* config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
[official-gcc.git] / gcc / ChangeLog
blob67299afada4eca143311e1f21f6888adbcaa43e0
1 2014-02-21  Nick Clifton  <nickc@redhat.com>
3         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
4         (pushhi1): Likewise.
5         (popqi1): Add mode to pre_dec.
6         (pophi1): Likewise.
8 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
10         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
11         mode for mask of V8SFmode permutation.
13 2014-02-20  Richard Henderson  <rth@redhat.com>
15         PR c++/60272
16         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
17         a new pseudo for OLDVAL.
19 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
21         PR target/57896
22         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
23         gen_reg_rtx if d->testing_p.
24         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
25         if d->testing_p and we will certainly return true.
26         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
27         if d->testing_p.
29 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
31         * emit-rtl.c (gen_reg_rtx): Assert that
32         crtl->emit.regno_pointer_align_length is non-zero.
34 2014-02-20  Richard Henderson  <rth@redhat.com>
36         PR c++/60272
37         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
38         on failure the store back into EXPECT.
40 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
41             Sandra Loosemore  <sandra@codesourcery.com>
43         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
44         * config/nios2/nios2.c (nios2_function_profiler):
45         Add -fPIC (flag_pic == 2) support.
46         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
47         (nios2_large_offset_p): New function.
48         (nios2_unspec_reloc_p): Move up position, update to use
49         nios2_large_offset_p.
50         (nios2_unspec_address): Remove function.
51         (nios2_unspec_offset): New function.
52         (nios2_large_got_address): New function.
53         (nios2_got_address): Add large offset support.
54         (nios2_legitimize_tls_address): Update usage of removed and new
55         functions.
56         (nios2_symbol_binds_local_p): New function.
57         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
58         (nios2_legitimize_address): Update to use nios2_large_offset_p.
59         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
60         (nios2_print_operand): Merge H/L processing, add hiadj/lo
61         processing for (const (unspec ...)).
62         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
64 2014-02-20  Richard Biener  <rguenther@suse.de>
66         * tree-cfg.c (replace_uses_by): Mark altered BBs before
67         doing the substitution.
68         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
70 2014-02-20  Martin Jambor  <mjambor@suse.cz>
72         PR ipa/55260
73         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
74         info when checking whether lattices are bottom.
76 2014-02-20  Richard Biener  <rguenther@suse.de>
78         PR middle-end/60221
79         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
80         regions at -O0.
82 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
84         PR ipa/58555
85         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
86         parameter specifying the scaling.
87         (inline_call): Update.
88         (want_inline_recursively): Guard division by zero.
89         (recursive_inlining): Update.
90         * ipa-inline.h (clone_inlined_nodes): Update.
92 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
94         PR target/60204
95         * config/i386/i386.c (classify_argument): Pass structures of size
96         64 bytes or less in register.
98 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
99             Kirill Yukhin  <kirill.yukhin@intel.com>
101         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
102         (_mm_rcp28_round_ss): Ditto.
103         (_mm_rsqrt28_round_sd): Ditto.
104         (_mm_rsqrt28_round_ss): Ditto.
105         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
106         (_mm_rcp14_round_ss): Ditto.
107         (_mm_rsqrt14_round_sd): Ditto.
108         (_mm_rsqrt14_round_ss): Ditto.
109         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
110         the first input operand, get rid of match_dup.
111         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
112         attribute to sse.
113         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
114         Ditto.
115         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
116         operand as the first input operand, set type attribute.
117         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
118         Set type attribute.
119         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
120         operand as the first input operand, set type attribute.
122 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
124         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
125         bit of zero.
127 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
129         PR target/60207
130         * config/i386/i386.c (construct_container): Remove TFmode check
131         for X86_64_INTEGER_CLASS.
133 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
135         PR target/59794
136         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
137         only when -Wpsabi is enabled.
139 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
141          PR target/59799
142         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
143         passing arrays in registers are the same as for structs, so remove the
144         special case for them.
146 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
148         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
149         destination type, extract only the valid bits if the source type is not
150         integral and has a different mode.
152 2014-02-19  Richard Biener  <rguenther@suse.de>
154         PR ipa/60243
155         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
156         for all calls.
158 2014-02-19  Richard Biener  <rguenther@suse.de>
160         PR ipa/60243
161         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
162         (ipa_modify_call_arguments): Emit an argument load explicitely and
163         preserve virtual SSA form there and for the replacement call.
164         Do not update SSA form nor free dominance info.
166 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
168         * ipa.c (function_and_variable_visibility): Also clear WEAK
169         flag when disolving COMDAT_GROUP.
171 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
173         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
174         * ipa-prop.c (ipa_set_jf_known_type): Return early when
175         not devirtualizing.
176         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
177         do more sanity checks.
178         (detect_type_change): Return true when giving up early.
179         (compute_complex_assign_jump_func): Fix type parameter of
180         ipa_set_ancestor_jf.
181         (compute_complex_ancestor_jump_func): Likewise.
182         (update_jump_functions_after_inlining): Fix updating of
183         ancestor function.
184         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
186 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
188         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
189         inline clones when edge disappears.
191 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
193         PR target/60203
194         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
195         Split 64-bit moves into 2 patterns.  Do not allow the use of
196         direct move for TDmode in little endian, since the decimal value
197         has little endian bytes within a word, but the 64-bit pieces are
198         ordered in a big endian fashion, and normal subreg's of TDmode are
199         not allowed.
200         (mov<mode>_64bit_dm): Likewise.
201         (movtd_64bit_nodm): Likewise.
203 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
205         PR tree-optimization/60174
206         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
207         statement of an SSA_NAME that occurs in an abnormal PHI node.
209 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
211         PR sanitizer/60142
212         * final.c (SEEN_BB): Remove.
213         (SEEN_NOTE, SEEN_EMITTED): Renumber.
214         (final_scan_insn): Don't force_source_line on second
215         NOTE_INSN_BASIC_BLOCK.
217 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
219         PR target/60205
220         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
221         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
222         (type_natural_mode): Warn ABI change when %zmm register is not
223         available for AVX512F vector value passing.
225 2014-02-18  Kai Tietz  <ktietz@redhat.com>
227         PR target/60193
228         * config/i386/i386.c (ix86_expand_prologue): Use value in
229         rax register as displacement when restoring %r10 or %rax.
230         Fix wrong offset when restoring both registers.
232 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
234         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
235         assertion with conditional return.
237 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
238             Uros Bizjak  <ubizjak@gmail.com>
240         PR driver/60233
241         * config/i386/driver-i386.c (host_detect_local_cpu): If
242         YMM state is not saved by the OS, also clear has_f16c.  Move
243         CPUID 0x80000001 handling before YMM state saving checking.
245 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
247         PR rtl-optimization/58960
248         * haifa-sched.c (alloc_global_sched_pressure_data): New,
249         factored out from ...
250         (sched_init): ... here.
251         (free_global_sched_pressure_data): New, factored out from ...
252         (sched_finish): ... here.
253         * sched-int.h (free_global_sched_pressure_data): Declare.
254         * sched-rgn.c (nr_regions_initial): New static global.
255         (haifa_find_rgns): Initialize it.
256         (schedule_region): Disable sched-pressure for the newly
257         generated regions.
259 2014-02-17  Richard Biener  <rguenther@suse.de>
261         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
262         release SSA defs of pattern stmts.
264 2014-02-17  Richard Biener  <rguenther@suse.de>
266         * tree-inline.c (expand_call_inline): Release the virtual
267         operand defined by the call we are about to inline.
269 2014-02-17  Richard Biener  <rguenther@suse.de>
271         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
273 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
274             Ilya Tocar  <ilya.tocar@intel.com>
276         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
277         arguments order in builtin.
278         (_mm512_permutexvar_epi64): Ditto.
279         (_mm512_mask_permutexvar_epi64): Ditto
280         (_mm512_maskz_permutexvar_epi32): Ditto
281         (_mm512_permutexvar_epi32): Ditto
282         (_mm512_mask_permutexvar_epi32): Ditto
284 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
286         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
287         (p8_vmrgow): Likewise.
289 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
291         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
292         endian targets.
294 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
296         PR target/60203
297         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
298         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
299         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
300         using direct move instructions on ISA 2.07.  Also adjust
301         instruction length for 64-bit.
302         (mov<mode>_64bit, TFmode/TDmode): Likewise.
303         (mov<mode>_32bit, TFmode/TDmode): Likewise.
305 2014-02-15  Alan Modra  <amodra@gmail.com>
307         PR target/58675
308         PR target/57935
309         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
310         find_replacement on parts of insn rtl that might be reloaded.
312 2014-02-15  Richard Biener  <rguenther@suse.de>
314         PR tree-optimization/60183
315         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
316         (tree_ssa_phiprop): Calculate and free post-dominators.
318 2014-02-14  Jeff Law  <law@redhat.com>
320         PR rtl-optimization/60131
321         * ree.c (get_extended_src_reg): New function.
322         (combine_reaching_defs): Use it rather than assuming location of REG.
323         (find_and_remove_re): Verify first operand of extension is
324         a REG before adding the insns to the copy list.
326 2014-02-14  Roland McGrath  <mcgrathr@google.com>
328         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
329         * configure: Regenerated.
330         * config.in: Regenerated.
331         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
332         instead of ASM_SHORT.
334 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
335             Richard Earnshaw  <rearnsha@arm.com>
337         PR rtl-optimization/59535
338         * lra-constraints.c (process_alt_operands): Encourage alternative
339         when unassigned pseudo class is superset of the alternative class.
340         (inherit_reload_reg): Don't inherit when optimizing for code size.
341         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
342         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
343         modes not less than 4 for Thumb1.
345 2014-02-14  Kyle McMartin  <kyle@redhat.com>
347         PR pch/60010
348         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
350 2014-02-14  Richard Biener  <rguenther@suse.de>
352         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
353         (get_frame_arg): Drop the assert with langhook types_compatible_p.
354         Do not strip INDIRECT_REFs.
356 2014-02-14  Richard Biener  <rguenther@suse.de>
358         PR lto/60179
359         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
360         DECL_FUNCTION_SPECIFIC_TARGET.
361         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
362         * tree-streamer-out.c (pack_ts_target_option): Remove.
363         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
364         (write_ts_function_decl_tree_pointers): Do not stream
365         DECL_FUNCTION_SPECIFIC_TARGET.
366         * tree-streamer-in.c (unpack_ts_target_option): Remove.
367         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
368         (lto_input_ts_function_decl_tree_pointers): Do not stream
369         DECL_FUNCTION_SPECIFIC_TARGET.
371 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
373         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
374         (get_initial_def_for_induction, vectorizable_induction): Ignore
375         debug stmts when looking for exit_phi.
376         (vectorizable_live_operation): Fix up condition.
378 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
380         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
381         nreverse() because it changes the content of original tree list.
383 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
385         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
386         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
388 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
390         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
391         GNU coding standards.
393 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
395         PR debug/60152
396         * dwarf2out.c (gen_subprogram_die): Don't call
397         add_calling_convention_attribute if subr_die is old_die.
399 2014-02-13  Sharad Singhai  <singhai@google.com>
401         * doc/optinfo.texi: Fix order of nodes.
403 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
405         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
406         operands[2], not operands[3].
408 2014-02-13  Richard Biener  <rguenther@suse.de>
410         PR bootstrap/59878
411         * doc/install.texi (ISL): Update recommended version to 0.12.2,
412         mention the possibility of an in-tree build.
413         (CLooG): Update recommended version to 0.18.1, mention the
414         possibility of an in-tree build and clarify that the ISL
415         bundled with CLooG does not work.
417 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
419         PR target/43546
420         * expr.c (compress_float_constant): If x is a hard register,
421         extend into a pseudo and then move to x.
423 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
425         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
426         caused by bad second argument to warning_at() with -mhotpatch and
427         nested functions (e.g. with gfortran).
429 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
431         * opts.c (option_name): Remove "enabled by default" rider.
433 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
435         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
437 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
438             Uros Bizjak  <ubizjak@gmail.com>
440         PR target/60151
441         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
442         * configure: Regenerated.
444 2014-02-12  Richard Biener  <rguenther@suse.de>
446         * vec.c (vec_prefix::calculate_allocation): Move as
447         inline variant to vec.h.
448         (vec_prefix::calculate_allocation_1): New out-of-line version.
449         * vec.h (vec_prefix::calculate_allocation_1): Declare.
450         (vec_prefix::m_has_auto_buf): Rename to ...
451         (vec_prefix::m_using_auto_storage): ... this.
452         (vec_prefix::calculate_allocation): Inline the easy cases
453         and dispatch to calculate_allocation_1 which doesn't need the
454         prefix address.
455         (va_heap::reserve): Use gcc_checking_assert.
456         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
457         m_using_auto_storage.
458         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
459         member and adjust.
460         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
461         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
462         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
464 2014-02-12  Richard Biener  <rguenther@suse.de>
466         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
467         when we found a dependence.
469 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
471         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
472         common code...
473         (maybe_fold_stmt): ... into this new function.
474         * omp-low.c (lower_omp): Update comment.
476         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
477         last use.
479         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
480         dereference.
482 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
484         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
485         identifiers in comments.
486         (cortexa53_extra_costs): Likewise.
487         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
488         (cortexa7_extra_costs): Likewise.
489         (cortexa12_extra_costs): Likewise.
490         (cortexa15_extra_costs): Likewise.
491         (v7m_extra_costs): Likewise.
493 2014-02-12  Richard Biener  <rguenther@suse.de>
495         PR middle-end/60092
496         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
497         of posix_memalign being successful.
498         (lower_stmt): Restrict lowering of posix_memalign to when
499         -ftree-bit-ccp is enabled.
501 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
503         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
504         arg_loc.
505         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
507 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
509         PR rtl-optimization/60116
510         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
511         other_insn once the combination has been validated.
