2014-03-18 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
bloba971bbedd0db06d0b2ed6fdab3580e65c3246b9a
1 2014-03-18  Richard Biener  <rguenther@suse.de>
3         * doc/lto.texi (fresolution): Fix typo.
5 2014-03-18  Richard Biener  <rguenther@suse.de>
7         * doc/invoke.texi (flto): Update for changes in 4.9.
9 2014-03-18  Richard Biener  <rguenther@suse.de>
11         * doc/loop.texi: Remove section on the removed lambda framework.
12         Update loop docs with recent changes in preserving loop structure.
14 2014-03-18  Richard Biener  <rguenther@suse.de>
16         * doc/lto.texi (-fresolution): Document.
18 2014-03-18  Richard Biener  <rguenther@suse.de>
20         * doc/contrib.texi: Adjust my name.
22 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
24         PR ipa/58721
25         * internal-fn.c: Include diagnostic-core.h.
26         (expand_BUILTIN_EXPECT): New function.
27         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
28         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
29         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
30         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
31         IFN_BUILTIN_EXPECT.
32         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
33         Revert 3 argument __builtin_expect code.
34         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
35         * gimple-fold.c (gimple_fold_call): Likewise.
36         * tree.h (fold_builtin_expect): New prototype.
37         * builtins.c (build_builtin_expect_predicate): Add predictor
38         argument, if non-NULL, create 3 argument __builtin_expect.
39         (fold_builtin_expect): No longer static.  Add ARG2 argument,
40         pass it through to build_builtin_expect_predicate.
41         (fold_builtin_2): Adjust caller.
42         (fold_builtin_3): Handle BUILT_IN_EXPECT.
43         * internal-fn.def (BUILTIN_EXPECT): New.
45 2014-03-18  Tobias Burnus  <burnus@net-b.de>
47         PR ipa/58721
48         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
49         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
50         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
52 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
54         PR ipa/58721
55         * predict.c (combine_predictions_for_bb): Fix up formatting.
56         (expr_expected_value_1, expr_expected_value): Add predictor argument,
57         fill what it points to if non-NULL.
58         (tree_predict_by_opcode): Adjust caller, use the predictor.
59         * predict.def (PRED_COMPARE_AND_SWAP): Add.
61 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
63         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
64         proper constant for the store mode.
66 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
68         * symtab.c (change_decl_assembler_name): Fix transparent alias
69         chain construction.
71 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
73         * config/aarch64/aarch64.c: Correct the comments about the
74         aarch64 stack layout.
76 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
78         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
79         check for GF_OMP_FOR_KIND_FOR.
81 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
83         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
84         ymm and zmm register names.
86 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
88         PR target/60516
89         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
90         note creation for the 2010-08-31 changes.
92 2014-03-17  Marek Polacek  <polacek@redhat.com>
94         PR middle-end/60534
95         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
96         as -fno-tree-loop-vectorize.
97         (expand_omp_simd): Likewise.
99 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
101         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
102         (eligible_for_call_delay): New prototype.
103         * config/sparc/sparc.c (tls_call_delay): Rename into...
104         (eligible_for_call_delay): ...this.  Return false if the instruction
105         cannot be put in the delay slot of a branch.
106         (eligible_for_restore_insn): Simplify.
107         (eligible_for_return_delay): Return false if the instruction cannot be
108         put in the delay slot of a branch and simplify.
109         (eligible_for_sibcall_delay): Return false if the instruction cannot be
110         put in the delay slot of a branch.
111         * config/sparc/sparc.md (fix_ut699): New attribute.
112         (tls_call_delay): Delete.
113         (in_call_delay): Reimplement.
114         (eligible_for_sibcall_delay): Rename into...
115         (in_sibcall_delay): ...this.
116         (eligible_for_return_delay): Rename into...
117         (in_return_delay): ...this.
118         (in_branch_delay): Reimplement.
119         (in_uncond_branch_delay): Delete.
120         (in_annul_branch_delay): Delete.
122 2014-03-14  Richard Henderson  <rth@redhat.com>
124         PR target/60525
125         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
126         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
127         (*floathi<X87MODEF>2_i387_with_temp): Remove.
128         (floathi splitters): Remove.
129         (float<SWI48x>xf2): New pattern.
130         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
131         code that tried to handle DImode for 32-bit, but which was excluded
132         by the pattern's condition.  Drop allocation of stack temporary.
133         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
134         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
135         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
136         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
137         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
138         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
139         (*float<SWI48><MODEF>2_sse_interunit): Remove.
140         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
141         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
142         (*float<SWI48x><X87MODEF>2_i387): Remove.
143         (all float _with_temp splitters): Remove.
144         (*float<SWI48x><MODEF>2_i387): New pattern.
145         (*float<SWI48><MODEF>2_sse): New pattern.
146         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
147         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
149 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
150             Marek Polacek  <polacek@redhat.com>
152         PR middle-end/60484
153         * common.opt (dump_base_name_prefixed): New Variable.
154         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
155         if x_dump_base_name_prefixed is already set, set it at the end.
157 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
159         PR rtl-optimization/60508
160         * lra-constraints.c (get_reload_reg): Add new parameter
161         in_subreg_p.
162         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
163         Pass the new parameter values.
165 2014-03-14  Richard Biener  <rguenther@suse.de>
167         * common.opt: Revert unintented changes from r205065.
168         * opts.c: Likewise.
170 2014-03-14  Richard Biener  <rguenther@suse.de>
172         PR middle-end/60518
173         * cfghooks.c (split_block): Properly adjust all loops the
174         block was a latch of.
176 2014-03-14  Martin Jambor  <mjambor@suse.cz>
178         PR lto/60461
179         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
180         and simplify it.
182 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
184         PR target/59396
185         * config/avr/avr.c (avr_set_current_function): Pass function name
186         through default_strip_name_encoding before sanity checking instead
187         of skipping the first char of the assembler name.
189 2014-03-13  Richard Henderson  <rth@redhat.com>
191         PR debug/60438
192         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
193         (ix86_force_to_memory, ix86_free_from_memory): Remove.
194         * config/i386/i386-protos.h: Likewise.
195         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
196         in the expander instead of a splitter.
197         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
198         any possibility of requiring a memory.
199         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
200         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
201         (fp branch splitters): Update for ix86_split_fp_branch.
202         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
203         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
204         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
205         (*fop_<MODEF>_2_i387): Remove f/r alternative.
206         (*fop_<MODEF>_3_i387): Likewise.
207         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
208         (splitters for the fop_* register patterns): Remove.
209         (fscalexf4_i387): Rename from *fscalexf4_i387.
210         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
211         
212 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
214         PR tree-optimization/59779
215         * tree-dfa.c (get_ref_base_and_extent): Use double_int
216         type for bitsize and maxsize instead of HOST_WIDE_INT.
218 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
220         PR rtl-optimization/57320
221         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
222         the CFG after thread_prologue_and_epilogue_insns.
224 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
226         PR rtl-optimization/57189
227         * lra-constraints.c (process_alt_operands): Disfavor spilling
228         vector pseudos.
230 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
232         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
233         messages.
235 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
237         PR tree-optimization/59025
238         PR middle-end/60418
239         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
240         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
242 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
244         PR target/60486
245         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
246         calls of avr_out_plus_1.
248 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
250         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
251         BB's single pred and update the father loop's latch info later.
253 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
255         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
256         (VEC_M): Likewise.
257         (VEC_N): Likewise.
258         (VEC_R): Likewise.
259         (VEC_base): Likewise.
260         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
261         registers, we need to swap double words in little endian mode.
263         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
264         to be a container mode for 128-bit integer operations added in ISA
265         2.07.  Unlike TImode and PTImode, the preferred register set is
266         the Altivec/VMX registers for the 128-bit operations.
268         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
269         declarations.
270         (rs6000_split_128bit_ok_p): Likewise.
272         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
273         macros for creating ISA 2.07 normal and overloaded builtin
274         functions with 3 arguments.
275         (BU_P8V_OVERLOAD_3): Likewise.
276         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
277         for use as overloaded functions.
278         (VPERM_1TI_UNS): Likewise.
279         (VSEL_1TI): Likewise.
280         (VSEL_1TI_UNS): Likewise.
281         (ST_INTERNAL_1ti): Likewise.
282         (LD_INTERNAL_1ti): Likewise.
283         (XXSEL_1TI): Likewise.
284         (XXSEL_1TI_UNS): Likewise.
285         (VPERM_1TI): Likewise.
286         (VPERM_1TI_UNS): Likewise.
287         (XXPERMDI_1TI): Likewise.
288         (SET_1TI): Likewise.
289         (LXVD2X_V1TI): Likewise.
290         (STXVD2X_V1TI): Likewise.
291         (VEC_INIT_V1TI): Likewise.
292         (VEC_SET_V1TI): Likewise.
293         (VEC_EXT_V1TI): Likewise.
294         (EQV_V1TI): Likewise.
295         (NAND_V1TI): Likewise.
296         (ORC_V1TI): Likewise.
297         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
298         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
299         overloaded builtin.
300         (VADDUQM): Likewise.
301         (VSUBCUQ): Likewise.
302         (VADDEUQM): Likewise.
303         (VADDECUQ): Likewise.
304         (VSUBEUQM): Likewise.
305         (VSUBECUQ): Likewise.
307         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
308         __int128_t and __uint128_t types.
309         (__uint128_type): Likewise.
310         (altivec_categorize_keyword): Add support for vector __int128_t,
311         vector __uint128_t, vector __int128, and vector unsigned __int128
312         as a container type for TImode operations that need to be done in
313         VSX/Altivec registers.
314         (rs6000_macro_to_expand): Likewise.
315         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
316         to support 128-bit integer instructions vaddcuq, vadduqm,
317         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
318         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
320         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
321         for V1TImode, and set up preferences to use VSX/Altivec registers.
322         Setup VSX reload handlers.
323         (rs6000_debug_reg_global): Likewise.
324         (rs6000_init_hard_regno_mode_ok): Likewise.
325         (rs6000_preferred_simd_mode): Likewise.
326         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
327         (easy_altivec_constant): Likewise.
328         (output_vec_const_move): Likewise.
329         (rs6000_expand_vector_set): Convert V1TImode set and extract to
330         simple move.
331         (rs6000_expand_vector_extract): Likewise.
332         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
333         addressing.
334         (rs6000_const_vec): Add support for V1TImode.
335         (rs6000_emit_le_vsx_load): Swap double words when loading or
336         storing TImode/V1TImode.
337         (rs6000_emit_le_vsx_store): Likewise.
338         (rs6000_emit_le_vsx_move): Likewise.
339         (rs6000_emit_move): Add support for V1TImode.
340         (altivec_expand_ld_builtin): Likewise.
341         (altivec_expand_st_builtin): Likewise.
342         (altivec_expand_vec_init_builtin): Likewise.
343         (altivec_expand_builtin): Likewise.
344         (rs6000_init_builtins): Add support for V1TImode type.  Add
345         support for ISA 2.07 128-bit integer builtins.  Define type names
346         for the VSX/Altivec vector types.
347         (altivec_init_builtins): Add support for overloaded vector
348         functions with V1TImode type.
349         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
350         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
351         external function.
352         (rs6000_split_128bit_ok_p): Likewise.
353         (rs6000_handle_altivec_attribute): Create V1TImode from vector
354         __int128_t and vector __uint128_t.
356         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
357         and mode attributes.
358         (VSX_M): Likewise.
359         (VSX_M2): Likewise.
360         (VSm): Likewise.
361         (VSs): Likewise.
362         (VSr): Likewise.
363         (VSv): Likewise.
364         (VS_scalar): Likewise.
365         (VS_double): Likewise.
366         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
368         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
369         we support the ISA 2.07 128-bit integer arithmetic instructions.
370         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
371         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
372         and TImode types for use with the builtin functions.
373         (V1TI_type_node): Likewise.
374         (unsigned_V1TI_type_node): Likewise.
375         (intTI_type_internal_node): Likewise.
376         (uintTI_type_internal_node): Likewise.
378         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
379         128-bit builtin functions.
380         (UNSPEC_VADDEUQM): Likewise.
381         (UNSPEC_VADDECUQ): Likewise.
382         (UNSPEC_VSUBCUQ): Likewise.
383         (UNSPEC_VSUBEUQM): Likewise.
384         (UNSPEC_VSUBECUQ): Likewise.
385         (VM): Add V1TImode to vector mode iterators.
386         (VM2): Likewise.
387         (VI_unit): Likewise.
388         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
389         (altivec_vaddcuq): Likewise.
390         (altivec_vsubuqm): Likewise.
391         (altivec_vsubcuq): Likewise.
392         (altivec_vaddeuqm): Likewise.
393         (altivec_vaddecuq): Likewise.
394         (altivec_vsubeuqm): Likewise.
395         (altivec_vsubecuq): Likewise.
397         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
398         mode iterators.
399         (BOOL_128): Likewise.
400         (BOOL_REGS_OUTPUT): Likewise.
401         (BOOL_REGS_OP1): Likewise.
402         (BOOL_REGS_OP2): Likewise.
403         (BOOL_REGS_UNARY): Likewise.
404         (BOOL_REGS_AND_CR0): Likewise.
406         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
407         128-bit integer builtin support.
408         (vec_vadduqm): Likewise.
409         (vec_vaddecuq): Likewise.
410         (vec_vaddeuqm): Likewise.
411         (vec_vsubecuq): Likewise.
412         (vec_vsubeuqm): Likewise.
413         (vec_vsubcuq): Likewise.
414         (vec_vsubuqm): Likewise.
416         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
417         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
418         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
419         128-bit integer add/subtract to ISA 2.07.
421 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
423         * config/arc/arc.c (arc_predicate_delay_insns):
424         Fix third argument passed to conditionalize_nonjump.
426 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
428         * config/aarch64/aarch64-builtins.c
429         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
430         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
431         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
432         instead of __builtin_lfloor.
433         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
435 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
437         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
438         (tree_ssa_ifcombine_bb_1): New function.
439         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
440         is an empty forwarder block to then_bb or vice versa and then_bb
441         and else_bb are effectively swapped.
443 2014-03-12  Christian Bruel  <christian.bruel@st.com>
445         PR target/60264
446         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
447         REG_CFA_DEF_CFA note.
448         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
449         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
451 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
453         PR tree-optimization/60454
454         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
456 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
458         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
459         Do not define target_cpu_default2 to generic.
460         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
461         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
462         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
464 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
465             Marc Glisse  <marc.glisse@inria.fr>
467         PR tree-optimization/60502
468         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
469         instead of build_low_bits_mask.
471 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
473         PR middle-end/60482
474         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
475         if there are multiple uses, but op doesn't live on E edge.
476         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
477         clobber stmts before __builtin_unreachable.
479 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
481         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
482         hard_frame_pointer_rtx.
483         * cse.c (cse_insn): Remove volatile check.
484         * cselib.c (cselib_process_insn): Likewise.
485         * dse.c (scan_insn): Likewise.
487 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
489         * config/arc/arc.c (conditionalize_nonjump): New function,
490         broken out of ...
491         (arc_ifcvt): ... this.
492         (arc_predicate_delay_insns): Use it.
494 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
496         * config/arc/predicates.md (extend_operand): During/after reload,
497         allow const_int_operand.
498         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
499         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
500         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
501         to "i".
502         (umulsi3_highpart_i): Likewise.
504 2014-03-11  Richard Biener  <rguenther@suse.de>
506         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
507         Add asserts to guard possible wrong-code bugs.
509 2014-03-11  Richard Biener  <rguenther@suse.de>
511         PR tree-optimization/60429
512         PR tree-optimization/60485
513         * tree-ssa-structalias.c (set_union_with_increment): Properly
514         take into account all fields that overlap the shifted vars.
515         (do_sd_constraint): Likewise.
516         (do_ds_constraint): Likewise.
517         (get_constraint_for_ptr_offset): Likewise.
519 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
521         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
522         (nios2_compute_frame_layout):
523         Add calculation of cfun->machine->fp_save_offset.
524         (nios2_expand_prologue): Correct setting of frame pointer register
525         in prologue.
526         (nios2_expand_epilogue): Update recovery of stack pointer from
527         frame pointer accordingly.
528         (nios2_initial_elimination_offset): Update calculation of offset
529         for eliminating to HARD_FRAME_POINTER_REGNUM.
531 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
533         PR ipa/60457
534         * ipa.c (symtab_remove_unreachable_nodes): Don't call
535         cgraph_get_create_node on VAR_DECLs.
537 2014-03-10  Richard Biener  <rguenther@suse.de>
539         PR middle-end/60474
540         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
542 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
544         * config/vms/vms.opt (vms_float_format): New variable.
546 2014-03-08  Tobias Burnus  <burnus@net-b.de>
548         * doc/invoke.texi (-fcilkplus): Update implementation status.
550 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
551             Richard Biener  <rguenther@suse.de>
553         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
554         consistently accross all TUs.
555         (run_gcc): Enable -fshort-double automatically at link at link-time
556         and disallow override.
558 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
560         PR target/58271
561         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
562         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
563         if they can't be used.
565 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
567         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
568         for Solaris 11/x86 ld.
569         * configure: Regenerate.
571 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
573         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
574         (LIB_TLS_SPEC): Save as ld_tls_libs.
575         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
576         (HAVE_AS_IX86_TLSLDM): New test.
577         * configure, config.in: Regenerate.
578         * config/i386/i386.c (legitimize_tls_address): Fall back to
579         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
580         cannot support TLS_MODEL_LOCAL_DYNAMIC.
581         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
582         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
584 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
586         * common.opt (fira-loop-pressure): Mark as optimization.
588 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
590         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
591         an OpenMP mappable type.
