2014-03-18 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blobc4b3b0ab842300dd09164e40032c28013cf624ec
1 2014-03-18  Richard Biener  <rguenther@suse.de>
3         * doc/loop.texi: Remove section on the removed lambda framework.
4         Update loop docs with recent changes in preserving loop structure.
6 2014-03-18  Richard Biener  <rguenther@suse.de>
8         * doc/lto.texi (-fresolution): Document.
10 2014-03-18  Richard Biener  <rguenther@suse.de>
12         * doc/contrib.texi: Adjust my name.
14 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
16         PR ipa/58721
17         * internal-fn.c: Include diagnostic-core.h.
18         (expand_BUILTIN_EXPECT): New function.
19         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
20         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
21         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
22         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
23         IFN_BUILTIN_EXPECT.
24         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
25         Revert 3 argument __builtin_expect code.
26         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
27         * gimple-fold.c (gimple_fold_call): Likewise.
28         * tree.h (fold_builtin_expect): New prototype.
29         * builtins.c (build_builtin_expect_predicate): Add predictor
30         argument, if non-NULL, create 3 argument __builtin_expect.
31         (fold_builtin_expect): No longer static.  Add ARG2 argument,
32         pass it through to build_builtin_expect_predicate.
33         (fold_builtin_2): Adjust caller.
34         (fold_builtin_3): Handle BUILT_IN_EXPECT.
35         * internal-fn.def (BUILTIN_EXPECT): New.
37 2014-03-18  Tobias Burnus  <burnus@net-b.de>
39         PR ipa/58721
40         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
41         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
42         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
44 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
46         PR ipa/58721
47         * predict.c (combine_predictions_for_bb): Fix up formatting.
48         (expr_expected_value_1, expr_expected_value): Add predictor argument,
49         fill what it points to if non-NULL.
50         (tree_predict_by_opcode): Adjust caller, use the predictor.
51         * predict.def (PRED_COMPARE_AND_SWAP): Add.
53 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
55         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
56         proper constant for the store mode.
58 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
60         * symtab.c (change_decl_assembler_name): Fix transparent alias
61         chain construction.
63 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
65         * config/aarch64/aarch64.c: Correct the comments about the
66         aarch64 stack layout.
68 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
70         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
71         check for GF_OMP_FOR_KIND_FOR.
73 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
75         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
76         ymm and zmm register names.
78 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
80         PR target/60516
81         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
82         note creation for the 2010-08-31 changes.
84 2014-03-17  Marek Polacek  <polacek@redhat.com>
86         PR middle-end/60534
87         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
88         as -fno-tree-loop-vectorize.
89         (expand_omp_simd): Likewise.
91 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
93         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
94         (eligible_for_call_delay): New prototype.
95         * config/sparc/sparc.c (tls_call_delay): Rename into...
96         (eligible_for_call_delay): ...this.  Return false if the instruction
97         cannot be put in the delay slot of a branch.
98         (eligible_for_restore_insn): Simplify.
99         (eligible_for_return_delay): Return false if the instruction cannot be
100         put in the delay slot of a branch and simplify.
101         (eligible_for_sibcall_delay): Return false if the instruction cannot be
102         put in the delay slot of a branch.
103         * config/sparc/sparc.md (fix_ut699): New attribute.
104         (tls_call_delay): Delete.
105         (in_call_delay): Reimplement.
106         (eligible_for_sibcall_delay): Rename into...
107         (in_sibcall_delay): ...this.
108         (eligible_for_return_delay): Rename into...
109         (in_return_delay): ...this.
110         (in_branch_delay): Reimplement.
111         (in_uncond_branch_delay): Delete.
112         (in_annul_branch_delay): Delete.
114 2014-03-14  Richard Henderson  <rth@redhat.com>
116         PR target/60525
117         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
118         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
119         (*floathi<X87MODEF>2_i387_with_temp): Remove.
120         (floathi splitters): Remove.
121         (float<SWI48x>xf2): New pattern.
122         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
123         code that tried to handle DImode for 32-bit, but which was excluded
124         by the pattern's condition.  Drop allocation of stack temporary.
125         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
126         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
127         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
128         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
129         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
130         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
131         (*float<SWI48><MODEF>2_sse_interunit): Remove.
132         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
133         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
134         (*float<SWI48x><X87MODEF>2_i387): Remove.
135         (all float _with_temp splitters): Remove.
136         (*float<SWI48x><MODEF>2_i387): New pattern.
137         (*float<SWI48><MODEF>2_sse): New pattern.
138         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
139         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
141 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
142             Marek Polacek  <polacek@redhat.com>
144         PR middle-end/60484
145         * common.opt (dump_base_name_prefixed): New Variable.
146         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
147         if x_dump_base_name_prefixed is already set, set it at the end.
149 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
151         PR rtl-optimization/60508
152         * lra-constraints.c (get_reload_reg): Add new parameter
153         in_subreg_p.
154         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
155         Pass the new parameter values.
157 2014-03-14  Richard Biener  <rguenther@suse.de>
159         * common.opt: Revert unintented changes from r205065.
160         * opts.c: Likewise.
162 2014-03-14  Richard Biener  <rguenther@suse.de>
164         PR middle-end/60518
165         * cfghooks.c (split_block): Properly adjust all loops the
166         block was a latch of.
168 2014-03-14  Martin Jambor  <mjambor@suse.cz>
170         PR lto/60461
171         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
172         and simplify it.
174 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
176         PR target/59396
177         * config/avr/avr.c (avr_set_current_function): Pass function name
178         through default_strip_name_encoding before sanity checking instead
179         of skipping the first char of the assembler name.
181 2014-03-13  Richard Henderson  <rth@redhat.com>
183         PR debug/60438
184         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
185         (ix86_force_to_memory, ix86_free_from_memory): Remove.
186         * config/i386/i386-protos.h: Likewise.
187         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
188         in the expander instead of a splitter.
189         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
190         any possibility of requiring a memory.
191         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
192         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
193         (fp branch splitters): Update for ix86_split_fp_branch.
194         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
195         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
196         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
197         (*fop_<MODEF>_2_i387): Remove f/r alternative.
198         (*fop_<MODEF>_3_i387): Likewise.
199         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
200         (splitters for the fop_* register patterns): Remove.
201         (fscalexf4_i387): Rename from *fscalexf4_i387.
202         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
203         
204 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
206         PR tree-optimization/59779
207         * tree-dfa.c (get_ref_base_and_extent): Use double_int
208         type for bitsize and maxsize instead of HOST_WIDE_INT.
210 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
212         PR rtl-optimization/57320
213         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
214         the CFG after thread_prologue_and_epilogue_insns.
216 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
218         PR rtl-optimization/57189
219         * lra-constraints.c (process_alt_operands): Disfavor spilling
220         vector pseudos.
222 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
224         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
225         messages.
227 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
229         PR tree-optimization/59025
230         PR middle-end/60418
231         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
232         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
234 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
236         PR target/60486
237         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
238         calls of avr_out_plus_1.
240 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
242         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
243         BB's single pred and update the father loop's latch info later.
245 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
247         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
248         (VEC_M): Likewise.
249         (VEC_N): Likewise.
250         (VEC_R): Likewise.
251         (VEC_base): Likewise.
252         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
253         registers, we need to swap double words in little endian mode.
255         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
256         to be a container mode for 128-bit integer operations added in ISA
257         2.07.  Unlike TImode and PTImode, the preferred register set is
258         the Altivec/VMX registers for the 128-bit operations.
260         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
261         declarations.
262         (rs6000_split_128bit_ok_p): Likewise.
264         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
265         macros for creating ISA 2.07 normal and overloaded builtin
266         functions with 3 arguments.
267         (BU_P8V_OVERLOAD_3): Likewise.
268         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
269         for use as overloaded functions.
270         (VPERM_1TI_UNS): Likewise.
271         (VSEL_1TI): Likewise.
272         (VSEL_1TI_UNS): Likewise.
273         (ST_INTERNAL_1ti): Likewise.
274         (LD_INTERNAL_1ti): Likewise.
275         (XXSEL_1TI): Likewise.
276         (XXSEL_1TI_UNS): Likewise.
277         (VPERM_1TI): Likewise.
278         (VPERM_1TI_UNS): Likewise.
279         (XXPERMDI_1TI): Likewise.
280         (SET_1TI): Likewise.
281         (LXVD2X_V1TI): Likewise.
282         (STXVD2X_V1TI): Likewise.
283         (VEC_INIT_V1TI): Likewise.
284         (VEC_SET_V1TI): Likewise.
285         (VEC_EXT_V1TI): Likewise.
286         (EQV_V1TI): Likewise.
287         (NAND_V1TI): Likewise.
288         (ORC_V1TI): Likewise.
289         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
290         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
291         overloaded builtin.
292         (VADDUQM): Likewise.
293         (VSUBCUQ): Likewise.
294         (VADDEUQM): Likewise.
295         (VADDECUQ): Likewise.
296         (VSUBEUQM): Likewise.
297         (VSUBECUQ): Likewise.
299         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
300         __int128_t and __uint128_t types.
301         (__uint128_type): Likewise.
302         (altivec_categorize_keyword): Add support for vector __int128_t,
303         vector __uint128_t, vector __int128, and vector unsigned __int128
304         as a container type for TImode operations that need to be done in
305         VSX/Altivec registers.
306         (rs6000_macro_to_expand): Likewise.
307         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
308         to support 128-bit integer instructions vaddcuq, vadduqm,
309         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
310         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
312         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
313         for V1TImode, and set up preferences to use VSX/Altivec registers.
314         Setup VSX reload handlers.
315         (rs6000_debug_reg_global): Likewise.
316         (rs6000_init_hard_regno_mode_ok): Likewise.
317         (rs6000_preferred_simd_mode): Likewise.
318         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
319         (easy_altivec_constant): Likewise.
320         (output_vec_const_move): Likewise.
321         (rs6000_expand_vector_set): Convert V1TImode set and extract to
322         simple move.
323         (rs6000_expand_vector_extract): Likewise.
324         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
325         addressing.
326         (rs6000_const_vec): Add support for V1TImode.
327         (rs6000_emit_le_vsx_load): Swap double words when loading or
328         storing TImode/V1TImode.
329         (rs6000_emit_le_vsx_store): Likewise.
330         (rs6000_emit_le_vsx_move): Likewise.
331         (rs6000_emit_move): Add support for V1TImode.
332         (altivec_expand_ld_builtin): Likewise.
333         (altivec_expand_st_builtin): Likewise.
334         (altivec_expand_vec_init_builtin): Likewise.
335         (altivec_expand_builtin): Likewise.