513 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
515         PR lto/59468
516         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
517         and wrapper.
518         * ipa-devirt.c: Include demangle.h
519         (odr_violation_reported): New static variable.
520         (add_type_duplicate): Update odr_violations.
521         (maybe_record_node): Add completep parameter; update it.
522         (record_target_from_binfo): Add COMPLETEP parameter;
523         update it as needed.
524         (possible_polymorphic_call_targets_1): Likewise.
525         (struct polymorphic_call_target_d): Add nonconstruction_targets;
526         rename FINAL to COMPLETE.
527         (record_targets_from_bases): Sanity check we found the binfo;
528         fix COMPLETEP updating.
529         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
530         parameter, fix computing of COMPLETEP.
531         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
532         at LTO time do demangling.
533         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
534         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
535         parameter.
536         (gimple_get_virt_method_for_binfo): Likewise.
537         * gimple-fold.h (gimple_get_virt_method_for_binfo,
538         gimple_get_virt_method_for_vtable): Update prototypes.
540 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
542         PR target/49008
543         * genautomata.c (add_presence_absence): Fix typo with
544         {final_}presence_list.
546 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
548         PR target/60137
549         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
550         for VSX/Altivec vectors that land in GPR registers.
552 2014-02-11  Richard Henderson  <rth@redhat.com>
553             Jakub Jelinek  <jakub@redhat.com>
555         PR debug/59776
556         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
557         around drhs if type conversion to lacc->type is not useless.
559 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
561         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
562         tuning struct.
563         (cortex-a57.cortex-a53): Likewise.
564         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
566 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
568         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
569         arm_restrict_it.
571 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
573         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
574         add_options_for_arm_vfp3.
576 2014-02-11  Jeff Law  <law@redhat.com>
578         PR middle-end/54041
579         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
580         object with an undesirable mode.
582 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
584         PR libgomp/60107
585         * config/i386/sol2-9.h: New file.
586         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
587         *-*-solaris2.9*): Use it.
589 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
591         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
592         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
594 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
596         * config/microblaze/microblaze.c: Extend mcpu version format
598 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
600         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
602 2014-02-10  Richard Henderson  <rth@redhat.com>
604         PR target/59927
605         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
606         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
607         ms-abi vs -mno-accumulate-outgoing-args.
608         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
609         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
610         respect to ms-abi.
612 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
614         PR middle-end/60080
615         * cfgexpand.c (expand_asm_operands): Attach source location to
616         ASM_INPUT rtx objects.
617         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
619 2014-02-10  Nick Clifton  <nickc@redhat.com>
621         * config/mn10300/mn10300.c (popcount): New function.
622         (mn10300_expand_prologue): Include saved registers in stack usage
623         count.
625 2014-02-10  Jeff Law  <law@redhat.com>
627         PR middle-end/52306
628         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
629         when changing the SET_DEST of a prior insn to avoid an input reload.
631 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
633         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
634         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
635         -mcall-openbsd, or -mcall-linux.
636         (CC1_ENDIAN_BIG_SPEC): Remove.
637         (CC1_ENDIAN_LITTLE_SPEC): Remove.
638         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
639         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
640         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
641         and %cc1_endian_default.
642         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
644 2014-02-10  Richard Biener  <rguenther@suse.de>
646         PR tree-optimization/60115
647         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
648         MEM_REF handling.  Properly verify that the accesses are not
649         out of the objects bound.
651 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
653         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
654         coretex to cortex.
656 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
658         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
659         proper constants and fix formatting.
660         (possible_polymorphic_call_targets): Fix formatting.
662 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
663             Ilya Tocar  <ilya.tocar@intel.com>
665         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
666         (_mm512_loadu_epi32): Renamed into...
667         (_mm512_loadu_si512): This.
668         (_mm512_storeu_epi32): Renamed into...
669         (_mm512_storeu_si512): This.
670         (_mm512_maskz_ceil_ps): Removed.
671         (_mm512_maskz_ceil_pd): Ditto.
672         (_mm512_maskz_floor_ps): Ditto.
673         (_mm512_maskz_floor_pd): Ditto.
674         (_mm512_floor_round_ps): Ditto.
675         (_mm512_floor_round_pd): Ditto.
676         (_mm512_ceil_round_ps): Ditto.
677         (_mm512_ceil_round_pd): Ditto.
678         (_mm512_mask_floor_round_ps): Ditto.
679         (_mm512_mask_floor_round_pd): Ditto.
680         (_mm512_mask_ceil_round_ps): Ditto.
681         (_mm512_mask_ceil_round_pd): Ditto.
682         (_mm512_maskz_floor_round_ps): Ditto.
683         (_mm512_maskz_floor_round_pd): Ditto.
684         (_mm512_maskz_ceil_round_ps): Ditto.
685         (_mm512_maskz_ceil_round_pd): Ditto.
686         (_mm512_expand_pd): Ditto.
687         (_mm512_expand_ps): Ditto.
688         * config/i386/i386.c (ix86_builtins): Remove
689         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
690         (bdesc_args): Ditto.
691         * config/i386/predicates.md (const1256_operand): New.
692         (const_1_to_2_operand): Ditto.
693         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
694         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
695         (*avx512pf_gatherpf<mode>sf): Ditto.
696         (avx512pf_gatherpf<mode>df): Ditto.
697         (*avx512pf_gatherpf<mode>df_mask): Ditto.
698         (*avx512pf_gatherpf<mode>df): Ditto.
699         (avx512pf_scatterpf<mode>sf): Ditto.
700         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
701         (*avx512pf_scatterpf<mode>sf): Ditto.
702         (avx512pf_scatterpf<mode>df): Ditto.
703         (*avx512pf_scatterpf<mode>df_mask): Ditto.
704         (*avx512pf_scatterpf<mode>df): Ditto.
705         (avx512f_expand<mode>): Removed.
706         (<shift_insn><mode>3<mask_name>): Change predicate type.
708 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
710         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
711         not at the end of datarefs vector use ordered_remove to avoid
712         reordering datarefs vector.
714         PR c/59984
715         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
716         mark local addressable non-static vars as GOVD_PRIVATE
717         instead of GOVD_LOCAL.
718         * omp-low.c (lower_omp_for): Move gimple_bind_vars
719         and BLOCK_VARS of gimple_bind_block to new_stmt rather
720         than copying them.
722         PR middle-end/60092
723         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
724         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
725         assume_aligned or alloc_align attributes.
726         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
727         arguments.  Handle also assume_aligned and alloc_align attributes.
728         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
729         calls to functions with assume_aligned or alloc_align attributes.
730         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
732 2014-02-08  Terry Guo  <terry.guo@arm.com>
734         * doc/invoke.texi: Document ARM -march=armv7e-m.
736 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
738         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
739         flag on __cilkrts_rethrow builtin.
741         PR ipa/60026
742         * ipa-cp.c (determine_versionability): Fail at -O0
743         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
744         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
746         Revert:
747         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
749         PR ipa/60026
750         * tree-inline.c (copy_forbidden): Fail for
751         __attribute__((optimize (0))) functions.
753 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
755         * varpool.c: Include pointer-set.h.
756         (varpool_remove_unreferenced_decls): Variables in other partitions
757         will not be output; be however careful to not lose information
758         about partitioning.
760 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
762         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
763         lookup in the vtable constructor.
765 2014-02-07  Jeff Law  <law@redhat.com>
767         PR target/40977
768         * config/m68k/m68k.md (ashldi_extsi): Turn into a
769         define_insn_and_split.
771         * ipa-inline.c (inline_small_functions): Fix typos.
773 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
775         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
776         (s390_can_use_return_insn): Declare.
777         * config/s390/s390.h (EPILOGUE_USES): Define.
778         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
779         instructions.
780         (s390_chunkify_start): Handle return JUMP_LABELs.
781         (s390_early_mach): Emit a main_pool instruction on the entry edge.
782         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
783         (s390_can_use_return_insn): New functions.
784         (s390_fix_long_loop_prediction): Handle conditional returns.
785         (TARGET_SET_UP_BY_PROLOGUE): Define.
786         * config/s390/s390.md (ANY_RETURN): New code iterator.
787         (*creturn, *csimple_return, return, simple_return): New patterns.
789 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
791         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
792         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
793         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
794         REG_CFA_RESTORE list when deciding not to restore a register.
796 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
798         * config/s390/s390.c: Include tree-pass.h and context.h.
799         (s390_early_mach): New function, split out from...
800         (s390_emit_prologue): ...here.
801         (pass_data_s390_early_mach): New pass structure.
802         (pass_s390_early_mach): New class.
803         (s390_option_override): Create and register early_mach pass.
804         Move to end of file.
806 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
808         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
809         to match for the exit block.
811 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
813         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
814         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
815         Reject misaligned operands.
817 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
819         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
821 2014-02-07  Richard Biener  <rguenther@suse.de>
823         PR middle-end/60092
824         * gimple-low.c (lower_builtin_posix_memalign): New function.
825         (lower_stmt): Call it to lower posix_memalign in a way
826         to make alignment info accessible.
828 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
830         PR c++/60082
831         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
832         __builtin_setjmp_receiver.
834 2014-02-07  Richard Biener  <rguenther@suse.de>
836         PR middle-end/60092
837         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
838         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
839         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
840         Handle BUILT_IN_POSIX_MEMALIGN.
841         (find_func_clobbers): Likewise.
842         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
843         (call_may_clobber_ref_p_1): Likewise.
845 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
847         PR ipa/59918
848         * ipa-devirt.c (record_target_from_binfo): Remove overactive
849         sanity check.
851 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
853         PR ipa/59469
854         * lto-cgraph.c (lto_output_node): Use
855         symtab_get_symbol_partitioning_class.
856         (lto_output_varpool_node): likewise.
857         (symtab_get_symbol_partitioning_class): Move here from
858         lto/lto-partition.c
859         * cgraph.h (symbol_partitioning_class): Likewise.
860         (symtab_get_symbol_partitioning_class): Declare.
862 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
864         * ggc.h (ggc_internal_cleared_alloc): New macro.
865         * vec.h (vec_safe_copy): Handle memory stats.
866         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
867         * target-globals.c (save_target_globals): Likewise.
869 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
871         PR target/60077
872         * expr.c (emit_move_resolve_push): Export; be bit more selective
873         on when to clear alias set.
874         * expr.h (emit_move_resolve_push): Declare.
875         * function.h (struct function): Add tail_call_marked.
876         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
877         * config/i386/i386-protos.h (ix86_expand_push): Remove.
878         * config/i386/i386.md (TImode move expander): De not call
879         ix86_expand_push.
880         (FP push expanders): Preserve memory attributes.
881         * config/i386/sse.md (push<mode>1): Remove.
882         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
883         (ix86_expand_push): Remove.
884         * config/i386/mmx.md (push<mode>1): Remove.
886 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
888         PR rtl-optimization/60030
889         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
890         lopart with paradoxical subreg before shifting it up by hprec.
892 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
894         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
895         Remove extra newline at end of file.
896         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
897         (arm_issue_rate): Handle cortexa57.
898         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
899         (cortex-a57.cortex-a53): Likewise.
901 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
903         PR target/59575
904         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
905         don't record in REG_FRAME_RELATED_EXPR registers not set in that
906         bitmask.
907         (arm_expand_prologue): Adjust all callers.
908         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
909         info, registers also at the lowest numbered registers side.  Use
910         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
911         XEXP.
913         PR debug/59992
914         * var-tracking.c (adjust_mems): Before adding a SET to
915         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
917 2014-02-06  Alan Modra  <amodra@gmail.com>
919         PR target/60032
920         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
921         change SDmode to DDmode when lra_in_progress.
923 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
925         PR middle-end/59150
926         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
927         free_data_ref on the dr first, and before goto again also set dr
928         to the next dr.  For simd_lane_access, free old datarefs[i] before
929         overwriting it.  For get_vectype_for_scalar_type failure, don't
930         free_data_ref if simd_lane_access.
932         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
934         PR target/60062
935         * tree.h (opts_for_fn): New inline function.
936         (opt_for_fn): Define.
937         * config/i386/i386.c (ix86_function_regparm): Use
938         opt_for_fn (decl, optimize) instead of optimize.
940 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
942         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
943         for SYMBOL_REF in large memory model.
945 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
947         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
948         and crypto support.
949         (cortex-a57): Likewise.
950         (cortex-a57.cortex-a53): Likewise.
952 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
953             Kugan Vivekanandarajah  <kuganv@linaro.org>
955         * config/arm/arm.c (arm_vector_alignment_reachable): Check
956         unaligned_access.
957         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
959 2014-02-06  Richard Biener  <rguenther@suse.de>
961         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
962         set_loop_copy and initialize_original_copy_tables.
964 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
966         * config/aarch64/aarch64-simd.md
967         (aarch64_ashr_simddi): Change QI to SI.
969 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
970             Jakub Jelinek  <jakub@redhat.com>
972         PR middle-end/60013
973         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
974         of the dataflow.
976 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
978         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
979         CODE_FOR_altivec_vpku[hw]um to
980         CODE_FOR_altivec_vpku[hw]um_direct.
981         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
982         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
983         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
984         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
986 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
988         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
989         generation for -maltivec=be.
990         (altivec_vsumsws): Simplify redundant test.
992 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
994         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
995         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
996         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
997         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
998         gen_altivec_vpkuwum.
999         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1000         BYTES_BIG_ENDIAN.
1001         (altivec_vpks<VI_char>ss): Likewise.
1002         (altivec_vpks<VI_char>us): Likewise.
1003         (altivec_vpku<VI_char>us): Likewise.
1004         (altivec_vpku<VI_char>um): Likewise.
1005         (altivec_vpku<VI_char>um_direct): New (copy of
1006         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1007         internal use).
1008         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1009         target is little endian and -maltivec=be is not specified.