593 2014-03-06  Matthias Klose  <doko@ubuntu.com>
595         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
596         MULTILIB_OSDIRNAMES is not defined.
598 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
599             Meador Inge  <meadori@codesourcery.com>
601         PR target/58595
602         * config/arm/arm.c (arm_tls_symbol_p): Remove.
603         (arm_legitimize_address): Call legitimize_tls_address for any
604         arm_tls_referenced_p expression, handle constant addend.  Call it
605         before testing for !TARGET_ARM.
606         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
608 2014-03-06  Richard Biener  <rguenther@suse.de>
610         PR middle-end/60445
611         PR lto/60424
612         PR lto/60427
613         Revert
614         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
616         * tree-streamer.c (record_common_node): Assert we don't record
617         nodes with type double.
618         (preload_common_node): Skip type double, complex double and double
619         pointer since it is now frontend dependent due to fshort-double option.
621 2014-03-06  Richard Biener  <rguenther@suse.de>
623         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
624         or -fno-lto is specified and the linker has full plugin support.
625         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
626         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
627         * lto-wrapper.c (merge_and_complain): Merge compile-time
628         optimization levels.
629         (run_gcc): And pass it through to the link options.
631 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
633         PR debug/60381
634         Revert:
635         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
636         PR debug/59992
637         * cselib.c (remove_useless_values): Skip to avoid quadratic
638         behavior if the condition moved from...
639         (cselib_process_insn): ... here holds.
641 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
643         PR plugins/59335
644         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
645         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
647         PR plugins/59335
648         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
649         (TM_H): Add x86-tune.def.
651 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
653         * config/aarch64/aarch64.c (generic_tunings):
654         Use cortexa57_extra_costs.
656 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
658         PR lto/60404
659         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
660         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
661         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
662         cost for in_lto_p.
664 2014-03-04  Heiher  <r@hev.cc>
666         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
667         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
669 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
671         * config/i386/predicates.md (const2356_operand): Change to ...
672         (const2367_operand): ... this.
673         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
674         const2367_operand.
675         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
676         (*avx512pf_scatterpf<mode>sf): Ditto.
677         (avx512pf_scatterpf<mode>df): Ditto.
678         (*avx512pf_scatterpf<mode>df_mask): Ditto.
679         (*avx512pf_scatterpf<mode>df): Ditto.
680         * config/i386/i386.c (ix86_expand_builtin): Update
681         incorrect hint operand error message.
683 2014-03-04  Richard Biener  <rguenther@suse.de>
685         * lto-section-in.c (lto_get_section_data): Fix const cast.
687 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
689         * tree-streamer.c (record_common_node): Assert we don't record
690         nodes with type double.
691         (preload_common_node): Skip type double, complex double and double
692         pointer since it is now frontend dependent due to fshort-double option.
694 2014-03-04  Richard Biener  <rguenther@suse.de>
696         PR lto/60405
697         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
698         (lto_input_toplevel_asms): Likewise.
699         * lto-section-in.c (lto_get_section_data): Instead do it here
700         for every section.
702 2014-03-04  Richard Biener  <rguenther@suse.de>
704         PR tree-optimization/60382
705         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
706         dead PHIs a reduction.
708 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
710         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
711         hint value.
712         (_mm_prefetch): Move out of GCC target("sse") pragma.
713         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
714         GCC target("prfchw") pragma.
715         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
716         for locality <= 2.
717         * config/i386/i386.c (ix86_option_override_internal): Enable
718         -mprfchw with -mprefetchwt1.
720 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
722         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
723         Mark as varying.
725 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
727         * opts.h (CL_PCH_IGNORE): Define.
728         * targhooks.c (option_affects_pch_p):
729         Return false for options that have CL_PCH_IGNORE set.
730         * opt-functions.awk: Process PchIgnore.
731         * doc/options.texi: Document PchIgnore.
733         * config/arc/arc.opt (misize): Add PchIgnore property.
735 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
737         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
738         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
739         constraint on constants to permit them being loaded into
740         GENERAL_REGS or BASE_REGS.
742 2014-03-03  Nick Clifton  <nickc@redhat.com>
744         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
745         anti-cacnonical alternatives.
746         (negandhi3_real): New pattern.
747         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
749 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
751         * config/avr/avr-mcus.def: Remove atxmega16x1.
752         * config/avr/avr-tables.opt: Regenerate.
753         * config/avr/t-multilib: Regenerate.
754         * doc/avr-mmcu.texi: Regenerate.
756 2014-03-03  Tobias Grosser  <tobias@grosser.es>
757             Mircea Namolaru  <mircea.namolaru@inria.fr>
759         PR tree-optimization/58028
760         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
761         scalar dimensions.
763 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
765         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
766         not handled by recognizers.
768 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
770         PR middle-end/60175
771         * function.c (expand_function_end): Don't emit
772         clobber_return_register sequence if clobber_after is a BARRIER.
773         * cfgexpand.c (construct_exit_block): Append instructions before
774         return_label to prev_bb.
776 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
778         * config/rs6000/constraints.md: Document reserved use of "wc".
780 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
782         PR ipa/60150
783         * ipa.c (function_and_variable_visibility): When dissolving comdat
784         group, also set all symbols to local.
786 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
788         PR ipa/60306
790         Revert:
791         2013-12-14   Jan Hubicka  <jh@suse.cz>
792         PR middle-end/58477
793         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
795 2014-03-02  Jon Beniston  <jon@beniston.com>
797         PR bootstrap/48230
798         PR bootstrap/50927
799         PR bootstrap/52466
800         PR target/46898
801         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
802           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
803         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
804         (simple_return, *simple_return): New patterns
805         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
806         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
808 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
810         * dwarf2out.c (gen_subprogram_die): Tidy.
812 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
814         PR target/60071
815         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
816         (*mov_t_msb_neg_negc): ... this new insn.
818 2014-02-28  Jason Merrill  <jason@redhat.com>
820         PR c++/58678
821         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
822         function.
824 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
826         PR c++/60314
827         * dwarf2out.c (decltype_auto_die): New static.
828         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
829         (gen_type_die_with_usage): Handle 'decltype(auto)'.
830         (is_cxx_auto): Likewise.
832 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
834         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
835         we are not using general regs only.
837 2014-02-28  Richard Biener  <rguenther@suse.de>
839         PR target/60280
840         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
841         previous fix and only allow to remove trivial pre-headers
842         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
843         (remove_forwarder_block): Properly update the latch of a loop.
845 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
847         PR debug/59992
848         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
849         (cselib_preserved_hash_table): New.
850         (preserve_constants_and_equivs): Move preserved vals to it.
851         (cselib_find_slot): Look it up first.
852         (cselib_init): Initialize it.
853         (cselib_finish): Release it.
854         (dump_cselib_table): Dump it.
856 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
858         PR debug/59992
859         * cselib.c (remove_useless_values): Skip to avoid quadratic
860         behavior if the condition moved from...
861         (cselib_process_insn): ... here holds.
863 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
865         PR debug/57232
866         * var-tracking.c (vt_initialize): Apply the same condition to
867         preserve the CFA base value.
869 2014-02-28  Joey Ye  <joey.ye@arm.com>
871         PR target/PR60169
872         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
873         if reload in progress or completed.
875 2014-02-28  Tobias Burnus  <burnus@net-b.de>
877         PR middle-end/60147
878         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
879         NAMELIST_DECL.
881 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
883         * doc/tm.texi.in (Condition Code Status): Update documention for
884         relative locations of cc0-setter and cc0-user.
886 2014-02-27  Jeff Law  <law@redhat.com>
888         PR rtl-optimization/52714
889         * combine.c (try_combine): When splitting an unrecognized PARALLEL
890         into two independent simple sets, if I3 is a jump, ensure the
891         pattern we place into I3 is a (set (pc) ...).
893 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
894             Jeff Law  <law@redhat.com>
896         PR rtl-optimization/49847
897         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
898         are in different blocks.
899         * doc/tm.texi (Condition Code Status): Update documention for
900         relative locations of cc0-setter and cc0-user.
902 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
904         PR target/59222
905         * lra.c (lra_emit_add): Check SUBREG too.
907 2014-02-27  Andreas Schwab  <schwab@suse.de>
909         * config/m68k/m68k.c (m68k_option_override): Disable
910         -flive-range-shrinkage for classic m68k.
911         (m68k_override_options_after_change): Likewise.
913 2014-02-27  Marek Polacek  <polacek@redhat.com>
915         PR middle-end/59223
916         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
917         -Wmaybe-uninitialized.
919 2014-02-27  Alan Modra  <amodra@gmail.com>
921         PR target/57936
922         * reload1.c (emit_input_reload_insns): When reload_override_in,
923         set old to rl->in_reg when rl->in_reg is a subreg.
925 2014-02-26  Richard Biener  <rguenther@suse.de>
927         PR bootstrap/60343
928         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
930 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
932         * common/config/i386/predicates.md (const1256_operand): Remove.
933         (const2356_operand): New.
934         (const_1_to_2_operand): Remove.
935         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
936         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
937         (*avx512pf_gatherpf<mode>sf): Ditto.
938         (avx512pf_gatherpf<mode>df): Ditto.
939         (*avx512pf_gatherpf<mode>df_mask): Ditto.
940         (*avx512pf_gatherpf<mode>df): Ditto.
941         (avx512pf_scatterpf<mode>sf): Ditto.
942         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
943         (*avx512pf_scatterpf<mode>sf): Ditto.
944         (avx512pf_scatterpf<mode>df): Ditto.
945         (*avx512pf_scatterpf<mode>df_mask): Ditto.
946         (*avx512pf_scatterpf<mode>df): Ditto.
947         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
949 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
951         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
952         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
953         (_mm512_mask_testn_epi64_mask): Move to ...
954         * config/i386/avx512cdintrin.h: Here.
955         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
956         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
957         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
958         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
959         TARGET_AVX512F from TARGET_AVX512CD.
961 2014-02-26  Richard Biener  <rguenther@suse.de>
963         PR ipa/60327
964         * ipa.c (walk_polymorphic_call_targets): Properly guard
965         call to inline_update_overall_summary.
967 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
969         PR target/60280
970         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
971         and latches only if requested.  Fix latch if it is removed.
972         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
973         LOOPS_HAVE_PREHEADERS.
975 2014-02-25  Andrew Pinski  <apinski@cavium.com>
977         * builtins.c (expand_builtin_thread_pointer): Create a new target
978         when the target is NULL.
980 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
982         PR rtl-optimization/60317
983         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
984         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
985         * lra-assigns.c: Include params.h.
986         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
987         other reload pseudos considerations.
989 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
991         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
992         to use canonical form for nor<mode>3.
994 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
996         PR target/55426
997         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
998         conversions.
1000 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1002         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1003         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1004         (ix86_handle_option): Handle OPT_mprefetchwt1.
1005         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1006         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1007         PREFETCHWT1 CPUID.
1008         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1009         OPTION_MASK_ISA_PREFETCHWT1.
1010         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1011         (PTA_PREFETCHWT1): New.
1012         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1013         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1014         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1015         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1016         (*prefetch_avx512pf_<mode>_: Change into ...
1017         (*prefetch_prefetchwt1_<mode>: This.
1018         * config/i386/i386.opt (mprefetchwt1): New.
1019         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1020         (_mm_prefetch): Handle intent to write.
1021         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1023 2014-02-25  Richard Biener  <rguenther@suse.de>
1025         PR middle-end/60291
1026         * emit-rtl.c (mem_attrs_htab): Remove.
1027         (mem_attrs_htab_hash): Likewise.
1028         (mem_attrs_htab_eq): Likewise.
1029         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1030         (init_emit_once): Do not allocate mem_attrs_htab.
1032 2014-02-25  Richard Biener  <rguenther@suse.de>
1034         PR lto/60319
1035         * lto-opts.c (lto_write_options): Output non-explicit conservative
1036         -fwrapv, -fno-trapv and -fno-strict-overflow.
1037         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1038         (run_gcc): And pass them through.
1040 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1042         * sel-sched.c (calculate_new_fences): New parameter ptime.
1043         Calculate it as a maximum over all fence cycles.
1044         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1045         Print the final schedule timing when sched_verbose.
1047 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1049         PR rtl-optimization/60292
1050         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1051         bit fot the fence instruction.
1053 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1055         * calls.h: Fix typo in comment.
1057 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1059         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1060         adjusting offsetable addresses.
1062 2014-02-24  Guozhi Wei  <carrot@google.com>
1064         * sparseset.h (sparseset_pop): Fix the wrong index.
1066 2014-02-24  Walter Lee  <walt@tilera.com>
1068         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1069         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1070         triplet.
1071         * common/config/tilegx/tilegx-common.c
1072         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1073         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1074         (LINK_SPEC): Ditto.
1075         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1076         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1077         (tilegx_gimplify_va_arg_expr): Handle big endian.
1078         (tilegx_expand_unaligned_load): Ditto.
1079         (tilegx_expand_unaligned_store): Ditto.
1080         (TARGET_RETURN_IN_MSB): New.
1081         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1082         (TARGET_ENDIAN_DEFAULT): New.
1083         (TARGET_BIG_ENDIAN): Handle big endian.
1084         (BYTES_BIG_ENDIAN): Ditto.
1085         (WORDS_BIG_ENDIAN): Ditto.
1086         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1087         (ENDIAN_SPEC): New.
1088         (EXTRA_SPECS): New.
1089         * config/tilegx/tilegx.md (extv): Handle big endian.
1090         (extzv): Ditto.
1091         (insn_st<n>): Ditto.
1092         (insn_st<n>_add<bitsuffix>): Ditto.
1093         (insn_stnt<n>): Ditto.
1094         (insn_stnt<n>_add<bitsuffix>):Ditto.
1095         (vec_interleave_highv8qi): Handle big endian.
1096         (vec_interleave_highv8qi_be): New.
1097         (vec_interleave_highv8qi_le): New.
1098         (insn_v1int_h): Handle big endian.
1099         (vec_interleave_lowv8qi): Handle big endian.
1100         (vec_interleave_lowv8qi_be): New.
1101         (vec_interleave_lowv8qi_le): New.
1102         (insn_v1int_l): Handle big endian.
1103         (vec_interleave_highv4hi): Handle big endian.
1104         (vec_interleave_highv4hi_be): New.
1105         (vec_interleave_highv4hi_le): New.
1106         (insn_v2int_h): Handle big endian.
1107         (vec_interleave_lowv4hi): Handle big endian.
1108         (vec_interleave_lowv4hi_be): New.
1109         (vec_interleave_lowv4hi_le): New.
1110         (insn_v2int_l): Handle big endian.
1111         (vec_interleave_highv2si): Handle big endian.
1112         (vec_interleave_highv2si_be): New.
1113         (vec_interleave_highv2si_le): New.
1114         (insn_v4int_h): Handle big endian.
1115         (vec_interleave_lowv2si): Handle big endian.
1116         (vec_interleave_lowv2si_be): New.
1117         (vec_interleave_lowv2si_le): New.
1118         (insn_v4int_l): Handle big endian.
1119         * config/tilegx/tilegx.opt (mbig-endian): New option.
1120         (mlittle-endian): New option.
1121         * doc/install.texi: Document tilegxbe-linux.
1122         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1124 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1126         PR ipa/60266
1127         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1128         there are no parameter descriptors.
1130 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1132         PR rtl-optimization/60268
1133         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1134         initialization to ...
1135         (sched_rgn_init): ... here.
1136         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1138 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1140         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1141         names.
1143 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1145         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1146         definition.
1148 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1150         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1151         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1153 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1155         * config/microblaze/predicates.md: Add cmp_op predicate.
1156         * config/microblaze/microblaze.md: Add branch_compare instruction
1157         which uses cmp_op predicate and emits cmp insn before branch.
1158         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1159         to microblaze_expand_conditional_branch and consolidate logic.
1160         (microblaze_expand_conditional_branch): emit branch_compare
1161         insn instead of handling cmp op separate from branch insn.
1163 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1165         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1166         to permit subregs.
1168 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1170         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1171         define_insn with define_expand and new define_insn
1172         *altivec_lve<VI_char>x_internal.
1173         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1174         and new define_insn *altivec_stve<VI_char>x_internal.
1175         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1176         prototype.
1177         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1178         lve*x built-ins.
1179         (altivec_expand_stvex_be): New function.
1181 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1183         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1184         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1185         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1186         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1188 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1190         PR target/60298
1191         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1192         instead of emit_move_insn.
1194 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1196         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1197         vspltw with vsldoi.
1198         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1199         gen_altivec_vsumsws.
1201 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1203         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1204         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1205         (altivec_lvxl_<mode>): New define_expand incorporating
1206         -maltivec=be semantics where needed.
1207         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1208         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1209         semantics where needed.
1210         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1211         (altivec_stvx_<mode>): New define_expand incorporating
1212         -maltivec=be semantics where needed.
1213         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1214         VM2 iterator instead of V4SI.
1215         (altivec_stvxl_<mode>): New define_expand incorporating
1216         -maltivec=be semantics where needed.
1217         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1218         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1219         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1220         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1221         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1222         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1223         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1224         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1225         ALTIVEC_BUILTIN_STVXL.
1226         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1227         (altivec_expand_stvx_be): Likewise.
1228         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1229         (altivec_expand_lvx_be): Likewise.
1230         (altivec_expand_stvx_be): Likewise.
1231         (altivec_expand_builtin): Add cases for
1232         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1233         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1234         (altivec_init_builtins): Add definitions for
1235         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1236         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1238 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1240         * doc/invoke.texi (mvirt, mno-virt): Document.
1241         * config/mips/mips.opt (mvirt): New option.