336         (rs6000_init_builtins): Add support for V1TImode type.  Add
337         support for ISA 2.07 128-bit integer builtins.  Define type names
338         for the VSX/Altivec vector types.
339         (altivec_init_builtins): Add support for overloaded vector
340         functions with V1TImode type.
341         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
342         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
343         external function.
344         (rs6000_split_128bit_ok_p): Likewise.
345         (rs6000_handle_altivec_attribute): Create V1TImode from vector
346         __int128_t and vector __uint128_t.
348         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
349         and mode attributes.
350         (VSX_M): Likewise.
351         (VSX_M2): Likewise.
352         (VSm): Likewise.
353         (VSs): Likewise.
354         (VSr): Likewise.
355         (VSv): Likewise.
356         (VS_scalar): Likewise.
357         (VS_double): Likewise.
358         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
360         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
361         we support the ISA 2.07 128-bit integer arithmetic instructions.
362         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
363         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
364         and TImode types for use with the builtin functions.
365         (V1TI_type_node): Likewise.
366         (unsigned_V1TI_type_node): Likewise.
367         (intTI_type_internal_node): Likewise.
368         (uintTI_type_internal_node): Likewise.
370         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
371         128-bit builtin functions.
372         (UNSPEC_VADDEUQM): Likewise.
373         (UNSPEC_VADDECUQ): Likewise.
374         (UNSPEC_VSUBCUQ): Likewise.
375         (UNSPEC_VSUBEUQM): Likewise.
376         (UNSPEC_VSUBECUQ): Likewise.
377         (VM): Add V1TImode to vector mode iterators.
378         (VM2): Likewise.
379         (VI_unit): Likewise.
380         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
381         (altivec_vaddcuq): Likewise.
382         (altivec_vsubuqm): Likewise.
383         (altivec_vsubcuq): Likewise.
384         (altivec_vaddeuqm): Likewise.
385         (altivec_vaddecuq): Likewise.
386         (altivec_vsubeuqm): Likewise.
387         (altivec_vsubecuq): Likewise.
389         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
390         mode iterators.
391         (BOOL_128): Likewise.
392         (BOOL_REGS_OUTPUT): Likewise.
393         (BOOL_REGS_OP1): Likewise.
394         (BOOL_REGS_OP2): Likewise.
395         (BOOL_REGS_UNARY): Likewise.
396         (BOOL_REGS_AND_CR0): Likewise.
398         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
399         128-bit integer builtin support.
400         (vec_vadduqm): Likewise.
401         (vec_vaddecuq): Likewise.
402         (vec_vaddeuqm): Likewise.
403         (vec_vsubecuq): Likewise.
404         (vec_vsubeuqm): Likewise.
405         (vec_vsubcuq): Likewise.
406         (vec_vsubuqm): Likewise.
408         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
409         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
410         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
411         128-bit integer add/subtract to ISA 2.07.
413 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
415         * config/arc/arc.c (arc_predicate_delay_insns):
416         Fix third argument passed to conditionalize_nonjump.
418 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
420         * config/aarch64/aarch64-builtins.c
421         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
422         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
423         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
424         instead of __builtin_lfloor.
425         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
427 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
429         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
430         (tree_ssa_ifcombine_bb_1): New function.
431         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
432         is an empty forwarder block to then_bb or vice versa and then_bb
433         and else_bb are effectively swapped.
435 2014-03-12  Christian Bruel  <christian.bruel@st.com>
437         PR target/60264
438         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
439         REG_CFA_DEF_CFA note.
440         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
441         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
443 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
445         PR tree-optimization/60454
446         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
448 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
450         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
451         Do not define target_cpu_default2 to generic.
452         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
453         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
454         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
456 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
457             Marc Glisse  <marc.glisse@inria.fr>
459         PR tree-optimization/60502
460         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
461         instead of build_low_bits_mask.
463 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
465         PR middle-end/60482
466         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
467         if there are multiple uses, but op doesn't live on E edge.
468         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
469         clobber stmts before __builtin_unreachable.
471 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
473         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
474         hard_frame_pointer_rtx.
475         * cse.c (cse_insn): Remove volatile check.
476         * cselib.c (cselib_process_insn): Likewise.
477         * dse.c (scan_insn): Likewise.
479 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
481         * config/arc/arc.c (conditionalize_nonjump): New function,
482         broken out of ...
483         (arc_ifcvt): ... this.
484         (arc_predicate_delay_insns): Use it.
486 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
488         * config/arc/predicates.md (extend_operand): During/after reload,
489         allow const_int_operand.
490         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
491         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
492         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
493         to "i".
494         (umulsi3_highpart_i): Likewise.
496 2014-03-11  Richard Biener  <rguenther@suse.de>
498         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
499         Add asserts to guard possible wrong-code bugs.
501 2014-03-11  Richard Biener  <rguenther@suse.de>
503         PR tree-optimization/60429
504         PR tree-optimization/60485
505         * tree-ssa-structalias.c (set_union_with_increment): Properly
506         take into account all fields that overlap the shifted vars.
507         (do_sd_constraint): Likewise.
508         (do_ds_constraint): Likewise.
509         (get_constraint_for_ptr_offset): Likewise.
511 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
513         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
514         (nios2_compute_frame_layout):
515         Add calculation of cfun->machine->fp_save_offset.
516         (nios2_expand_prologue): Correct setting of frame pointer register
517         in prologue.
518         (nios2_expand_epilogue): Update recovery of stack pointer from
519         frame pointer accordingly.
520         (nios2_initial_elimination_offset): Update calculation of offset
521         for eliminating to HARD_FRAME_POINTER_REGNUM.
523 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
525         PR ipa/60457
526         * ipa.c (symtab_remove_unreachable_nodes): Don't call
527         cgraph_get_create_node on VAR_DECLs.
529 2014-03-10  Richard Biener  <rguenther@suse.de>
531         PR middle-end/60474
532         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
534 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
536         * config/vms/vms.opt (vms_float_format): New variable.
538 2014-03-08  Tobias Burnus  <burnus@net-b.de>
540         * doc/invoke.texi (-fcilkplus): Update implementation status.
542 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
543             Richard Biener  <rguenther@suse.de>
545         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
546         consistently accross all TUs.
547         (run_gcc): Enable -fshort-double automatically at link at link-time
548         and disallow override.
550 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
552         PR target/58271
553         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
554         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
555         if they can't be used.
557 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
559         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
560         for Solaris 11/x86 ld.
561         * configure: Regenerate.
563 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
565         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
566         (LIB_TLS_SPEC): Save as ld_tls_libs.
567         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
568         (HAVE_AS_IX86_TLSLDM): New test.
569         * configure, config.in: Regenerate.
570         * config/i386/i386.c (legitimize_tls_address): Fall back to
571         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
572         cannot support TLS_MODEL_LOCAL_DYNAMIC.
573         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
574         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
576 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
578         * common.opt (fira-loop-pressure): Mark as optimization.
580 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
582         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
583         an OpenMP mappable type.
585 2014-03-06  Matthias Klose  <doko@ubuntu.com>
587         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
588         MULTILIB_OSDIRNAMES is not defined.
590 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
591             Meador Inge  <meadori@codesourcery.com>
593         PR target/58595
594         * config/arm/arm.c (arm_tls_symbol_p): Remove.
595         (arm_legitimize_address): Call legitimize_tls_address for any
596         arm_tls_referenced_p expression, handle constant addend.  Call it
597         before testing for !TARGET_ARM.
598         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
600 2014-03-06  Richard Biener  <rguenther@suse.de>
602         PR middle-end/60445
603         PR lto/60424
604         PR lto/60427
605         Revert
606         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
608         * tree-streamer.c (record_common_node): Assert we don't record
609         nodes with type double.
610         (preload_common_node): Skip type double, complex double and double
611         pointer since it is now frontend dependent due to fshort-double option.
613 2014-03-06  Richard Biener  <rguenther@suse.de>
615         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
616         or -fno-lto is specified and the linker has full plugin support.
617         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
618         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
619         * lto-wrapper.c (merge_and_complain): Merge compile-time
620         optimization levels.
621         (run_gcc): And pass it through to the link options.
623 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
625         PR debug/60381
626         Revert:
627         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
628         PR debug/59992
629         * cselib.c (remove_useless_values): Skip to avoid quadratic
630         behavior if the condition moved from...
631         (cselib_process_insn): ... here holds.
633 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
635         PR plugins/59335
636         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
637         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
639         PR plugins/59335
640         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
641         (TM_H): Add x86-tune.def.
643 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
645         * config/aarch64/aarch64.c (generic_tunings):
646         Use cortexa57_extra_costs.
648 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
650         PR lto/60404
651         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
652         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
653         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
654         cost for in_lto_p.
656 2014-03-04  Heiher  <r@hev.cc>
658         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
659         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
661 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
663         * config/i386/predicates.md (const2356_operand): Change to ...
664         (const2367_operand): ... this.
665         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
666         const2367_operand.
667         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
668         (*avx512pf_scatterpf<mode>sf): Ditto.
669         (avx512pf_scatterpf<mode>df): Ditto.
670         (*avx512pf_scatterpf<mode>df_mask): Ditto.
671         (*avx512pf_scatterpf<mode>df): Ditto.
672         * config/i386/i386.c (ix86_expand_builtin): Update
673         incorrect hint operand error message.
675 2014-03-04  Richard Biener  <rguenther@suse.de>
677         * lto-section-in.c (lto_get_section_data): Fix const cast.
679 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
681         * tree-streamer.c (record_common_node): Assert we don't record
682         nodes with type double.
683         (preload_common_node): Skip type double, complex double and double
684         pointer since it is now frontend dependent due to fshort-double option.
686 2014-03-04  Richard Biener  <rguenther@suse.de>
688         PR lto/60405
689         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
690         (lto_input_toplevel_asms): Likewise.
691         * lto-section-in.c (lto_get_section_data): Instead do it here
692         for every section.
694 2014-03-04  Richard Biener  <rguenther@suse.de>
696         PR tree-optimization/60382
697         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
698         dead PHIs a reduction.
700 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
702         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
703         hint value.
704         (_mm_prefetch): Move out of GCC target("sse") pragma.
705         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
706         GCC target("prfchw") pragma.
707         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
708         for locality <= 2.
709         * config/i386/i386.c (ix86_option_override_internal): Enable
710         -mprfchw with -mprefetchwt1.
712 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
714         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
715         Mark as varying.
717 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
719         * opts.h (CL_PCH_IGNORE): Define.
720         * targhooks.c (option_affects_pch_p):
721         Return false for options that have CL_PCH_IGNORE set.
722         * opt-functions.awk: Process PchIgnore.