1010         (*altivec_vupkhs<VU_char>_direct): New (copy of
1011         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1012         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1013         target is little endian and -maltivec=be is not specified.
1014         (*altivec_vupkls<VU_char>_direct): New (copy of
1015         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1016         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1017         little endian and -maltivec=be is not specified.
1018         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1019         little endian and -maltivec=be is not specified.
1021 2014-02-05  Richard Henderson  <rth@redhat.com>
1023         PR debug/52727
1024         * combine-stack-adj.c: Revert r206943.
1025         * sched-int.h (struct deps_desc): Add last_args_size.
1026         * sched-deps.c (init_deps): Initialize it.
1027         (sched_analyze_insn): Add OUTPUT dependencies between insns that
1028         contain REG_ARGS_SIZE notes.
1030 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
1032         * lto-cgraph.c (asm_nodes_output): Make global.
1033         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1034         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1035         (driver_handle_option): Handle OPT_fwpa.
1037 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
1039         PR ipa/59947
1040         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1041         a comment typo and formatting issue.  If odr_hash hasn't been
1042         created, return vNULL and set *completep to false.
1044         PR middle-end/57499
1045         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1046         bb with no successors.
1048 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
1050         PR target/59718
1051         * doc/invoke.texi (-march): Clarify documentation for ARM.
1052         (-mtune): Likewise.
1053         (-mcpu): Likewise.
1055 2014-02-05  Richard Biener  <rguenther@suse.de>
1057         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1058         when not vectorizing because of too many alias checks.
1059         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1060         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1062 2014-02-05  Nick Clifton  <nickc@redhat.com>
1064         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1065         accept extended registers in any mode when compiling for the MN10300.
1067 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
1069         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1070         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1071         sanitization attributes.
1072         (can_inline_edge_p): Likewise.
1073         (sanitize_attrs_match_for_inline_p): New function.
1075 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1077         * ipa-prop.c (detect_type_change): Shor circuit testing of
1078         type changes on THIS pointer.
1080 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
1082         PR target/59777
1083         * config/pa/pa.c (legitimize_tls_address): Return original address
1084         if not passed a SYMBOL_REF rtx.
1085         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1086         addresses.
1087         (pa_emit_move_sequence): Simplify TLS source operands.
1088         (pa_legitimate_constant_p): Reject all TLS constants.
1089         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1090         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1092 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1094         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1095         groups when we know they are controlled by LTO.
1096         * varasm.c (default_binds_local_p_1): If object is in other partition,
1097         it will be resolved locally.
1099 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1101         * config/host-linux.c (linux_gt_pch_use_address): Don't
1102         use SSIZE_MAX because it is not always defined.
1104 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
1106         PR bootstrap/59913
1107         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1108         threshold for pseudo splitting.
1109         (update_ebb_live_info): Process call argument hard registers and
1110         hard registers from insn definition too.
1111         (max_small_class_regs_num): New constant.
1112         (inherit_in_ebb): Update live hard regs through EBBs.  Update
1113         reloads_num only for small register classes.  Don't split for
1114         outputs of jumps.
1116 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
1118         PR ipa/60058
1119         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1120         is non-null.
1122 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1124         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1125         visibility is safe.
1127 2014-02-04  Marek Polacek  <polacek@redhat.com>
1129         * gdbinit.in (pel): Define.
1131 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1133         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1134         behavior.
1136 2014-02-04  Richard Biener  <rguenther@suse.de>
1138         PR lto/59723
1139         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1140         in function context local.
1141         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1142         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1143         similar to LTO_imported_decl_ref.
1145 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1147         PR tree-optimization/60002
1148         * cgraphclones.c (build_function_decl_skip_args): Clear
1149         DECL_LANG_SPECIFIC.
1151         PR tree-optimization/60023
1152         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1153         false to gsi_replace.
1154         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1155         has been in some EH region and vec_stmt could throw, add
1156         vec_stmt into the same EH region.
1157         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1158         has no lhs, ignore it.
1159         * internal-fn.c (expand_MASK_LOAD): Likewise.
1161         PR ipa/60026
1162         * tree-inline.c (copy_forbidden): Fail for
1163         __attribute__((optimize (0))) functions.
1165         PR other/58712
1166         * omp-low.c (simd_clone_struct_copy): If from->inbranch
1167         is set, copy one less argument.
1168         (expand_simd_clones): Don't subtract clone_info->inbranch
1169         from simd_clone_struct_alloc argument.
1171         PR rtl-optimization/57915
1172         * recog.c (simplify_while_replacing): If all unary/binary/relational
1173         operation arguments are constant, attempt to simplify those.
1175         PR middle-end/59261
1176         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1177         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1179 2014-02-04  Richard Biener  <rguenther@suse.de>
1181         PR tree-optimization/60012
1182         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1183         TBAA disambiguation to all DDRs.
1185 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1187         PR target/59788
1188         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1189         (LINK_SPEC): Use it for -shared, -shared-libgcc.
1191 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1193         PR ipa/59882
1194         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1196 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1198         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1199         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1201 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1203         PR ipa/59831
1204         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1205         to figure out targets of polymorphic calls with known decl.
1206         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1207         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1208         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1209         (get_polymorphic_call_info): ... here.
1210         (get_polymorphic_call_info_from_invariant): New function.
1212 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1214         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1215         lookup via vtable pointer; check for type consistency
1216         and turn inconsitent facts into UNREACHABLE.
1217         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1218         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1219         type inconsistent querries; return UNREACHABLE instead.
1221 2014-02-03  Richard Henderson  <rth@twiddle.net>
1223         PR tree-opt/59924
1224         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1225         already processed this node.
1226         (normalize_one_pred_1): Pass along mark_set.
1227         (normalize_one_pred): Create and destroy a pointer_set_t.
1228         (normalize_one_pred_chain): Likewise.
1230 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
1232         PR gcov-profile/58602
1233         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1235 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1237         PR ipa/59831
1238         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1239         -fno-devirtualize; try to devirtualize by the knowledge of
1240         virtual table pointer given by aggregate propagation.
1241         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1242         (ipa_print_node_jump_functions): Dump also offset that
1243         is relevant for polymorphic calls.
1244         (determine_known_aggregate_parts): Add arg_type parameter; use it
1245         instead of determining the type from pointer type.
1246         (ipa_compute_jump_functions_for_edge): Update call of
1247         determine_known_aggregate_parts.
1248         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1249         (gimple_get_virt_method_for_binfo): ... here; simplify using
1250         vtable_pointer_value_to_vtable.
1251         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1252         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1253         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1254         (vtable_pointer_value_to_vtable): Break out from ...; handle also
1255         POINTER_PLUS_EXPR.
1256         (vtable_pointer_value_to_binfo): ... here.
1257         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1259 2014-02-03  Teresa Johnson  <tejohnson@google.com>
1261         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1262         redef of outer loop index variable.
1264 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
1266         PR c++/53017
1267         PR c++/59211
1268         * doc/extend.texi (Function Attributes): Typo.
1270 2014-02-03  Cong Hou  <congh@google.com>
1272         PR tree-optimization/60000
1273         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1274         if the vectorized statement is a store.  A store statement can only
1275         appear at the end of pattern statements.
1277 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1279         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1280         (ix86_option_override_internal): Default long double to 64-bit for
1281         32-bit Bionic and to 128-bit for 64-bit Bionic.
1283         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1284         TARGET_LONG_DOUBLE_128 is true.
1285         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1287         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1288         (mlong-double-64): Negate -mlong-double-128.
1289         (mlong-double-128): New option.
1291         * config/i386/i386-c.c (ix86_target_macros): Define
1292         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1294         * doc/invoke.texi: Document -mlong-double-128.
1296 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1298         PR rtl-optimization/60024
1299         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1301 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
1303         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1305 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1307         PR rtl-optimization/57662
1308         * sel-sched.c (code_motion_path_driver): Do not mark already not
1309         existing blocks in the visiting bitmap.
1311 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1313         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1314         on the insn being emitted.
1316 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
1317             Will Deacon  <will.deacon@arm.com>
1319         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1321 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1323         * config/arm/arm-tables.opt: Regenerate.
1325 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1327         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1328         for vector types other than V16QImode.
1329         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1330         define_expand, and call altivec_expand_vec_perm_le when producing
1331         code with little endian element order.
1332         (*altivec_vperm_<mode>_internal): New insn having previous
1333         behavior of altivec_vperm_<mode>.
1334         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1335         altivec_expand_vec_perm_le when producing code with little endian
1336         element order.
1337         (*altivec_vperm_<mode>_uns_internal): New insn having previous
1338         behavior of altivec_vperm_<mode>_uns.
1340 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1342         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1343         (altivec_vsumsws): Add handling for -maltivec=be with a little
1344         endian target.
1345         (altivec_vsumsws_direct): New.
1346         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1347         gen_altivec_vsumsws.
1349 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1351         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1352         vtable_pointer_value_to_binfo): New functions.
1353         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1354         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1356 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
1358         * config/nios2/nios2.md (load_got_register): Initialize GOT
1359         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1360         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1362 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1364         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1365         preserverd by passthrough, do not propagate the type.
1367 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1369         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1370         (mips_atomic_assign_expand_fenv): New function.
1371         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1373 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1375         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1376         (__builtin_mips_set_fcsr): Likewise.
1377         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1378         MIPS_USI_FTYPE_VOID.
1379         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1380         (mips16_expand_set_fcsr): Likewise.
1381         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1382         (mips16_set_fcsr_stub): Likewise.
1383         (mips16_get_fcsr_one_only_stub): New class.
1384         (mips16_set_fcsr_one_only_stub): Likewise.
1385         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1386         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1387         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1388         (hard_float): New availability predicate.
1389         (mips_builtins): Add get_fcsr and set_fcsr.
1390         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1391         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1392         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1393         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1394         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1395         patterns.
1397 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
1399         * config/mips/mips.c (mips_one_only_stub): New class.
1400         (mips_need_mips16_rdhwr_p): Replace with...
1401         (mips16_rdhwr_stub): ...this new variable.
1402         (mips16_stub_call_address): New function.
1403         (mips16_rdhwr_one_only_stub): New class.
1404         (mips_expand_thread_pointer): Use mips16_stub_call_address.
1405         (mips_output_mips16_rdhwr): Delete.
1406         (mips_finish_stub): New function.
1407         (mips_code_end): Use it to handle rdhwr stubs.
1409 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
1411         PR target/60017
1412         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1413         when calculating size of integer atomic types.
1415 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
1417         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1419 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
1421         PR tree-optimization/60003
1422         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1423         * profile.c (branch_prob): Use gimple_call_builtin_p
1424         to check for BUILT_IN_SETJMP_RECEIVER.
1425         * tree-inline.c (copy_bb): Call notice_special_calls.
1427 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
1429         PR bootstrap/59985
1430         * lra-constraints.c (process_alt_operands): Update reload_sum only
1431         on the first pass.
1433 2014-01-31  Richard Henderson  <rth@redhat.com>
1435         PR middle-end/60004
1436         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1437         until after else_eh is processed.
1439 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1441         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1442         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1443         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1444         in smmintrin.h, remove them.
1445         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1446         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1447         * config/i386/i386.md (ROUND_SAE): Fix value.
1448         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1449         (const48_operand): New.
1450         * config/i386/subst.md (round), (round_expand): Use
1451         const_4_or_8_to_11_operand.
1452         (round_saeonly), (round_saeonly_expand): Use const48_operand.
1454 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
1456         * config/i386/constraints.md (Yk): Swap meaning with k.
1457         * config/i386/i386.md (movhi_internal): Change Yk to k.
1458         (movqi_internal): Ditto.
1459         (*k<logic><mode>): Ditto.
1460         (*andhi_1): Ditto.
1461         (*andqi_1): Ditto.
1462         (kandn<mode>): Ditto.
1463         (*<code>hi_1): Ditto.
1464         (*<code>qi_1): Ditto.
1465         (kxnor<mode>): Ditto.
1466         (kortestzhi): Ditto.
1467         (kortestchi): Ditto.
1468         (kunpckhi): Ditto.
1469         (*one_cmplhi2_1): Ditto.
1470         (*one_cmplqi2_1): Ditto.
1471         * config/i386/sse.md (): Change k to Yk.
1472         (avx512f_load<mode>_mask): Ditto.
1473         (avx512f_blendm<mode>): Ditto.
1474         (avx512f_store<mode>_mask): Ditto.
1475         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1476         (avx512f_storedqu<mode>_mask): Ditto.
1477         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1478         Ditto.
1479         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1480         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1481         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1482         (avx512f_maskcmp<mode>3): Ditto.
1483         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1484         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1485         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1486         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1487         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1488         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1489         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1490         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1491         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1492         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1493         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1494         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1495         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1496         (vec_extract_lo_<mode>_maskm): Ditto.
1497         (vec_extract_hi_<mode>_maskm): Ditto.
1498         (avx512f_vternlog<mode>_mask): Ditto.
1499         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1500         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1501         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1502         (avx512f_<code>v8div16qi2_mask): Ditto.
1503         (avx512f_<code>v8div16qi2_mask_store): Ditto.
1504         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1505         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1506         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1507         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1508         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1509         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1510         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1511         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1512         (avx512cd_maskb_vec_dupv8di): Ditto.
1513         (avx512cd_maskw_vec_dupv16si): Ditto.
1514         (avx512f_vpermi2var<mode>3_maskz): Ditto.
1515         (avx512f_vpermi2var<mode>3_mask): Ditto.
1516         (avx512f_vpermi2var<mode>3_mask): Ditto.
1517         (avx512f_vpermt2var<mode>3_maskz): Ditto.
1518         (*avx512f_gathersi<mode>): Ditto.
1519         (*avx512f_gathersi<mode>_2): Ditto.
1520         (*avx512f_gatherdi<mode>): Ditto.
1521         (*avx512f_gatherdi<mode>_2): Ditto.