1242         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1244 2014-02-21  Richard Biener  <rguenther@suse.de>
1246         PR tree-optimization/60276
1247         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1248         (STMT_VINFO_MIN_NEG_DIST): New macro.
1249         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1250         STMT_VINFO_MIN_NEG_DIST.
1251         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1252         made for negative dependence distances still hold.
1254 2014-02-21  Richard Biener  <rguenther@suse.de>
1256         PR middle-end/60291
1257         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1258         DECL_INITIAL for globals not in the current function context.
1260 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1262         PR tree-optimization/56490
1263         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1264         * tree-ssa-uninit.c: Include params.h.
1265         (compute_control_dep_chain): Add num_calls argument, return false
1266         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1267         num_calls to recursive call.
1268         (find_predicates): Change dep_chain into normal array,
1269         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1270         variable and adjust compute_control_dep_chain caller.
1271         (find_def_preds): Likewise.
1273 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1275         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1276         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1278 2014-02-21  Nick Clifton  <nickc@redhat.com>
1280         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1281         (pushhi1): Likewise.
1282         (popqi1): Add mode to pre_dec.
1283         (pophi1): Likewise.
1285 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1287         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1288         mode for mask of V8SFmode permutation.
1290 2014-02-20  Richard Henderson  <rth@redhat.com>
1292         PR c++/60272
1293         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1294         a new pseudo for OLDVAL.
1296 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1298         PR target/57896
1299         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1300         gen_reg_rtx if d->testing_p.
1301         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1302         if d->testing_p and we will certainly return true.
1303         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1304         if d->testing_p.
1306 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1308         * emit-rtl.c (gen_reg_rtx): Assert that
1309         crtl->emit.regno_pointer_align_length is non-zero.
1311 2014-02-20  Richard Henderson  <rth@redhat.com>
1313         PR c++/60272
1314         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1315         on failure the store back into EXPECT.
1317 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1318             Sandra Loosemore  <sandra@codesourcery.com>
1320         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1321         * config/nios2/nios2.c (nios2_function_profiler): Add
1322         -fPIC (flag_pic == 2) support.
1323         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1324         (nios2_large_offset_p): New function.
1325         (nios2_unspec_reloc_p): Move up position, update to use
1326         nios2_large_offset_p.
1327         (nios2_unspec_address): Remove function.
1328         (nios2_unspec_offset): New function.
1329         (nios2_large_got_address): New function.
1330         (nios2_got_address): Add large offset support.
1331         (nios2_legitimize_tls_address): Update usage of removed and new
1332         functions.
1333         (nios2_symbol_binds_local_p): New function.
1334         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1335         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1336         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1337         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1338         processing for (const (unspec ...)).
1339         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1341 2014-02-20  Richard Biener  <rguenther@suse.de>
1343         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1344         doing the substitution.
1345         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1347 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1349         PR ipa/55260
1350         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1351         info when checking whether lattices are bottom.
1353 2014-02-20  Richard Biener  <rguenther@suse.de>
1355         PR middle-end/60221
1356         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1357         regions at -O0.
1359 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1361         PR ipa/58555
1362         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1363         parameter specifying the scaling.
1364         (inline_call): Update.
1365         (want_inline_recursively): Guard division by zero.
1366         (recursive_inlining): Update.
1367         * ipa-inline.h (clone_inlined_nodes): Update.
1369 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1371         PR target/60204
1372         * config/i386/i386.c (classify_argument): Pass structures of size
1373         64 bytes or less in register.
1375 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1376             Kirill Yukhin  <kirill.yukhin@intel.com>
1378         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1379         (_mm_rcp28_round_ss): Ditto.
1380         (_mm_rsqrt28_round_sd): Ditto.
1381         (_mm_rsqrt28_round_ss): Ditto.
1382         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1383         (_mm_rcp14_round_ss): Ditto.
1384         (_mm_rsqrt14_round_sd): Ditto.
1385         (_mm_rsqrt14_round_ss): Ditto.
1386         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1387         the first input operand, get rid of match_dup.
1388         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1389         attribute to sse.
1390         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1391         Ditto.
1392         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1393         operand as the first input operand, set type attribute.
1394         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1395         Set type attribute.
1396         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1397         operand as the first input operand, set type attribute.
1399 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1401         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1402         bit of zero.
1404 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1406         PR target/60207
1407         * config/i386/i386.c (construct_container): Remove TFmode check
1408         for X86_64_INTEGER_CLASS.
1410 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1412         PR target/59794
1413         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1414         only when -Wpsabi is enabled.
1416 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1418          PR target/59799
1419         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1420         passing arrays in registers are the same as for structs, so remove the
1421         special case for them.
1423 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1425         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1426         destination type, extract only the valid bits if the source type is not
1427         integral and has a different mode.
1429 2014-02-19  Richard Biener  <rguenther@suse.de>
1431         PR ipa/60243
1432         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1433         for all calls.
1435 2014-02-19  Richard Biener  <rguenther@suse.de>
1437         PR ipa/60243
1438         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1439         (ipa_modify_call_arguments): Emit an argument load explicitely and
1440         preserve virtual SSA form there and for the replacement call.
1441         Do not update SSA form nor free dominance info.
1443 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1445         * ipa.c (function_and_variable_visibility): Also clear WEAK
1446         flag when disolving COMDAT_GROUP.
1448 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1450         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1451         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1452         not devirtualizing.
1453         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1454         do more sanity checks.
1455         (detect_type_change): Return true when giving up early.
1456         (compute_complex_assign_jump_func): Fix type parameter of
1457         ipa_set_ancestor_jf.
1458         (compute_complex_ancestor_jump_func): Likewise.
1459         (update_jump_functions_after_inlining): Fix updating of
1460         ancestor function.
1461         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1463 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1465         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1466         inline clones when edge disappears.
1468 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1470         PR target/60203
1471         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1472         Split 64-bit moves into 2 patterns.  Do not allow the use of
1473         direct move for TDmode in little endian, since the decimal value
1474         has little endian bytes within a word, but the 64-bit pieces are
1475         ordered in a big endian fashion, and normal subreg's of TDmode are
1476         not allowed.
1477         (mov<mode>_64bit_dm): Likewise.
1478         (movtd_64bit_nodm): Likewise.
1480 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1482         PR tree-optimization/60174
1483         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1484         statement of an SSA_NAME that occurs in an abnormal PHI node.
1486 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1488         PR sanitizer/60142
1489         * final.c (SEEN_BB): Remove.
1490         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1491         (final_scan_insn): Don't force_source_line on second
1492         NOTE_INSN_BASIC_BLOCK.
1494 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1496         PR target/60205
1497         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1498         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1499         (type_natural_mode): Warn ABI change when %zmm register is not
1500         available for AVX512F vector value passing.
1502 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1504         PR target/60193
1505         * config/i386/i386.c (ix86_expand_prologue): Use value in
1506         rax register as displacement when restoring %r10 or %rax.
1507         Fix wrong offset when restoring both registers.
1509 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1511         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1512         assertion with conditional return.
1514 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1515             Uros Bizjak  <ubizjak@gmail.com>
1517         PR driver/60233
1518         * config/i386/driver-i386.c (host_detect_local_cpu): If
1519         YMM state is not saved by the OS, also clear has_f16c.  Move
1520         CPUID 0x80000001 handling before YMM state saving checking.
1522 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1524         PR rtl-optimization/58960
1525         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1526         factored out from ...
1527         (sched_init): ... here.
1528         (free_global_sched_pressure_data): New, factored out from ...
1529         (sched_finish): ... here.
1530         * sched-int.h (free_global_sched_pressure_data): Declare.
1531         * sched-rgn.c (nr_regions_initial): New static global.
1532         (haifa_find_rgns): Initialize it.
1533         (schedule_region): Disable sched-pressure for the newly
1534         generated regions.
1536 2014-02-17  Richard Biener  <rguenther@suse.de>
1538         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1539         release SSA defs of pattern stmts.
1541 2014-02-17  Richard Biener  <rguenther@suse.de>
1543         * tree-inline.c (expand_call_inline): Release the virtual
1544         operand defined by the call we are about to inline.
1546 2014-02-17  Richard Biener  <rguenther@suse.de>
1548         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1550 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1551             Ilya Tocar  <ilya.tocar@intel.com>
1553         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1554         arguments order in builtin.
1555         (_mm512_permutexvar_epi64): Ditto.
1556         (_mm512_mask_permutexvar_epi64): Ditto
1557         (_mm512_maskz_permutexvar_epi32): Ditto
1558         (_mm512_permutexvar_epi32): Ditto
1559         (_mm512_mask_permutexvar_epi32): Ditto
1561 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1563         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1564         (p8_vmrgow): Likewise.
1566 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1568         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1569         endian targets.
1571 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1573         PR target/60203
1574         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1575         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1576         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1577         using direct move instructions on ISA 2.07.  Also adjust
1578         instruction length for 64-bit.
1579         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1580         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1582 2014-02-15  Alan Modra  <amodra@gmail.com>
1584         PR target/58675
1585         PR target/57935
1586         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1587         find_replacement on parts of insn rtl that might be reloaded.
1589 2014-02-15  Richard Biener  <rguenther@suse.de>
1591         PR tree-optimization/60183
1592         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1593         (tree_ssa_phiprop): Calculate and free post-dominators.
1595 2014-02-14  Jeff Law  <law@redhat.com>
1597         PR rtl-optimization/60131
1598         * ree.c (get_extended_src_reg): New function.
1599         (combine_reaching_defs): Use it rather than assuming location of REG.
1600         (find_and_remove_re): Verify first operand of extension is
1601         a REG before adding the insns to the copy list.
1603 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1605         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1606         * configure: Regenerated.
1607         * config.in: Regenerated.
1608         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1609         instead of ASM_SHORT.
1611 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1612             Richard Earnshaw  <rearnsha@arm.com>
1614         PR rtl-optimization/59535
1615         * lra-constraints.c (process_alt_operands): Encourage alternative
1616         when unassigned pseudo class is superset of the alternative class.
1617         (inherit_reload_reg): Don't inherit when optimizing for code size.
1618         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1619         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1620         modes not less than 4 for Thumb1.
1622 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1624         PR pch/60010
1625         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1627 2014-02-14  Richard Biener  <rguenther@suse.de>
1629         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1630         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1631         Do not strip INDIRECT_REFs.
1633 2014-02-14  Richard Biener  <rguenther@suse.de>
1635         PR lto/60179
1636         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1637         DECL_FUNCTION_SPECIFIC_TARGET.
1638         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1639         * tree-streamer-out.c (pack_ts_target_option): Remove.
1640         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1641         (write_ts_function_decl_tree_pointers): Do not stream
1642         DECL_FUNCTION_SPECIFIC_TARGET.
1643         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1644         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1645         (lto_input_ts_function_decl_tree_pointers): Do not stream
1646         DECL_FUNCTION_SPECIFIC_TARGET.
1648 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1650         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1651         (get_initial_def_for_induction, vectorizable_induction): Ignore
1652         debug stmts when looking for exit_phi.
1653         (vectorizable_live_operation): Fix up condition.
1655 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1657         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1658         nreverse() because it changes the content of original tree list.
1660 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1662         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1663         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1665 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1667         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1668         GNU coding standards.
1670 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1672         PR debug/60152
1673         * dwarf2out.c (gen_subprogram_die): Don't call
1674         add_calling_convention_attribute if subr_die is old_die.
1676 2014-02-13  Sharad Singhai  <singhai@google.com>
1678         * doc/optinfo.texi: Fix order of nodes.
1680 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1682         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1683         operands[2], not operands[3].
1685 2014-02-13  Richard Biener  <rguenther@suse.de>
1687         PR bootstrap/59878
1688         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1689         mention the possibility of an in-tree build.
1690         (CLooG): Update recommended version to 0.18.1, mention the
1691         possibility of an in-tree build and clarify that the ISL
1692         bundled with CLooG does not work.
1694 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1696         PR target/43546
1697         * expr.c (compress_float_constant): If x is a hard register,
1698         extend into a pseudo and then move to x.
1700 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1702         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1703         caused by bad second argument to warning_at() with -mhotpatch and
1704         nested functions (e.g. with gfortran).
1706 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1708         * opts.c (option_name): Remove "enabled by default" rider.
1710 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1712         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1714 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1715             Uros Bizjak  <ubizjak@gmail.com>
1717         PR target/60151
1718         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1719         * configure: Regenerated.
1721 2014-02-12  Richard Biener  <rguenther@suse.de>
1723         * vec.c (vec_prefix::calculate_allocation): Move as
1724         inline variant to vec.h.
1725         (vec_prefix::calculate_allocation_1): New out-of-line version.
1726         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1727         (vec_prefix::m_has_auto_buf): Rename to ...
1728         (vec_prefix::m_using_auto_storage): ... this.
1729         (vec_prefix::calculate_allocation): Inline the easy cases
1730         and dispatch to calculate_allocation_1 which doesn't need the
1731         prefix address.
1732         (va_heap::reserve): Use gcc_checking_assert.
1733         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1734         m_using_auto_storage.
1735         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1736         member and adjust.
1737         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1738         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1739         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1741 2014-02-12  Richard Biener  <rguenther@suse.de>
1743         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1744         when we found a dependence.
1746 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1748         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1749         common code...
1750         (maybe_fold_stmt): ... into this new function.
1751         * omp-low.c (lower_omp): Update comment.
1753         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1754         last use.
1756         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1757         dereference.
1759 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1761         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1762         identifiers in comments.
1763         (cortexa53_extra_costs): Likewise.
1764         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1765         (cortexa7_extra_costs): Likewise.
1766         (cortexa12_extra_costs): Likewise.
1767         (cortexa15_extra_costs): Likewise.
1768         (v7m_extra_costs): Likewise.
1770 2014-02-12  Richard Biener  <rguenther@suse.de>
1772         PR middle-end/60092
1773         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1774         of posix_memalign being successful.
1775         (lower_stmt): Restrict lowering of posix_memalign to when
1776         -ftree-bit-ccp is enabled.
1778 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1780         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1781         arg_loc.
1782         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1784 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1786         PR rtl-optimization/60116
1787         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1788         other_insn once the combination has been validated.
1790 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1792         PR lto/59468
1793         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1794         and wrapper.
1795         * ipa-devirt.c: Include demangle.h
1796         (odr_violation_reported): New static variable.
1797         (add_type_duplicate): Update odr_violations.
1798         (maybe_record_node): Add completep parameter; update it.
1799         (record_target_from_binfo): Add COMPLETEP parameter;
1800         update it as needed.
1801         (possible_polymorphic_call_targets_1): Likewise.
1802         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1803         rename FINAL to COMPLETE.
1804         (record_targets_from_bases): Sanity check we found the binfo;
1805         fix COMPLETEP updating.
1806         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1807         parameter, fix computing of COMPLETEP.
1808         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1809         at LTO time do demangling.
1810         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1811         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1812         parameter.
1813         (gimple_get_virt_method_for_binfo): Likewise.
1814         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1815         gimple_get_virt_method_for_vtable): Update prototypes.
1817 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1819         PR target/49008
1820         * genautomata.c (add_presence_absence): Fix typo with
1821         {final_}presence_list.
1823 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1825         PR target/60137
1826         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1827         for VSX/Altivec vectors that land in GPR registers.
1829 2014-02-11  Richard Henderson  <rth@redhat.com>
1830             Jakub Jelinek  <jakub@redhat.com>
1832         PR debug/59776
1833         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1834         around drhs if type conversion to lacc->type is not useless.
1836 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1838         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1839         tuning struct.
1840         (cortex-a57.cortex-a53): Likewise.
1841         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1843 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1845         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1846         arm_restrict_it.
1848 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1850         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1851         add_options_for_arm_vfp3.
1853 2014-02-11  Jeff Law  <law@redhat.com>
1855         PR middle-end/54041
1856         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1857         object with an undesirable mode.
1859 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1861         PR libgomp/60107
1862         * config/i386/sol2-9.h: New file.
1863         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1864         *-*-solaris2.9*): Use it.
1866 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1868         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1869         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1871 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1873         * config/microblaze/microblaze.c: Extend mcpu version format
1875 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
1877         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1879 2014-02-10  Richard Henderson  <rth@redhat.com>
1881         PR target/59927
1882         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1883         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1884         ms-abi vs -mno-accumulate-outgoing-args.
1885         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1886         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1887         respect to ms-abi.
1889 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1891         PR middle-end/60080
1892         * cfgexpand.c (expand_asm_operands): Attach source location to
1893         ASM_INPUT rtx objects.
1894         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1896 2014-02-10  Nick Clifton  <nickc@redhat.com>
1898         * config/mn10300/mn10300.c (popcount): New function.
1899         (mn10300_expand_prologue): Include saved registers in stack usage
1900         count.
1902 2014-02-10  Jeff Law  <law@redhat.com>
1904         PR middle-end/52306
1905         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1906         when changing the SET_DEST of a prior insn to avoid an input reload.
1908 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
1910         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1911         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1912         -mcall-openbsd, or -mcall-linux.
1913         (CC1_ENDIAN_BIG_SPEC): Remove.
1914         (CC1_ENDIAN_LITTLE_SPEC): Remove.
1915         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1916         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1917         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1918         and %cc1_endian_default.
1919         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1921 2014-02-10  Richard Biener  <rguenther@suse.de>
1923         PR tree-optimization/60115
1924         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1925         MEM_REF handling.  Properly verify that the accesses are not
1926         out of the objects bound.
1928 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1930         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1931         coretex to cortex.
1933 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
1935         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1936         proper constants and fix formatting.
1937         (possible_polymorphic_call_targets): Fix formatting.
1939 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
1940             Ilya Tocar  <ilya.tocar@intel.com>
1942         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1943         (_mm512_loadu_epi32): Renamed into...
