree.c (find_and_remove_re): Properly handle case where a second
3265         eliminated extension requires widening a copy created for elimination
3266         of a prior extension.
3267         (combine_set_extension): Ensure that the number of hard regs needed
3268         for a destination register does not change when we widen it.
3270 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3272         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3273         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3274         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3275         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3276         (avr-*-rtems*): Likewise.
3277         (bfin*-rtems*): Likewise.
3278         (moxie-*-rtems*): Likewise.
3279         (h8300-*-rtems*): Likewise.
3280         (i[34567]86-*-rtems*): Likewise.
3281         (lm32-*-rtems*): Likewise.
3282         (m32r-*-rtems*): Likewise.
3283         (m68k-*-rtems*): Likewise.
3284         (microblaze*-*-rtems*): Likewise.
3285         (mips*-*-rtems*): Likewise.
3286         (powerpc-*-rtems*): Likewise.
3287         (sh-*-rtems*): Likewise.
3288         (sparc-*-rtems*): Likewise.
3289         (sparc64-*-rtems*): Likewise.
3290         (v850-*-rtems*): Likewise.
3291         (m32c-*-rtems*): Likewise.
3293 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
3295         PR rtl-optimization/59511
3296         * ira.c (ira_init_register_move_cost): Use memory costs for some
3297         cases of register move cost calculations.
3298         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3299         instead of BB frequency.
3300         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3301         * lra-assigns.c (find_hard_regno_for): Ditto.
3303 2014-01-15  Richard Biener  <rguenther@suse.de>
3305         PR tree-optimization/59822
3306         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3307         (vectorizable_load): Use it to hoist defs of uses of invariant
3308         loads out of the loop.
3310 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
3311             Kugan Vivekanandarajah  <kuganv@linaro.org>
3313         PR target/59695
3314         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3315         truncation.
3317 2014-01-15  Richard Biener  <rguenther@suse.de>
3319         PR rtl-optimization/59802
3320         * lcm.c (compute_available): Use inverted postorder to seed
3321         the initial worklist.
3323 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3325         PR target/59803
3326         * config/s390/s390.c (s390_preferred_reload_class): Don't return
3327         ADDR_REGS for invalid symrefs in non-PIC code.
3329 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
3331         PR other/58712
3332         * builtins.c (determine_block_size): Initialize *probable_max_size
3333         even if len_rtx is CONST_INT.
3335 2014-01-14  Andrew Pinski  <apinski@cavium.com>
3337         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3338         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3339         (cortexa53_tunings): Likewise.
3340         (aarch64_sched_issue_rate): New function.
3341         (TARGET_SCHED_ISSUE_RATE): Define.
3343 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3345         * ira-costs.c (find_costs_and_classes): Add missed
3346         ira_init_register_move_cost_if_necessary.
3348 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3350         PR target/59787
3351         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3353 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
3355         PR target/59794
3356         * config/i386/i386.c (type_natural_mode): Add a bool parameter
3357         to indicate if type is used for function return value.  Warn ABI
3358         change if the vector mode isn't available for function return value.
3359         (ix86_function_arg_advance): Pass false to type_natural_mode.
3360         (ix86_function_arg): Likewise.
3361         (ix86_gimplify_va_arg): Likewise.
3362         (function_arg_32): Don't warn ABI change.
3363         (ix86_function_value): Pass true to type_natural_mode.
3364         (ix86_return_in_memory): Likewise.
3365         (ix86_struct_value_rtx): Removed.
3366         (TARGET_STRUCT_VALUE_RTX): Likewise.
3368 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3370         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3371         converting a conditional jump into a conditional return.
3373 2014-01-14  Richard Biener  <rguenther@suse.de>
3375         PR tree-optimization/58921
3376         PR tree-optimization/59006
3377         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3378         hoisting invariant stmts.
3379         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3380         invariant loads on the preheader edge if possible.
3382 2014-01-14  Joey Ye  <joey.ye@arm.com>
3384         * doc/plugin.texi (Building GCC plugins): Update to C++.
3386 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
3388         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3389         (_mm_rcp28_round_ss): Ditto.
3390         (_mm_rsqrt28_round_sd): Ditto.
3391         (_mm_rsqrt28_round_ss): Ditto.