1522         (*avx512f_scattersi<mode>): Ditto.
1523         (*avx512f_scatterdi<mode>): Ditto.
1524         (avx512f_compress<mode>_mask): Ditto.
1525         (avx512f_compressstore<mode>_mask): Ditto.
1526         (avx512f_expand<mode>_mask): Ditto.
1527         * config/i386/subst.md (mask): Change k to Yk.
1528         (mask_scalar_merge): Ditto.
1529         (sd): Ditto.
1531 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
1533         * doc/extend.texi (Vector Extensions): Document ?: in C++.
1535 2014-01-31  Richard Biener  <rguenther@suse.de>
1537         PR middle-end/59990
1538         * builtins.c (fold_builtin_memory_op): Make sure to not
1539         use a floating-point mode or a boolean or enumeral type for
1540         the copy operation.
1542 2014-01-30  DJ Delorie  <dj@redhat.com>
1544         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1545         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1546         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1547         whenever main() has an epilogue.
1549 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1551         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1552         unused variable "field".
1553         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1554         (vsx_mergeh_<mode>): Likewise.
1555         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1556         (altivec_vmrghh): Likewise.
1557         (altivec_vmrghw): Likewise.
1558         (altivec_vmrglb): Likewise.
1559         (altivec_vmrglh): Likewise.
1560         (altivec_vmrglw): Likewise.
1561         (altivec_vspltb): Add missing uses.
1562         (altivec_vsplth): Likewise.
1563         (altivec_vspltw): Likewise.
1564         (altivec_vspltsf): Likewise.
1566 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1568         PR target/59923
1569         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1570         frame related instructions.
1572 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
1574         PR rtl-optimization/59959
1575         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1576         any reload of register whose subreg is invalid.
1578 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
1580         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1581         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1582         Add missing return type - void.
1584 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1586         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1587         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1588         remove element index adjustment for endian (now handled in vsx.md
1589         and altivec.md).
1590         (altivec_expand_vec_perm_const): Use
1591         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1592         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1593         (vsx_xxspltw_<mode>): Adjust element index for little endian.
1594         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1595         define_expand and a new define_insn *altivec_vspltb_internal;
1596         adjust for -maltivec=be on a little endian target.
1597         (altivec_vspltb_direct): New.
1598         (altivec_vsplth): Divide into a define_expand and a new
1599         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1600         little endian target.
1601         (altivec_vsplth_direct): New.
1602         (altivec_vspltw): Divide into a define_expand and a new
1603         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1604         little endian target.
1605         (altivec_vspltw_direct): New.
1606         (altivec_vspltsf): Divide into a define_expand and a new
1607         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1608         a little endian target.
1610 2014-01-30  Richard Biener  <rguenther@suse.de>
1612         PR tree-optimization/59993
1613         * tree-ssa-forwprop.c (associate_pointerplus): Check we
1614         can propagate form the earlier stmt and avoid the transform
1615         when the intermediate result is needed.
1617 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
1619         * README.Portability: Fix typo.
1621 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
1623         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1624         comparison_operator with ordered_comparison_operator.
1626 2014-01-30  Nick Clifton  <nickc@redhat.com>
1628         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1629         Rename to mn10300_store_multiple_regs.
1630         * config/mn10300/mn10300.c: Likewise.
1631         * config/mn10300/mn10300.md (store_movm): Fix typo: call
1632         store_multiple_regs.
1633         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1634         Call mn10300_store_multiple_regs.
1636 2014-01-30  Nick Clifton  <nickc@redhat.com>
1637             DJ Delorie  <dj@redhat.com>
1639         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1640         %fp 2 to keep registers after it properly word-aligned.
1641         (rl78_alloc_physical_registers_umul): Handle the case where both
1642         input operands are the same.
1644 2014-01-30  Richard Biener  <rguenther@suse.de>
1646         PR tree-optimization/59903
1647         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1648         check properly.
1650 2014-01-30  Jason Merrill  <jason@redhat.com>
1652         PR c++/59633
1653         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1655         PR c++/59645
1656         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1658 2014-01-30  Richard Biener  <rguenther@suse.de>
1660         PR tree-optimization/59951
1661         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1663 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
1665         PR target/59784
1666         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1667         SFmode to DFmode case.
1669 2014-01-29  DJ Delorie  <dj@redhat.com>
1671         * config/msp430/msp430.opt (-minrt): New.
1672         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1673         if -minrt given.
1674         (ENDFILE_SPEC): Likewise.
1676 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
1678         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1679         (estimate_function_body_sizes): Use it.
1681 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
1683         PR c++/58561
1684         * dwarf2out.c (is_cxx_auto): New.
1685         (is_base_type): Use it.
1686         (gen_type_die_with_usage): Likewise.
1688 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1690         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
1691         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1692         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1693         -maltivec=be with LE targets.
1694         (vsx_mergeh_<mode>): Likewise.
1695         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1696         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1697         (altivec_vmrghb): Replace with define_expand and new
1698         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1699         (altivec_vmrghb_direct): New define_insn.
1700         (altivec_vmrghh): Replace with define_expand and new
1701         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1702         (altivec_vmrghh_direct): New define_insn.
1703         (altivec_vmrghw): Replace with define_expand and new
1704         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1705         (altivec_vmrghw_direct): New define_insn.
1706         (*altivec_vmrghsf): Adjust for endianness.
1707         (altivec_vmrglb): Replace with define_expand and new
1708         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1709         (altivec_vmrglb_direct): New define_insn.
1710         (altivec_vmrglh): Replace with define_expand and new
1711         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1712         (altivec_vmrglh_direct): New define_insn.
1713         (altivec_vmrglw): Replace with define_expand and new
1714         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1715         (altivec_vmrglw_direct): New define_insn.
1716         (*altivec_vmrglsf): Adjust for endianness.
1717         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1718         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1719         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1720         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1721         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1722         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1723         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1724         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1726 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1728         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1729         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1730         whitespace.
1732 2014-01-29  Richard Biener  <rguenther@suse.de>
1734         PR tree-optimization/58742
1735         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1736         associate_pointerplus_align.
1737         (associate_pointerplus_diff): New function.
1738         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
1739         and associate_pointerplus_diff.
1741 2014-01-29  Richard Biener  <rguenther@suse.de>
1743         * lto-streamer.h (LTO_major_version): Bump to 3.
1744         (LTO_minor_version): Reset to 0.
1746 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
1748         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1749         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1750         (arm_file_start): Generate correct asm header for armv7ve.
1751         * config/arm/bpabi.h: Add multilib support for armv7ve.
1752         * config/arm/driver-arm.c: Change the architectures of cortex-a7
1753         and cortex-a15 to armv7ve.
1754         * config/arm/t-aprofile: Add multilib support for armv7ve.
1755         * doc/invoke.texi: Document -march=armv7ve.
1757 2014-01-29  Richard Biener  <rguenther@suse.de>
1759         PR tree-optimization/58742
1760         * tree-ssa-forwprop.c (associate_plusminus): Return true
1761         if we changed sth, defer EH cleanup to ...
1762         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
1763         (simplify_mult): New function.
1765 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1767         PR middle-end/59917
1768         PR tree-optimization/59920
1769         * tree.c (build_common_builtin_nodes): Remove
1770         __builtin_setjmp_dispatcher initialization.
1771         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1772         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1773         instead of gsi_after_labels + manually skipping debug stmts.
1774         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1775         ignore bbs with IFN_ABNORMAL_DISPATCHER.
1776         * tree-inline.c (copy_edges_for_bb): Remove
1777         can_make_abnormal_goto argument, instead add abnormal_goto_dest
1778         argument.  Ignore computed_goto_p stmts.  Don't call
1779         make_abnormal_goto_edges.  If a call might need abnormal edges
1780         for non-local gotos, see if it already has an edge to
1781         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1782         with true argument, don't do anything then, otherwise add
1783         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1784         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1785         caller.
1786         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1787         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1788         (lower_stmt): Don't set data->calls_builtin_setjmp.
1789         (lower_builtin_setjmp): Adjust comment.
1790         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1791         * tree-cfg.c (found_computed_goto): Remove.
1792         (factor_computed_gotos): Remove.
1793         (make_goto_expr_edges): Return bool, true for computed gotos.
1794         Don't call make_abnormal_goto_edges.
1795         (build_gimple_cfg): Don't set found_computed_goto, don't call
1796         factor_computed_gotos.
1797         (computed_goto_p): No longer static.
1798         (make_blocks): Don't set found_computed_goto.
1799         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1800         (make_edges): If make_goto_expr_edges returns true, push bb
1801         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1802         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1803         vector.  Record mapping between bbs and OpenMP regions if there
1804         are any, adjust make_gimple_omp_edges caller.  Call
1805         handle_abnormal_edges.
1806         (make_abnormal_goto_edges): Remove.
1807         * tree-cfg.h (make_abnormal_goto_edges): Remove.
1808         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1809         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1810         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1811         * internal-fn.def (ABNORMAL_DISPATCHER): New.
1812         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1813         filling *region also set *region_idx to (*region)->entry->index.
1815         PR other/58712
1816         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1817         For REGs set ORIGINAL_REGNO.
1819 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
1821         * doc/md.texi: Mention that a target shouldn't implement
1822         vec_widen_(s|u)mul_even/odd pair if it is less efficient
1823         than hi/lo pair.
1825 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
1827         PR tree-optimization/59594
1828         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1829         a copy of the datarefs vector rather than the vector itself.
1831 2014-01-28  Jason Merrill  <jason@redhat.com>
1833         PR c++/53756
1834         * dwarf2out.c (auto_die): New static.
1835         (gen_type_die_with_usage): Handle C++1y 'auto'.
1836         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1837         on definition.
1839 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
1841         PR target/59672
1842         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1843         (SPEC_X32): Likewise.
1844         (SPEC_64): Likewise.
1845         * config/i386/i386.c (ix86_option_override_internal): Turn off
1846         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1847         for TARGET_16BIT.
1848         (x86_file_start): Output .code16gcc for TARGET_16BIT.
1849         * config/i386/i386.h (TARGET_16BIT): New macro.
1850         (TARGET_16BIT_P): Likewise.
1851         * config/i386/i386.opt: Add m16.
1852         * doc/invoke.texi: Document -m16.
1854 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
1856         PR preprocessor/59935
1857         * input.c (location_get_source_line): Bail out on when line number
1858         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1860 2014-01-28  Richard Biener  <rguenther@suse.de>
1862         PR tree-optimization/58742
1863         * tree-ssa-forwprop.c (associate_plusminus): Handle
1864         pointer subtraction of the form (T)(P + A) - (T)P.
1866 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1868         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1869         at const_int_cost.
1871 2014-01-28  Richard Biener  <rguenther@suse.de>
1873         Revert
1874         2014-01-28  Richard Biener  <rguenther@suse.de>
1876         PR rtl-optimization/45364
1877         PR rtl-optimization/59890
1878         * var-tracking.c (local_get_addr_clear_given_value): Handle
1879         already cleared slot.
1880         (val_reset): Handle not allocated local_get_addr_cache.
1881         (vt_find_locations): Use post-order on the inverted CFG.
1883 2014-01-28  Richard Biener  <rguenther@suse.de>
1885         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1887 2014-01-28  Richard Biener  <rguenther@suse.de>
1889         PR rtl-optimization/45364
1890         PR rtl-optimization/59890
1891         * var-tracking.c (local_get_addr_clear_given_value): Handle
1892         already cleared slot.
1893         (val_reset): Handle not allocated local_get_addr_cache.
1894         (vt_find_locations): Use post-order on the inverted CFG.
1896 2014-01-28  Alan Modra  <amodra@gmail.com>
1898         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1899         * configure.ac <recursive call for build != host>: Define
1900         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1901         and LD_FOR_BUILD too.
1902         * configure: Regenerate.
1904 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
1906         * config/i386/i386.c (get_builtin_code_for_version): Separate
1907         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1908         Broadwell from Haswell.
1910 2014-01-27  Steve Ellcey  <sellcey@mips.com>
1912         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1913         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1914         * config/mips/mips.c (mips_option_override): Change setting
1915         of TARGET_DSP.
1916         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1917         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1918         Change from Mask to Var.
1920 2014-01-27  Jeff Law  <law@redhat.com>
1922         * ipa-inline.c (inline_small_functions): Fix typo.
1924 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
1926         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1927         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1928         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1929         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1930         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1931         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1932         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1933         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1934         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1935         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1936         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1937         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1938         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1939         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1940         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1941         (_mm512_storeu_epi64): Ditto.
1942         (_mm512_cmpge_epi32_mask): Ditto.
1943         (_mm512_cmpge_epu32_mask): Ditto.
1944         (_mm512_cmpge_epi64_mask): Ditto.
1945         (_mm512_cmpge_epu64_mask): Ditto.
1946         (_mm512_cmple_epi32_mask): Ditto.
1947         (_mm512_cmple_epu32_mask): Ditto.
1948         (_mm512_cmple_epi64_mask): Ditto.
1949         (_mm512_cmple_epu64_mask): Ditto.
1950         (_mm512_cmplt_epi32_mask): Ditto.
1951         (_mm512_cmplt_epu32_mask): Ditto.
1952         (_mm512_cmplt_epi64_mask): Ditto.
1953         (_mm512_cmplt_epu64_mask): Ditto.
1954         (_mm512_cmpneq_epi32_mask): Ditto.
1955         (_mm512_cmpneq_epu32_mask): Ditto.
1956         (_mm512_cmpneq_epi64_mask): Ditto.
1957         (_mm512_cmpneq_epu64_mask): Ditto.
1958         (_mm512_expand_pd): Ditto.
1959         (_mm512_expand_ps): Ditto.
1960         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1961         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1962         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1963         * config/i386/i386.c (ix86_builtins): Add
1964         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1965         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1966         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1967         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1968         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1969         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1970         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1971         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1972         IX86_BUILTIN_PMOVUSQW512_MEM.