723         * doc/options.texi: Document PchIgnore.
725         * config/arc/arc.opt (misize): Add PchIgnore property.
727 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
729         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
730         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
731         constraint on constants to permit them being loaded into
732         GENERAL_REGS or BASE_REGS.
734 2014-03-03  Nick Clifton  <nickc@redhat.com>
736         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
737         anti-cacnonical alternatives.
738         (negandhi3_real): New pattern.
739         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
741 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
743         * config/avr/avr-mcus.def: Remove atxmega16x1.
744         * config/avr/avr-tables.opt: Regenerate.
745         * config/avr/t-multilib: Regenerate.
746         * doc/avr-mmcu.texi: Regenerate.
748 2014-03-03  Tobias Grosser  <tobias@grosser.es>
749             Mircea Namolaru  <mircea.namolaru@inria.fr>
751         PR tree-optimization/58028
752         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
753         scalar dimensions.
755 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
757         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
758         not handled by recognizers.
760 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
762         PR middle-end/60175
763         * function.c (expand_function_end): Don't emit
764         clobber_return_register sequence if clobber_after is a BARRIER.
765         * cfgexpand.c (construct_exit_block): Append instructions before
766         return_label to prev_bb.
768 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
770         * config/rs6000/constraints.md: Document reserved use of "wc".
772 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
774         PR ipa/60150
775         * ipa.c (function_and_variable_visibility): When dissolving comdat
776         group, also set all symbols to local.
778 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
780         PR ipa/60306
782         Revert:
783         2013-12-14   Jan Hubicka  <jh@suse.cz>
784         PR middle-end/58477
785         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
787 2014-03-02  Jon Beniston  <jon@beniston.com>
789         PR bootstrap/48230
790         PR bootstrap/50927
791         PR bootstrap/52466
792         PR target/46898
793         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
794           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
795         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
796         (simple_return, *simple_return): New patterns
797         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
798         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
800 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
802         * dwarf2out.c (gen_subprogram_die): Tidy.
804 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
806         PR target/60071
807         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
808         (*mov_t_msb_neg_negc): ... this new insn.
810 2014-02-28  Jason Merrill  <jason@redhat.com>
812         PR c++/58678
813         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
814         function.
816 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
818         PR c++/60314
819         * dwarf2out.c (decltype_auto_die): New static.
820         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
821         (gen_type_die_with_usage): Handle 'decltype(auto)'.
822         (is_cxx_auto): Likewise.
824 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
826         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
827         we are not using general regs only.
829 2014-02-28  Richard Biener  <rguenther@suse.de>
831         PR target/60280
832         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
833         previous fix and only allow to remove trivial pre-headers
834         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
835         (remove_forwarder_block): Properly update the latch of a loop.
837 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
839         PR debug/59992
840         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
841         (cselib_preserved_hash_table): New.
842         (preserve_constants_and_equivs): Move preserved vals to it.
843         (cselib_find_slot): Look it up first.
844         (cselib_init): Initialize it.
845         (cselib_finish): Release it.
846         (dump_cselib_table): Dump it.
848 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
850         PR debug/59992
851         * cselib.c (remove_useless_values): Skip to avoid quadratic
852         behavior if the condition moved from...
853         (cselib_process_insn): ... here holds.
855 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
857         PR debug/57232
858         * var-tracking.c (vt_initialize): Apply the same condition to
859         preserve the CFA base value.
861 2014-02-28  Joey Ye  <joey.ye@arm.com>
863         PR target/PR60169
864         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
865         if reload in progress or completed.
867 2014-02-28  Tobias Burnus  <burnus@net-b.de>
869         PR middle-end/60147
870         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
871         NAMELIST_DECL.
873 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
875         * doc/tm.texi.in (Condition Code Status): Update documention for
876         relative locations of cc0-setter and cc0-user.
878 2014-02-27  Jeff Law  <law@redhat.com>
880         PR rtl-optimization/52714
881         * combine.c (try_combine): When splitting an unrecognized PARALLEL
882         into two independent simple sets, if I3 is a jump, ensure the
883         pattern we place into I3 is a (set (pc) ...).
885 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
886             Jeff Law  <law@redhat.com>
888         PR rtl-optimization/49847
889         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
890         are in different blocks.
891         * doc/tm.texi (Condition Code Status): Update documention for
892         relative locations of cc0-setter and cc0-user.
894 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
896         PR target/59222
897         * lra.c (lra_emit_add): Check SUBREG too.
899 2014-02-27  Andreas Schwab  <schwab@suse.de>
901         * config/m68k/m68k.c (m68k_option_override): Disable
902         -flive-range-shrinkage for classic m68k.
903         (m68k_override_options_after_change): Likewise.
905 2014-02-27  Marek Polacek  <polacek@redhat.com>
907         PR middle-end/59223
908         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
909         -Wmaybe-uninitialized.
911 2014-02-27  Alan Modra  <amodra@gmail.com>
913         PR target/57936
914         * reload1.c (emit_input_reload_insns): When reload_override_in,
915         set old to rl->in_reg when rl->in_reg is a subreg.
917 2014-02-26  Richard Biener  <rguenther@suse.de>
919         PR bootstrap/60343
920         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
922 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
924         * common/config/i386/predicates.md (const1256_operand): Remove.
925         (const2356_operand): New.
926         (const_1_to_2_operand): Remove.
927         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
928         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
929         (*avx512pf_gatherpf<mode>sf): Ditto.
930         (avx512pf_gatherpf<mode>df): Ditto.
931         (*avx512pf_gatherpf<mode>df_mask): Ditto.
932         (*avx512pf_gatherpf<mode>df): Ditto.
933         (avx512pf_scatterpf<mode>sf): Ditto.
934         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
935         (*avx512pf_scatterpf<mode>sf): Ditto.
936         (avx512pf_scatterpf<mode>df): Ditto.
937         (*avx512pf_scatterpf<mode>df_mask): Ditto.
938         (*avx512pf_scatterpf<mode>df): Ditto.
939         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
941 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
943         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
944         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
945         (_mm512_mask_testn_epi64_mask): Move to ...
946         * config/i386/avx512cdintrin.h: Here.
947         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
948         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
949         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
950         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
951         TARGET_AVX512F from TARGET_AVX512CD.
953 2014-02-26  Richard Biener  <rguenther@suse.de>
955         PR ipa/60327
956         * ipa.c (walk_polymorphic_call_targets): Properly guard
957         call to inline_update_overall_summary.
959 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
961         PR target/60280
962         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
963         and latches only if requested.  Fix latch if it is removed.
964         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
965         LOOPS_HAVE_PREHEADERS.
967 2014-02-25  Andrew Pinski  <apinski@cavium.com>
969         * builtins.c (expand_builtin_thread_pointer): Create a new target
970         when the target is NULL.
972 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
974         PR rtl-optimization/60317
975         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
976         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
977         * lra-assigns.c: Include params.h.
978         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
979         other reload pseudos considerations.
981 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
983         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
984         to use canonical form for nor<mode>3.
986 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
988         PR target/55426
989         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
990         conversions.
992 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
994         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
995         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
996         (ix86_handle_option): Handle OPT_mprefetchwt1.
997         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
998         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
999         PREFETCHWT1 CPUID.
1000         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1001         OPTION_MASK_ISA_PREFETCHWT1.
1002         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1003         (PTA_PREFETCHWT1): New.
1004         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1005         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1006         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1007         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1008         (*prefetch_avx512pf_<mode>_: Change into ...
1009         (*prefetch_prefetchwt1_<mode>: This.
1010         * config/i386/i386.opt (mprefetchwt1): New.
1011         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1012         (_mm_prefetch): Handle intent to write.
1013         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1015 2014-02-25  Richard Biener  <rguenther@suse.de>
1017         PR middle-end/60291
1018         * emit-rtl.c (mem_attrs_htab): Remove.
1019         (mem_attrs_htab_hash): Likewise.
1020         (mem_attrs_htab_eq): Likewise.
1021         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1022         (init_emit_once): Do not allocate mem_attrs_htab.
1024 2014-02-25  Richard Biener  <rguenther@suse.de>
1026         PR lto/60319
1027         * lto-opts.c (lto_write_options): Output non-explicit conservative
1028         -fwrapv, -fno-trapv and -fno-strict-overflow.
1029         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1030         (run_gcc): And pass them through.
1032 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1034         * sel-sched.c (calculate_new_fences): New parameter ptime.
1035         Calculate it as a maximum over all fence cycles.
1036         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1037         Print the final schedule timing when sched_verbose.
1039 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1041         PR rtl-optimization/60292
1042         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1043         bit fot the fence instruction.
1045 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1047         * calls.h: Fix typo in comment.
1049 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1051         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1052         adjusting offsetable addresses.
1054 2014-02-24  Guozhi Wei  <carrot@google.com>
1056         * sparseset.h (sparseset_pop): Fix the wrong index.
1058 2014-02-24  Walter Lee  <walt@tilera.com>
1060         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1061         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1062         triplet.
1063         * common/config/tilegx/tilegx-common.c
1064         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1065         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1066         (LINK_SPEC): Ditto.
1067         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1068         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1069         (tilegx_gimplify_va_arg_expr): Handle big endian.
1070         (tilegx_expand_unaligned_load): Ditto.
1071         (tilegx_expand_unaligned_store): Ditto.
1072         (TARGET_RETURN_IN_MSB): New.
1073         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1074         (TARGET_ENDIAN_DEFAULT): New.
1075         (TARGET_BIG_ENDIAN): Handle big endian.
1076         (BYTES_BIG_ENDIAN): Ditto.
1077         (WORDS_BIG_ENDIAN): Ditto.
1078         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1079         (ENDIAN_SPEC): New.
1080         (EXTRA_SPECS): New.
1081         * config/tilegx/tilegx.md (extv): Handle big endian.
1082         (extzv): Ditto.
1083         (insn_st<n>): Ditto.
1084         (insn_st<n>_add<bitsuffix>): Ditto.
1085         (insn_stnt<n>): Ditto.
1086         (insn_stnt<n>_add<bitsuffix>):Ditto.
1087         (vec_interleave_highv8qi): Handle big endian.
1088         (vec_interleave_highv8qi_be): New.
1089         (vec_interleave_highv8qi_le): New.
1090         (insn_v1int_h): Handle big endian.
1091         (vec_interleave_lowv8qi): Handle big endian.
1092         (vec_interleave_lowv8qi_be): New.
1093         (vec_interleave_lowv8qi_le): New.
1094         (insn_v1int_l): Handle big endian.