3392         (_mm_rcp28_sd): Ditto.
3393         (_mm_rcp28_ss): Ditto.
3394         (_mm_rsqrt28_sd): Ditto.
3395         (_mm_rsqrt28_ss): Ditto.
3396         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3397         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3398         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3399         (IX86_BUILTIN_RCP28SD): Ditto.
3400         (IX86_BUILTIN_RCP28SS): Ditto.
3401         (IX86_BUILTIN_RSQRT28SD): Ditto.
3402         (IX86_BUILTIN_RSQRT28SS): Ditto.
3403         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3404         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3405         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3406         (ix86_expand_special_args_builtin): Expand new FTYPE.
3407         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3408         (srcp14<mode>): Make insn unary.
3409         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3410         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3411         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3412         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3413         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3414         Fix rounding: make it SAE only.
3415         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3416         Ditto.
3417         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3418         Ditto.
3419         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3420         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3421         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3422         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3423         (round_saeonly_mask_scalar_operand4): Ditto.
3424         (round_saeonly_mask_scalar_op3): Ditto.
3425         (round_saeonly_mask_scalar_op4): Ditto.
3427 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3429         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3430         Implement -maltivec=be for vec_insert and vec_extract.
3432 2014-01-10  DJ Delorie  <dj@redhat.com>
3434         * config/msp430/msp430.md (call_internal): Don't allow memory
3435         references with SP as the base register.
3436         (call_value_internal): Likewise.
3437         * config/msp430/constraints.md (Yc): New.  For memory references
3438         that don't use SP as a base register.
3440         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3441         "an integer without a # prefix"
3442         * config/msp430/msp430.md (epilogue_helper): Use it.
3444 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3446         PR target/59617
3447         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3448         AVX512F gather builtins.
3449         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3450         on gather decls with INTEGER_TYPE masktype.
3451         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3452         directly into the builtin rather than hoisting it before loop.
3454         PR tree-optimization/59387
3455         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3456         (scev_const_prop): If folded_casts and type has undefined overflow,
3457         use force_gimple_operand instead of force_gimple_operand_gsi and
3458         for each added stmt if it is assign with
3459         arith_code_with_undefined_signed_overflow, call
3460         rewrite_to_defined_overflow.
3461         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3462         gimple-fold.h instead.
3463         (arith_code_with_undefined_signed_overflow,
3464         rewrite_to_defined_overflow): Moved to ...
3465         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3466         rewrite_to_defined_overflow): ... here.  No longer static.
3467         Include gimplify-me.h.
3468         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3469         rewrite_to_defined_overflow): New prototypes.
3471 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3473         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3475 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
3477         * builtins.c (get_object_alignment_2): Minor tweak.
3478         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3480 2014-01-13  Christian Bruel  <christian.bruel@st.com>
3482         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3483         optimized non constant lengths.
3485 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3487         PR libgomp/59194
3488         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3489         load as __atomic_load_N if possible.
3491 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3493         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3494         target parameter.
3495         (rs6000_expand_builtin): Adjust call.
3497 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3499         PR target/58115
3500         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3501         * config/rs6000/rs6000.c: Include target-globals.h.
3502         (rs6000_set_current_function): Instead of doing target_reinit
3503         unconditionally, use save_target_globals_default_opts and
3504         restore_target_globals.
3506         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3507         FPSCR.
3508         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3509         (rs6000_expand_builtin): Handle mffs and mtfsf.
3510         (rs6000_init_builtins): Define mffs and mtfsf.
3511         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3512         (rs6000_mffs): New pattern.
3513         (rs6000_mtfsf): New pattern.
3515 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
3517         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3518         Start narrowing with START.  Apply candidate-use pair
3519         and check overall cost in narrowing.
3520         (iv_ca_prune): Pass new argument.
3522 2014-01-10  Jeff Law  <law@redhat.com>
3524         PR middle-end/59743
3525         * ree.c (combine_reaching_defs): Ensure the defining statement
3526         occurs before the extension when optimizing extensions with
3527         different source and destination hard registers.
3529 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3531         PR ipa/58585
3532         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3533         vtables into the type inheritance graph.
3535 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3537         PR rtl-optimization/59754
3538         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3539         modes in the REGNO != REGNO case.