1973         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1974         __builtin_ia32_pmovsqd512mem_mask,
1975         __builtin_ia32_pmovqd512mem_mask,
1976         __builtin_ia32_pmovusqw512mem_mask,
1977         __builtin_ia32_pmovsqw512mem_mask,
1978         __builtin_ia32_pmovqw512mem_mask,
1979         __builtin_ia32_pmovusdw512mem_mask,
1980         __builtin_ia32_pmovsdw512mem_mask,
1981         __builtin_ia32_pmovdw512mem_mask,
1982         __builtin_ia32_pmovqb512mem_mask,
1983         __builtin_ia32_pmovusqb512mem_mask,
1984         __builtin_ia32_pmovsqb512mem_mask,
1985         __builtin_ia32_pmovusdb512mem_mask,
1986         __builtin_ia32_pmovsdb512mem_mask,
1987         __builtin_ia32_pmovdb512mem_mask.
1988         (bdesc_args): Add __builtin_ia32_expanddf512,
1989         __builtin_ia32_expandsf512.
1990         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1991         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1992         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1993         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1994         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1995         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1996         (avx512f_<code>v8div16qi2_mask_store): This.
1997         (avx512f_expand<mode>): New.
1999 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
2001         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2002         New.
2003         (_mm512_mask_prefetch_i64gather_pd): Ditto.
2004         (_mm512_prefetch_i32scatter_pd): Ditto.
2005         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2006         (_mm512_prefetch_i64scatter_pd): Ditto.
2007         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2008         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2009         (_mm512_mask_prefetch_i64gather_ps): Ditto.
2010         (_mm512_prefetch_i32scatter_ps): Ditto.
2011         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2012         (_mm512_prefetch_i64scatter_ps): Ditto.
2013         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2014         * config/i386/i386-builtin-types.def: Define
2015         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2016         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2017         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2018         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2019         IX86_BUILTIN_SCATTERPFQPD.
2020         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2021         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2022         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2023         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2024         __builtin_ia32_scatterpfqps.
2025         (ix86_expand_builtin): Expand new built-ins.
2026         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2027         fix memory access data type.
2028         (*avx512pf_gatherpf<mode>_mask): Ditto.
2029         (*avx512pf_gatherpf<mode>): Ditto.
2030         (avx512pf_scatterpf<mode>): Ditto.
2031         (*avx512pf_scatterpf<mode>_mask): Ditto.
2032         (*avx512pf_scatterpf<mode>): Ditto.
2033         (GATHER_SCATTER_SF_MEM_MODE): New.
2034         (avx512pf_gatherpf<mode>df): Ditto.
2035         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2036         (*avx512pf_scatterpf<mode>df): Ditto.
2038 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
2040         PR bootstrap/59934
2041         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2042         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2043         reached.
2045 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2047         * common/config/arm/arm-common.c
2048         (arm_rewrite_mcpu): Handle multiple names.
2049         * config/arm/arm.h
2050         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2052 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2054         * gimple-builder.h (create_gimple_tmp): Delete.
2056 2014-01-27  Christian Bruel  <christian.bruel@st.com>
2058         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2059         words comparisons.
2061 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
2063         * config/pa/pa.md (call): Generate indirect long calls to non-local
2064         functions when outputing 32-bit code.
2065         (call_value): Likewise except for special call to buggy powf function.
2067         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2068         portable runtime and PIC indirect calls.
2069         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2070         and PIC call sequences.  Use ldo instead of blr to set return register
2071         in PIC call sequence.
2073 2014-01-25  Walter Lee  <walt@tilera.com>
2075         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2076         avoid clobbering a live register.
2078 2014-01-25  Walter Lee  <walt@tilera.com>
2080         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2081         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2082         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2083         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2085 2014-01-25  Walter Lee  <walt@tilera.com>
2087         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2088         arguments on even registers.
2089         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2090         STACK_BOUNDARY.
2091         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2092         (BIGGEST_ALIGNMENT): Ditto.
2093         (BIGGEST_FIELD_ALIGNMENT): Ditto.
2095 2014-01-25  Walter Lee  <walt@tilera.com>
2097         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2098         insns before bundling.
2099         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2101 2014-01-25  Walter Lee  <walt@tilera.com>
2103         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2104         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2105         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2107 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2109         * config/mips/constraints.md (kl): Delete.
2110         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2111         define expands, using...
2112         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2113         instructions for MIPS16.
2114         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2115         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2117 2014-01-25  Walter Lee  <walt@tilera.com>
2119         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2120         (clzdi2): Ditto.
2121         (ffsdi2): Ditto.
2123 2014-01-25  Walter Lee  <walt@tilera.com>
2125         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2126         (TARGET_EXPAND_TO_RTL_HOOK): Define.
2128 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2130         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2131         Handle XOR.
2133 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
2135         * print-rtl.c (in_call_function_usage): New var.
2136         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2137         EXPR_LIST mode as mode and not as reg note name.
2139         PR middle-end/59561
2140         * cfgloopmanip.c (copy_loop_info): If
2141         loop->warned_aggressive_loop_optimizations, make sure
2142         the flag is set in target loop too.
2144 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
2146         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2147         flag_cilkplus.
2148         * builtins.def: Likewise.
2149         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2150         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2151         * ira.c (ira_setup_eliminable_regset): Likewise.
2152         * omp-low.c (gate_expand_omp): Likewise.
2153         (execute_lower_omp): Likewise.
2154         (diagnose_sb_0): Likewise.
2155         (gate_diagnose_omp_blocks): Likewise.
2156         (simd_clone_clauses_extract): Likewise.
2157         (gate): Likewise.
2159 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2161         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2162         correction for little endian...
2163         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2164         here.
2166 2014-01-24  Jeff Law  <law@redhat.com>
2168         PR tree-optimization/59919
2169         * tree-vrp.c (find_assert_locations_1): Do not register asserts
2170         for non-returning calls.
2172 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
2174         * common/config/aarch64/aarch64-common.c
2175         (aarch64_rewrite_mcpu): Handle multiple names.
2176         * config/aarch64/aarch64.h
2177         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2179 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
2181         * input.c (add_file_to_cache_tab): Handle the case where fopen
2182         returns NULL.
2184 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
2186         PR target/59929
2187         * config/i386/i386.md (pushsf splitter): Get stack adjustment
2188         from push operand if code of push isn't PRE_DEC.
2190 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
2192         PR target/59909
2193         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2194         -mquad-memory-atomic.  Update -mquad-memory documentation to say
2195         it is only used for non-atomic loads/stores.
2197         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2198         -mquad-memory or -mquad-memory-atomic switches.
2200         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2201         -mquad-memory-atomic to ISA 2.07 support.
2203         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2204         to separate support of normal quad word memory operations (ldq, stq)
2205         from the atomic quad word memory operations.
2207         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2208         support to separate non-atomic quad word operations from atomic
2209         quad word operations.  Disable non-atomic quad word operations in
2210         little endian mode so that we don't have to swap words after the
2211         load and before the store.
2212         (quad_load_store_p): Add comment about atomic quad word support.
2213         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2214         options printed with -mdebug=reg.
2216         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2217         -mquad-memory-atomic as the test for whether we have quad word
2218         atomic instructions.
2219         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2220         or -mp8-vector are used, allow byte/half-word atomic operations.
2222         * config/rs6000/sync.md (load_lockedti): Insure that the address
2223         is a proper indexed or indirect address for the lqarx instruction.
2224         On little endian systems, swap the hi/lo registers after the lqarx
2225         instruction.
2226         (load_lockedpti): Use indexed_or_indirect_operand predicate to
2227         insure the address is valid for the lqarx instruction.
2228         (store_conditionalti): Insure that the address is a proper indexed
2229         or indirect address for the stqcrx. instruction.  On little endian
2230         systems, swap the hi/lo registers before doing the stqcrx.
2231         instruction.
2232         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2233         insure the address is valid for the stqcrx. instruction.
2235         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2236         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2237         type of quad memory support is available.
2239 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
2241         PR regression/59915
2242         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2243         there is a danger of looping.
2245 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
2247         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2248         force flag_ira_loop_pressure if set via command line.
2250 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2252         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2253         (ashr_simd): New builtin handling DI mode.
2254         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2255         (aarch64_sshr_simddi): New match pattern.
2256         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2257         (vshrd_n_s64): Likewise.
2258         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2260 2014-01-23  Nick Clifton  <nickc@redhat.com>
2262         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2263         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2264         favour of mcu specific scripts.
2265         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2266         430x multilibs.
2268 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2269             Alex Velenko  <Alex.Velenko@arm.com>
2271         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2272         (vaddv_s16): Likewise.
2273         (vaddv_s32): Likewise.
2274         (vaddv_u8): Likewise.
2275         (vaddv_u16): Likewise.
2276         (vaddv_u32): Likewise.
2277         (vaddvq_s8): Likewise.
2278         (vaddvq_s16): Likewise.
2279         (vaddvq_s32): Likewise.
2280         (vaddvq_s64): Likewise.
2281         (vaddvq_u8): Likewise.
2282         (vaddvq_u16): Likewise.
2283         (vaddvq_u32): Likewise.
2284         (vaddvq_u64): Likewise.
2285         (vaddv_f32): Likewise.
2286         (vaddvq_f32): Likewise.
2287         (vaddvq_f64): Likewise.
2288         (vmaxv_f32): Likewise.
2289         (vmaxv_s8): Likewise.
2290         (vmaxv_s16): Likewise.
2291         (vmaxv_s32): Likewise.
2292         (vmaxv_u8): Likewise.
2293         (vmaxv_u16): Likewise.
2294         (vmaxv_u32): Likewise.
2295         (vmaxvq_f32): Likewise.
2296         (vmaxvq_f64): Likewise.
2297         (vmaxvq_s8): Likewise.
2298         (vmaxvq_s16): Likewise.
2299         (vmaxvq_s32): Likewise.
2300         (vmaxvq_u8): Likewise.
2301         (vmaxvq_u16): Likewise.
2302         (vmaxvq_u32): Likewise.
2303         (vmaxnmv_f32): Likewise.
2304         (vmaxnmvq_f32): Likewise.
2305         (vmaxnmvq_f64): Likewise.
2306         (vminv_f32): Likewise.
2307         (vminv_s8): Likewise.
2308         (vminv_s16): Likewise.
2309         (vminv_s32): Likewise.
2310         (vminv_u8): Likewise.
2311         (vminv_u16): Likewise.
2312         (vminv_u32): Likewise.
2313         (vminvq_f32): Likewise.
2314         (vminvq_f64): Likewise.
2315         (vminvq_s8): Likewise.
2316         (vminvq_s16): Likewise.
2317         (vminvq_s32): Likewise.
2318         (vminvq_u8): Likewise.
2319         (vminvq_u16): Likewise.
2320         (vminvq_u32): Likewise.
2321         (vminnmv_f32): Likewise.
2322         (vminnmvq_f32): Likewise.
2323         (vminnmvq_f64): Likewise.
2325 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2327         * config/aarch64/aarch64-simd.md
2328         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2329         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2330         (*aarch64_mul3_elt<mode>): Likewise.
2331         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2332         (*aarch64_mul3_elt_to_64v2df): Likewise.
2333         (*aarch64_mla_elt<mode>): Likewise.
2334         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2335         (*aarch64_mls_elt<mode>): Likewise.
2336         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2337         (*aarch64_fma4_elt<mode>): Likewise.
2338         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2339         (*aarch64_fma4_elt_to_64v2df): Likewise.
2340         (*aarch64_fnma4_elt<mode>): Likewise.
2341         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2342         (*aarch64_fnma4_elt_to_64v2df): Likewise.
2343         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2344         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2345         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2346         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2347         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2348         (aarch64_sqdmull_lane<mode>_internal): Likewise.
2349         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2351 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2353         * config/aarch64/aarch64-simd.md
2354         (aarch64_be_checked_get_lane<mode>): New define_expand.
2355         * config/aarch64/aarch64-simd-builtins.def
2356         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2357         New builtin definition.
2358         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2359         Use new safe be builtin.
2361 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2363         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2364         New define_insn.
2365         (aarch64_be_st1<mode>): Likewise.
2366         (aarch_ld1<VALL:mode>): Define_expand modified.
2367         (aarch_st1<VALL:mode>): Likewise.
2368         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2369         (UNSPEC_ST1): Likewise.
2371 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
2373         * config/microblaze/microblaze.md: Add trap insn and attribute
2375 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
2377         PR preprocessor/58580
2378         * input.h (location_get_source_line): Take an additional line_size
2379         parameter.
2380         (void diagnostics_file_cache_fini): Declare new function.
2381         * input.c (struct fcache): New type.
2382         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2383         New static constants.
2384         (diagnostic_file_cache_init, total_lines_num)
2385         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2386         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2387         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2388         (get_next_line, read_next_line, goto_next_line, read_line_num):
2389         New static function definitions.
2390         (diagnostic_file_cache_fini): New function.
2391         (location_get_source_line): Take an additional output line_len
2392         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
2393         read_line_num.
2394         * diagnostic.c (diagnostic_finish): Call
2395         diagnostic_file_cache_fini.
2396         (adjust_line): Take an additional input parameter for the length
2397         of the line, rather than calculating it with strlen.
2398         (diagnostic_show_locus): Adjust the use of
2399         location_get_source_line and adjust_line with respect to their new
2400         signature.  While displaying a line now, do not stop at the first
2401         null byte.  Rather, display the zero byte as a space and keep
2402         going until we reach the size of the line.
2403         * Makefile.in: Add vec.o to OBJS-libcommon
2405 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2406             Ilya Tocar     <ilya.tocar@intel.com>
2408         * config/i386/avx512fintrin.h (_mm512_kmov): New.