1095         (vec_interleave_highv4hi): Handle big endian.
1096         (vec_interleave_highv4hi_be): New.
1097         (vec_interleave_highv4hi_le): New.
1098         (insn_v2int_h): Handle big endian.
1099         (vec_interleave_lowv4hi): Handle big endian.
1100         (vec_interleave_lowv4hi_be): New.
1101         (vec_interleave_lowv4hi_le): New.
1102         (insn_v2int_l): Handle big endian.
1103         (vec_interleave_highv2si): Handle big endian.
1104         (vec_interleave_highv2si_be): New.
1105         (vec_interleave_highv2si_le): New.
1106         (insn_v4int_h): Handle big endian.
1107         (vec_interleave_lowv2si): Handle big endian.
1108         (vec_interleave_lowv2si_be): New.
1109         (vec_interleave_lowv2si_le): New.
1110         (insn_v4int_l): Handle big endian.
1111         * config/tilegx/tilegx.opt (mbig-endian): New option.
1112         (mlittle-endian): New option.
1113         * doc/install.texi: Document tilegxbe-linux.
1114         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1116 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1118         PR ipa/60266
1119         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1120         there are no parameter descriptors.
1122 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1124         PR rtl-optimization/60268
1125         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1126         initialization to ...
1127         (sched_rgn_init): ... here.
1128         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1130 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1132         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1133         names.
1135 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1137         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1138         definition.
1140 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1142         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1143         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1145 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1147         * config/microblaze/predicates.md: Add cmp_op predicate.
1148         * config/microblaze/microblaze.md: Add branch_compare instruction
1149         which uses cmp_op predicate and emits cmp insn before branch.
1150         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1151         to microblaze_expand_conditional_branch and consolidate logic.
1152         (microblaze_expand_conditional_branch): emit branch_compare
1153         insn instead of handling cmp op separate from branch insn.
1155 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1157         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1158         to permit subregs.
1160 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1162         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1163         define_insn with define_expand and new define_insn
1164         *altivec_lve<VI_char>x_internal.
1165         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1166         and new define_insn *altivec_stve<VI_char>x_internal.
1167         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1168         prototype.
1169         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1170         lve*x built-ins.
1171         (altivec_expand_stvex_be): New function.
1173 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1175         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1176         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1177         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1178         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1180 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1182         PR target/60298
1183         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1184         instead of emit_move_insn.
1186 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1188         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1189         vspltw with vsldoi.
1190         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1191         gen_altivec_vsumsws.
1193 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1195         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1196         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1197         (altivec_lvxl_<mode>): New define_expand incorporating
1198         -maltivec=be semantics where needed.
1199         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1200         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1201         semantics where needed.
1202         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1203         (altivec_stvx_<mode>): New define_expand incorporating
1204         -maltivec=be semantics where needed.
1205         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1206         VM2 iterator instead of V4SI.
1207         (altivec_stvxl_<mode>): New define_expand incorporating
1208         -maltivec=be semantics where needed.
1209         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1210         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1211         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1212         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1213         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1214         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1215         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1216         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1217         ALTIVEC_BUILTIN_STVXL.
1218         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1219         (altivec_expand_stvx_be): Likewise.
1220         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1221         (altivec_expand_lvx_be): Likewise.
1222         (altivec_expand_stvx_be): Likewise.
1223         (altivec_expand_builtin): Add cases for
1224         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1225         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1226         (altivec_init_builtins): Add definitions for
1227         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1228         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1230 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1232         * doc/invoke.texi (mvirt, mno-virt): Document.
1233         * config/mips/mips.opt (mvirt): New option.
1234         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1236 2014-02-21  Richard Biener  <rguenther@suse.de>
1238         PR tree-optimization/60276
1239         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1240         (STMT_VINFO_MIN_NEG_DIST): New macro.
1241         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1242         STMT_VINFO_MIN_NEG_DIST.
1243         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1244         made for negative dependence distances still hold.
1246 2014-02-21  Richard Biener  <rguenther@suse.de>
1248         PR middle-end/60291
1249         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1250         DECL_INITIAL for globals not in the current function context.
1252 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1254         PR tree-optimization/56490
1255         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1256         * tree-ssa-uninit.c: Include params.h.
1257         (compute_control_dep_chain): Add num_calls argument, return false
1258         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1259         num_calls to recursive call.
1260         (find_predicates): Change dep_chain into normal array,
1261         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1262         variable and adjust compute_control_dep_chain caller.
1263         (find_def_preds): Likewise.
1265 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1267         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1268         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1270 2014-02-21  Nick Clifton  <nickc@redhat.com>
1272         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1273         (pushhi1): Likewise.
1274         (popqi1): Add mode to pre_dec.
1275         (pophi1): Likewise.
1277 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1279         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1280         mode for mask of V8SFmode permutation.
1282 2014-02-20  Richard Henderson  <rth@redhat.com>
1284         PR c++/60272
1285         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1286         a new pseudo for OLDVAL.
1288 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1290         PR target/57896
1291         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1292         gen_reg_rtx if d->testing_p.
1293         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1294         if d->testing_p and we will certainly return true.
1295         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1296         if d->testing_p.
1298 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1300         * emit-rtl.c (gen_reg_rtx): Assert that
1301         crtl->emit.regno_pointer_align_length is non-zero.
1303 2014-02-20  Richard Henderson  <rth@redhat.com>
1305         PR c++/60272
1306         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1307         on failure the store back into EXPECT.
1309 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1310             Sandra Loosemore  <sandra@codesourcery.com>
1312         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1313         * config/nios2/nios2.c (nios2_function_profiler): Add
1314         -fPIC (flag_pic == 2) support.
1315         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1316         (nios2_large_offset_p): New function.
1317         (nios2_unspec_reloc_p): Move up position, update to use
1318         nios2_large_offset_p.
1319         (nios2_unspec_address): Remove function.
1320         (nios2_unspec_offset): New function.
1321         (nios2_large_got_address): New function.
1322         (nios2_got_address): Add large offset support.
1323         (nios2_legitimize_tls_address): Update usage of removed and new
1324         functions.
1325         (nios2_symbol_binds_local_p): New function.
1326         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1327         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1328         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1329         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1330         processing for (const (unspec ...)).
1331         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1333 2014-02-20  Richard Biener  <rguenther@suse.de>
1335         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1336         doing the substitution.
1337         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1339 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1341         PR ipa/55260
1342         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1343         info when checking whether lattices are bottom.
1345 2014-02-20  Richard Biener  <rguenther@suse.de>
1347         PR middle-end/60221
1348         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1349         regions at -O0.
1351 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1353         PR ipa/58555
1354         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1355         parameter specifying the scaling.
1356         (inline_call): Update.
1357         (want_inline_recursively): Guard division by zero.
1358         (recursive_inlining): Update.
1359         * ipa-inline.h (clone_inlined_nodes): Update.
1361 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1363         PR target/60204
1364         * config/i386/i386.c (classify_argument): Pass structures of size
1365         64 bytes or less in register.
1367 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1368             Kirill Yukhin  <kirill.yukhin@intel.com>
1370         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1371         (_mm_rcp28_round_ss): Ditto.
1372         (_mm_rsqrt28_round_sd): Ditto.
1373         (_mm_rsqrt28_round_ss): Ditto.
1374         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1375         (_mm_rcp14_round_ss): Ditto.
1376         (_mm_rsqrt14_round_sd): Ditto.
1377         (_mm_rsqrt14_round_ss): Ditto.
1378         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1379         the first input operand, get rid of match_dup.
1380         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1381         attribute to sse.
1382         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1383         Ditto.
1384         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1385         operand as the first input operand, set type attribute.
1386         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1387         Set type attribute.
1388         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1389         operand as the first input operand, set type attribute.
1391 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1393         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1394         bit of zero.
1396 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1398         PR target/60207
1399         * config/i386/i386.c (construct_container): Remove TFmode check
1400         for X86_64_INTEGER_CLASS.
1402 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1404         PR target/59794
1405         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1406         only when -Wpsabi is enabled.
1408 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1410          PR target/59799
1411         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1412         passing arrays in registers are the same as for structs, so remove the
1413         special case for them.
1415 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1417         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1418         destination type, extract only the valid bits if the source type is not
1419         integral and has a different mode.
1421 2014-02-19  Richard Biener  <rguenther@suse.de>
1423         PR ipa/60243
1424         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1425         for all calls.
1427 2014-02-19  Richard Biener  <rguenther@suse.de>
1429         PR ipa/60243
1430         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1431         (ipa_modify_call_arguments): Emit an argument load explicitely and
1432         preserve virtual SSA form there and for the replacement call.
1433         Do not update SSA form nor free dominance info.
1435 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1437         * ipa.c (function_and_variable_visibility): Also clear WEAK
1438         flag when disolving COMDAT_GROUP.
1440 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1442         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1443         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1444         not devirtualizing.
1445         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1446         do more sanity checks.
1447         (detect_type_change): Return true when giving up early.
1448         (compute_complex_assign_jump_func): Fix type parameter of
1449         ipa_set_ancestor_jf.
1450         (compute_complex_ancestor_jump_func): Likewise.
1451         (update_jump_functions_after_inlining): Fix updating of
1452         ancestor function.
1453         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1455 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1457         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1458         inline clones when edge disappears.
1460 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1462         PR target/60203
1463         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1464         Split 64-bit moves into 2 patterns.  Do not allow the use of
1465         direct move for TDmode in little endian, since the decimal value
1466         has little endian bytes within a word, but the 64-bit pieces are
1467         ordered in a big endian fashion, and normal subreg's of TDmode are
1468         not allowed.
1469         (mov<mode>_64bit_dm): Likewise.
1470         (movtd_64bit_nodm): Likewise.
1472 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1474         PR tree-optimization/60174
1475         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1476         statement of an SSA_NAME that occurs in an abnormal PHI node.
1478 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1480         PR sanitizer/60142
1481         * final.c (SEEN_BB): Remove.
1482         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1483         (final_scan_insn): Don't force_source_line on second
1484         NOTE_INSN_BASIC_BLOCK.
1486 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1488         PR target/60205
1489         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1490         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1491         (type_natural_mode): Warn ABI change when %zmm register is not
1492         available for AVX512F vector value passing.
1494 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1496         PR target/60193
1497         * config/i386/i386.c (ix86_expand_prologue): Use value in
1498         rax register as displacement when restoring %r10 or %rax.
1499         Fix wrong offset when restoring both registers.
1501 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1503         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1504         assertion with conditional return.