3541 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3543         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3545 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3547         PR tree-optimization/59745
3548         * tree-predcom.c (tree_predictive_commoning_loop): Call
3549         free_affine_expand_cache if giving up because components is NULL.
3551         * target-globals.c (save_target_globals): Allocate < 4KB structs using
3552         GC in payload of target_globals struct instead of allocating them on
3553         the heap and the larger structs separately using GC.
3554         * target-globals.h (struct target_globals): Make regs, hard_regs,
3555         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3556         of GTY((skip)) and change type to void *.
3557         (reset_target_globals): Cast loads from those fields to corresponding
3558         types.
3560 2014-01-10  Steve Ellcey  <sellcey@mips.com>
3562         PR plugins/59335
3563         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3564         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3565         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3567 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
3569         PR target/59744
3570         * aarch64-modes.def (CC_Zmode): New flags mode.
3571         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3572         represents an equality.
3573         (aarch64_get_condition_code): Handle CC_Zmode.
3574         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3576 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3578         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3579         extraction in good case.
3581 2014-01-10  Richard Biener  <rguenther@suse.de>
3583         PR tree-optimization/59374
3584         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3585         checking after SLP discovery.  Mark stmts not participating
3586         in any SLP instance properly.
3588 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3590         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3591         when handling a SET rtx.
3593 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3595         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3596         (cortex-a57): Likewise.
3597         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3599 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3601         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3602         non-iwmmxt builtins.
3604 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3606         PR ipa/58252
3607         PR ipa/59226
3608         * ipa-devirt.c record_target_from_binfo): Take as argument
3609         stack of binfos and lookup matching one for virtual inheritance.
3610         (possible_polymorphic_call_targets_1): Update.
3612 2014-01-10  Huacai Chen  <chenhc@lemote.com>
3614         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3615         kernel strings for Loongson-2E/2F/3A.
3617 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3619         PR middle-end/59670
3620         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3621         is_gimple_call before calling gimple_call_internal_p.
3623 2014-01-09  Steve Ellcey  <sellcey@mips.com>
3625         * Makefile.in (TREE_FLOW_H): Remove.
3626         (TREE_SSA_H): Add file names from tree-flow.h.
3627         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3628         * tree.h: Remove tree-flow.h reference.
3629         * hash-table.h: Remove tree-flow.h reference.
3630         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3631         reference with tree-ssa-loop.h.
3633 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3635         * doc/invoke.texi: Add -maltivec={be,le} options, and document
3636         default element-order behavior for -maltivec.
3637         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3638         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3639         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3640         when targeting big endian, at least for now.
3641         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3643 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3645         PR middle-end/47735
3646         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3647         var satisfies use_register_for_decl, just take into account type
3648         alignment, rather than decl alignment.
3650         PR tree-optimization/59622
3651         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
3652         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3653         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3654         Don't devirtualize for inplace at all.  For targets.length () == 1,
3655         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3657 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3659         * config/i386/i386.md (cpu): Remove the unused btver1.
3661 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3663         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3665 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3667         PR target/58115
3668         * tree-core.h (struct target_globals): New forward declaration.
3669         (struct tree_target_option): Add globals field.
3670         * tree.h (TREE_TARGET_GLOBALS): Define.
3671         (prepare_target_option_nodes_for_pch): New prototype.
3672         * target-globals.h (struct target_globals): Define even if
3673         !SWITCHABLE_TARGET.
3674         * tree.c (prepare_target_option_node_for_pch,
3675         prepare_target_option_nodes_for_pch): New functions.
3676         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3677         * config/i386/i386.c: Include target-globals.h.
3678         (ix86_set_current_function): Instead of doing target_reinit
3679         unconditionally, use save_target_globals_default_opts and
3680         restore_target_globals.
3682 2014-01-09  Richard Biener  <rguenther@suse.de>
3684         PR tree-optimization/59715
3685         * tree-cfg.h (split_critical_edges): Declare.
3686         * tree-cfg.c (split_critical_edges): Export.
3687         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3689 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
3691         * cfgexpand.c (expand_stack_vars): Optionally disable
3692         asan stack protection.
3693         (expand_used_vars): Likewise.