2409         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2410         (__builtin_ia32_kmov16): Ditto.
2411         * config/i386/i386.md (UNSPEC_KMOV): New.
2412         (kmovw): Ditto.
2414 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
2416         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2417         (_mm512_storeu_si512): Ditto.
2419 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
2421         PR target/52125
2422         * rtl.h (get_referenced_operands): Declare.
2423         * recog.c (get_referenced_operands): New function.
2424         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2425         operands have been referenced when recording LO_SUM references.
2427 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
2429         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2431 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2433         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2434         Enable for generic and recent AMD targets.
2436 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
2438         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2439         ARG_SIZE note when adjustment was eliminated.
2441 2014-01-22  Jeff Law  <law@redhat.com>
2443         PR tree-optimization/59597
2444         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2445         in file.  Accept new argument REGISTERING and use it to modify
2446         dump output appropriately.
2447         (register_jump_thread): Corresponding changes.
2448         (mark_threaded_blocks): Reinstate code to cancel unprofitable
2449         thread paths involving joiner blocks.  Add code to dump cancelled
2450         jump threading paths.
2452 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
2454         PR rtl-optimization/59477
2455         * lra-constraints.c (inherit_in_ebb): Process call for living hard
2456         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
2458 2014-01-22  Tom Tromey  <tromey@redhat.com>
2460         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2461         PARAMS.
2462         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2464 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2466         PR rtl-optimization/59896
2467         * lra-constraints.c (process_alt_operands): Check unused note for
2468         matched operands of insn with no output reloads.
2470 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
2472         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2473         (mips_move_from_gpr_cost): Likewise.
2475 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
2477         PR rtl-optimization/59858
2478         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2479         ira_class_hard_regs_num.
2480         (process_alt_operands): Increase reject for dying matched operand.
2482 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
2484         PR target/59003
2485         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2486         smaller than size, perform several stores or loads and stores
2487         at dst + count - size to store or copy all of size bytes, rather
2488         than just last modesize bytes.
2490 2014-01-20  DJ Delorie  <dj@redhat.com>
2492         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2493         that CLOBBERs are REGs before propogating their values.
2495 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
2497         PR middle-end/59789
2498         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2499         (cgraph_inline_failed_type): New function.
2500         * cgraph.h (DEFCIFCODE): Add type.
2501         (cgraph_inline_failed_type_t): New enum.
2502         (cgraph_inline_failed_type): New prototype.
2503         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2504         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2505         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2506         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2507         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2508         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2509         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2510         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2511         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2512         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2513         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2514         OPTIMIZATION_MISMATCH.
2515         * tree-inline.c (expand_call_inline): Emit errors during
2516         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2518 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2520         PR target/59685
2521         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2522         mode attribute in insn output.
2524 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
2526         * output.h (output_constant): Delete.
2527         * varasm.c (output_constant): Make private.
2529 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
2531         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2533 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2535         PR middle-end/59860
2536         * tree.h (fold_builtin_strcat): New prototype.
2537         * builtins.c (fold_builtin_strcat): No longer static.  Add len
2538         argument, if non-NULL, don't call c_strlen.  Optimize
2539         directly into __builtin_memcpy instead of __builtin_strcpy.
2540         (fold_builtin_2): Adjust fold_builtin_strcat caller.
2541         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2543 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
2545         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2546         for SImode_address_operand operands, having only a REG argument.
2548 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2550         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2551         loader name using mbig-endian.
2552         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2554 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2556         * doc/invoke.texi (-march): Clarify documentation for AArch64.
2557         (-mtune): Likewise.
2558         (-mcpu): Likewise.
2560 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
2562         * config/aarch64/aarch64-protos.h
2563         (aarch64_cannot_change_mode_class_ptr): Declare.
2564         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2565         aarch64_cannot_change_mode_class_ptr): New.
2566         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2567         backend hook aarch64_cannot_change_mode_class.
2569 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
2571         * common/config/aarch64/aarch64-common.c
2572         (aarch64_handle_option): Don't handle any option order logic here.
2573         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2574         selected_cpu, warn on architecture version mismatch.
2575         (aarch64_override_options): Fix parsing order for option strings.
2577 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2578             Iain Sandoe  <iain@codesourcery.com>
2580         PR bootstrap/59496
2581         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2582         warning.  Amend comment to reflect current functionality.
2584 2014-01-20  Richard Biener  <rguenther@suse.de>
2586         PR middle-end/59860
2587         * builtins.c (fold_builtin_strcat): Remove case better handled
2588         by tree-ssa-strlen.c.
2590 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
2592         * config/aarch64/aarch64.opt
2593         (mcpu, march, mtune): Make case-insensitive.
2595 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
2597         PR target/59880
2598         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2599         if operands[1] is a REG or ZERO_EXTEND of a REG.
2601 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
2603         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2605 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
2607         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2608         long non-pic millicode calls.
2610 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2612         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2614 2014-01-19  Kito Cheng  <kito@0xlab.org>
2616         * builtins.c (expand_movstr): Check movstr expand done or fail.
2618 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2619             H.J. Lu  <hongjiu.lu@intel.com>
2621         PR target/59379
2622         * config/i386/i386.md (*lea<mode>): Zero-extend return register
2623         to DImode for zero-extended addresses.
2625 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
2627         PR rtl-optimization/57763
2628         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2629         on the new indirect jump_insn and increment LABEL_NUSES (label).
2631 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
2633         PR bootstrap/59580
2634         PR bootstrap/59583
2635         * config.gcc (x86_archs): New variable.
2636         (x86_64_archs): Likewise.
2637         (x86_cpus): Likewise.
2638         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2639         --with-arch/--with-cpu= options.
2640         Support --with-arch=/--with-cpu={nehalem,westmere,
2641         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2643 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2645         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2646         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
2648 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
2650         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2652 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
2654         PR target/58944
2655         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2656         clear cpp_get_options (parse_in)->warn_unused_macros for
2657         ix86_target_macros_internal with cpp_define.
2659 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
2661         * jump.c (delete_related_insns): Keep (use (insn))s.
2662         * reorg.c (redundant_insn): Check for barriers too.
2664 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2666         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2668 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
2670         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2671         call to $$dyncall when TARGET_LONG_CALLS is true.
2673 2014-01-17  Jeff Law  <law@redhat.com>
2675         * ree.c (combine_set_extension): Temporarily disable test for
2676         changing number of hard registers.
2678 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2680         PR middle-end/58125
2681         * ipa-inline-analysis.c (inline_free_summary):
2682         Do not free summary of aliases.
2684 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2686         PR middle-end/59706
2687         * gimplify.c (gimplify_expr): Use create_tmp_var
2688         instead of create_tmp_var_raw.  If cond doesn't have
2689         integral type, don't add the IFN_ANNOTATE builtin at all.
2691 2014-01-17  Martin Jambor  <mjambor@suse.cz>
2693         PR ipa/59736
2694         * ipa-cp.c (prev_edge_clone): New variable.
2695         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2696         Also resize prev_edge_clone vector.
2697         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2698         (ipcp_edge_removal_hook): New function.
2699         (ipcp_driver): Register ipcp_edge_removal_hook.
2701 2014-01-17  Andrew Pinski <apinski@cavium.com>
2702             Steve Ellcey  <sellcey@mips.com>
2704         PR target/59462
2705         * config/mips/mips.c (mips_print_operand): Check operand mode instead
2706         of operator mode.
2708 2014-01-17  Jeff Law  <law@redhat.com>
2710         PR middle-end/57904
2711         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2712         so that pass_ccp runs first.
2714 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2716         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2717         (ix86_adjust_cost): Use !TARGET_XXX.
2718         (do_reorder_for_imul): Likewise.
2719         (swap_top_of_ready_list): Likewise.
2720         (ix86_sched_reorder): Likewise.
2722 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
2724         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2725         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
2726         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
2727         (intel_memset): New.  Duplicate slm_memset.
2728         (intel_cost): New.  Duplicate slm_cost.
2729         (m_INTEL): New macro.
2730         (processor_target_table): Add "intel".
2731         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2732         with PROCESSOR_INTEL for "intel".
2733         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
2734         PROCESSOR_SILVERMONT.
2735         (ix86_issue_rate): Likewise.
2736         (ix86_adjust_cost): Likewise.
2737         (ia32_multipass_dfa_lookahead): Likewise.
2738         (swap_top_of_ready_list): Likewise.
2739         (ix86_sched_reorder): Likewise.
2740         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2741         instead of TARGET_OPT_AGU.
2742         * config/i386/i386.h (TARGET_INTEL): New.
2743         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2744         (processor_type): Add PROCESSOR_INTEL.
2745         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2746         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2748 2014-01-17  Marek Polacek  <polacek@redhat.com>
2750         PR c/58346
2751         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2752         size is zero.
2754 2014-01-17  Richard Biener  <rguenther@suse.de>
2756         PR tree-optimization/46590
2757         * opts.c (default_options_table): Add entries for
2758         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2759         all enabled at -O1 but not for -Og.
2760         * common.opt (fbranch-count-reg): Remove Init(1).
2761         (fmove-loop-invariants): Likewise.
2762         (ftree-pta): Likewise.
2764 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
2766         * config/i386/i386.c (ix86_data_alignment): For compatibility with
2767         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2768         decls to at least the GCC 4.8 used alignments.
2770         PR fortran/59440
2771         * tree-nested.c (convert_nonlocal_reference_stmt,
2772         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2773         of GIMPLE_BIND stmts, adjust associated decls.
2775 2014-01-17  Richard Biener  <rguenther@suse.de>
2777         PR tree-optimization/46590
2778         * vec.h (vec<>::bseach): New member function implementing
2779         binary search according to C89 bsearch.
2780         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2781         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2782         bitmap pointer again.  Make accesses_in_loop a flat array.
2783         (mem_ref_obstack): New global.
2784         (outermost_indep_loop): Adjust for mem_ref->stored changes.
2785         (mark_ref_stored): Likewise.
2786         (ref_indep_loop_p_2): Likewise.
2787         (set_ref_stored_in_loop): New helper function.
2788         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2789         (memref_free): Adjust.
2790         (record_mem_ref_loc): Simplify.
2791         (gather_mem_refs_stmt): Adjust.
2792         (sort_locs_in_loop_postorder_cmp): New function.
2793         (analyze_memory_references): Sort accesses_in_loop after
2794         loop postorder number.
2795         (find_ref_loc_in_loop_cmp): New function.
2796         (for_all_locs_in_loop): Find relevant cluster of locs in
2797         accesses_in_loop and iterate without recursion.
2798         (execute_sm): Avoid uninit warning.
2799         (struct ref_always_accessed): Simplify.
2800         (ref_always_accessed::operator ()): Likewise.
2801         (ref_always_accessed_p): Likewise.
2802         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2803         loop postorder numbers here.
2804         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2805         numbers.
2807 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
2809         PR c++/57945
2810         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2811         on decls for which assemble_alias has been called.
2813 2014-01-17  Nick Clifton  <nickc@redhat.com>
2815         * config/msp430/msp430.opt: (mcpu): New option.
2816         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2817         (msp430_option_override): Parse target_cpu.  If the MCU name
2818         matches a generic string, clear target_mcu.
2819         (msp430_attr): Allow numeric interrupt values up to 63.
2820         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2821         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2822         option.
2823         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2824         Add mcpu matches.
2825         * config/msp430/msp430.md (popm): Use %J rather than %I.
2826         (addsi3): Use msp430_nonimmediate_operand for operand 2.
2827         (addhi_cy_i): Use immediate_operand for operand 2.
2828         * doc/invoke.texi: Document -mcpu option.
2830 2014-01-17  Richard Biener  <rguenther@suse.de>
2832         PR rtl-optimization/38518
2833         * df.h (df_analyze_loop): Declare.
2834         * df-core.c: Include cfgloop.h.
2835         (df_analyze_1): Split out main part of df_analyze.
2836         (df_analyze): Adjust.
2837         (loop_inverted_post_order_compute): New function.
2838         (loop_post_order_compute): Likewise.
2839         (df_analyze_loop): New function avoiding whole-function
2840         postorder computes.
2841         * loop-invariant.c (find_defs): Use df_analyze_loop.
2842         (find_invariants): Adjust.
2843         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2845 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
2847         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2848         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2850 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
2852         * ipa-ref.c (ipa_remove_stmt_references): Fix references
2853         traversal when removing references.
2855 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
2857         PR ipa/59775
2858         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2860 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
2862         PR middle-end/56791
2863         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2864         pushing a reload for an autoinc when we had previously reloaded an
2865         inner part of the address.
2867 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
2869         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2870         field.
2871         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2872         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2873         when not giving up or versioning for alias only because of
2874         loop->safelen.
2875         (vect_analyze_data_ref_dependences): Set to true.
2876         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2877         is a GIMPLE_PHI.
2878         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2879         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2880         to the condition.
2882         PR middle-end/58344
2883         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2885         PR target/59839
2886         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2887         operand 0 predicate for gathers, use a new pseudo as subtarget.
2889 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2891         PR middle-end/59609
2892         * lra-constraints.c (process_alt_operands): Add printing debug info.
2893         Check absence of input/output reloads for matched operands too.
2895 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
2897         PR rtl-optimization/59835
2898         * ira.c (ira_init_register_move_cost): Increase cost for
2899         impossible modes.
2901 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
2903         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2905 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
2907         PR target/59780
2908         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2909         non-register objects.  Use gen_(high/low)part more consistently.
2910         Fix assertions.
2912 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
2914         PR target/59844
2915         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2916         endian support, remove tests for WORDS_BIG_ENDIAN.
2917         (p8_mfvsrd_3_<mode>): Likewise.
2918         (reload_gpr_from_vsx<mode>): Likewise.
2919         (reload_gpr_from_vsxsf): Likewise.
2920         (p8_mfvsrd_4_disf): Likewise.