1944         (_mm512_loadu_si512): This.
1945         (_mm512_storeu_epi32): Renamed into...
1946         (_mm512_storeu_si512): This.
1947         (_mm512_maskz_ceil_ps): Removed.
1948         (_mm512_maskz_ceil_pd): Ditto.
1949         (_mm512_maskz_floor_ps): Ditto.
1950         (_mm512_maskz_floor_pd): Ditto.
1951         (_mm512_floor_round_ps): Ditto.
1952         (_mm512_floor_round_pd): Ditto.
1953         (_mm512_ceil_round_ps): Ditto.
1954         (_mm512_ceil_round_pd): Ditto.
1955         (_mm512_mask_floor_round_ps): Ditto.
1956         (_mm512_mask_floor_round_pd): Ditto.
1957         (_mm512_mask_ceil_round_ps): Ditto.
1958         (_mm512_mask_ceil_round_pd): Ditto.
1959         (_mm512_maskz_floor_round_ps): Ditto.
1960         (_mm512_maskz_floor_round_pd): Ditto.
1961         (_mm512_maskz_ceil_round_ps): Ditto.
1962         (_mm512_maskz_ceil_round_pd): Ditto.
1963         (_mm512_expand_pd): Ditto.
1964         (_mm512_expand_ps): Ditto.
1965         * config/i386/i386.c (ix86_builtins): Remove
1966         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1967         (bdesc_args): Ditto.
1968         * config/i386/predicates.md (const1256_operand): New.
1969         (const_1_to_2_operand): Ditto.
1970         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1971         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1972         (*avx512pf_gatherpf<mode>sf): Ditto.
1973         (avx512pf_gatherpf<mode>df): Ditto.
1974         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1975         (*avx512pf_gatherpf<mode>df): Ditto.
1976         (avx512pf_scatterpf<mode>sf): Ditto.
1977         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1978         (*avx512pf_scatterpf<mode>sf): Ditto.
1979         (avx512pf_scatterpf<mode>df): Ditto.
1980         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1981         (*avx512pf_scatterpf<mode>df): Ditto.
1982         (avx512f_expand<mode>): Removed.
1983         (<shift_insn><mode>3<mask_name>): Change predicate type.
1985 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1987         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1988         not at the end of datarefs vector use ordered_remove to avoid
1989         reordering datarefs vector.
1991         PR c/59984
1992         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1993         mark local addressable non-static vars as GOVD_PRIVATE
1994         instead of GOVD_LOCAL.
1995         * omp-low.c (lower_omp_for): Move gimple_bind_vars
1996         and BLOCK_VARS of gimple_bind_block to new_stmt rather
1997         than copying them.
1999         PR middle-end/60092
2000         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2001         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2002         assume_aligned or alloc_align attributes.
2003         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2004         arguments.  Handle also assume_aligned and alloc_align attributes.
2005         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2006         calls to functions with assume_aligned or alloc_align attributes.
2007         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2009 2014-02-08  Terry Guo  <terry.guo@arm.com>
2011         * doc/invoke.texi: Document ARM -march=armv7e-m.
2013 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2015         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2016         flag on __cilkrts_rethrow builtin.
2018         PR ipa/60026
2019         * ipa-cp.c (determine_versionability): Fail at -O0
2020         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2021         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2023         Revert:
2024         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2026         PR ipa/60026
2027         * tree-inline.c (copy_forbidden): Fail for
2028         __attribute__((optimize (0))) functions.
2030 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2032         * varpool.c: Include pointer-set.h.
2033         (varpool_remove_unreferenced_decls): Variables in other partitions
2034         will not be output; be however careful to not lose information
2035         about partitioning.
2037 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2039         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2040         lookup in the vtable constructor.
2042 2014-02-07  Jeff Law  <law@redhat.com>
2044         PR target/40977
2045         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2046         define_insn_and_split.
2048         * ipa-inline.c (inline_small_functions): Fix typos.
2050 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2052         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2053         (s390_can_use_return_insn): Declare.
2054         * config/s390/s390.h (EPILOGUE_USES): Define.
2055         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2056         instructions.
2057         (s390_chunkify_start): Handle return JUMP_LABELs.
2058         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2059         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2060         (s390_can_use_return_insn): New functions.
2061         (s390_fix_long_loop_prediction): Handle conditional returns.
2062         (TARGET_SET_UP_BY_PROLOGUE): Define.
2063         * config/s390/s390.md (ANY_RETURN): New code iterator.
2064         (*creturn, *csimple_return, return, simple_return): New patterns.
2066 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2068         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2069         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2070         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2071         REG_CFA_RESTORE list when deciding not to restore a register.
2073 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2075         * config/s390/s390.c: Include tree-pass.h and context.h.
2076         (s390_early_mach): New function, split out from...
2077         (s390_emit_prologue): ...here.
2078         (pass_data_s390_early_mach): New pass structure.
2079         (pass_s390_early_mach): New class.
2080         (s390_option_override): Create and register early_mach pass.
2081         Move to end of file.
2083 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2085         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2086         to match for the exit block.
2088 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2090         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2091         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2092         Reject misaligned operands.
2094 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2096         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2098 2014-02-07  Richard Biener  <rguenther@suse.de>
2100         PR middle-end/60092
2101         * gimple-low.c (lower_builtin_posix_memalign): New function.
2102         (lower_stmt): Call it to lower posix_memalign in a way
2103         to make alignment info accessible.
2105 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2107         PR c++/60082
2108         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2109         __builtin_setjmp_receiver.
2111 2014-02-07  Richard Biener  <rguenther@suse.de>
2113         PR middle-end/60092
2114         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2115         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2116         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2117         Handle BUILT_IN_POSIX_MEMALIGN.
2118         (find_func_clobbers): Likewise.
2119         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2120         (call_may_clobber_ref_p_1): Likewise.
2122 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2124         PR ipa/59918
2125         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2126         sanity check.
2128 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2130         PR ipa/59469
2131         * lto-cgraph.c (lto_output_node): Use
2132         symtab_get_symbol_partitioning_class.
2133         (lto_output_varpool_node): likewise.
2134         (symtab_get_symbol_partitioning_class): Move here from
2135         lto/lto-partition.c
2136         * cgraph.h (symbol_partitioning_class): Likewise.
2137         (symtab_get_symbol_partitioning_class): Declare.
2139 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2141         * ggc.h (ggc_internal_cleared_alloc): New macro.
2142         * vec.h (vec_safe_copy): Handle memory stats.
2143         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2144         * target-globals.c (save_target_globals): Likewise.
2146 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2148         PR target/60077
2149         * expr.c (emit_move_resolve_push): Export; be bit more selective
2150         on when to clear alias set.
2151         * expr.h (emit_move_resolve_push): Declare.
2152         * function.h (struct function): Add tail_call_marked.
2153         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2154         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2155         * config/i386/i386.md (TImode move expander): De not call
2156         ix86_expand_push.
2157         (FP push expanders): Preserve memory attributes.
2158         * config/i386/sse.md (push<mode>1): Remove.
2159         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2160         (ix86_expand_push): Remove.
2161         * config/i386/mmx.md (push<mode>1): Remove.
2163 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2165         PR rtl-optimization/60030
2166         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2167         lopart with paradoxical subreg before shifting it up by hprec.
2169 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2171         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2172         Remove extra newline at end of file.
2173         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2174         (arm_issue_rate): Handle cortexa57.
2175         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2176         (cortex-a57.cortex-a53): Likewise.
2178 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2180         PR target/59575
2181         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2182         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2183         bitmask.
2184         (arm_expand_prologue): Adjust all callers.
2185         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2186         info, registers also at the lowest numbered registers side.  Use
2187         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2188         XEXP.
2190         PR debug/59992
2191         * var-tracking.c (adjust_mems): Before adding a SET to
2192         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2194 2014-02-06  Alan Modra  <amodra@gmail.com>
2196         PR target/60032
2197         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2198         change SDmode to DDmode when lra_in_progress.
2200 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2202         PR middle-end/59150
2203         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2204         free_data_ref on the dr first, and before goto again also set dr
2205         to the next dr.  For simd_lane_access, free old datarefs[i] before
2206         overwriting it.  For get_vectype_for_scalar_type failure, don't
2207         free_data_ref if simd_lane_access.
2209         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2211         PR target/60062
2212         * tree.h (opts_for_fn): New inline function.
2213         (opt_for_fn): Define.
2214         * config/i386/i386.c (ix86_function_regparm): Use
2215         opt_for_fn (decl, optimize) instead of optimize.
2217 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2219         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2220         for SYMBOL_REF in large memory model.
2222 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2224         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2225         and crypto support.
2226         (cortex-a57): Likewise.
2227         (cortex-a57.cortex-a53): Likewise.
2229 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2230             Kugan Vivekanandarajah  <kuganv@linaro.org>
2232         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2233         unaligned_access.
2234         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2236 2014-02-06  Richard Biener  <rguenther@suse.de>
2238         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2239         set_loop_copy and initialize_original_copy_tables.
2241 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2243         * config/aarch64/aarch64-simd.md
2244         (aarch64_ashr_simddi): Change QI to SI.
2246 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2247             Jakub Jelinek  <jakub@redhat.com>
2249         PR middle-end/60013
2250         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2251         of the dataflow.
2253 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2255         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2256         CODE_FOR_altivec_vpku[hw]um to
2257         CODE_FOR_altivec_vpku[hw]um_direct.
2258         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2259         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2260         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2261         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2263 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2265         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2266         generation for -maltivec=be.
2267         (altivec_vsumsws): Simplify redundant test.
2269 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2271         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2272         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2273         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2274         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2275         gen_altivec_vpkuwum.
2276         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2277         BYTES_BIG_ENDIAN.
2278         (altivec_vpks<VI_char>ss): Likewise.
2279         (altivec_vpks<VI_char>us): Likewise.
2280         (altivec_vpku<VI_char>us): Likewise.
2281         (altivec_vpku<VI_char>um): Likewise.
2282         (altivec_vpku<VI_char>um_direct): New (copy of
2283         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2284         internal use).
2285         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2286         target is little endian and -maltivec=be is not specified.
2287         (*altivec_vupkhs<VU_char>_direct): New (copy of
2288         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2289         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2290         target is little endian and -maltivec=be is not specified.
2291         (*altivec_vupkls<VU_char>_direct): New (copy of
2292         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2293         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2294         little endian and -maltivec=be is not specified.
2295         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2296         little endian and -maltivec=be is not specified.
2298 2014-02-05  Richard Henderson  <rth@redhat.com>
2300         PR debug/52727
2301         * combine-stack-adj.c: Revert r206943.
2302         * sched-int.h (struct deps_desc): Add last_args_size.
2303         * sched-deps.c (init_deps): Initialize it.
2304         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2305         contain REG_ARGS_SIZE notes.
2307 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2309         * lto-cgraph.c (asm_nodes_output): Make global.
2310         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2311         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2312         (driver_handle_option): Handle OPT_fwpa.
2314 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2316         PR ipa/59947
2317         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2318         a comment typo and formatting issue.  If odr_hash hasn't been
2319         created, return vNULL and set *completep to false.
2321         PR middle-end/57499
2322         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2323         bb with no successors.
2325 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2327         PR target/59718
2328         * doc/invoke.texi (-march): Clarify documentation for ARM.
2329         (-mtune): Likewise.
2330         (-mcpu): Likewise.
2332 2014-02-05  Richard Biener  <rguenther@suse.de>
2334         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2335         when not vectorizing because of too many alias checks.
2336         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2337         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2339 2014-02-05  Nick Clifton  <nickc@redhat.com>
2341         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2342         accept extended registers in any mode when compiling for the MN10300.
2344 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2346         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2347         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2348         sanitization attributes.
2349         (can_inline_edge_p): Likewise.
2350         (sanitize_attrs_match_for_inline_p): New function.
2352 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2354         * ipa-prop.c (detect_type_change): Shor circuit testing of
2355         type changes on THIS pointer.
2357 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2359         PR target/59777
2360         * config/pa/pa.c (legitimize_tls_address): Return original address
2361         if not passed a SYMBOL_REF rtx.
2362         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2363         addresses.
2364         (pa_emit_move_sequence): Simplify TLS source operands.
2365         (pa_legitimate_constant_p): Reject all TLS constants.
2366         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2367         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2369 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2371         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2372         groups when we know they are controlled by LTO.
2373         * varasm.c (default_binds_local_p_1): If object is in other partition,
2374         it will be resolved locally.
2376 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2378         * config/host-linux.c (linux_gt_pch_use_address): Don't
2379         use SSIZE_MAX because it is not always defined.
2381 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2383         PR bootstrap/59913
2384         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2385         threshold for pseudo splitting.
2386         (update_ebb_live_info): Process call argument hard registers and
2387         hard registers from insn definition too.
2388         (max_small_class_regs_num): New constant.
2389         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2390         reloads_num only for small register classes.  Don't split for
2391         outputs of jumps.
2393 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2395         PR ipa/60058
2396         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2397         is non-null.
2399 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2401         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2402         visibility is safe.
2404 2014-02-04  Marek Polacek  <polacek@redhat.com>
2406         * gdbinit.in (pel): Define.
2408 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2410         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2411         behavior.
2413 2014-02-04  Richard Biener  <rguenther@suse.de>
2415         PR lto/59723
2416         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2417         in function context local.
2418         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2419         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2420         similar to LTO_imported_decl_ref.
2422 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2424         PR tree-optimization/60002
2425         * cgraphclones.c (build_function_decl_skip_args): Clear
2426         DECL_LANG_SPECIFIC.
2428         PR tree-optimization/60023
2429         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2430         false to gsi_replace.
2431         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2432         has been in some EH region and vec_stmt could throw, add
2433         vec_stmt into the same EH region.
2434         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2435         has no lhs, ignore it.
2436         * internal-fn.c (expand_MASK_LOAD): Likewise.
2438         PR ipa/60026
2439         * tree-inline.c (copy_forbidden): Fail for
2440         __attribute__((optimize (0))) functions.
2442         PR other/58712
2443         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2444         is set, copy one less argument.
2445         (expand_simd_clones): Don't subtract clone_info->inbranch
2446         from simd_clone_struct_alloc argument.
2448         PR rtl-optimization/57915
2449         * recog.c (simplify_while_replacing): If all unary/binary/relational
2450         operation arguments are constant, attempt to simplify those.
2452         PR middle-end/59261
2453         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2454         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2456 2014-02-04  Richard Biener  <rguenther@suse.de>
2458         PR tree-optimization/60012
2459         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2460         TBAA disambiguation to all DDRs.
2462 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2464         PR target/59788
2465         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2466         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2468 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2470         PR ipa/59882
2471         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2473 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2475         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2476         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2478 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2480         PR ipa/59831
2481         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2482         to figure out targets of polymorphic calls with known decl.
2483         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2484         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2485         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2486         (get_polymorphic_call_info): ... here.
2487         (get_polymorphic_call_info_from_invariant): New function.
2489 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2491         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2492         lookup via vtable pointer; check for type consistency
2493         and turn inconsitent facts into UNREACHABLE.
2494         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2495         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2496         type inconsistent querries; return UNREACHABLE instead.
2498 2014-02-03  Richard Henderson  <rth@twiddle.net>
2500         PR tree-opt/59924
2501         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2502         already processed this node.
2503         (normalize_one_pred_1): Pass along mark_set.
2504         (normalize_one_pred): Create and destroy a pointer_set_t.
2505         (normalize_one_pred_chain): Likewise.
2507 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2509         PR gcov-profile/58602
2510         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2512 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2514         PR ipa/59831
2515         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2516         -fno-devirtualize; try to devirtualize by the knowledge of
2517         virtual table pointer given by aggregate propagation.
2518         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2519         (ipa_print_node_jump_functions): Dump also offset that
2520         is relevant for polymorphic calls.
2521         (determine_known_aggregate_parts): Add arg_type parameter; use it
2522         instead of determining the type from pointer type.
2523         (ipa_compute_jump_functions_for_edge): Update call of
2524         determine_known_aggregate_parts.
2525         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2526         (gimple_get_virt_method_for_binfo): ... here; simplify using
2527         vtable_pointer_value_to_vtable.
2528         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2529         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2530         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2531         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2532         POINTER_PLUS_EXPR.
2533         (vtable_pointer_value_to_binfo): ... here.
2534         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2536 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2538         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2539         redef of outer loop index variable.
2541 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2543         PR c++/53017
2544         PR c++/59211
2545         * doc/extend.texi (Function Attributes): Typo.
2547 2014-02-03  Cong Hou  <congh@google.com>
2549         PR tree-optimization/60000
2550         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2551         if the vectorized statement is a store.  A store statement can only
2552         appear at the end of pattern statements.
2554 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2556         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2557         (ix86_option_override_internal): Default long double to 64-bit for
2558         32-bit Bionic and to 128-bit for 64-bit Bionic.
2560         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2561         TARGET_LONG_DOUBLE_128 is true.
2562         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2564         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2565         (mlong-double-64): Negate -mlong-double-128.
2566         (mlong-double-128): New option.
2568         * config/i386/i386-c.c (ix86_target_macros): Define
2569         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2571         * doc/invoke.texi: Document -mlong-double-128.
2573 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2575         PR rtl-optimization/60024
2576         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2578 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2580         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2582 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2584         PR rtl-optimization/57662
2585         * sel-sched.c (code_motion_path_driver): Do not mark already not
2586         existing blocks in the visiting bitmap.
2588 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2590         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2591         on the insn being emitted.
2593 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2594             Will Deacon  <will.deacon@arm.com>
2596         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2598 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2600         * config/arm/arm-tables.opt: Regenerate.
2602 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2604         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2605         for vector types other than V16QImode.