3694         (partition_stack_vars): Likewise.
3695         * asan.c (asan_emit_stack_protection): Optionally disable
3696         after return stack usage.
3697         (instrument_derefs): Optionally disable memory access instrumentation.
3698         (instrument_builtin_call): Likewise.
3699         (instrument_strlen_call): Likewise.
3700         (asan_protect_global): Optionally disable global variables protection.
3701         * doc/invoke.texi: Added doc for new options.
3702         * params.def: Added new options.
3703         * params.h: Likewise.
3705 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3707         PR rtl-optimization/59724
3708         * ifcvt.c (cond_exec_process_if_block): Don't call
3709         flow_find_head_matching_sequence with 0 longest_match.
3710         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3711         non-active insns if !stop_after.
3712         (try_head_merge_bb): Revert 2014-01-07 changes.
3714 2014-01-08  Jeff Law  <law@redhat.com>
3716         * ree.c (get_sub_rtx): New function, extracted from...
3717         (merge_def_and_ext): Here.
3718         (combine_reaching_defs): Use get_sub_rtx.
3720 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
3722         * cgraph.h (varpool_variable_node): Do not choke on null node.
3724 2014-01-08  Catherine Moore  <clm@codesourcery.com>
3726         * config/mips/mips.md (simple_return): Attempt to use JRC
3727         for microMIPS.
3728         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3730 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3732         PR rtl-optimization/59137
3733         * reorg.c (steal_delay_list_from_target): Call update_block for
3734         elided insns.
3735         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3737 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3739         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3740         two duplicate entries.
3742 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3744         Revert:
3745         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
3747         * config/mips/mips.c (mips_truncated_op_cost): New function.
3748         (mips_rtx_costs): Adjust test for BADDU.
3749         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3751         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
3753         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3754         (*baddu_si): ...this new pattern.
3756 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3758         PR ipa/59722
3759         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3761 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3763         PR middle-end/57748
3764         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3765         inner_reference_p.
3766         (expand_expr, expand_normal): Adjust.
3767         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3768         inner_reference_p. Use inner_reference_p to expand inner references.
3769         (store_expr): Adjust.
3770         * cfgexpand.c (expand_call_stmt): Adjust.
3772 2014-01-08  Rong Xu  <xur@google.com>
3774         * gcov-io.c (gcov_var): Move from gcov-io.h.
3775         (gcov_position): Ditto.
3776         (gcov_is_error): Ditto.
3777         (gcov_rewrite): Ditto.
3778         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3779         only part to libgcc/libgcov.h.
3781 2014-01-08  Marek Polacek  <polacek@redhat.com>
3783         PR middle-end/59669
3784         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3786 2014-01-08  Marek Polacek  <polacek@redhat.com>
3788         PR sanitizer/59667
3789         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3791 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3793         PR rtl-optimization/59649
3794         * stor-layout.c (get_mode_bounds): For BImode return
3795         0 and STORE_FLAG_VALUE.
3797 2014-01-08  Richard Biener  <rguenther@suse.de>
3799         PR middle-end/59630
3800         * gimple.h (is_gimple_builtin_call): Remove.
3801         (gimple_builtin_call_types_compatible_p): New.
3802         (gimple_call_builtin_p): New overload.
3803         * gimple.c (is_gimple_builtin_call): Remove.
3804         (validate_call): Rename to ...
3805         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
3806         check return types.
3807         (validate_type): New static function.
3808         (gimple_call_builtin_p): New overload and adjust.
3809         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3810         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
3811         (gimple_fold_stmt_to_constant_1): Likewise.
3812         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3814 2014-01-08  Richard Biener  <rguenther@suse.de>
3816         PR middle-end/59471
3817         * gimplify.c (gimplify_expr): Gimplify register-register type
3818         VIEW_CONVERT_EXPRs to separate stmts.
3820 2014-01-07  Jeff Law  <law@redhat.com>
3822         PR middle-end/53623
3823         * ree.c (combine_set_extension): Handle case where source
3824         and destination registers in an extension insn are different.
3825         (combine_reaching_defs): Allow source and destination registers
3826         in extension to be different under limited circumstances.
3827         (add_removable_extension): Remove restriction that the
3828         source and destination registers in the extension are the same.