2922 2014-01-16  Richard Biener  <rguenther@suse.de>
2924         PR rtl-optimization/46590
2925         * lcm.c (compute_antinout_edge): Use postorder iteration.
2926         (compute_laterin): Use inverted postorder iteration.
2928 2014-01-16  Nick Clifton  <nickc@redhat.com>
2930         PR middle-end/28865
2931         * varasm.c (output_constant): Return the number of bytes actually
2932         emitted.
2933         (output_constructor_array_range): Update the field size with the
2934         number of bytes emitted by output_constant.
2935         (output_constructor_regular_field): Likewise.  Also do not
2936         complain if the total number of bytes emitted is now greater
2937         than the expected fieldpos.
2938         * output.h (output_constant): Update prototype and descriptive comment.
2940 2014-01-16  Marek Polacek  <polacek@redhat.com>
2942         PR middle-end/59827
2943         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2944         it is error_mark_node.
2946 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
2948         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2949         VALID_AVX256_REG_OR_OI_MODE.
2951 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
2953         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2954         current procedure should be profiled.
2956 2014-01-15  Andrew Pinski  <apinski@cavium.com>
2958         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2959         of moving from/to the STACK_REG register class.
2961 2014-01-15  Richard Henderson  <rth@redhat.com>
2963         PR debug/54694
2964         * reginfo.c (global_regs_decl): Globalize.
2965         * rtl.h (global_regs_decl): Declare.
2966         * ira.c (do_reload): Diagnose frame_pointer_needed and it
2967         reserved via global_regs.
2969 2014-01-15  Teresa Johnson  <tejohnson@google.com>
2971         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2973 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
2975         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2976         and vmulosh rather than call gen_vec_widen_smult_*.
2977         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2978         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2979         (vec_widen_smult_even_v16qi): Likewise.
2980         (vec_widen_umult_even_v8hi): Likewise.
2981         (vec_widen_smult_even_v8hi): Likewise.
2982         (vec_widen_umult_odd_v16qi): Likewise.
2983         (vec_widen_smult_odd_v16qi): Likewise.
2984         (vec_widen_umult_odd_v8hi): Likewise.
2985         (vec_widen_smult_odd_v8hi): Likewise.
2986         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2987         vmuloub rather than call gen_vec_widen_umult_*.
2988         (vec_widen_umult_lo_v16qi): Likewise.
2989         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2990         vmulosb rather than call gen_vec_widen_smult_*.
2991         (vec_widen_smult_lo_v16qi): Likewise.
2992         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2993         rather than call gen_vec_widen_umult_*.
2994         (vec_widen_umult_lo_v8hi): Likewise.
2995         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2996         rather than call gen_vec_widen_smult_*.
2997         (vec_widen_smult_lo_v8hi): Likewise.
2999 2014-01-15  Jeff Law  <law@redhat.com>
3001         PR tree-optimization/59747
3002         * ree.c (find_and_remove_re): Properly handle case where a second
3003         eliminated extension requires widening a copy created for elimination
3004         of a prior extension.
3005         (combine_set_extension): Ensure that the number of hard regs needed
3006         for a destination register does not change when we widen it.
3008 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3010         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3011         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3012         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3013         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3014         (avr-*-rtems*): Likewise.
3015         (bfin*-rtems*): Likewise.
3016         (moxie-*-rtems*): Likewise.
3017         (h8300-*-rtems*): Likewise.
3018         (i[34567]86-*-rtems*): Likewise.
3019         (lm32-*-rtems*): Likewise.
3020         (m32r-*-rtems*): Likewise.
3021         (m68k-*-rtems*): Likewise.
3022         (microblaze*-*-rtems*): Likewise.
3023         (mips*-*-rtems*): Likewise.
3024         (powerpc-*-rtems*): Likewise.
3025         (sh-*-rtems*): Likewise.
3026         (sparc-*-rtems*): Likewise.
3027         (sparc64-*-rtems*): Likewise.
3028         (v850-*-rtems*): Likewise.
3029         (m32c-*-rtems*): Likewise.
3031 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
3033         PR rtl-optimization/59511
3034         * ira.c (ira_init_register_move_cost): Use memory costs for some
3035         cases of register move cost calculations.
3036         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3037         instead of BB frequency.
3038         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3039         * lra-assigns.c (find_hard_regno_for): Ditto.
3041 2014-01-15  Richard Biener  <rguenther@suse.de>
3043         PR tree-optimization/59822
3044         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3045         (vectorizable_load): Use it to hoist defs of uses of invariant
3046         loads out of the loop.
3048 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
3049             Kugan Vivekanandarajah  <kuganv@linaro.org>
3051         PR target/59695
3052         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3053         truncation.
3055 2014-01-15  Richard Biener  <rguenther@suse.de>
3057         PR rtl-optimization/59802
3058         * lcm.c (compute_available): Use inverted postorder to seed
3059         the initial worklist.
3061 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3063         PR target/59803
3064         * config/s390/s390.c (s390_preferred_reload_class): Don't return
3065         ADDR_REGS for invalid symrefs in non-PIC code.
3067 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
3069         PR other/58712
3070         * builtins.c (determine_block_size): Initialize *probable_max_size
3071         even if len_rtx is CONST_INT.
3073 2014-01-14  Andrew Pinski  <apinski@cavium.com>
3075         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3076         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3077         (cortexa53_tunings): Likewise.
3078         (aarch64_sched_issue_rate): New function.
3079         (TARGET_SCHED_ISSUE_RATE): Define.
3081 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3083         * ira-costs.c (find_costs_and_classes): Add missed
3084         ira_init_register_move_cost_if_necessary.
3086 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3088         PR target/59787
3089         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3091 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
3093         PR target/59794
3094         * config/i386/i386.c (type_natural_mode): Add a bool parameter
3095         to indicate if type is used for function return value.  Warn ABI
3096         change if the vector mode isn't available for function return value.
3097         (ix86_function_arg_advance): Pass false to type_natural_mode.
3098         (ix86_function_arg): Likewise.
3099         (ix86_gimplify_va_arg): Likewise.
3100         (function_arg_32): Don't warn ABI change.
3101         (ix86_function_value): Pass true to type_natural_mode.
3102         (ix86_return_in_memory): Likewise.
3103         (ix86_struct_value_rtx): Removed.
3104         (TARGET_STRUCT_VALUE_RTX): Likewise.
3106 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3108         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3109         converting a conditional jump into a conditional return.
3111 2014-01-14  Richard Biener  <rguenther@suse.de>
3113         PR tree-optimization/58921
3114         PR tree-optimization/59006
3115         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3116         hoisting invariant stmts.
3117         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3118         invariant loads on the preheader edge if possible.
3120 2014-01-14  Joey Ye  <joey.ye@arm.com>
3122         * doc/plugin.texi (Building GCC plugins): Update to C++.
3124 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
3126         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3127         (_mm_rcp28_round_ss): Ditto.
3128         (_mm_rsqrt28_round_sd): Ditto.
3129         (_mm_rsqrt28_round_ss): Ditto.
3130         (_mm_rcp28_sd): Ditto.
3131         (_mm_rcp28_ss): Ditto.
3132         (_mm_rsqrt28_sd): Ditto.
3133         (_mm_rsqrt28_ss): Ditto.
3134         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3135         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3136         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3137         (IX86_BUILTIN_RCP28SD): Ditto.
3138         (IX86_BUILTIN_RCP28SS): Ditto.
3139         (IX86_BUILTIN_RSQRT28SD): Ditto.
3140         (IX86_BUILTIN_RSQRT28SS): Ditto.
3141         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3142         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3143         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3144         (ix86_expand_special_args_builtin): Expand new FTYPE.
3145         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3146         (srcp14<mode>): Make insn unary.
3147         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3148         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3149         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3150         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3151         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3152         Fix rounding: make it SAE only.
3153         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3154         Ditto.
3155         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3156         Ditto.
3157         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3158         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3159         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3160         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3161         (round_saeonly_mask_scalar_operand4): Ditto.
3162         (round_saeonly_mask_scalar_op3): Ditto.
3163         (round_saeonly_mask_scalar_op4): Ditto.
3165 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3167         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3168         Implement -maltivec=be for vec_insert and vec_extract.
3170 2014-01-10  DJ Delorie  <dj@redhat.com>
3172         * config/msp430/msp430.md (call_internal): Don't allow memory
3173         references with SP as the base register.
3174         (call_value_internal): Likewise.
3175         * config/msp430/constraints.md (Yc): New.  For memory references
3176         that don't use SP as a base register.
3178         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3179         "an integer without a # prefix"
3180         * config/msp430/msp430.md (epilogue_helper): Use it.
3182 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3184         PR target/59617
3185         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3186         AVX512F gather builtins.
3187         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3188         on gather decls with INTEGER_TYPE masktype.
3189         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3190         directly into the builtin rather than hoisting it before loop.
3192         PR tree-optimization/59387
3193         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3194         (scev_const_prop): If folded_casts and type has undefined overflow,
3195         use force_gimple_operand instead of force_gimple_operand_gsi and
3196         for each added stmt if it is assign with
3197         arith_code_with_undefined_signed_overflow, call
3198         rewrite_to_defined_overflow.
3199         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3200         gimple-fold.h instead.
3201         (arith_code_with_undefined_signed_overflow,
3202         rewrite_to_defined_overflow): Moved to ...
3203         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3204         rewrite_to_defined_overflow): ... here.  No longer static.
3205         Include gimplify-me.h.
3206         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3207         rewrite_to_defined_overflow): New prototypes.
3209 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3211         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3213 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
3215         * builtins.c (get_object_alignment_2): Minor tweak.
3216         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3218 2014-01-13  Christian Bruel  <christian.bruel@st.com>
3220         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3221         optimized non constant lengths.
3223 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3225         PR libgomp/59194
3226         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3227         load as __atomic_load_N if possible.
3229 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3231         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3232         target parameter.
3233         (rs6000_expand_builtin): Adjust call.
3235 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3237         PR target/58115
3238         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3239         * config/rs6000/rs6000.c: Include target-globals.h.
3240         (rs6000_set_current_function): Instead of doing target_reinit
3241         unconditionally, use save_target_globals_default_opts and
3242         restore_target_globals.
3244         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3245         FPSCR.
3246         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3247         (rs6000_expand_builtin): Handle mffs and mtfsf.
3248         (rs6000_init_builtins): Define mffs and mtfsf.
3249         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3250         (rs6000_mffs): New pattern.
3251         (rs6000_mtfsf): New pattern.
3253 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
3255         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3256         Start narrowing with START.  Apply candidate-use pair
3257         and check overall cost in narrowing.
3258         (iv_ca_prune): Pass new argument.
3260 2014-01-10  Jeff Law  <law@redhat.com>
3262         PR middle-end/59743
3263         * ree.c (combine_reaching_defs): Ensure the defining statement
3264         occurs before the extension when optimizing extensions with
3265         different source and destination hard registers.
3267 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3269         PR ipa/58585
3270         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3271         vtables into the type inheritance graph.
3273 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3275         PR rtl-optimization/59754
3276         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3277         modes in the REGNO != REGNO case.
3279 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3281         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3283 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3285         PR tree-optimization/59745
3286         * tree-predcom.c (tree_predictive_commoning_loop): Call
3287         free_affine_expand_cache if giving up because components is NULL.
3289         * target-globals.c (save_target_globals): Allocate < 4KB structs using
3290         GC in payload of target_globals struct instead of allocating them on
3291         the heap and the larger structs separately using GC.
3292         * target-globals.h (struct target_globals): Make regs, hard_regs,
3293         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3294         of GTY((skip)) and change type to void *.
3295         (reset_target_globals): Cast loads from those fields to corresponding
3296         types.
3298 2014-01-10  Steve Ellcey  <sellcey@mips.com>
3300         PR plugins/59335
3301         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3302         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3303         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3305 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
3307         PR target/59744
3308         * aarch64-modes.def (CC_Zmode): New flags mode.
3309         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3310         represents an equality.
3311         (aarch64_get_condition_code): Handle CC_Zmode.
3312         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3314 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3316         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3317         extraction in good case.
3319 2014-01-10  Richard Biener  <rguenther@suse.de>
3321         PR tree-optimization/59374
3322         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3323         checking after SLP discovery.  Mark stmts not participating
3324         in any SLP instance properly.
3326 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3328         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3329         when handling a SET rtx.
3331 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3333         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3334         (cortex-a57): Likewise.
3335         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3337 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3339         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3340         non-iwmmxt builtins.
3342 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3344         PR ipa/58252
3345         PR ipa/59226
3346         * ipa-devirt.c record_target_from_binfo): Take as argument
3347         stack of binfos and lookup matching one for virtual inheritance.
3348         (possible_polymorphic_call_targets_1): Update.
3350 2014-01-10  Huacai Chen  <chenhc@lemote.com>
3352         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3353         kernel strings for Loongson-2E/2F/3A.
3355 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3357         PR middle-end/59670
3358         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3359         is_gimple_call before calling gimple_call_internal_p.
3361 2014-01-09  Steve Ellcey  <sellcey@mips.com>
3363         * Makefile.in (TREE_FLOW_H): Remove.
3364         (TREE_SSA_H): Add file names from tree-flow.h.
3365         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3366         * tree.h: Remove tree-flow.h reference.
3367         * hash-table.h: Remove tree-flow.h reference.
3368         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3369         reference with tree-ssa-loop.h.
3371 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3373         * doc/invoke.texi: Add -maltivec={be,le} options, and document
3374         default element-order behavior for -maltivec.
3375         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3376         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3377         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3378         when targeting big endian, at least for now.
3379         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3381 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3383         PR middle-end/47735
3384         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3385         var satisfies use_register_for_decl, just take into account type
3386         alignment, rather than decl alignment.