1506 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1507             Uros Bizjak  <ubizjak@gmail.com>
1509         PR driver/60233
1510         * config/i386/driver-i386.c (host_detect_local_cpu): If
1511         YMM state is not saved by the OS, also clear has_f16c.  Move
1512         CPUID 0x80000001 handling before YMM state saving checking.
1514 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1516         PR rtl-optimization/58960
1517         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1518         factored out from ...
1519         (sched_init): ... here.
1520         (free_global_sched_pressure_data): New, factored out from ...
1521         (sched_finish): ... here.
1522         * sched-int.h (free_global_sched_pressure_data): Declare.
1523         * sched-rgn.c (nr_regions_initial): New static global.
1524         (haifa_find_rgns): Initialize it.
1525         (schedule_region): Disable sched-pressure for the newly
1526         generated regions.
1528 2014-02-17  Richard Biener  <rguenther@suse.de>
1530         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1531         release SSA defs of pattern stmts.
1533 2014-02-17  Richard Biener  <rguenther@suse.de>
1535         * tree-inline.c (expand_call_inline): Release the virtual
1536         operand defined by the call we are about to inline.
1538 2014-02-17  Richard Biener  <rguenther@suse.de>
1540         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1542 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1543             Ilya Tocar  <ilya.tocar@intel.com>
1545         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1546         arguments order in builtin.
1547         (_mm512_permutexvar_epi64): Ditto.
1548         (_mm512_mask_permutexvar_epi64): Ditto
1549         (_mm512_maskz_permutexvar_epi32): Ditto
1550         (_mm512_permutexvar_epi32): Ditto
1551         (_mm512_mask_permutexvar_epi32): Ditto
1553 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1555         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1556         (p8_vmrgow): Likewise.
1558 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1560         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1561         endian targets.
1563 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1565         PR target/60203
1566         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1567         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1568         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1569         using direct move instructions on ISA 2.07.  Also adjust
1570         instruction length for 64-bit.
1571         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1572         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1574 2014-02-15  Alan Modra  <amodra@gmail.com>
1576         PR target/58675
1577         PR target/57935
1578         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1579         find_replacement on parts of insn rtl that might be reloaded.
1581 2014-02-15  Richard Biener  <rguenther@suse.de>
1583         PR tree-optimization/60183
1584         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1585         (tree_ssa_phiprop): Calculate and free post-dominators.
1587 2014-02-14  Jeff Law  <law@redhat.com>
1589         PR rtl-optimization/60131
1590         * ree.c (get_extended_src_reg): New function.
1591         (combine_reaching_defs): Use it rather than assuming location of REG.
1592         (find_and_remove_re): Verify first operand of extension is
1593         a REG before adding the insns to the copy list.
1595 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1597         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1598         * configure: Regenerated.
1599         * config.in: Regenerated.
1600         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1601         instead of ASM_SHORT.
1603 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1604             Richard Earnshaw  <rearnsha@arm.com>
1606         PR rtl-optimization/59535
1607         * lra-constraints.c (process_alt_operands): Encourage alternative
1608         when unassigned pseudo class is superset of the alternative class.
1609         (inherit_reload_reg): Don't inherit when optimizing for code size.
1610         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1611         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1612         modes not less than 4 for Thumb1.
1614 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1616         PR pch/60010
1617         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1619 2014-02-14  Richard Biener  <rguenther@suse.de>
1621         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1622         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1623         Do not strip INDIRECT_REFs.
1625 2014-02-14  Richard Biener  <rguenther@suse.de>
1627         PR lto/60179
1628         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1629         DECL_FUNCTION_SPECIFIC_TARGET.
1630         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1631         * tree-streamer-out.c (pack_ts_target_option): Remove.
1632         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1633         (write_ts_function_decl_tree_pointers): Do not stream
1634         DECL_FUNCTION_SPECIFIC_TARGET.
1635         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1636         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1637         (lto_input_ts_function_decl_tree_pointers): Do not stream
1638         DECL_FUNCTION_SPECIFIC_TARGET.
1640 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1642         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1643         (get_initial_def_for_induction, vectorizable_induction): Ignore
1644         debug stmts when looking for exit_phi.
1645         (vectorizable_live_operation): Fix up condition.
1647 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1649         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1650         nreverse() because it changes the content of original tree list.
1652 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1654         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1655         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1657 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1659         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1660         GNU coding standards.
1662 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1664         PR debug/60152
1665         * dwarf2out.c (gen_subprogram_die): Don't call
1666         add_calling_convention_attribute if subr_die is old_die.
1668 2014-02-13  Sharad Singhai  <singhai@google.com>
1670         * doc/optinfo.texi: Fix order of nodes.
1672 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1674         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1675         operands[2], not operands[3].
1677 2014-02-13  Richard Biener  <rguenther@suse.de>
1679         PR bootstrap/59878
1680         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1681         mention the possibility of an in-tree build.
1682         (CLooG): Update recommended version to 0.18.1, mention the
1683         possibility of an in-tree build and clarify that the ISL
1684         bundled with CLooG does not work.
1686 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1688         PR target/43546
1689         * expr.c (compress_float_constant): If x is a hard register,
1690         extend into a pseudo and then move to x.
1692 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1694         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1695         caused by bad second argument to warning_at() with -mhotpatch and
1696         nested functions (e.g. with gfortran).
1698 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1700         * opts.c (option_name): Remove "enabled by default" rider.
1702 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1704         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1706 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1707             Uros Bizjak  <ubizjak@gmail.com>
1709         PR target/60151
1710         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1711         * configure: Regenerated.
1713 2014-02-12  Richard Biener  <rguenther@suse.de>
1715         * vec.c (vec_prefix::calculate_allocation): Move as
1716         inline variant to vec.h.
1717         (vec_prefix::calculate_allocation_1): New out-of-line version.
1718         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1719         (vec_prefix::m_has_auto_buf): Rename to ...
1720         (vec_prefix::m_using_auto_storage): ... this.
1721         (vec_prefix::calculate_allocation): Inline the easy cases
1722         and dispatch to calculate_allocation_1 which doesn't need the
1723         prefix address.
1724         (va_heap::reserve): Use gcc_checking_assert.
1725         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1726         m_using_auto_storage.
1727         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1728         member and adjust.
1729         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1730         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1731         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1733 2014-02-12  Richard Biener  <rguenther@suse.de>
1735         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1736         when we found a dependence.
1738 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1740         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1741         common code...
1742         (maybe_fold_stmt): ... into this new function.
1743         * omp-low.c (lower_omp): Update comment.
1745         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1746         last use.
1748         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1749         dereference.
1751 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1753         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1754         identifiers in comments.
1755         (cortexa53_extra_costs): Likewise.
1756         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1757         (cortexa7_extra_costs): Likewise.
1758         (cortexa12_extra_costs): Likewise.
1759         (cortexa15_extra_costs): Likewise.
1760         (v7m_extra_costs): Likewise.
1762 2014-02-12  Richard Biener  <rguenther@suse.de>
1764         PR middle-end/60092
1765         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1766         of posix_memalign being successful.
1767         (lower_stmt): Restrict lowering of posix_memalign to when
1768         -ftree-bit-ccp is enabled.
1770 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1772         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1773         arg_loc.
1774         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1776 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1778         PR rtl-optimization/60116
1779         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1780         other_insn once the combination has been validated.
1782 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1784         PR lto/59468
1785         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1786         and wrapper.
1787         * ipa-devirt.c: Include demangle.h
1788         (odr_violation_reported): New static variable.
1789         (add_type_duplicate): Update odr_violations.
1790         (maybe_record_node): Add completep parameter; update it.
1791         (record_target_from_binfo): Add COMPLETEP parameter;
1792         update it as needed.
1793         (possible_polymorphic_call_targets_1): Likewise.
1794         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1795         rename FINAL to COMPLETE.
1796         (record_targets_from_bases): Sanity check we found the binfo;
1797         fix COMPLETEP updating.
1798         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1799         parameter, fix computing of COMPLETEP.
1800         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1801         at LTO time do demangling.
1802         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1803         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1804         parameter.
1805         (gimple_get_virt_method_for_binfo): Likewise.
1806         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1807         gimple_get_virt_method_for_vtable): Update prototypes.
1809 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1811         PR target/49008
1812         * genautomata.c (add_presence_absence): Fix typo with
1813         {final_}presence_list.
1815 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1817         PR target/60137
1818         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1819         for VSX/Altivec vectors that land in GPR registers.
1821 2014-02-11  Richard Henderson  <rth@redhat.com>
1822             Jakub Jelinek  <jakub@redhat.com>
1824         PR debug/59776
1825         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1826         around drhs if type conversion to lacc->type is not useless.
1828 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1830         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1831         tuning struct.
1832         (cortex-a57.cortex-a53): Likewise.
1833         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1835 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1837         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1838         arm_restrict_it.
1840 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1842         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1843         add_options_for_arm_vfp3.
1845 2014-02-11  Jeff Law  <law@redhat.com>
1847         PR middle-end/54041
1848         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1849         object with an undesirable mode.
1851 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1853         PR libgomp/60107
1854         * config/i386/sol2-9.h: New file.
1855         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1856         *-*-solaris2.9*): Use it.
1858 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1860         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1861         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1863 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1865         * config/microblaze/microblaze.c: Extend mcpu version format
1867 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
1869         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1871 2014-02-10  Richard Henderson  <rth@redhat.com>
1873         PR target/59927
1874         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1875         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1876         ms-abi vs -mno-accumulate-outgoing-args.
1877         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1878         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1879         respect to ms-abi.
1881 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1883         PR middle-end/60080
1884         * cfgexpand.c (expand_asm_operands): Attach source location to
1885         ASM_INPUT rtx objects.
1886         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1888 2014-02-10  Nick Clifton  <nickc@redhat.com>
1890         * config/mn10300/mn10300.c (popcount): New function.
1891         (mn10300_expand_prologue): Include saved registers in stack usage
1892         count.
1894 2014-02-10  Jeff Law  <law@redhat.com>
1896         PR middle-end/52306
1897         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1898         when changing the SET_DEST of a prior insn to avoid an input reload.
1900 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
1902         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1903         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1904         -mcall-openbsd, or -mcall-linux.
1905         (CC1_ENDIAN_BIG_SPEC): Remove.
1906         (CC1_ENDIAN_LITTLE_SPEC): Remove.
1907         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1908         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1909         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1910         and %cc1_endian_default.
1911         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1913 2014-02-10  Richard Biener  <rguenther@suse.de>
1915         PR tree-optimization/60115
1916         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1917         MEM_REF handling.  Properly verify that the accesses are not
1918         out of the objects bound.
1920 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1922         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1923         coretex to cortex.
1925 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
1927         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1928         proper constants and fix formatting.