3829         (find_and_remove_re): Emit a copy from the extension's
3830         destination to its source after the defining insn if
3831         the source and destination registers are different.
3833         PR middle-end/59285
3834         * ifcvt.c (merge_if_block): If we are merging a block with more than
3835         one successor with a block with no successors, remove any BARRIER
3836         after the second block.
3838 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3840         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3842 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3844         PR target/59652
3845         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3846         for 14-bit register offsets when INT14_OK_STRICT is false.
3848 2014-01-07  Roland Stigge  <stigge@antcom.de>
3849             Michael Meissner  <meissner@linux.vnet.ibm.com>
3851         PR 57386/target
3852         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3853         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3855 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3857         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3858         -mcpu.
3860 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3862         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3863         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3864         rtx is const0_rtx or not.
3866 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3868         PR target/58115
3869         * target-globals.c (save_target_globals): Remove this_fn_optab
3870         handling.
3871         * toplev.c: Include optabs.h.
3872         (target_reinit): Temporarily restore the global options if another
3873         set of options are in force.
3875 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3877         PR rtl-optimization/58668
3878         * cfgcleanup.c (flow_find_cross_jump): Don't count
3879         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3880         to determine what is counted.
3881         (flow_find_head_matching_sequence): Use active_insn_p to determine
3882         what is counted.
3883         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3884         counting change.
3885         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3886         determine what is counted.
3888         PR tree-optimization/59643
3889         * tree-predcom.c (split_data_refs_to_components): If one dr is
3890         read and one write, determine_offset fails and the write isn't
3891         in the bad component, just put the read into the bad component.
3893 2014-01-07  Mike Stump  <mikestump@comcast.net>
3894             Jakub Jelinek  <jakub@redhat.com>
3896         PR pch/59436
3897         * tree-core.h (struct tree_optimization_option): Change optabs
3898         type from unsigned char * to void *.
3899         * optabs.c (init_tree_optimization_optabs): Adjust
3900         TREE_OPTIMIZATION_OPTABS initialization.
3902 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3904         PR target/59644
3905         * config/i386/i386.h (struct machine_function): Add
3906         no_drap_save_restore field.
3907         * config/i386/i386.c (ix86_save_reg): Use
3908         !cfun->machine->no_drap_save_restore instead of
3909         crtl->stack_realign_needed.
3910         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3911         this function clears frame_pointer_needed.  Set
3912         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3913         and DRAP reg is needed.
3915 2014-01-06  Marek Polacek  <polacek@redhat.com>
3917         PR c/57773
3918         * doc/implement-c.texi: Mention that other integer types are
3919         permitted as bit-field types in strictly conforming mode.
3921 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3923         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3924         is newly allocated.
3926 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3928         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3930 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3932         PR ipa/59008
3933         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3934         to int.
3935         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3937 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3939         PR debug/59350
3940         PR debug/59510
3941         * var-tracking.c (add_stores): Preserve the value of the source even if
3942         we don't record the store.
3944 2014-01-06  Terry Guo  <terry.guo@arm.com>
3946         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3948 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3950         PR bootstrap/59541
3951         * config/darwin.c (darwin_function_section): Adjust return values to
3952         correspond to optimisation changes made in r206070.
3954 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3956         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3957         from prefetch_block tune setting.
3958         (nocona_cost): Correct size of prefetch block to 64.
3960 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3962         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3963         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3964         used to save the static chain register in the computation of the offset
3965         from which the FP registers need to be restored.
3967 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3969         PR tree-optimization/59519
3970         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3971         ICE if get_current_def (current_new_name) is already non-NULL, as long
3972         as it is a phi result of some other phi in *new_exit_bb that has
3973         the same argument.
3975         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3976         or vmovdqu* for misaligned_operand.
3977         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3978         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3979         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3980         aligned_mem for AVX512F masked aligned load and store builtins and for
3981         non-temporal moves.
3983 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3985         PR tree-optimization/59651
3986         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3987         Address range for negative step should be added by TYPE_SIZE_UNIT.
3989 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3991         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3992         overlapping registers also for registers other than the stack pointer.
3994 2014-01-03  Marek Polacek  <polacek@redhat.com>
3996         PR other/59661
3997         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3998         __builtin_FILE.