3388         PR tree-optimization/59622
3389         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
3390         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3391         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3392         Don't devirtualize for inplace at all.  For targets.length () == 1,
3393         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3395 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3397         * config/i386/i386.md (cpu): Remove the unused btver1.
3399 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
3401         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3403 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3405         PR target/58115
3406         * tree-core.h (struct target_globals): New forward declaration.
3407         (struct tree_target_option): Add globals field.
3408         * tree.h (TREE_TARGET_GLOBALS): Define.
3409         (prepare_target_option_nodes_for_pch): New prototype.
3410         * target-globals.h (struct target_globals): Define even if
3411         !SWITCHABLE_TARGET.
3412         * tree.c (prepare_target_option_node_for_pch,
3413         prepare_target_option_nodes_for_pch): New functions.
3414         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3415         * config/i386/i386.c: Include target-globals.h.
3416         (ix86_set_current_function): Instead of doing target_reinit
3417         unconditionally, use save_target_globals_default_opts and
3418         restore_target_globals.
3420 2014-01-09  Richard Biener  <rguenther@suse.de>
3422         PR tree-optimization/59715
3423         * tree-cfg.h (split_critical_edges): Declare.
3424         * tree-cfg.c (split_critical_edges): Export.
3425         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3427 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
3429         * cfgexpand.c (expand_stack_vars): Optionally disable
3430         asan stack protection.
3431         (expand_used_vars): Likewise.
3432         (partition_stack_vars): Likewise.
3433         * asan.c (asan_emit_stack_protection): Optionally disable
3434         after return stack usage.
3435         (instrument_derefs): Optionally disable memory access instrumentation.
3436         (instrument_builtin_call): Likewise.
3437         (instrument_strlen_call): Likewise.
3438         (asan_protect_global): Optionally disable global variables protection.
3439         * doc/invoke.texi: Added doc for new options.
3440         * params.def: Added new options.
3441         * params.h: Likewise.
3443 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
3445         PR rtl-optimization/59724
3446         * ifcvt.c (cond_exec_process_if_block): Don't call
3447         flow_find_head_matching_sequence with 0 longest_match.
3448         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3449         non-active insns if !stop_after.
3450         (try_head_merge_bb): Revert 2014-01-07 changes.
3452 2014-01-08  Jeff Law  <law@redhat.com>
3454         * ree.c (get_sub_rtx): New function, extracted from...
3455         (merge_def_and_ext): Here.
3456         (combine_reaching_defs): Use get_sub_rtx.
3458 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
3460         * cgraph.h (varpool_variable_node): Do not choke on null node.
3462 2014-01-08  Catherine Moore  <clm@codesourcery.com>
3464         * config/mips/mips.md (simple_return): Attempt to use JRC
3465         for microMIPS.
3466         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3468 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3470         PR rtl-optimization/59137
3471         * reorg.c (steal_delay_list_from_target): Call update_block for
3472         elided insns.
3473         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3475 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3477         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3478         two duplicate entries.
3480 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
3482         Revert:
3483         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
3485         * config/mips/mips.c (mips_truncated_op_cost): New function.
3486         (mips_rtx_costs): Adjust test for BADDU.
3487         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3489         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
3491         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3492         (*baddu_si): ...this new pattern.
3494 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3496         PR ipa/59722
3497         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3499 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3501         PR middle-end/57748
3502         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3503         inner_reference_p.
3504         (expand_expr, expand_normal): Adjust.
3505         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3506         inner_reference_p. Use inner_reference_p to expand inner references.
3507         (store_expr): Adjust.
3508         * cfgexpand.c (expand_call_stmt): Adjust.
3510 2014-01-08  Rong Xu  <xur@google.com>
3512         * gcov-io.c (gcov_var): Move from gcov-io.h.
3513         (gcov_position): Ditto.
3514         (gcov_is_error): Ditto.
3515         (gcov_rewrite): Ditto.
3516         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3517         only part to libgcc/libgcov.h.
3519 2014-01-08  Marek Polacek  <polacek@redhat.com>
3521         PR middle-end/59669
3522         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3524 2014-01-08  Marek Polacek  <polacek@redhat.com>
3526         PR sanitizer/59667
3527         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3529 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
3531         PR rtl-optimization/59649
3532         * stor-layout.c (get_mode_bounds): For BImode return
3533         0 and STORE_FLAG_VALUE.
3535 2014-01-08  Richard Biener  <rguenther@suse.de>
3537         PR middle-end/59630
3538         * gimple.h (is_gimple_builtin_call): Remove.
3539         (gimple_builtin_call_types_compatible_p): New.
3540         (gimple_call_builtin_p): New overload.
3541         * gimple.c (is_gimple_builtin_call): Remove.
3542         (validate_call): Rename to ...
3543         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
3544         check return types.
3545         (validate_type): New static function.
3546         (gimple_call_builtin_p): New overload and adjust.
3547         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3548         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
3549         (gimple_fold_stmt_to_constant_1): Likewise.
3550         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3552 2014-01-08  Richard Biener  <rguenther@suse.de>
3554         PR middle-end/59471
3555         * gimplify.c (gimplify_expr): Gimplify register-register type
3556         VIEW_CONVERT_EXPRs to separate stmts.
3558 2014-01-07  Jeff Law  <law@redhat.com>
3560         PR middle-end/53623
3561         * ree.c (combine_set_extension): Handle case where source
3562         and destination registers in an extension insn are different.
3563         (combine_reaching_defs): Allow source and destination registers
3564         in extension to be different under limited circumstances.
3565         (add_removable_extension): Remove restriction that the
3566         source and destination registers in the extension are the same.
3567         (find_and_remove_re): Emit a copy from the extension's
3568         destination to its source after the defining insn if
3569         the source and destination registers are different.
3571         PR middle-end/59285
3572         * ifcvt.c (merge_if_block): If we are merging a block with more than
3573         one successor with a block with no successors, remove any BARRIER
3574         after the second block.
3576 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
3578         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3580 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
3582         PR target/59652
3583         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3584         for 14-bit register offsets when INT14_OK_STRICT is false.
3586 2014-01-07  Roland Stigge  <stigge@antcom.de>
3587             Michael Meissner  <meissner@linux.vnet.ibm.com>
3589         PR 57386/target
3590         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3591         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
3593 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
3595         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3596         -mcpu.
3598 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
3600         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3601         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3602         rtx is const0_rtx or not.
3604 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
3606         PR target/58115
3607         * target-globals.c (save_target_globals): Remove this_fn_optab
3608         handling.
3609         * toplev.c: Include optabs.h.
3610         (target_reinit): Temporarily restore the global options if another
3611         set of options are in force.
3613 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
3615         PR rtl-optimization/58668
3616         * cfgcleanup.c (flow_find_cross_jump): Don't count
3617         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
3618         to determine what is counted.
3619         (flow_find_head_matching_sequence): Use active_insn_p to determine
3620         what is counted.
3621         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3622         counting change.
3623         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3624         determine what is counted.
3626         PR tree-optimization/59643
3627         * tree-predcom.c (split_data_refs_to_components): If one dr is
3628         read and one write, determine_offset fails and the write isn't
3629         in the bad component, just put the read into the bad component.
3631 2014-01-07  Mike Stump  <mikestump@comcast.net>
3632             Jakub Jelinek  <jakub@redhat.com>
3634         PR pch/59436
3635         * tree-core.h (struct tree_optimization_option): Change optabs
3636         type from unsigned char * to void *.
3637         * optabs.c (init_tree_optimization_optabs): Adjust
3638         TREE_OPTIMIZATION_OPTABS initialization.
3640 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
3642         PR target/59644
3643         * config/i386/i386.h (struct machine_function): Add
3644         no_drap_save_restore field.
3645         * config/i386/i386.c (ix86_save_reg): Use
3646         !cfun->machine->no_drap_save_restore instead of
3647         crtl->stack_realign_needed.
3648         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3649         this function clears frame_pointer_needed.  Set
3650         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3651         and DRAP reg is needed.
3653 2014-01-06  Marek Polacek  <polacek@redhat.com>
3655         PR c/57773
3656         * doc/implement-c.texi: Mention that other integer types are
3657         permitted as bit-field types in strictly conforming mode.
3659 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
3661         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3662         is newly allocated.
3664 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
3666         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3668 2014-01-06  Martin Jambor  <mjambor@suse.cz>
3670         PR ipa/59008
3671         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3672         to int.
3673         * ipa-prop.c (ipa_print_node_params): Fix indentation.
3675 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
3677         PR debug/59350
3678         PR debug/59510
3679         * var-tracking.c (add_stores): Preserve the value of the source even if
3680         we don't record the store.
3682 2014-01-06  Terry Guo  <terry.guo@arm.com>
3684         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3686 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
3688         PR bootstrap/59541
3689         * config/darwin.c (darwin_function_section): Adjust return values to
3690         correspond to optimisation changes made in r206070.
3692 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
3694         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3695         from prefetch_block tune setting.
3696         (nocona_cost): Correct size of prefetch block to 64.
3698 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
3700         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3701         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3702         used to save the static chain register in the computation of the offset
3703         from which the FP registers need to be restored.
3705 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
3707         PR tree-optimization/59519
3708         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3709         ICE if get_current_def (current_new_name) is already non-NULL, as long
3710         as it is a phi result of some other phi in *new_exit_bb that has
3711         the same argument.
3713         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3714         or vmovdqu* for misaligned_operand.
3715         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3716         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3717         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3718         aligned_mem for AVX512F masked aligned load and store builtins and for
3719         non-temporal moves.
3721 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
3723         PR tree-optimization/59651
3724         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3725         Address range for negative step should be added by TYPE_SIZE_UNIT.
3727 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
3729         * config/m68k/m68k.c (handle_move_double): Handle pushes with
3730         overlapping registers also for registers other than the stack pointer.
3732 2014-01-03  Marek Polacek  <polacek@redhat.com>
3734         PR other/59661
3735         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3736         __builtin_FILE.
3738 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
3740         PR target/59625
3741         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3742         asm goto as jump.
3744         * config/i386/i386.md (MODE_SIZE): New mode attribute.
3745         (push splitter): Use <P:MODE_SIZE> instead of
3746         GET_MODE_SIZE (<P:MODE>mode).
3747         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3748         (mov -1, reg peephole2): Likewise.
3749         * config/i386/sse.md (*mov<mode>_internal,
3750         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3751         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3752         *<code><mode>3, *andnot<mode>3<mask_name>,
3753         <mask_codefor><code><mode>3<mask_name>): Likewise.
3754         * config/i386/subst.md (mask_mode512bit_condition,
3755         sd_mask_mode512bit_condition): Likewise.
3757 2014-01-02  Xinliang David Li  <davidxl@google.com>
3759         PR tree-optimization/59303
3760         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3761         (dump_predicates): Better output format.
3762         (pred_equal_p): New function.
3763         (is_neq_relop_p): Ditto.
3764         (is_neq_zero_form_p): Ditto.
3765         (pred_expr_equal_p): Ditto.
3766         (pred_neg_p): Ditto.
3767         (simplify_pred): Ditto.
3768         (simplify_preds_2): Ditto.
3769         (simplify_preds_3): Ditto.
3770         (simplify_preds_4): Ditto.
3771         (simplify_preds): Ditto.
3772         (push_pred): Ditto.
3773         (push_to_worklist): Ditto.
3774         (get_pred_info_from_cmp): Ditto.
3775         (is_degenerated_phi): Ditto.
3776         (normalize_one_pred_1): Ditto.
3777         (normalize_one_pred): Ditto.
3778         (normalize_one_pred_chain): Ditto.
3779         (normalize_preds): Ditto.
3780         (normalize_cond_1): Remove function.
3781         (normalize_cond): Ditto.
3782         (is_gcond_subset_of): Ditto.
3783         (is_subset_of_any): Ditto.
3784         (is_or_set_subset_of): Ditto.
3785         (is_and_set_subset_of): Ditto.
3786         (is_norm_cond_subset_of): Ditto.
3787         (pred_chain_length_cmp): Ditto.
3788         (convert_control_dep_chain_into_preds): Type change.
3789         (find_predicates): Ditto.
3790         (find_def_preds): Ditto.
3791         (destroy_predicates_vecs): Ditto.
3792         (find_matching_predicates_in_rest_chains): Ditto.
3793         (use_pred_not_overlap_with_undef_path_pred): Ditto.
3794         (is_pred_expr_subset): Ditto.
3795         (is_pred_chain_subset_of): Ditto.
3796         (is_included_in): Ditto.
3797         (is_superset_of): Ditto.
3799 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3801         Update copyright years.
3803 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
3805         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3806         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3807         config/arc/arc.md, config/arc/arc.opt,
3808         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3809         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3810         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3811         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3812         config/linux-protos.h, config/linux.c, config/winnt-c.c,
3813         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3814         vtable-verify.c, vtable-verify.h: Use the standard form for the
3815         copyright notice.
3817 2014-01-02  Tobias Burnus  <burnus@net-b.de>
3819         * gcc.c (process_command): Update copyright notice dates.
3820         * gcov-dump.c: Ditto.
3821         * gcov.c: Ditto.
3822         * doc/cpp.texi: Bump @copying's copyright year.
3823         * doc/cppinternals.texi: Ditto.
3824         * doc/gcc.texi: Ditto.
3825         * doc/gccint.texi: Ditto.
3826         * doc/gcov.texi: Ditto.
3827         * doc/install.texi: Ditto.
3828         * doc/invoke.texi: Ditto.
3830 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3832         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3834 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
3836         * config/i386/sse.md (*mov<mode>_internal): Guard
3837         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3839         PR rtl-optimization/59647
3840         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3841         new_rtx into UNSIGNED_FLOAT rtxes.
3843 Copyright (C) 2014 Free Software Foundation, Inc.
3845 Copying and distribution of this file, with or without modification,
3846 are permitted in any medium without royalty provided the copyright
3847 notice and this notice are preserved.