2606         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2607         define_expand, and call altivec_expand_vec_perm_le when producing
2608         code with little endian element order.
2609         (*altivec_vperm_<mode>_internal): New insn having previous
2610         behavior of altivec_vperm_<mode>.
2611         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2612         altivec_expand_vec_perm_le when producing code with little endian
2613         element order.
2614         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2615         behavior of altivec_vperm_<mode>_uns.
2617 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2619         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2620         (altivec_vsumsws): Add handling for -maltivec=be with a little
2621         endian target.
2622         (altivec_vsumsws_direct): New.
2623         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2624         gen_altivec_vsumsws.
2626 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2628         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2629         vtable_pointer_value_to_binfo): New functions.
2630         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2631         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2633 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2635         * config/nios2/nios2.md (load_got_register): Initialize GOT
2636         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2637         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2639 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2641         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2642         preserverd by passthrough, do not propagate the type.
2644 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2646         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2647         (mips_atomic_assign_expand_fenv): New function.
2648         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2650 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2652         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2653         (__builtin_mips_set_fcsr): Likewise.
2654         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2655         MIPS_USI_FTYPE_VOID.
2656         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2657         (mips16_expand_set_fcsr): Likewise.
2658         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2659         (mips16_set_fcsr_stub): Likewise.
2660         (mips16_get_fcsr_one_only_stub): New class.
2661         (mips16_set_fcsr_one_only_stub): Likewise.
2662         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2663         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2664         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2665         (hard_float): New availability predicate.
2666         (mips_builtins): Add get_fcsr and set_fcsr.
2667         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2668         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2669         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2670         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2671         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2672         patterns.
2674 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2676         * config/mips/mips.c (mips_one_only_stub): New class.
2677         (mips_need_mips16_rdhwr_p): Replace with...
2678         (mips16_rdhwr_stub): ...this new variable.
2679         (mips16_stub_call_address): New function.
2680         (mips16_rdhwr_one_only_stub): New class.
2681         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2682         (mips_output_mips16_rdhwr): Delete.
2683         (mips_finish_stub): New function.
2684         (mips_code_end): Use it to handle rdhwr stubs.
2686 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2688         PR target/60017
2689         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2690         when calculating size of integer atomic types.
2692 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2694         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2696 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2698         PR tree-optimization/60003
2699         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2700         * profile.c (branch_prob): Use gimple_call_builtin_p
2701         to check for BUILT_IN_SETJMP_RECEIVER.
2702         * tree-inline.c (copy_bb): Call notice_special_calls.
2704 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2706         PR bootstrap/59985
2707         * lra-constraints.c (process_alt_operands): Update reload_sum only
2708         on the first pass.
2710 2014-01-31  Richard Henderson  <rth@redhat.com>
2712         PR middle-end/60004
2713         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2714         until after else_eh is processed.
2716 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2718         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2719         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2720         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2721         in smmintrin.h, remove them.
2722         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2723         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2724         * config/i386/i386.md (ROUND_SAE): Fix value.
2725         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2726         (const48_operand): New.
2727         * config/i386/subst.md (round), (round_expand): Use
2728         const_4_or_8_to_11_operand.
2729         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2731 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2733         * config/i386/constraints.md (Yk): Swap meaning with k.
2734         * config/i386/i386.md (movhi_internal): Change Yk to k.
2735         (movqi_internal): Ditto.
2736         (*k<logic><mode>): Ditto.
2737         (*andhi_1): Ditto.
2738         (*andqi_1): Ditto.
2739         (kandn<mode>): Ditto.
2740         (*<code>hi_1): Ditto.
2741         (*<code>qi_1): Ditto.
2742         (kxnor<mode>): Ditto.
2743         (kortestzhi): Ditto.
2744         (kortestchi): Ditto.
2745         (kunpckhi): Ditto.
2746         (*one_cmplhi2_1): Ditto.
2747         (*one_cmplqi2_1): Ditto.
2748         * config/i386/sse.md (): Change k to Yk.
2749         (avx512f_load<mode>_mask): Ditto.
2750         (avx512f_blendm<mode>): Ditto.
2751         (avx512f_store<mode>_mask): Ditto.
2752         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2753         (avx512f_storedqu<mode>_mask): Ditto.
2754         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2755         Ditto.
2756         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2757         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2758         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2759         (avx512f_maskcmp<mode>3): Ditto.
2760         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2761         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2762         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2763         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2764         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2765         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2766         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2767         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2768         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2769         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2770         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2771         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2772         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2773         (vec_extract_lo_<mode>_maskm): Ditto.
2774         (vec_extract_hi_<mode>_maskm): Ditto.
2775         (avx512f_vternlog<mode>_mask): Ditto.
2776         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2777         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2778         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2779         (avx512f_<code>v8div16qi2_mask): Ditto.
2780         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2781         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2782         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2783         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2784         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2785         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2786         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2787         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2788         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2789         (avx512cd_maskb_vec_dupv8di): Ditto.
2790         (avx512cd_maskw_vec_dupv16si): Ditto.
2791         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2792         (avx512f_vpermi2var<mode>3_mask): Ditto.
2793         (avx512f_vpermi2var<mode>3_mask): Ditto.
2794         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2795         (*avx512f_gathersi<mode>): Ditto.
2796         (*avx512f_gathersi<mode>_2): Ditto.
2797         (*avx512f_gatherdi<mode>): Ditto.
2798         (*avx512f_gatherdi<mode>_2): Ditto.
2799         (*avx512f_scattersi<mode>): Ditto.
2800         (*avx512f_scatterdi<mode>): Ditto.
2801         (avx512f_compress<mode>_mask): Ditto.
2802         (avx512f_compressstore<mode>_mask): Ditto.
2803         (avx512f_expand<mode>_mask): Ditto.
2804         * config/i386/subst.md (mask): Change k to Yk.
2805         (mask_scalar_merge): Ditto.
2806         (sd): Ditto.
2808 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2810         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2812 2014-01-31  Richard Biener  <rguenther@suse.de>
2814         PR middle-end/59990
2815         * builtins.c (fold_builtin_memory_op): Make sure to not
2816         use a floating-point mode or a boolean or enumeral type for
2817         the copy operation.
2819 2014-01-30  DJ Delorie  <dj@redhat.com>
2821         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2822         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2823         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2824         whenever main() has an epilogue.
2826 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2828         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2829         unused variable "field".
2830         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2831         (vsx_mergeh_<mode>): Likewise.
2832         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2833         (altivec_vmrghh): Likewise.
2834         (altivec_vmrghw): Likewise.
2835         (altivec_vmrglb): Likewise.
2836         (altivec_vmrglh): Likewise.
2837         (altivec_vmrglw): Likewise.
2838         (altivec_vspltb): Add missing uses.
2839         (altivec_vsplth): Likewise.
2840         (altivec_vspltw): Likewise.
2841         (altivec_vspltsf): Likewise.
2843 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2845         PR target/59923
2846         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2847         frame related instructions.
2849 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2851         PR rtl-optimization/59959
2852         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2853         any reload of register whose subreg is invalid.
2855 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2857         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2858         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2859         Add missing return type - void.
2861 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2863         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2864         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2865         remove element index adjustment for endian (now handled in vsx.md
2866         and altivec.md).
2867         (altivec_expand_vec_perm_const): Use
2868         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2869         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2870         (vsx_xxspltw_<mode>): Adjust element index for little endian.
2871         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2872         define_expand and a new define_insn *altivec_vspltb_internal;
2873         adjust for -maltivec=be on a little endian target.
2874         (altivec_vspltb_direct): New.
2875         (altivec_vsplth): Divide into a define_expand and a new
2876         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2877         little endian target.
2878         (altivec_vsplth_direct): New.
2879         (altivec_vspltw): Divide into a define_expand and a new
2880         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2881         little endian target.
2882         (altivec_vspltw_direct): New.
2883         (altivec_vspltsf): Divide into a define_expand and a new
2884         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2885         a little endian target.
2887 2014-01-30  Richard Biener  <rguenther@suse.de>
2889         PR tree-optimization/59993
2890         * tree-ssa-forwprop.c (associate_pointerplus): Check we
2891         can propagate form the earlier stmt and avoid the transform
2892         when the intermediate result is needed.
2894 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
2896         * README.Portability: Fix typo.
2898 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
2900         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2901         comparison_operator with ordered_comparison_operator.
2903 2014-01-30  Nick Clifton  <nickc@redhat.com>
2905         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2906         Rename to mn10300_store_multiple_regs.
2907         * config/mn10300/mn10300.c: Likewise.
2908         * config/mn10300/mn10300.md (store_movm): Fix typo: call
2909         store_multiple_regs.
2910         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2911         Call mn10300_store_multiple_regs.
2913 2014-01-30  Nick Clifton  <nickc@redhat.com>
2914             DJ Delorie  <dj@redhat.com>
2916         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2917         %fp 2 to keep registers after it properly word-aligned.
2918         (rl78_alloc_physical_registers_umul): Handle the case where both
2919         input operands are the same.
2921 2014-01-30  Richard Biener  <rguenther@suse.de>
2923         PR tree-optimization/59903
2924         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2925         check properly.
2927 2014-01-30  Jason Merrill  <jason@redhat.com>
2929         PR c++/59633
2930         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2932         PR c++/59645
2933         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2935 2014-01-30  Richard Biener  <rguenther@suse.de>
2937         PR tree-optimization/59951
2938         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2940 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
2942         PR target/59784
2943         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2944         SFmode to DFmode case.
2946 2014-01-29  DJ Delorie  <dj@redhat.com>
2948         * config/msp430/msp430.opt (-minrt): New.
2949         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2950         if -minrt given.
2951         (ENDFILE_SPEC): Likewise.
2953 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
2955         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2956         (estimate_function_body_sizes): Use it.
2958 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
2960         PR c++/58561
2961         * dwarf2out.c (is_cxx_auto): New.
2962         (is_base_type): Use it.
2963         (gen_type_die_with_usage): Likewise.
2965 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2967         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
2968         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2969         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2970         -maltivec=be with LE targets.
2971         (vsx_mergeh_<mode>): Likewise.
2972         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2973         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2974         (altivec_vmrghb): Replace with define_expand and new
2975         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2976         (altivec_vmrghb_direct): New define_insn.
2977         (altivec_vmrghh): Replace with define_expand and new
2978         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2979         (altivec_vmrghh_direct): New define_insn.
2980         (altivec_vmrghw): Replace with define_expand and new
2981         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2982         (altivec_vmrghw_direct): New define_insn.
2983         (*altivec_vmrghsf): Adjust for endianness.
2984         (altivec_vmrglb): Replace with define_expand and new
2985         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2986         (altivec_vmrglb_direct): New define_insn.
2987         (altivec_vmrglh): Replace with define_expand and new
2988         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2989         (altivec_vmrglh_direct): New define_insn.
2990         (altivec_vmrglw): Replace with define_expand and new
2991         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2992         (altivec_vmrglw_direct): New define_insn.
2993         (*altivec_vmrglsf): Adjust for endianness.
2994         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2995         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2996         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2997         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2998         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2999         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3000         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3001         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3003 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3005         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3006         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3007         whitespace.
3009 2014-01-29  Richard Biener  <rguenther@suse.de>
3011         PR tree-optimization/58742
3012         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3013         associate_pointerplus_align.
3014         (associate_pointerplus_diff): New function.
3015         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3016         and associate_pointerplus_diff.
3018 2014-01-29  Richard Biener  <rguenther@suse.de>
3020         * lto-streamer.h (LTO_major_version): Bump to 3.
3021         (LTO_minor_version): Reset to 0.
3023 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3025         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3026         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3027         (arm_file_start): Generate correct asm header for armv7ve.
3028         * config/arm/bpabi.h: Add multilib support for armv7ve.
3029         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3030         and cortex-a15 to armv7ve.
3031         * config/arm/t-aprofile: Add multilib support for armv7ve.
3032         * doc/invoke.texi: Document -march=armv7ve.
3034 2014-01-29  Richard Biener  <rguenther@suse.de>
3036         PR tree-optimization/58742
3037         * tree-ssa-forwprop.c (associate_plusminus): Return true
3038         if we changed sth, defer EH cleanup to ...
3039         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3040         (simplify_mult): New function.
3042 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3044         PR middle-end/59917
3045         PR tree-optimization/59920
3046         * tree.c (build_common_builtin_nodes): Remove
3047         __builtin_setjmp_dispatcher initialization.
3048         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3049         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3050         instead of gsi_after_labels + manually skipping debug stmts.
3051         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3052         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3053         * tree-inline.c (copy_edges_for_bb): Remove
3054         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3055         argument.  Ignore computed_goto_p stmts.  Don't call
3056         make_abnormal_goto_edges.  If a call might need abnormal edges
3057         for non-local gotos, see if it already has an edge to
3058         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3059         with true argument, don't do anything then, otherwise add
3060         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3061         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3062         caller.
3063         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3064         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3065         (lower_stmt): Don't set data->calls_builtin_setjmp.
3066         (lower_builtin_setjmp): Adjust comment.
3067         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3068         * tree-cfg.c (found_computed_goto): Remove.
3069         (factor_computed_gotos): Remove.
3070         (make_goto_expr_edges): Return bool, true for computed gotos.
3071         Don't call make_abnormal_goto_edges.
3072         (build_gimple_cfg): Don't set found_computed_goto, don't call
3073         factor_computed_gotos.
3074         (computed_goto_p): No longer static.
3075         (make_blocks): Don't set found_computed_goto.
3076         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3077         (make_edges): If make_goto_expr_edges returns true, push bb
3078         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3079         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3080         vector.  Record mapping between bbs and OpenMP regions if there
3081         are any, adjust make_gimple_omp_edges caller.  Call
3082         handle_abnormal_edges.
3083         (make_abnormal_goto_edges): Remove.
3084         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3085         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3086         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3087         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3088         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3089         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3090         filling *region also set *region_idx to (*region)->entry->index.
3092         PR other/58712
3093         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3094         For REGs set ORIGINAL_REGNO.
3096 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3098         * doc/md.texi: Mention that a target shouldn't implement
3099         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3100         than hi/lo pair.
3102 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3104         PR tree-optimization/59594
3105         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3106         a copy of the datarefs vector rather than the vector itself.
3108 2014-01-28  Jason Merrill  <jason@redhat.com>
3110         PR c++/53756
3111         * dwarf2out.c (auto_die): New static.
3112         (gen_type_die_with_usage): Handle C++1y 'auto'.
3113         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3114         on definition.
3116 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3118         PR target/59672
3119         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3120         (SPEC_X32): Likewise.
3121         (SPEC_64): Likewise.
3122         * config/i386/i386.c (ix86_option_override_internal): Turn off
3123         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3124         for TARGET_16BIT.
3125         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3126         * config/i386/i386.h (TARGET_16BIT): New macro.
3127         (TARGET_16BIT_P): Likewise.
3128         * config/i386/i386.opt: Add m16.
3129         * doc/invoke.texi: Document -m16.
3131 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3133         PR preprocessor/59935
3134         * input.c (location_get_source_line): Bail out on when line number
3135         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3137 2014-01-28  Richard Biener  <rguenther@suse.de>
3139         PR tree-optimization/58742
3140         * tree-ssa-forwprop.c (associate_plusminus): Handle
3141         pointer subtraction of the form (T)(P + A) - (T)P.
3143 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3145         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3146         at const_int_cost.
3148 2014-01-28  Richard Biener  <rguenther@suse.de>
3150         Revert
3151         2014-01-28  Richard Biener  <rguenther@suse.de>
3153         PR rtl-optimization/45364
3154         PR rtl-optimization/59890
3155         * var-tracking.c (local_get_addr_clear_given_value): Handle
3156         already cleared slot.
3157         (val_reset): Handle not allocated local_get_addr_cache.
3158         (vt_find_locations): Use post-order on the inverted CFG.
3160 2014-01-28  Richard Biener  <rguenther@suse.de>
3162         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3164 2014-01-28  Richard Biener  <rguenther@suse.de>
3166         PR rtl-optimization/45364
3167         PR rtl-optimization/59890
3168         * var-tracking.c (local_get_addr_clear_given_value): Handle
3169         already cleared slot.
3170         (val_reset): Handle not allocated local_get_addr_cache.
3171         (vt_find_locations): Use post-order on the inverted CFG.
3173 2014-01-28  Alan Modra  <amodra@gmail.com>
3175         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3176         * configure.ac <recursive call for build != host>: Define
3177         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3178         and LD_FOR_BUILD too.
3179         * configure: Regenerate.
3181 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3183         * config/i386/i386.c (get_builtin_code_for_version): Separate
3184         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3185         Broadwell from Haswell.
3187 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3189         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3190         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3191         * config/mips/mips.c (mips_option_override): Change setting
3192         of TARGET_DSP.
3193         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3194         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3195         Change from Mask to Var.
3197 2014-01-27  Jeff Law  <law@redhat.com>
3199         * ipa-inline.c (inline_small_functions): Fix typo.
3201 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3203         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3204         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3205         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3206         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3207         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3208         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3209         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3210         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3211         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3212         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3213         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3214         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3215         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3216         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3217         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3218         (_mm512_storeu_epi64): Ditto.
3219         (_mm512_cmpge_epi32_mask): Ditto.
3220         (_mm512_cmpge_epu32_mask): Ditto.
3221         (_mm512_cmpge_epi64_mask): Ditto.
3222         (_mm512_cmpge_epu64_mask): Ditto.
3223         (_mm512_cmple_epi32_mask): Ditto.
3224         (_mm512_cmple_epu32_mask): Ditto.
3225         (_mm512_cmple_epi64_mask): Ditto.
3226         (_mm512_cmple_epu64_mask): Ditto.
3227         (_mm512_cmplt_epi32_mask): Ditto.