4000 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
4002         PR target/59625
4003         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4004         asm goto as jump.
4006         * config/i386/i386.md (MODE_SIZE): New mode attribute.
4007         (push splitter): Use <P:MODE_SIZE> instead of
4008         GET_MODE_SIZE (<P:MODE>mode).
4009         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
4010         (mov -1, reg peephole2): Likewise.
4011         * config/i386/sse.md (*mov<mode>_internal,
4012         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4013         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
4014         *<code><mode>3, *andnot<mode>3<mask_name>,
4015         <mask_codefor><code><mode>3<mask_name>): Likewise.
4016         * config/i386/subst.md (mask_mode512bit_condition,
4017         sd_mask_mode512bit_condition): Likewise.
4019 2014-01-02  Xinliang David Li  <davidxl@google.com>
4021         PR tree-optimization/59303
4022         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
4023         (dump_predicates): Better output format.
4024         (pred_equal_p): New function.
4025         (is_neq_relop_p): Ditto.
4026         (is_neq_zero_form_p): Ditto.
4027         (pred_expr_equal_p): Ditto.
4028         (pred_neg_p): Ditto.
4029         (simplify_pred): Ditto.
4030         (simplify_preds_2): Ditto.
4031         (simplify_preds_3): Ditto.
4032         (simplify_preds_4): Ditto.
4033         (simplify_preds): Ditto.
4034         (push_pred): Ditto.
4035         (push_to_worklist): Ditto.
4036         (get_pred_info_from_cmp): Ditto.
4037         (is_degenerated_phi): Ditto.
4038         (normalize_one_pred_1): Ditto.
4039         (normalize_one_pred): Ditto.
4040         (normalize_one_pred_chain): Ditto.
4041         (normalize_preds): Ditto.
4042         (normalize_cond_1): Remove function.
4043         (normalize_cond): Ditto.
4044         (is_gcond_subset_of): Ditto.
4045         (is_subset_of_any): Ditto.
4046         (is_or_set_subset_of): Ditto.
4047         (is_and_set_subset_of): Ditto.
4048         (is_norm_cond_subset_of): Ditto.
4049         (pred_chain_length_cmp): Ditto.
4050         (convert_control_dep_chain_into_preds): Type change.
4051         (find_predicates): Ditto.
4052         (find_def_preds): Ditto.
4053         (destroy_predicates_vecs): Ditto.
4054         (find_matching_predicates_in_rest_chains): Ditto.
4055         (use_pred_not_overlap_with_undef_path_pred): Ditto.
4056         (is_pred_expr_subset): Ditto.
4057         (is_pred_chain_subset_of): Ditto.
4058         (is_included_in): Ditto.
4059         (is_superset_of): Ditto.
4061 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
4063         Update copyright years.
4065 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
4067         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
4068         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
4069         config/arc/arc.md, config/arc/arc.opt,
4070         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
4071         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
4072         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
4073         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
4074         config/linux-protos.h, config/linux.c, config/winnt-c.c,
4075         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
4076         vtable-verify.c, vtable-verify.h: Use the standard form for the
4077         copyright notice.
4079 2014-01-02  Tobias Burnus  <burnus@net-b.de>
4081         * gcc.c (process_command): Update copyright notice dates.
4082         * gcov-dump.c: Ditto.
4083         * gcov.c: Ditto.
4084         * doc/cpp.texi: Bump @copying's copyright year.
4085         * doc/cppinternals.texi: Ditto.
4086         * doc/gcc.texi: Ditto.
4087         * doc/gccint.texi: Ditto.
4088         * doc/gcov.texi: Ditto.
4089         * doc/install.texi: Ditto.
4090         * doc/invoke.texi: Ditto.
4092 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4094         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
4096 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
4098         * config/i386/sse.md (*mov<mode>_internal): Guard
4099         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
4101         PR rtl-optimization/59647
4102         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4103         new_rtx into UNSIGNED_FLOAT rtxes.
4105 Copyright (C) 2014 Free Software Foundation, Inc.
4107 Copying and distribution of this file, with or without modification,
4108 are permitted in any medium without royalty provided the copyright
4109 notice and this notice are preserved.