1929         (possible_polymorphic_call_targets): Fix formatting.
1931 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
1932             Ilya Tocar  <ilya.tocar@intel.com>
1934         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1935         (_mm512_loadu_epi32): Renamed into...
1936         (_mm512_loadu_si512): This.
1937         (_mm512_storeu_epi32): Renamed into...
1938         (_mm512_storeu_si512): This.
1939         (_mm512_maskz_ceil_ps): Removed.
1940         (_mm512_maskz_ceil_pd): Ditto.
1941         (_mm512_maskz_floor_ps): Ditto.
1942         (_mm512_maskz_floor_pd): Ditto.
1943         (_mm512_floor_round_ps): Ditto.
1944         (_mm512_floor_round_pd): Ditto.
1945         (_mm512_ceil_round_ps): Ditto.
1946         (_mm512_ceil_round_pd): Ditto.
1947         (_mm512_mask_floor_round_ps): Ditto.
1948         (_mm512_mask_floor_round_pd): Ditto.
1949         (_mm512_mask_ceil_round_ps): Ditto.
1950         (_mm512_mask_ceil_round_pd): Ditto.
1951         (_mm512_maskz_floor_round_ps): Ditto.
1952         (_mm512_maskz_floor_round_pd): Ditto.
1953         (_mm512_maskz_ceil_round_ps): Ditto.
1954         (_mm512_maskz_ceil_round_pd): Ditto.
1955         (_mm512_expand_pd): Ditto.
1956         (_mm512_expand_ps): Ditto.
1957         * config/i386/i386.c (ix86_builtins): Remove
1958         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1959         (bdesc_args): Ditto.
1960         * config/i386/predicates.md (const1256_operand): New.
1961         (const_1_to_2_operand): Ditto.
1962         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1963         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1964         (*avx512pf_gatherpf<mode>sf): Ditto.
1965         (avx512pf_gatherpf<mode>df): Ditto.
1966         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1967         (*avx512pf_gatherpf<mode>df): Ditto.
1968         (avx512pf_scatterpf<mode>sf): Ditto.
1969         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1970         (*avx512pf_scatterpf<mode>sf): Ditto.
1971         (avx512pf_scatterpf<mode>df): Ditto.
1972         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1973         (*avx512pf_scatterpf<mode>df): Ditto.
1974         (avx512f_expand<mode>): Removed.
1975         (<shift_insn><mode>3<mask_name>): Change predicate type.
1977 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1979         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1980         not at the end of datarefs vector use ordered_remove to avoid
1981         reordering datarefs vector.
1983         PR c/59984
1984         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1985         mark local addressable non-static vars as GOVD_PRIVATE
1986         instead of GOVD_LOCAL.
1987         * omp-low.c (lower_omp_for): Move gimple_bind_vars
1988         and BLOCK_VARS of gimple_bind_block to new_stmt rather
1989         than copying them.
1991         PR middle-end/60092
1992         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
1993         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
1994         assume_aligned or alloc_align attributes.
1995         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
1996         arguments.  Handle also assume_aligned and alloc_align attributes.
1997         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
1998         calls to functions with assume_aligned or alloc_align attributes.
1999         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2001 2014-02-08  Terry Guo  <terry.guo@arm.com>
2003         * doc/invoke.texi: Document ARM -march=armv7e-m.
2005 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2007         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2008         flag on __cilkrts_rethrow builtin.
2010         PR ipa/60026
2011         * ipa-cp.c (determine_versionability): Fail at -O0
2012         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2013         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2015         Revert:
2016         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2018         PR ipa/60026
2019         * tree-inline.c (copy_forbidden): Fail for
2020         __attribute__((optimize (0))) functions.
2022 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2024         * varpool.c: Include pointer-set.h.
2025         (varpool_remove_unreferenced_decls): Variables in other partitions
2026         will not be output; be however careful to not lose information
2027         about partitioning.
2029 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2031         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2032         lookup in the vtable constructor.
2034 2014-02-07  Jeff Law  <law@redhat.com>
2036         PR target/40977
2037         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2038         define_insn_and_split.
2040         * ipa-inline.c (inline_small_functions): Fix typos.
2042 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2044         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2045         (s390_can_use_return_insn): Declare.
2046         * config/s390/s390.h (EPILOGUE_USES): Define.
2047         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2048         instructions.
2049         (s390_chunkify_start): Handle return JUMP_LABELs.
2050         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2051         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2052         (s390_can_use_return_insn): New functions.
2053         (s390_fix_long_loop_prediction): Handle conditional returns.
2054         (TARGET_SET_UP_BY_PROLOGUE): Define.
2055         * config/s390/s390.md (ANY_RETURN): New code iterator.
2056         (*creturn, *csimple_return, return, simple_return): New patterns.
2058 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2060         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2061         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2062         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2063         REG_CFA_RESTORE list when deciding not to restore a register.
2065 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2067         * config/s390/s390.c: Include tree-pass.h and context.h.
2068         (s390_early_mach): New function, split out from...
2069         (s390_emit_prologue): ...here.
2070         (pass_data_s390_early_mach): New pass structure.
2071         (pass_s390_early_mach): New class.
2072         (s390_option_override): Create and register early_mach pass.
2073         Move to end of file.
2075 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2077         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2078         to match for the exit block.
2080 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2082         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2083         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2084         Reject misaligned operands.
2086 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2088         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2090 2014-02-07  Richard Biener  <rguenther@suse.de>
2092         PR middle-end/60092
2093         * gimple-low.c (lower_builtin_posix_memalign): New function.
2094         (lower_stmt): Call it to lower posix_memalign in a way
2095         to make alignment info accessible.
2097 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2099         PR c++/60082
2100         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2101         __builtin_setjmp_receiver.
2103 2014-02-07  Richard Biener  <rguenther@suse.de>
2105         PR middle-end/60092
2106         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2107         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2108         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2109         Handle BUILT_IN_POSIX_MEMALIGN.
2110         (find_func_clobbers): Likewise.
2111         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2112         (call_may_clobber_ref_p_1): Likewise.
2114 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2116         PR ipa/59918
2117         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2118         sanity check.
2120 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2122         PR ipa/59469
2123         * lto-cgraph.c (lto_output_node): Use
2124         symtab_get_symbol_partitioning_class.
2125         (lto_output_varpool_node): likewise.
2126         (symtab_get_symbol_partitioning_class): Move here from
2127         lto/lto-partition.c
2128         * cgraph.h (symbol_partitioning_class): Likewise.
2129         (symtab_get_symbol_partitioning_class): Declare.
2131 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2133         * ggc.h (ggc_internal_cleared_alloc): New macro.
2134         * vec.h (vec_safe_copy): Handle memory stats.
2135         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2136         * target-globals.c (save_target_globals): Likewise.
2138 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2140         PR target/60077
2141         * expr.c (emit_move_resolve_push): Export; be bit more selective
2142         on when to clear alias set.
2143         * expr.h (emit_move_resolve_push): Declare.
2144         * function.h (struct function): Add tail_call_marked.
2145         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2146         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2147         * config/i386/i386.md (TImode move expander): De not call
2148         ix86_expand_push.
2149         (FP push expanders): Preserve memory attributes.
2150         * config/i386/sse.md (push<mode>1): Remove.
2151         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2152         (ix86_expand_push): Remove.
2153         * config/i386/mmx.md (push<mode>1): Remove.
2155 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2157         PR rtl-optimization/60030
2158         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2159         lopart with paradoxical subreg before shifting it up by hprec.
2161 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2163         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2164         Remove extra newline at end of file.
2165         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2166         (arm_issue_rate): Handle cortexa57.
2167         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2168         (cortex-a57.cortex-a53): Likewise.
2170 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2172         PR target/59575
2173         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2174         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2175         bitmask.
2176         (arm_expand_prologue): Adjust all callers.
2177         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2178         info, registers also at the lowest numbered registers side.  Use
2179         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2180         XEXP.
2182         PR debug/59992
2183         * var-tracking.c (adjust_mems): Before adding a SET to
2184         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2186 2014-02-06  Alan Modra  <amodra@gmail.com>
2188         PR target/60032
2189         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2190         change SDmode to DDmode when lra_in_progress.
2192 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2194         PR middle-end/59150
2195         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2196         free_data_ref on the dr first, and before goto again also set dr
2197         to the next dr.  For simd_lane_access, free old datarefs[i] before
2198         overwriting it.  For get_vectype_for_scalar_type failure, don't
2199         free_data_ref if simd_lane_access.
2201         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2203         PR target/60062
2204         * tree.h (opts_for_fn): New inline function.
2205         (opt_for_fn): Define.
2206         * config/i386/i386.c (ix86_function_regparm): Use
2207         opt_for_fn (decl, optimize) instead of optimize.
2209 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2211         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2212         for SYMBOL_REF in large memory model.
2214 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2216         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2217         and crypto support.
2218         (cortex-a57): Likewise.
2219         (cortex-a57.cortex-a53): Likewise.
2221 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2222             Kugan Vivekanandarajah  <kuganv@linaro.org>
2224         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2225         unaligned_access.
2226         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2228 2014-02-06  Richard Biener  <rguenther@suse.de>
2230         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2231         set_loop_copy and initialize_original_copy_tables.
2233 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2235         * config/aarch64/aarch64-simd.md
2236         (aarch64_ashr_simddi): Change QI to SI.
2238 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2239             Jakub Jelinek  <jakub@redhat.com>
2241         PR middle-end/60013
2242         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2243         of the dataflow.
2245 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2247         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2248         CODE_FOR_altivec_vpku[hw]um to
2249         CODE_FOR_altivec_vpku[hw]um_direct.
2250         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2251         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2252         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2253         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2255 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2257         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2258         generation for -maltivec=be.
2259         (altivec_vsumsws): Simplify redundant test.
2261 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2263         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2264         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2265         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2266         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2267         gen_altivec_vpkuwum.
2268         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2269         BYTES_BIG_ENDIAN.
2270         (altivec_vpks<VI_char>ss): Likewise.
2271         (altivec_vpks<VI_char>us): Likewise.
2272         (altivec_vpku<VI_char>us): Likewise.
2273         (altivec_vpku<VI_char>um): Likewise.
2274         (altivec_vpku<VI_char>um_direct): New (copy of
2275         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2276         internal use).
2277         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2278         target is little endian and -maltivec=be is not specified.
2279         (*altivec_vupkhs<VU_char>_direct): New (copy of
2280         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2281         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2282         target is little endian and -maltivec=be is not specified.
2283         (*altivec_vupkls<VU_char>_direct): New (copy of
2284         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2285         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2286         little endian and -maltivec=be is not specified.