3228         (_mm512_cmplt_epu32_mask): Ditto.
3229         (_mm512_cmplt_epi64_mask): Ditto.
3230         (_mm512_cmplt_epu64_mask): Ditto.
3231         (_mm512_cmpneq_epi32_mask): Ditto.
3232         (_mm512_cmpneq_epu32_mask): Ditto.
3233         (_mm512_cmpneq_epi64_mask): Ditto.
3234         (_mm512_cmpneq_epu64_mask): Ditto.
3235         (_mm512_expand_pd): Ditto.
3236         (_mm512_expand_ps): Ditto.
3237         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3238         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3239         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3240         * config/i386/i386.c (ix86_builtins): Add
3241         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3242         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3243         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3244         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3245         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3246         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3247         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3248         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3249         IX86_BUILTIN_PMOVUSQW512_MEM.
3250         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3251         __builtin_ia32_pmovsqd512mem_mask,
3252         __builtin_ia32_pmovqd512mem_mask,
3253         __builtin_ia32_pmovusqw512mem_mask,
3254         __builtin_ia32_pmovsqw512mem_mask,
3255         __builtin_ia32_pmovqw512mem_mask,
3256         __builtin_ia32_pmovusdw512mem_mask,
3257         __builtin_ia32_pmovsdw512mem_mask,
3258         __builtin_ia32_pmovdw512mem_mask,
3259         __builtin_ia32_pmovqb512mem_mask,
3260         __builtin_ia32_pmovusqb512mem_mask,
3261         __builtin_ia32_pmovsqb512mem_mask,
3262         __builtin_ia32_pmovusdb512mem_mask,
3263         __builtin_ia32_pmovsdb512mem_mask,
3264         __builtin_ia32_pmovdb512mem_mask.
3265         (bdesc_args): Add __builtin_ia32_expanddf512,
3266         __builtin_ia32_expandsf512.
3267         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3268         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3269         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3270         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3271         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3272         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3273         (avx512f_<code>v8div16qi2_mask_store): This.
3274         (avx512f_expand<mode>): New.
3276 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3278         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3279         New.
3280         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3281         (_mm512_prefetch_i32scatter_pd): Ditto.
3282         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3283         (_mm512_prefetch_i64scatter_pd): Ditto.
3284         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3285         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3286         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3287         (_mm512_prefetch_i32scatter_ps): Ditto.
3288         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3289         (_mm512_prefetch_i64scatter_ps): Ditto.
3290         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3291         * config/i386/i386-builtin-types.def: Define
3292         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3293         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3294         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3295         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3296         IX86_BUILTIN_SCATTERPFQPD.
3297         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3298         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3299         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3300         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3301         __builtin_ia32_scatterpfqps.
3302         (ix86_expand_builtin): Expand new built-ins.
3303         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3304         fix memory access data type.
3305         (*avx512pf_gatherpf<mode>_mask): Ditto.
3306         (*avx512pf_gatherpf<mode>): Ditto.
3307         (avx512pf_scatterpf<mode>): Ditto.
3308         (*avx512pf_scatterpf<mode>_mask): Ditto.
3309         (*avx512pf_scatterpf<mode>): Ditto.
3310         (GATHER_SCATTER_SF_MEM_MODE): New.
3311         (avx512pf_gatherpf<mode>df): Ditto.
3312         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3313         (*avx512pf_scatterpf<mode>df): Ditto.
3315 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3317         PR bootstrap/59934
3318         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3319         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3320         reached.
3322 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3324         * common/config/arm/arm-common.c
3325         (arm_rewrite_mcpu): Handle multiple names.
3326         * config/arm/arm.h
3327         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3329 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3331         * gimple-builder.h (create_gimple_tmp): Delete.
3333 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3335         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3336         words comparisons.
3338 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3340         * config/pa/pa.md (call): Generate indirect long calls to non-local
3341         functions when outputing 32-bit code.
3342         (call_value): Likewise except for special call to buggy powf function.
3344         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3345         portable runtime and PIC indirect calls.
3346         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3347         and PIC call sequences.  Use ldo instead of blr to set return register
3348         in PIC call sequence.
3350 2014-01-25  Walter Lee  <walt@tilera.com>
3352         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3353         avoid clobbering a live register.
3355 2014-01-25  Walter Lee  <walt@tilera.com>
3357         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3358         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3359         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3360         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3362 2014-01-25  Walter Lee  <walt@tilera.com>
3364         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3365         arguments on even registers.
3366         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3367         STACK_BOUNDARY.
3368         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3369         (BIGGEST_ALIGNMENT): Ditto.
3370         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3372 2014-01-25  Walter Lee  <walt@tilera.com>
3374         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3375         insns before bundling.
3376         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3378 2014-01-25  Walter Lee  <walt@tilera.com>
3380         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3381         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3382         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3384 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3386         * config/mips/constraints.md (kl): Delete.
3387         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3388         define expands, using...
3389         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3390         instructions for MIPS16.
3391         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3392         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3394 2014-01-25  Walter Lee  <walt@tilera.com>
3396         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3397         (clzdi2): Ditto.
3398         (ffsdi2): Ditto.
3400 2014-01-25  Walter Lee  <walt@tilera.com>
3402         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3403         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3405 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3407         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3408         Handle XOR.
3410 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3412         * print-rtl.c (in_call_function_usage): New var.
3413         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3414         EXPR_LIST mode as mode and not as reg note name.
3416         PR middle-end/59561
3417         * cfgloopmanip.c (copy_loop_info): If
3418         loop->warned_aggressive_loop_optimizations, make sure
3419         the flag is set in target loop too.
3421 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3423         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3424         flag_cilkplus.
3425         * builtins.def: Likewise.
3426         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3427         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3428         * ira.c (ira_setup_eliminable_regset): Likewise.
3429         * omp-low.c (gate_expand_omp): Likewise.
3430         (execute_lower_omp): Likewise.
3431         (diagnose_sb_0): Likewise.
3432         (gate_diagnose_omp_blocks): Likewise.
3433         (simd_clone_clauses_extract): Likewise.
3434         (gate): Likewise.
3436 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3438         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3439         correction for little endian...
3440         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3441         here.
3443 2014-01-24  Jeff Law  <law@redhat.com>
3445         PR tree-optimization/59919
3446         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3447         for non-returning calls.
3449 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3451         * common/config/aarch64/aarch64-common.c
3452         (aarch64_rewrite_mcpu): Handle multiple names.
3453         * config/aarch64/aarch64.h
3454         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3456 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3458         * input.c (add_file_to_cache_tab): Handle the case where fopen
3459         returns NULL.
3461 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3463         PR target/59929
3464         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3465         from push operand if code of push isn't PRE_DEC.
3467 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3469         PR target/59909
3470         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3471         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3472         it is only used for non-atomic loads/stores.
3474         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3475         -mquad-memory or -mquad-memory-atomic switches.
3477         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3478         -mquad-memory-atomic to ISA 2.07 support.
3480         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3481         to separate support of normal quad word memory operations (ldq, stq)
3482         from the atomic quad word memory operations.
3484         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3485         support to separate non-atomic quad word operations from atomic
3486         quad word operations.  Disable non-atomic quad word operations in
3487         little endian mode so that we don't have to swap words after the
3488         load and before the store.
3489         (quad_load_store_p): Add comment about atomic quad word support.
3490         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3491         options printed with -mdebug=reg.
3493         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3494         -mquad-memory-atomic as the test for whether we have quad word
3495         atomic instructions.
3496         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3497         or -mp8-vector are used, allow byte/half-word atomic operations.
3499         * config/rs6000/sync.md (load_lockedti): Insure that the address
3500         is a proper indexed or indirect address for the lqarx instruction.
3501         On little endian systems, swap the hi/lo registers after the lqarx
3502         instruction.
3503         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3504         insure the address is valid for the lqarx instruction.
3505         (store_conditionalti): Insure that the address is a proper indexed
3506         or indirect address for the stqcrx. instruction.  On little endian
3507         systems, swap the hi/lo registers before doing the stqcrx.
3508         instruction.
3509         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3510         insure the address is valid for the stqcrx. instruction.
3512         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3513         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3514         type of quad memory support is available.
3516 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3518         PR regression/59915
3519         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3520         there is a danger of looping.
3522 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3524         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3525         force flag_ira_loop_pressure if set via command line.
3527 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3529         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3530         (ashr_simd): New builtin handling DI mode.
3531         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3532         (aarch64_sshr_simddi): New match pattern.
3533         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3534         (vshrd_n_s64): Likewise.
3535         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3537 2014-01-23  Nick Clifton  <nickc@redhat.com>
3539         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3540         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3541         favour of mcu specific scripts.
3542         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3543         430x multilibs.
3545 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3546             Alex Velenko  <Alex.Velenko@arm.com>
3548         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3549         (vaddv_s16): Likewise.
3550         (vaddv_s32): Likewise.
3551         (vaddv_u8): Likewise.
3552         (vaddv_u16): Likewise.
3553         (vaddv_u32): Likewise.
3554         (vaddvq_s8): Likewise.
3555         (vaddvq_s16): Likewise.
3556         (vaddvq_s32): Likewise.
3557         (vaddvq_s64): Likewise.
3558         (vaddvq_u8): Likewise.
3559         (vaddvq_u16): Likewise.
3560         (vaddvq_u32): Likewise.
3561         (vaddvq_u64): Likewise.
3562         (vaddv_f32): Likewise.
3563         (vaddvq_f32): Likewise.
3564         (vaddvq_f64): Likewise.
3565         (vmaxv_f32): Likewise.
3566         (vmaxv_s8): Likewise.
3567         (vmaxv_s16): Likewise.
3568         (vmaxv_s32): Likewise.
3569         (vmaxv_u8): Likewise.
3570         (vmaxv_u16): Likewise.
3571         (vmaxv_u32): Likewise.
3572         (vmaxvq_f32): Likewise.
3573         (vmaxvq_f64): Likewise.
3574         (vmaxvq_s8): Likewise.
3575         (vmaxvq_s16): Likewise.
3576         (vmaxvq_s32): Likewise.
3577         (vmaxvq_u8): Likewise.
3578         (vmaxvq_u16): Likewise.
3579         (vmaxvq_u32): Likewise.
3580         (vmaxnmv_f32): Likewise.
3581         (vmaxnmvq_f32): Likewise.
3582         (vmaxnmvq_f64): Likewise.
3583         (vminv_f32): Likewise.
3584         (vminv_s8): Likewise.
3585         (vminv_s16): Likewise.
3586         (vminv_s32): Likewise.
3587         (vminv_u8): Likewise.
3588         (vminv_u16): Likewise.
3589         (vminv_u32): Likewise.
3590         (vminvq_f32): Likewise.
3591         (vminvq_f64): Likewise.
3592         (vminvq_s8): Likewise.
3593         (vminvq_s16): Likewise.
3594         (vminvq_s32): Likewise.
3595         (vminvq_u8): Likewise.
3596         (vminvq_u16): Likewise.
3597         (vminvq_u32): Likewise.
3598         (vminnmv_f32): Likewise.
3599         (vminnmvq_f32): Likewise.
3600         (vminnmvq_f64): Likewise.
3602 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3604         * config/aarch64/aarch64-simd.md
3605         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3606         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3607         (*aarch64_mul3_elt<mode>): Likewise.
3608         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3609         (*aarch64_mul3_elt_to_64v2df): Likewise.
3610         (*aarch64_mla_elt<mode>): Likewise.
3611         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3612         (*aarch64_mls_elt<mode>): Likewise.
3613         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3614         (*aarch64_fma4_elt<mode>): Likewise.
3615         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3616         (*aarch64_fma4_elt_to_64v2df): Likewise.
3617         (*aarch64_fnma4_elt<mode>): Likewise.
3618         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3619         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3620         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3621         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3622         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3623         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3624         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3625         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3626         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3628 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3630         * config/aarch64/aarch64-simd.md
3631         (aarch64_be_checked_get_lane<mode>): New define_expand.
3632         * config/aarch64/aarch64-simd-builtins.def
3633         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3634         New builtin definition.
3635         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3636         Use new safe be builtin.
3638 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3640         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3641         New define_insn.
3642         (aarch64_be_st1<mode>): Likewise.
3643         (aarch_ld1<VALL:mode>): Define_expand modified.
3644         (aarch_st1<VALL:mode>): Likewise.
3645         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3646         (UNSPEC_ST1): Likewise.
3648 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3650         * config/microblaze/microblaze.md: Add trap insn and attribute
3652 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3654         PR preprocessor/58580
3655         * input.h (location_get_source_line): Take an additional line_size
3656         parameter.
3657         (void diagnostics_file_cache_fini): Declare new function.
3658         * input.c (struct fcache): New type.
3659         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3660         New static constants.
3661         (diagnostic_file_cache_init, total_lines_num)
3662         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3663         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3664         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3665         (get_next_line, read_next_line, goto_next_line, read_line_num):
3666         New static function definitions.
3667         (diagnostic_file_cache_fini): New function.
3668         (location_get_source_line): Take an additional output line_len
3669         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3670         read_line_num.
3671         * diagnostic.c (diagnostic_finish): Call
3672         diagnostic_file_cache_fini.
3673         (adjust_line): Take an additional input parameter for the length
3674         of the line, rather than calculating it with strlen.
3675         (diagnostic_show_locus): Adjust the use of
3676         location_get_source_line and adjust_line with respect to their new
3677         signature.  While displaying a line now, do not stop at the first
3678         null byte.  Rather, display the zero byte as a space and keep
3679         going until we reach the size of the line.
3680         * Makefile.in: Add vec.o to OBJS-libcommon
3682 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3683             Ilya Tocar     <ilya.tocar@intel.com>
3685         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3686         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3687         (__builtin_ia32_kmov16): Ditto.
3688         * config/i386/i386.md (UNSPEC_KMOV): New.
3689         (kmovw): Ditto.
3691 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3693         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3694         (_mm512_storeu_si512): Ditto.
3696 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3698         PR target/52125
3699         * rtl.h (get_referenced_operands): Declare.
3700         * recog.c (get_referenced_operands): New function.
3701         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3702         operands have been referenced when recording LO_SUM references.
3704 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3706         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3708 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3710         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3711         Enable for generic and recent AMD targets.
3713 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3715         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3716         ARG_SIZE note when adjustment was eliminated.
3718 2014-01-22  Jeff Law  <law@redhat.com>
3720         PR tree-optimization/59597
3721         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3722         in file.  Accept new argument REGISTERING and use it to modify
3723         dump output appropriately.
3724         (register_jump_thread): Corresponding changes.
3725         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3726         thread paths involving joiner blocks.  Add code to dump cancelled
3727         jump threading paths.
3729 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3731         PR rtl-optimization/59477
3732         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3733         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3735 2014-01-22  Tom Tromey  <tromey@redhat.com>
3737         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3738         PARAMS.
3739         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3741 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3743         PR rtl-optimization/59896
3744         * lra-constraints.c (process_alt_operands): Check unused note for
3745         matched operands of insn with no output reloads.
3747 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3749         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3750         (mips_move_from_gpr_cost): Likewise.
3752 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3754         PR rtl-optimization/59858
3755         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3756         ira_class_hard_regs_num.
3757         (process_alt_operands): Increase reject for dying matched operand.
3759 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3761         PR target/59003
3762         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3763         smaller than size, perform several stores or loads and stores
3764         at dst + count - size to store or copy all of size bytes, rather
3765         than just last modesize bytes.
3767 2014-01-20  DJ Delorie  <dj@redhat.com>
3769         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3770         that CLOBBERs are REGs before propogating their values.
3772 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3774         PR middle-end/59789
3775         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3776         (cgraph_inline_failed_type): New function.
3777         * cgraph.h (DEFCIFCODE): Add type.
3778         (cgraph_inline_failed_type_t): New enum.
3779         (cgraph_inline_failed_type): New prototype.
3780         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3781         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3782         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3783         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3784         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3785         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3786         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3787         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3788         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3789         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3790         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3791         OPTIMIZATION_MISMATCH.
3792         * tree-inline.c (expand_call_inline): Emit errors during
3793         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3795 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3797         PR target/59685
3798         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3799         mode attribute in insn output.
3801 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3803         * output.h (output_constant): Delete.
3804         * varasm.c (output_constant): Make private.
3806 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3808         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3810 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3812         PR middle-end/59860
3813         * tree.h (fold_builtin_strcat): New prototype.
3814         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3815         argument, if non-NULL, don't call c_strlen.  Optimize
3816         directly into __builtin_memcpy instead of __builtin_strcpy.
3817         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3818         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3820 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3822         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3823         for SImode_address_operand operands, having only a REG argument.
3825 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3827         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3828         loader name using mbig-endian.
3829         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3831 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3833         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3834         (-mtune): Likewise.
3835         (-mcpu): Likewise.
3837 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3839         * config/aarch64/aarch64-protos.h
3840         (aarch64_cannot_change_mode_class_ptr): Declare.
3841         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3842         aarch64_cannot_change_mode_class_ptr): New.
3843         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3844         backend hook aarch64_cannot_change_mode_class.
3846 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3848         * common/config/aarch64/aarch64-common.c
3849         (aarch64_handle_option): Don't handle any option order logic here.
3850         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3851         selected_cpu, warn on architecture version mismatch.
3852         (aarch64_override_options): Fix parsing order for option strings.
3854 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3855             Iain Sandoe  <iain@codesourcery.com>
3857         PR bootstrap/59496
3858         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3859         warning.  Amend comment to reflect current functionality.
3861 2014-01-20  Richard Biener  <rguenther@suse.de>
3863         PR middle-end/59860
3864         * builtins.c (fold_builtin_strcat): Remove case better handled
3865         by tree-ssa-strlen.c.