2287         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2288         little endian and -maltivec=be is not specified.
2290 2014-02-05  Richard Henderson  <rth@redhat.com>
2292         PR debug/52727
2293         * combine-stack-adj.c: Revert r206943.
2294         * sched-int.h (struct deps_desc): Add last_args_size.
2295         * sched-deps.c (init_deps): Initialize it.
2296         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2297         contain REG_ARGS_SIZE notes.
2299 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2301         * lto-cgraph.c (asm_nodes_output): Make global.
2302         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2303         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2304         (driver_handle_option): Handle OPT_fwpa.
2306 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2308         PR ipa/59947
2309         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2310         a comment typo and formatting issue.  If odr_hash hasn't been
2311         created, return vNULL and set *completep to false.
2313         PR middle-end/57499
2314         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2315         bb with no successors.
2317 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2319         PR target/59718
2320         * doc/invoke.texi (-march): Clarify documentation for ARM.
2321         (-mtune): Likewise.
2322         (-mcpu): Likewise.
2324 2014-02-05  Richard Biener  <rguenther@suse.de>
2326         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2327         when not vectorizing because of too many alias checks.
2328         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2329         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2331 2014-02-05  Nick Clifton  <nickc@redhat.com>
2333         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2334         accept extended registers in any mode when compiling for the MN10300.
2336 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2338         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2339         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2340         sanitization attributes.
2341         (can_inline_edge_p): Likewise.
2342         (sanitize_attrs_match_for_inline_p): New function.
2344 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2346         * ipa-prop.c (detect_type_change): Shor circuit testing of
2347         type changes on THIS pointer.
2349 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2351         PR target/59777
2352         * config/pa/pa.c (legitimize_tls_address): Return original address
2353         if not passed a SYMBOL_REF rtx.
2354         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2355         addresses.
2356         (pa_emit_move_sequence): Simplify TLS source operands.
2357         (pa_legitimate_constant_p): Reject all TLS constants.
2358         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2359         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2361 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2363         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2364         groups when we know they are controlled by LTO.
2365         * varasm.c (default_binds_local_p_1): If object is in other partition,
2366         it will be resolved locally.
2368 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2370         * config/host-linux.c (linux_gt_pch_use_address): Don't
2371         use SSIZE_MAX because it is not always defined.
2373 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2375         PR bootstrap/59913
2376         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2377         threshold for pseudo splitting.
2378         (update_ebb_live_info): Process call argument hard registers and
2379         hard registers from insn definition too.
2380         (max_small_class_regs_num): New constant.
2381         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2382         reloads_num only for small register classes.  Don't split for
2383         outputs of jumps.
2385 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2387         PR ipa/60058
2388         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2389         is non-null.
2391 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2393         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2394         visibility is safe.
2396 2014-02-04  Marek Polacek  <polacek@redhat.com>
2398         * gdbinit.in (pel): Define.
2400 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2402         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2403         behavior.
2405 2014-02-04  Richard Biener  <rguenther@suse.de>
2407         PR lto/59723
2408         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2409         in function context local.
2410         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2411         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2412         similar to LTO_imported_decl_ref.
2414 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2416         PR tree-optimization/60002
2417         * cgraphclones.c (build_function_decl_skip_args): Clear
2418         DECL_LANG_SPECIFIC.
2420         PR tree-optimization/60023
2421         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2422         false to gsi_replace.
2423         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2424         has been in some EH region and vec_stmt could throw, add
2425         vec_stmt into the same EH region.
2426         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2427         has no lhs, ignore it.
2428         * internal-fn.c (expand_MASK_LOAD): Likewise.
2430         PR ipa/60026
2431         * tree-inline.c (copy_forbidden): Fail for
2432         __attribute__((optimize (0))) functions.
2434         PR other/58712
2435         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2436         is set, copy one less argument.
2437         (expand_simd_clones): Don't subtract clone_info->inbranch
2438         from simd_clone_struct_alloc argument.
2440         PR rtl-optimization/57915
2441         * recog.c (simplify_while_replacing): If all unary/binary/relational
2442         operation arguments are constant, attempt to simplify those.
2444         PR middle-end/59261
2445         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2446         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2448 2014-02-04  Richard Biener  <rguenther@suse.de>
2450         PR tree-optimization/60012
2451         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2452         TBAA disambiguation to all DDRs.
2454 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2456         PR target/59788
2457         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2458         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2460 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2462         PR ipa/59882
2463         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2465 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2467         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2468         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2470 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2472         PR ipa/59831
2473         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2474         to figure out targets of polymorphic calls with known decl.
2475         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2476         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2477         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2478         (get_polymorphic_call_info): ... here.
2479         (get_polymorphic_call_info_from_invariant): New function.
2481 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2483         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2484         lookup via vtable pointer; check for type consistency
2485         and turn inconsitent facts into UNREACHABLE.
2486         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2487         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2488         type inconsistent querries; return UNREACHABLE instead.
2490 2014-02-03  Richard Henderson  <rth@twiddle.net>
2492         PR tree-opt/59924
2493         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2494         already processed this node.
2495         (normalize_one_pred_1): Pass along mark_set.
2496         (normalize_one_pred): Create and destroy a pointer_set_t.
2497         (normalize_one_pred_chain): Likewise.
2499 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2501         PR gcov-profile/58602
2502         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2504 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2506         PR ipa/59831
2507         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2508         -fno-devirtualize; try to devirtualize by the knowledge of
2509         virtual table pointer given by aggregate propagation.
2510         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2511         (ipa_print_node_jump_functions): Dump also offset that
2512         is relevant for polymorphic calls.
2513         (determine_known_aggregate_parts): Add arg_type parameter; use it
2514         instead of determining the type from pointer type.
2515         (ipa_compute_jump_functions_for_edge): Update call of
2516         determine_known_aggregate_parts.
2517         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2518         (gimple_get_virt_method_for_binfo): ... here; simplify using
2519         vtable_pointer_value_to_vtable.
2520         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2521         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2522         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2523         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2524         POINTER_PLUS_EXPR.
2525         (vtable_pointer_value_to_binfo): ... here.
2526         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2528 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2530         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2531         redef of outer loop index variable.
2533 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2535         PR c++/53017
2536         PR c++/59211
2537         * doc/extend.texi (Function Attributes): Typo.
2539 2014-02-03  Cong Hou  <congh@google.com>
2541         PR tree-optimization/60000
2542         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2543         if the vectorized statement is a store.  A store statement can only
2544         appear at the end of pattern statements.
2546 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2548         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2549         (ix86_option_override_internal): Default long double to 64-bit for
2550         32-bit Bionic and to 128-bit for 64-bit Bionic.
2552         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2553         TARGET_LONG_DOUBLE_128 is true.
2554         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2556         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2557         (mlong-double-64): Negate -mlong-double-128.
2558         (mlong-double-128): New option.
2560         * config/i386/i386-c.c (ix86_target_macros): Define
2561         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2563         * doc/invoke.texi: Document -mlong-double-128.
2565 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2567         PR rtl-optimization/60024
2568         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2570 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2572         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2574 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2576         PR rtl-optimization/57662
2577         * sel-sched.c (code_motion_path_driver): Do not mark already not
2578         existing blocks in the visiting bitmap.
2580 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2582         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2583         on the insn being emitted.
2585 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2586             Will Deacon  <will.deacon@arm.com>
2588         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2590 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2592         * config/arm/arm-tables.opt: Regenerate.
2594 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2596         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2597         for vector types other than V16QImode.
2598         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2599         define_expand, and call altivec_expand_vec_perm_le when producing
2600         code with little endian element order.
2601         (*altivec_vperm_<mode>_internal): New insn having previous
2602         behavior of altivec_vperm_<mode>.
2603         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2604         altivec_expand_vec_perm_le when producing code with little endian
2605         element order.
2606         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2607         behavior of altivec_vperm_<mode>_uns.
2609 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2611         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2612         (altivec_vsumsws): Add handling for -maltivec=be with a little
2613         endian target.
2614         (altivec_vsumsws_direct): New.
2615         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2616         gen_altivec_vsumsws.
2618 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2620         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2621         vtable_pointer_value_to_binfo): New functions.
2622         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2623         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2625 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2627         * config/nios2/nios2.md (load_got_register): Initialize GOT
2628         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2629         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2631 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2633         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2634         preserverd by passthrough, do not propagate the type.
2636 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2638         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2639         (mips_atomic_assign_expand_fenv): New function.
2640         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2642 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2644         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2645         (__builtin_mips_set_fcsr): Likewise.
2646         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2647         MIPS_USI_FTYPE_VOID.
2648         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2649         (mips16_expand_set_fcsr): Likewise.
2650         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2651         (mips16_set_fcsr_stub): Likewise.
2652         (mips16_get_fcsr_one_only_stub): New class.
2653         (mips16_set_fcsr_one_only_stub): Likewise.
2654         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2655         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2656         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2657         (hard_float): New availability predicate.
2658         (mips_builtins): Add get_fcsr and set_fcsr.
2659         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2660         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2661         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2662         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2663         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2664         patterns.
2666 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2668         * config/mips/mips.c (mips_one_only_stub): New class.
2669         (mips_need_mips16_rdhwr_p): Replace with...
2670         (mips16_rdhwr_stub): ...this new variable.
2671         (mips16_stub_call_address): New function.
2672         (mips16_rdhwr_one_only_stub): New class.
2673         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2674         (mips_output_mips16_rdhwr): Delete.
2675         (mips_finish_stub): New function.
2676         (mips_code_end): Use it to handle rdhwr stubs.
2678 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2680         PR target/60017
2681         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2682         when calculating size of integer atomic types.
2684 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2686         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2688 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2690         PR tree-optimization/60003
2691         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2692         * profile.c (branch_prob): Use gimple_call_builtin_p
2693         to check for BUILT_IN_SETJMP_RECEIVER.
2694         * tree-inline.c (copy_bb): Call notice_special_calls.
2696 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2698         PR bootstrap/59985
2699         * lra-constraints.c (process_alt_operands): Update reload_sum only
2700         on the first pass.
2702 2014-01-31  Richard Henderson  <rth@redhat.com>
2704         PR middle-end/60004
2705         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2706         until after else_eh is processed.
2708 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2710         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2711         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2712         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2713         in smmintrin.h, remove them.
2714         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2715         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2716         * config/i386/i386.md (ROUND_SAE): Fix value.
2717         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2718         (const48_operand): New.
2719         * config/i386/subst.md (round), (round_expand): Use
2720         const_4_or_8_to_11_operand.