3867 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
3869         * config/aarch64/aarch64.opt
3870         (mcpu, march, mtune): Make case-insensitive.
3872 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3874         PR target/59880
3875         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3876         if operands[1] is a REG or ZERO_EXTEND of a REG.
3878 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
3880         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3882 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
3884         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3885         long non-pic millicode calls.
3887 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3889         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3891 2014-01-19  Kito Cheng  <kito@0xlab.org>
3893         * builtins.c (expand_movstr): Check movstr expand done or fail.
3895 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3896             H.J. Lu  <hongjiu.lu@intel.com>
3898         PR target/59379
3899         * config/i386/i386.md (*lea<mode>): Zero-extend return register
3900         to DImode for zero-extended addresses.
3902 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
3904         PR rtl-optimization/57763
3905         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3906         on the new indirect jump_insn and increment LABEL_NUSES (label).
3908 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
3910         PR bootstrap/59580
3911         PR bootstrap/59583
3912         * config.gcc (x86_archs): New variable.
3913         (x86_64_archs): Likewise.
3914         (x86_cpus): Likewise.
3915         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3916         --with-arch/--with-cpu= options.
3917         Support --with-arch=/--with-cpu={nehalem,westmere,
3918         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3920 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3922         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3923         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
3925 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3927         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3929 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
3931         PR target/58944
3932         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3933         clear cpp_get_options (parse_in)->warn_unused_macros for
3934         ix86_target_macros_internal with cpp_define.
3936 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
3938         * jump.c (delete_related_insns): Keep (use (insn))s.
3939         * reorg.c (redundant_insn): Check for barriers too.
3941 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3943         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3945 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
3947         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3948         call to $$dyncall when TARGET_LONG_CALLS is true.
3950 2014-01-17  Jeff Law  <law@redhat.com>
3952         * ree.c (combine_set_extension): Temporarily disable test for
3953         changing number of hard registers.
3955 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3957         PR middle-end/58125
3958         * ipa-inline-analysis.c (inline_free_summary):
3959         Do not free summary of aliases.
3961 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3963         PR middle-end/59706
3964         * gimplify.c (gimplify_expr): Use create_tmp_var
3965         instead of create_tmp_var_raw.  If cond doesn't have
3966         integral type, don't add the IFN_ANNOTATE builtin at all.
3968 2014-01-17  Martin Jambor  <mjambor@suse.cz>
3970         PR ipa/59736
3971         * ipa-cp.c (prev_edge_clone): New variable.
3972         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3973         Also resize prev_edge_clone vector.
3974         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3975         (ipcp_edge_removal_hook): New function.
3976         (ipcp_driver): Register ipcp_edge_removal_hook.
3978 2014-01-17  Andrew Pinski <apinski@cavium.com>
3979             Steve Ellcey  <sellcey@mips.com>
3981         PR target/59462
3982         * config/mips/mips.c (mips_print_operand): Check operand mode instead
3983         of operator mode.
3985 2014-01-17  Jeff Law  <law@redhat.com>
3987         PR middle-end/57904
3988         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3989         so that pass_ccp runs first.
3991 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3993         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3994         (ix86_adjust_cost): Use !TARGET_XXX.
3995         (do_reorder_for_imul): Likewise.
3996         (swap_top_of_ready_list): Likewise.
3997         (ix86_sched_reorder): Likewise.
3999 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4001         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4002         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4003         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4004         (intel_memset): New.  Duplicate slm_memset.
4005         (intel_cost): New.  Duplicate slm_cost.
4006         (m_INTEL): New macro.
4007         (processor_target_table): Add "intel".
4008         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4009         with PROCESSOR_INTEL for "intel".
4010         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4011         PROCESSOR_SILVERMONT.
4012         (ix86_issue_rate): Likewise.
4013         (ix86_adjust_cost): Likewise.
4014         (ia32_multipass_dfa_lookahead): Likewise.
4015         (swap_top_of_ready_list): Likewise.
4016         (ix86_sched_reorder): Likewise.
4017         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4018         instead of TARGET_OPT_AGU.
4019         * config/i386/i386.h (TARGET_INTEL): New.
4020         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4021         (processor_type): Add PROCESSOR_INTEL.
4022         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4023         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4025 2014-01-17  Marek Polacek  <polacek@redhat.com>
4027         PR c/58346
4028         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4029         size is zero.
4031 2014-01-17  Richard Biener  <rguenther@suse.de>
4033         PR tree-optimization/46590
4034         * opts.c (default_options_table): Add entries for
4035         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4036         all enabled at -O1 but not for -Og.
4037         * common.opt (fbranch-count-reg): Remove Init(1).
4038         (fmove-loop-invariants): Likewise.
4039         (ftree-pta): Likewise.
4041 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4043         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4044         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4045         decls to at least the GCC 4.8 used alignments.
4047         PR fortran/59440
4048         * tree-nested.c (convert_nonlocal_reference_stmt,
4049         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4050         of GIMPLE_BIND stmts, adjust associated decls.
4052 2014-01-17  Richard Biener  <rguenther@suse.de>
4054         PR tree-optimization/46590
4055         * vec.h (vec<>::bseach): New member function implementing
4056         binary search according to C89 bsearch.
4057         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4058         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4059         bitmap pointer again.  Make accesses_in_loop a flat array.
4060         (mem_ref_obstack): New global.
4061         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4062         (mark_ref_stored): Likewise.
4063         (ref_indep_loop_p_2): Likewise.
4064         (set_ref_stored_in_loop): New helper function.
4065         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4066         (memref_free): Adjust.
4067         (record_mem_ref_loc): Simplify.
4068         (gather_mem_refs_stmt): Adjust.
4069         (sort_locs_in_loop_postorder_cmp): New function.
4070         (analyze_memory_references): Sort accesses_in_loop after
4071         loop postorder number.
4072         (find_ref_loc_in_loop_cmp): New function.
4073         (for_all_locs_in_loop): Find relevant cluster of locs in
4074         accesses_in_loop and iterate without recursion.
4075         (execute_sm): Avoid uninit warning.
4076         (struct ref_always_accessed): Simplify.
4077         (ref_always_accessed::operator ()): Likewise.
4078         (ref_always_accessed_p): Likewise.
4079         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4080         loop postorder numbers here.
4081         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4082         numbers.
4084 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4086         PR c++/57945
4087         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4088         on decls for which assemble_alias has been called.
4090 2014-01-17  Nick Clifton  <nickc@redhat.com>
4092         * config/msp430/msp430.opt: (mcpu): New option.
4093         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4094         (msp430_option_override): Parse target_cpu.  If the MCU name
4095         matches a generic string, clear target_mcu.
4096         (msp430_attr): Allow numeric interrupt values up to 63.
4097         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4098         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4099         option.
4100         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4101         Add mcpu matches.
4102         * config/msp430/msp430.md (popm): Use %J rather than %I.
4103         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4104         (addhi_cy_i): Use immediate_operand for operand 2.
4105         * doc/invoke.texi: Document -mcpu option.
4107 2014-01-17  Richard Biener  <rguenther@suse.de>
4109         PR rtl-optimization/38518
4110         * df.h (df_analyze_loop): Declare.
4111         * df-core.c: Include cfgloop.h.
4112         (df_analyze_1): Split out main part of df_analyze.
4113         (df_analyze): Adjust.
4114         (loop_inverted_post_order_compute): New function.
4115         (loop_post_order_compute): Likewise.
4116         (df_analyze_loop): New function avoiding whole-function
4117         postorder computes.
4118         * loop-invariant.c (find_defs): Use df_analyze_loop.
4119         (find_invariants): Adjust.
4120         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4122 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4124         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4125         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4127 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4129         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4130         traversal when removing references.
4132 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4134         PR ipa/59775
4135         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4137 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4139         PR middle-end/56791
4140         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4141         pushing a reload for an autoinc when we had previously reloaded an
4142         inner part of the address.
4144 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4146         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4147         field.
4148         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4149         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4150         when not giving up or versioning for alias only because of
4151         loop->safelen.
4152         (vect_analyze_data_ref_dependences): Set to true.
4153         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4154         is a GIMPLE_PHI.
4155         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4156         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4157         to the condition.
4159         PR middle-end/58344
4160         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4162         PR target/59839
4163         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4164         operand 0 predicate for gathers, use a new pseudo as subtarget.
4166 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4168         PR middle-end/59609
4169         * lra-constraints.c (process_alt_operands): Add printing debug info.
4170         Check absence of input/output reloads for matched operands too.
4172 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4174         PR rtl-optimization/59835
4175         * ira.c (ira_init_register_move_cost): Increase cost for
4176         impossible modes.
4178 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4180         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4182 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4184         PR target/59780
4185         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4186         non-register objects.  Use gen_(high/low)part more consistently.
4187         Fix assertions.
4189 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4191         PR target/59844
4192         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4193         endian support, remove tests for WORDS_BIG_ENDIAN.
4194         (p8_mfvsrd_3_<mode>): Likewise.
4195         (reload_gpr_from_vsx<mode>): Likewise.
4196         (reload_gpr_from_vsxsf): Likewise.
4197         (p8_mfvsrd_4_disf): Likewise.
4199 2014-01-16  Richard Biener  <rguenther@suse.de>
4201         PR rtl-optimization/46590
4202         * lcm.c (compute_antinout_edge): Use postorder iteration.
4203         (compute_laterin): Use inverted postorder iteration.
4205 2014-01-16  Nick Clifton  <nickc@redhat.com>
4207         PR middle-end/28865
4208         * varasm.c (output_constant): Return the number of bytes actually
4209         emitted.
4210         (output_constructor_array_range): Update the field size with the
4211         number of bytes emitted by output_constant.
4212         (output_constructor_regular_field): Likewise.  Also do not
4213         complain if the total number of bytes emitted is now greater
4214         than the expected fieldpos.
4215         * output.h (output_constant): Update prototype and descriptive comment.
4217 2014-01-16  Marek Polacek  <polacek@redhat.com>
4219         PR middle-end/59827
4220         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4221         it is error_mark_node.
4223 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4225         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4226         VALID_AVX256_REG_OR_OI_MODE.
4228 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4230         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4231         current procedure should be profiled.
4233 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4235         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4236         of moving from/to the STACK_REG register class.
4238 2014-01-15  Richard Henderson  <rth@redhat.com>
4240         PR debug/54694
4241         * reginfo.c (global_regs_decl): Globalize.
4242         * rtl.h (global_regs_decl): Declare.
4243         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4244         reserved via global_regs.
4246 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4248         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4250 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4252         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4253         and vmulosh rather than call gen_vec_widen_smult_*.
4254         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4255         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4256         (vec_widen_smult_even_v16qi): Likewise.
4257         (vec_widen_umult_even_v8hi): Likewise.
4258         (vec_widen_smult_even_v8hi): Likewise.
4259         (vec_widen_umult_odd_v16qi): Likewise.
4260         (vec_widen_smult_odd_v16qi): Likewise.
4261         (vec_widen_umult_odd_v8hi): Likewise.
4262         (vec_widen_smult_odd_v8hi): Likewise.
4263         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4264         vmuloub rather than call gen_vec_widen_umult_*.
4265         (vec_widen_umult_lo_v16qi): Likewise.
4266         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4267         vmulosb rather than call gen_vec_widen_smult_*.
4268         (vec_widen_smult_lo_v16qi): Likewise.
4269         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4270         rather than call gen_vec_widen_umult_*.
4271         (vec_widen_umult_lo_v8hi): Likewise.
4272         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4273         rather than call gen_vec_widen_smult_*.
4274         (vec_widen_smult_lo_v8hi): Likewise.
4276 2014-01-15  Jeff Law  <law@redhat.com>
4278         PR tree-optimization/59747
4279         * ree.c (find_and_remove_re): Properly handle case where a second
4280         eliminated extension requires widening a copy created for elimination
4281         of a prior extension.
4282         (combine_set_extension): Ensure that the number of hard regs needed
4283         for a destination register does not change when we widen it.
4285 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4287         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4288         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4289         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4290         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4291         (avr-*-rtems*): Likewise.
4292         (bfin*-rtems*): Likewise.
4293         (moxie-*-rtems*): Likewise.
4294         (h8300-*-rtems*): Likewise.
4295         (i[34567]86-*-rtems*): Likewise.
4296         (lm32-*-rtems*): Likewise.
4297         (m32r-*-rtems*): Likewise.
4298         (m68k-*-rtems*): Likewise.
4299         (microblaze*-*-rtems*): Likewise.
4300         (mips*-*-rtems*): Likewise.
4301         (powerpc-*-rtems*): Likewise.
4302         (sh-*-rtems*): Likewise.
4303         (sparc-*-rtems*): Likewise.
4304         (sparc64-*-rtems*): Likewise.
4305         (v850-*-rtems*): Likewise.
4306         (m32c-*-rtems*): Likewise.
4308 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4310         PR rtl-optimization/59511
4311         * ira.c (ira_init_register_move_cost): Use memory costs for some
4312         cases of register move cost calculations.
4313         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4314         instead of BB frequency.
4315         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4316         * lra-assigns.c (find_hard_regno_for): Ditto.
4318 2014-01-15  Richard Biener  <rguenther@suse.de>
4320         PR tree-optimization/59822
4321         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4322         (vectorizable_load): Use it to hoist defs of uses of invariant
4323         loads out of the loop.
4325 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4326             Kugan Vivekanandarajah  <kuganv@linaro.org>
4328         PR target/59695
4329         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4330         truncation.
4332 2014-01-15  Richard Biener  <rguenther@suse.de>
4334         PR rtl-optimization/59802
4335         * lcm.c (compute_available): Use inverted postorder to seed
4336         the initial worklist.
4338 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4340         PR target/59803
4341         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4342         ADDR_REGS for invalid symrefs in non-PIC code.
4344 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4346         PR other/58712
4347         * builtins.c (determine_block_size): Initialize *probable_max_size
4348         even if len_rtx is CONST_INT.
4350 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4352         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4353         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4354         (cortexa53_tunings): Likewise.
4355         (aarch64_sched_issue_rate): New function.
4356         (TARGET_SCHED_ISSUE_RATE): Define.
4358 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4360         * ira-costs.c (find_costs_and_classes): Add missed
4361         ira_init_register_move_cost_if_necessary.
4363 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4365         PR target/59787
4366         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4368 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4370         PR target/59794
4371         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4372         to indicate if type is used for function return value.  Warn ABI
4373         change if the vector mode isn't available for function return value.
4374         (ix86_function_arg_advance): Pass false to type_natural_mode.
4375         (ix86_function_arg): Likewise.
4376         (ix86_gimplify_va_arg): Likewise.
4377         (function_arg_32): Don't warn ABI change.
4378         (ix86_function_value): Pass true to type_natural_mode.
4379         (ix86_return_in_memory): Likewise.
4380         (ix86_struct_value_rtx): Removed.
4381         (TARGET_STRUCT_VALUE_RTX): Likewise.
4383 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4385         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4386         converting a conditional jump into a conditional return.
4388 2014-01-14  Richard Biener  <rguenther@suse.de>
4390         PR tree-optimization/58921
4391         PR tree-optimization/59006
4392         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4393         hoisting invariant stmts.
4394         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4395         invariant loads on the preheader edge if possible.
4397 2014-01-14  Joey Ye  <joey.ye@arm.com>
4399         * doc/plugin.texi (Building GCC plugins): Update to C++.
4401 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4403         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4404         (_mm_rcp28_round_ss): Ditto.
4405         (_mm_rsqrt28_round_sd): Ditto.
4406         (_mm_rsqrt28_round_ss): Ditto.
4407         (_mm_rcp28_sd): Ditto.
4408         (_mm_rcp28_ss): Ditto.
4409         (_mm_rsqrt28_sd): Ditto.
4410         (_mm_rsqrt28_ss): Ditto.
4411         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4412         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4413         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4414         (IX86_BUILTIN_RCP28SD): Ditto.
4415         (IX86_BUILTIN_RCP28SS): Ditto.
4416         (IX86_BUILTIN_RSQRT28SD): Ditto.
4417         (IX86_BUILTIN_RSQRT28SS): Ditto.
4418         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4419         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4420         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4421         (ix86_expand_special_args_builtin): Expand new FTYPE.
4422         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4423         (srcp14<mode>): Make insn unary.
4424         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4425         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4426         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4427         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4428         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4429         Fix rounding: make it SAE only.
4430         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4431         Ditto.
4432         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4433         Ditto.
4434         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4435         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4436         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4437         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4438         (round_saeonly_mask_scalar_operand4): Ditto.
4439         (round_saeonly_mask_scalar_op3): Ditto.
4440         (round_saeonly_mask_scalar_op4): Ditto.
4442 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4444         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4445         Implement -maltivec=be for vec_insert and vec_extract.
4447 2014-01-10  DJ Delorie  <dj@redhat.com>
4449         * config/msp430/msp430.md (call_internal): Don't allow memory
4450         references with SP as the base register.
4451         (call_value_internal): Likewise.
4452         * config/msp430/constraints.md (Yc): New.  For memory references
4453         that don't use SP as a base register.
4455         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4456         "an integer without a # prefix"
4457         * config/msp430/msp430.md (epilogue_helper): Use it.
4459 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4461         PR target/59617
4462         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4463         AVX512F gather builtins.
4464         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4465         on gather decls with INTEGER_TYPE masktype.
4466         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4467         directly into the builtin rather than hoisting it before loop.
4469         PR tree-optimization/59387
4470         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4471         (scev_const_prop): If folded_casts and type has undefined overflow,
4472         use force_gimple_operand instead of force_gimple_operand_gsi and
4473         for each added stmt if it is assign with
4474         arith_code_with_undefined_signed_overflow, call
4475         rewrite_to_defined_overflow.
4476         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4477         gimple-fold.h instead.
4478         (arith_code_with_undefined_signed_overflow,
4479         rewrite_to_defined_overflow): Moved to ...
4480         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4481         rewrite_to_defined_overflow): ... here.  No longer static.
4482         Include gimplify-me.h.
4483         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4484         rewrite_to_defined_overflow): New prototypes.
4486 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4488         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4490 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4492         * builtins.c (get_object_alignment_2): Minor tweak.
4493         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4495 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4497         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4498         optimized non constant lengths.
4500 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4502         PR libgomp/59194
4503         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4504         load as __atomic_load_N if possible.
4506 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4508         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4509         target parameter.
4510         (rs6000_expand_builtin): Adjust call.
4512 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4514         PR target/58115
4515         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4516         * config/rs6000/rs6000.c: Include target-globals.h.
4517         (rs6000_set_current_function): Instead of doing target_reinit
4518         unconditionally, use save_target_globals_default_opts and
4519         restore_target_globals.
4521         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4522         FPSCR.
4523         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4524         (rs6000_expand_builtin): Handle mffs and mtfsf.
4525         (rs6000_init_builtins): Define mffs and mtfsf.
4526         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4527         (rs6000_mffs): New pattern.
4528         (rs6000_mtfsf): New pattern.
4530 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4532         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4533         Start narrowing with START.  Apply candidate-use pair
4534         and check overall cost in narrowing.
4535         (iv_ca_prune): Pass new argument.
4537 2014-01-10  Jeff Law  <law@redhat.com>
4539         PR middle-end/59743
4540         * ree.c (combine_reaching_defs): Ensure the defining statement
4541         occurs before the extension when optimizing extensions with
4542         different source and destination hard registers.
4544 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4546         PR ipa/58585
4547         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4548         vtables into the type inheritance graph.
4550 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4552         PR rtl-optimization/59754
4553         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4554         modes in the REGNO != REGNO case.
4556 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4558         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4560 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4562         PR tree-optimization/59745
4563         * tree-predcom.c (tree_predictive_commoning_loop): Call
4564         free_affine_expand_cache if giving up because components is NULL.
4566         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4567         GC in payload of target_globals struct instead of allocating them on
4568         the heap and the larger structs separately using GC.
4569         * target-globals.h (struct target_globals): Make regs, hard_regs,
4570         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4571         of GTY((skip)) and change type to void *.
4572         (reset_target_globals): Cast loads from those fields to corresponding
4573         types.
4575 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4577         PR plugins/59335
4578         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4579         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4580         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4582 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4584         PR target/59744
4585         * aarch64-modes.def (CC_Zmode): New flags mode.
4586         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4587         represents an equality.
4588         (aarch64_get_condition_code): Handle CC_Zmode.
4589         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4591 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4593         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4594         extraction in good case.
4596 2014-01-10  Richard Biener  <rguenther@suse.de>
4598         PR tree-optimization/59374
4599         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4600         checking after SLP discovery.  Mark stmts not participating
4601         in any SLP instance properly.
4603 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4605         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4606         when handling a SET rtx.
4608 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4610         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4611         (cortex-a57): Likewise.
4612         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4614 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4616         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4617         non-iwmmxt builtins.
4619 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4621         PR ipa/58252
4622         PR ipa/59226
4623         * ipa-devirt.c record_target_from_binfo): Take as argument
4624         stack of binfos and lookup matching one for virtual inheritance.
4625         (possible_polymorphic_call_targets_1): Update.
4627 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4629         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4630         kernel strings for Loongson-2E/2F/3A.
4632 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4634         PR middle-end/59670
4635         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4636         is_gimple_call before calling gimple_call_internal_p.
4638 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4640         * Makefile.in (TREE_FLOW_H): Remove.
4641         (TREE_SSA_H): Add file names from tree-flow.h.
4642         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4643         * tree.h: Remove tree-flow.h reference.
4644         * hash-table.h: Remove tree-flow.h reference.
4645         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4646         reference with tree-ssa-loop.h.
4648 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4650         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4651         default element-order behavior for -maltivec.
4652         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4653         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4654         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4655         when targeting big endian, at least for now.
4656         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4658 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4660         PR middle-end/47735
4661         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4662         var satisfies use_register_for_decl, just take into account type
4663         alignment, rather than decl alignment.
4665         PR tree-optimization/59622
4666         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4667         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4668         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4669         Don't devirtualize for inplace at all.  For targets.length () == 1,
4670         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4672 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4674         * config/i386/i386.md (cpu): Remove the unused btver1.
4676 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4678         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4680 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4682         PR target/58115
4683         * tree-core.h (struct target_globals): New forward declaration.
4684         (struct tree_target_option): Add globals field.
4685         * tree.h (TREE_TARGET_GLOBALS): Define.
4686         (prepare_target_option_nodes_for_pch): New prototype.
4687         * target-globals.h (struct target_globals): Define even if
4688         !SWITCHABLE_TARGET.
4689         * tree.c (prepare_target_option_node_for_pch,
4690         prepare_target_option_nodes_for_pch): New functions.
4691         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4692         * config/i386/i386.c: Include target-globals.h.
4693         (ix86_set_current_function): Instead of doing target_reinit
4694         unconditionally, use save_target_globals_default_opts and
4695         restore_target_globals.
4697 2014-01-09  Richard Biener  <rguenther@suse.de>
4699         PR tree-optimization/59715
4700         * tree-cfg.h (split_critical_edges): Declare.
4701         * tree-cfg.c (split_critical_edges): Export.
4702         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4704 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4706         * cfgexpand.c (expand_stack_vars): Optionally disable
4707         asan stack protection.
4708         (expand_used_vars): Likewise.
4709         (partition_stack_vars): Likewise.
4710         * asan.c (asan_emit_stack_protection): Optionally disable
4711         after return stack usage.
4712         (instrument_derefs): Optionally disable memory access instrumentation.
4713         (instrument_builtin_call): Likewise.
4714         (instrument_strlen_call): Likewise.
4715         (asan_protect_global): Optionally disable global variables protection.
4716         * doc/invoke.texi: Added doc for new options.
4717         * params.def: Added new options.
4718         * params.h: Likewise.
4720 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4722         PR rtl-optimization/59724
4723         * ifcvt.c (cond_exec_process_if_block): Don't call
4724         flow_find_head_matching_sequence with 0 longest_match.
4725         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4726         non-active insns if !stop_after.
4727         (try_head_merge_bb): Revert 2014-01-07 changes.
4729 2014-01-08  Jeff Law  <law@redhat.com>
4731         * ree.c (get_sub_rtx): New function, extracted from...
4732         (merge_def_and_ext): Here.
4733         (combine_reaching_defs): Use get_sub_rtx.
4735 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4737         * cgraph.h (varpool_variable_node): Do not choke on null node.
4739 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4741         * config/mips/mips.md (simple_return): Attempt to use JRC
4742         for microMIPS.
4743         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4745 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4747         PR rtl-optimization/59137
4748         * reorg.c (steal_delay_list_from_target): Call update_block for
4749         elided insns.
4750         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4752 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4754         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4755         two duplicate entries.
4757 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4759         Revert:
4760         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4762         * config/mips/mips.c (mips_truncated_op_cost): New function.
4763         (mips_rtx_costs): Adjust test for BADDU.
4764         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4766         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4768         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4769         (*baddu_si): ...this new pattern.
4771 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4773         PR ipa/59722
4774         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4776 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4778         PR middle-end/57748
4779         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4780         inner_reference_p.
4781         (expand_expr, expand_normal): Adjust.
4782         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4783         inner_reference_p. Use inner_reference_p to expand inner references.
4784         (store_expr): Adjust.
4785         * cfgexpand.c (expand_call_stmt): Adjust.
4787 2014-01-08  Rong Xu  <xur@google.com>
4789         * gcov-io.c (gcov_var): Move from gcov-io.h.
4790         (gcov_position): Ditto.
4791         (gcov_is_error): Ditto.
4792         (gcov_rewrite): Ditto.
4793         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4794         only part to libgcc/libgcov.h.
4796 2014-01-08  Marek Polacek  <polacek@redhat.com>
4798         PR middle-end/59669
4799         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4801 2014-01-08  Marek Polacek  <polacek@redhat.com>
4803         PR sanitizer/59667
4804         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4806 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4808         PR rtl-optimization/59649
4809         * stor-layout.c (get_mode_bounds): For BImode return
4810         0 and STORE_FLAG_VALUE.
4812 2014-01-08  Richard Biener  <rguenther@suse.de>
4814         PR middle-end/59630
4815         * gimple.h (is_gimple_builtin_call): Remove.
4816         (gimple_builtin_call_types_compatible_p): New.
4817         (gimple_call_builtin_p): New overload.
4818         * gimple.c (is_gimple_builtin_call): Remove.
4819         (validate_call): Rename to ...
4820         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4821         check return types.
4822         (validate_type): New static function.
4823         (gimple_call_builtin_p): New overload and adjust.
4824         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4825         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4826         (gimple_fold_stmt_to_constant_1): Likewise.
4827         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4829 2014-01-08  Richard Biener  <rguenther@suse.de>
4831         PR middle-end/59471
4832         * gimplify.c (gimplify_expr): Gimplify register-register type
4833         VIEW_CONVERT_EXPRs to separate stmts.
4835 2014-01-07  Jeff Law  <law@redhat.com>
4837         PR middle-end/53623
4838         * ree.c (combine_set_extension): Handle case where source
4839         and destination registers in an extension insn are different.
4840         (combine_reaching_defs): Allow source and destination registers
4841         in extension to be different under limited circumstances.
4842         (add_removable_extension): Remove restriction that the
4843         source and destination registers in the extension are the same.
4844         (find_and_remove_re): Emit a copy from the extension's
4845         destination to its source after the defining insn if
4846         the source and destination registers are different.
4848         PR middle-end/59285
4849         * ifcvt.c (merge_if_block): If we are merging a block with more than
4850         one successor with a block with no successors, remove any BARRIER
4851         after the second block.
4853 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4855         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4857 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4859         PR target/59652
4860         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4861         for 14-bit register offsets when INT14_OK_STRICT is false.
4863 2014-01-07  Roland Stigge  <stigge@antcom.de>
4864             Michael Meissner  <meissner@linux.vnet.ibm.com>
4866         PR 57386/target
4867         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4868         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
4870 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
4872         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4873         -mcpu.
4875 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
4877         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4878         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4879         rtx is const0_rtx or not.
4881 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
4883         PR target/58115
4884         * target-globals.c (save_target_globals): Remove this_fn_optab
4885         handling.
4886         * toplev.c: Include optabs.h.
4887         (target_reinit): Temporarily restore the global options if another
4888         set of options are in force.
4890 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
4892         PR rtl-optimization/58668
4893         * cfgcleanup.c (flow_find_cross_jump): Don't count
4894         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
4895         to determine what is counted.
4896         (flow_find_head_matching_sequence): Use active_insn_p to determine
4897         what is counted.
4898         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4899         counting change.
4900         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4901         determine what is counted.
4903         PR tree-optimization/59643
4904         * tree-predcom.c (split_data_refs_to_components): If one dr is
4905         read and one write, determine_offset fails and the write isn't
4906         in the bad component, just put the read into the bad component.
4908 2014-01-07  Mike Stump  <mikestump@comcast.net>
4909             Jakub Jelinek  <jakub@redhat.com>
4911         PR pch/59436
4912         * tree-core.h (struct tree_optimization_option): Change optabs
4913         type from unsigned char * to void *.
4914         * optabs.c (init_tree_optimization_optabs): Adjust
4915         TREE_OPTIMIZATION_OPTABS initialization.
4917 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
4919         PR target/59644
4920         * config/i386/i386.h (struct machine_function): Add
4921         no_drap_save_restore field.
4922         * config/i386/i386.c (ix86_save_reg): Use
4923         !cfun->machine->no_drap_save_restore instead of
4924         crtl->stack_realign_needed.
4925         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4926         this function clears frame_pointer_needed.  Set
4927         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4928         and DRAP reg is needed.
4930 2014-01-06  Marek Polacek  <polacek@redhat.com>
4932         PR c/57773
4933         * doc/implement-c.texi: Mention that other integer types are
4934         permitted as bit-field types in strictly conforming mode.
4936 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
4938         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4939         is newly allocated.
4941 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
4943         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4945 2014-01-06  Martin Jambor  <mjambor@suse.cz>
4947         PR ipa/59008
4948         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4949         to int.
4950         * ipa-prop.c (ipa_print_node_params): Fix indentation.
4952 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
4954         PR debug/59350
4955         PR debug/59510
4956         * var-tracking.c (add_stores): Preserve the value of the source even if
4957         we don't record the store.
4959 2014-01-06  Terry Guo  <terry.guo@arm.com>
4961         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4963 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
4965         PR bootstrap/59541
4966         * config/darwin.c (darwin_function_section): Adjust return values to
4967         correspond to optimisation changes made in r206070.
4969 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
4971         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4972         from prefetch_block tune setting.
4973         (nocona_cost): Correct size of prefetch block to 64.
4975 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
4977         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4978         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4979         used to save the static chain register in the computation of the offset
4980         from which the FP registers need to be restored.
4982 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
4984         PR tree-optimization/59519
4985         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4986         ICE if get_current_def (current_new_name) is already non-NULL, as long
4987         as it is a phi result of some other phi in *new_exit_bb that has
4988         the same argument.
4990         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4991         or vmovdqu* for misaligned_operand.
4992         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4993         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4994         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4995         aligned_mem for AVX512F masked aligned load and store builtins and for
4996         non-temporal moves.
4998 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5000         PR tree-optimization/59651
5001         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5002         Address range for negative step should be added by TYPE_SIZE_UNIT.
5004 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5006         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5007         overlapping registers also for registers other than the stack pointer.
5009 2014-01-03  Marek Polacek  <polacek@redhat.com>
5011         PR other/59661
5012         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5013         __builtin_FILE.
5015 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5017         PR target/59625
5018         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5019         asm goto as jump.
5021         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5022         (push splitter): Use <P:MODE_SIZE> instead of
5023         GET_MODE_SIZE (<P:MODE>mode).
5024         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5025         (mov -1, reg peephole2): Likewise.
5026         * config/i386/sse.md (*mov<mode>_internal,
5027         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5028         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5029         *<code><mode>3, *andnot<mode>3<mask_name>,
5030         <mask_codefor><code><mode>3<mask_name>): Likewise.
5031         * config/i386/subst.md (mask_mode512bit_condition,
5032         sd_mask_mode512bit_condition): Likewise.
5034 2014-01-02  Xinliang David Li  <davidxl@google.com>
5036         PR tree-optimization/59303
5037         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5038         (dump_predicates): Better output format.
5039         (pred_equal_p): New function.
5040         (is_neq_relop_p): Ditto.
5041         (is_neq_zero_form_p): Ditto.
5042         (pred_expr_equal_p): Ditto.
5043         (pred_neg_p): Ditto.
5044         (simplify_pred): Ditto.
5045         (simplify_preds_2): Ditto.
5046         (simplify_preds_3): Ditto.
5047         (simplify_preds_4): Ditto.
5048         (simplify_preds): Ditto.
5049         (push_pred): Ditto.
5050         (push_to_worklist): Ditto.
5051         (get_pred_info_from_cmp): Ditto.
5052         (is_degenerated_phi): Ditto.
5053         (normalize_one_pred_1): Ditto.
5054         (normalize_one_pred): Ditto.
5055         (normalize_one_pred_chain): Ditto.
5056         (normalize_preds): Ditto.
5057         (normalize_cond_1): Remove function.
5058         (normalize_cond): Ditto.
5059         (is_gcond_subset_of): Ditto.
5060         (is_subset_of_any): Ditto.
5061         (is_or_set_subset_of): Ditto.
5062         (is_and_set_subset_of): Ditto.
5063         (is_norm_cond_subset_of): Ditto.
5064         (pred_chain_length_cmp): Ditto.
5065         (convert_control_dep_chain_into_preds): Type change.
5066         (find_predicates): Ditto.
5067         (find_def_preds): Ditto.
5068         (destroy_predicates_vecs): Ditto.
5069         (find_matching_predicates_in_rest_chains): Ditto.
5070         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5071         (is_pred_expr_subset): Ditto.
5072         (is_pred_chain_subset_of): Ditto.
5073         (is_included_in): Ditto.
5074         (is_superset_of): Ditto.
5076 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5078         Update copyright years.
5080 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5082         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5083         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5084         config/arc/arc.md, config/arc/arc.opt,
5085         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5086         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5087         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5088         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5089         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5090         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5091         vtable-verify.c, vtable-verify.h: Use the standard form for the
5092         copyright notice.
5094 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5096         * gcc.c (process_command): Update copyright notice dates.
5097         * gcov-dump.c: Ditto.
5098         * gcov.c: Ditto.
5099         * doc/cpp.texi: Bump @copying's copyright year.
5100         * doc/cppinternals.texi: Ditto.
5101         * doc/gcc.texi: Ditto.
5102         * doc/gccint.texi: Ditto.
5103         * doc/gcov.texi: Ditto.
5104         * doc/install.texi: Ditto.
5105         * doc/invoke.texi: Ditto.
5107 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5109         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5111 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5113         * config/i386/sse.md (*mov<mode>_internal): Guard
5114         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5116         PR rtl-optimization/59647
5117         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5118         new_rtx into UNSIGNED_FLOAT rtxes.
5120 Copyright (C) 2014 Free Software Foundation, Inc.
5122 Copying and distribution of this file, with or without modification,
5123 are permitted in any medium without royalty provided the copyright
5124 notice and this notice are preserved.