2721         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2723 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2725         * config/i386/constraints.md (Yk): Swap meaning with k.
2726         * config/i386/i386.md (movhi_internal): Change Yk to k.
2727         (movqi_internal): Ditto.
2728         (*k<logic><mode>): Ditto.
2729         (*andhi_1): Ditto.
2730         (*andqi_1): Ditto.
2731         (kandn<mode>): Ditto.
2732         (*<code>hi_1): Ditto.
2733         (*<code>qi_1): Ditto.
2734         (kxnor<mode>): Ditto.
2735         (kortestzhi): Ditto.
2736         (kortestchi): Ditto.
2737         (kunpckhi): Ditto.
2738         (*one_cmplhi2_1): Ditto.
2739         (*one_cmplqi2_1): Ditto.
2740         * config/i386/sse.md (): Change k to Yk.
2741         (avx512f_load<mode>_mask): Ditto.
2742         (avx512f_blendm<mode>): Ditto.
2743         (avx512f_store<mode>_mask): Ditto.
2744         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2745         (avx512f_storedqu<mode>_mask): Ditto.
2746         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2747         Ditto.
2748         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2749         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2750         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2751         (avx512f_maskcmp<mode>3): Ditto.
2752         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2753         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2754         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2755         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2756         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2757         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2758         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2759         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2760         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2761         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2762         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2763         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2764         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2765         (vec_extract_lo_<mode>_maskm): Ditto.
2766         (vec_extract_hi_<mode>_maskm): Ditto.
2767         (avx512f_vternlog<mode>_mask): Ditto.
2768         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2769         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2770         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2771         (avx512f_<code>v8div16qi2_mask): Ditto.
2772         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2773         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2774         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2775         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2776         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2777         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2778         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2779         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2780         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2781         (avx512cd_maskb_vec_dupv8di): Ditto.
2782         (avx512cd_maskw_vec_dupv16si): Ditto.
2783         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2784         (avx512f_vpermi2var<mode>3_mask): Ditto.
2785         (avx512f_vpermi2var<mode>3_mask): Ditto.
2786         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2787         (*avx512f_gathersi<mode>): Ditto.
2788         (*avx512f_gathersi<mode>_2): Ditto.
2789         (*avx512f_gatherdi<mode>): Ditto.
2790         (*avx512f_gatherdi<mode>_2): Ditto.
2791         (*avx512f_scattersi<mode>): Ditto.
2792         (*avx512f_scatterdi<mode>): Ditto.
2793         (avx512f_compress<mode>_mask): Ditto.
2794         (avx512f_compressstore<mode>_mask): Ditto.
2795         (avx512f_expand<mode>_mask): Ditto.
2796         * config/i386/subst.md (mask): Change k to Yk.
2797         (mask_scalar_merge): Ditto.
2798         (sd): Ditto.
2800 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2802         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2804 2014-01-31  Richard Biener  <rguenther@suse.de>
2806         PR middle-end/59990
2807         * builtins.c (fold_builtin_memory_op): Make sure to not
2808         use a floating-point mode or a boolean or enumeral type for
2809         the copy operation.
2811 2014-01-30  DJ Delorie  <dj@redhat.com>
2813         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2814         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2815         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2816         whenever main() has an epilogue.
2818 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2820         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2821         unused variable "field".
2822         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2823         (vsx_mergeh_<mode>): Likewise.
2824         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2825         (altivec_vmrghh): Likewise.
2826         (altivec_vmrghw): Likewise.
2827         (altivec_vmrglb): Likewise.
2828         (altivec_vmrglh): Likewise.
2829         (altivec_vmrglw): Likewise.
2830         (altivec_vspltb): Add missing uses.
2831         (altivec_vsplth): Likewise.
2832         (altivec_vspltw): Likewise.
2833         (altivec_vspltsf): Likewise.
2835 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2837         PR target/59923
2838         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2839         frame related instructions.
2841 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2843         PR rtl-optimization/59959
2844         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2845         any reload of register whose subreg is invalid.
2847 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2849         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2850         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2851         Add missing return type - void.
2853 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2855         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2856         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2857         remove element index adjustment for endian (now handled in vsx.md
2858         and altivec.md).
2859         (altivec_expand_vec_perm_const): Use
2860         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2861         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2862         (vsx_xxspltw_<mode>): Adjust element index for little endian.
2863         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2864         define_expand and a new define_insn *altivec_vspltb_internal;
2865         adjust for -maltivec=be on a little endian target.
2866         (altivec_vspltb_direct): New.
2867         (altivec_vsplth): Divide into a define_expand and a new
2868         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2869         little endian target.
2870         (altivec_vsplth_direct): New.
2871         (altivec_vspltw): Divide into a define_expand and a new
2872         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2873         little endian target.
2874         (altivec_vspltw_direct): New.
2875         (altivec_vspltsf): Divide into a define_expand and a new
2876         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2877         a little endian target.
2879 2014-01-30  Richard Biener  <rguenther@suse.de>
2881         PR tree-optimization/59993
2882         * tree-ssa-forwprop.c (associate_pointerplus): Check we
2883         can propagate form the earlier stmt and avoid the transform
2884         when the intermediate result is needed.
2886 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
2888         * README.Portability: Fix typo.
2890 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
2892         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2893         comparison_operator with ordered_comparison_operator.
2895 2014-01-30  Nick Clifton  <nickc@redhat.com>
2897         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2898         Rename to mn10300_store_multiple_regs.
2899         * config/mn10300/mn10300.c: Likewise.
2900         * config/mn10300/mn10300.md (store_movm): Fix typo: call
2901         store_multiple_regs.
2902         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2903         Call mn10300_store_multiple_regs.
2905 2014-01-30  Nick Clifton  <nickc@redhat.com>
2906             DJ Delorie  <dj@redhat.com>
2908         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2909         %fp 2 to keep registers after it properly word-aligned.
2910         (rl78_alloc_physical_registers_umul): Handle the case where both
2911         input operands are the same.
2913 2014-01-30  Richard Biener  <rguenther@suse.de>
2915         PR tree-optimization/59903
2916         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2917         check properly.
2919 2014-01-30  Jason Merrill  <jason@redhat.com>
2921         PR c++/59633
2922         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2924         PR c++/59645
2925         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2927 2014-01-30  Richard Biener  <rguenther@suse.de>
2929         PR tree-optimization/59951
2930         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2932 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
2934         PR target/59784
2935         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2936         SFmode to DFmode case.
2938 2014-01-29  DJ Delorie  <dj@redhat.com>
2940         * config/msp430/msp430.opt (-minrt): New.
2941         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2942         if -minrt given.
2943         (ENDFILE_SPEC): Likewise.
2945 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
2947         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2948         (estimate_function_body_sizes): Use it.
2950 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
2952         PR c++/58561
2953         * dwarf2out.c (is_cxx_auto): New.
2954         (is_base_type): Use it.
2955         (gen_type_die_with_usage): Likewise.
2957 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2959         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
2960         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2961         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2962         -maltivec=be with LE targets.
2963         (vsx_mergeh_<mode>): Likewise.
2964         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2965         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2966         (altivec_vmrghb): Replace with define_expand and new
2967         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2968         (altivec_vmrghb_direct): New define_insn.
2969         (altivec_vmrghh): Replace with define_expand and new
2970         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2971         (altivec_vmrghh_direct): New define_insn.
2972         (altivec_vmrghw): Replace with define_expand and new
2973         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2974         (altivec_vmrghw_direct): New define_insn.
2975         (*altivec_vmrghsf): Adjust for endianness.
2976         (altivec_vmrglb): Replace with define_expand and new
2977         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2978         (altivec_vmrglb_direct): New define_insn.
2979         (altivec_vmrglh): Replace with define_expand and new
2980         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2981         (altivec_vmrglh_direct): New define_insn.
2982         (altivec_vmrglw): Replace with define_expand and new
2983         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2984         (altivec_vmrglw_direct): New define_insn.
2985         (*altivec_vmrglsf): Adjust for endianness.
2986         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2987         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2988         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2989         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2990         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2991         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2992         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2993         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2995 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2997         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
2998         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
2999         whitespace.
3001 2014-01-29  Richard Biener  <rguenther@suse.de>
3003         PR tree-optimization/58742
3004         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3005         associate_pointerplus_align.
3006         (associate_pointerplus_diff): New function.
3007         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3008         and associate_pointerplus_diff.
3010 2014-01-29  Richard Biener  <rguenther@suse.de>
3012         * lto-streamer.h (LTO_major_version): Bump to 3.
3013         (LTO_minor_version): Reset to 0.
3015 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3017         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3018         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3019         (arm_file_start): Generate correct asm header for armv7ve.
3020         * config/arm/bpabi.h: Add multilib support for armv7ve.
3021         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3022         and cortex-a15 to armv7ve.
3023         * config/arm/t-aprofile: Add multilib support for armv7ve.
3024         * doc/invoke.texi: Document -march=armv7ve.
3026 2014-01-29  Richard Biener  <rguenther@suse.de>
3028         PR tree-optimization/58742
3029         * tree-ssa-forwprop.c (associate_plusminus): Return true
3030         if we changed sth, defer EH cleanup to ...
3031         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3032         (simplify_mult): New function.
3034 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3036         PR middle-end/59917
3037         PR tree-optimization/59920
3038         * tree.c (build_common_builtin_nodes): Remove
3039         __builtin_setjmp_dispatcher initialization.
3040         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3041         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3042         instead of gsi_after_labels + manually skipping debug stmts.
3043         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3044         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3045         * tree-inline.c (copy_edges_for_bb): Remove
3046         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3047         argument.  Ignore computed_goto_p stmts.  Don't call
3048         make_abnormal_goto_edges.  If a call might need abnormal edges
3049         for non-local gotos, see if it already has an edge to
3050         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3051         with true argument, don't do anything then, otherwise add
3052         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3053         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3054         caller.
3055         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3056         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3057         (lower_stmt): Don't set data->calls_builtin_setjmp.
3058         (lower_builtin_setjmp): Adjust comment.
3059         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3060         * tree-cfg.c (found_computed_goto): Remove.
3061         (factor_computed_gotos): Remove.
3062         (make_goto_expr_edges): Return bool, true for computed gotos.
3063         Don't call make_abnormal_goto_edges.
3064         (build_gimple_cfg): Don't set found_computed_goto, don't call
3065         factor_computed_gotos.
3066         (computed_goto_p): No longer static.
3067         (make_blocks): Don't set found_computed_goto.
3068         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3069         (make_edges): If make_goto_expr_edges returns true, push bb
3070         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3071         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3072         vector.  Record mapping between bbs and OpenMP regions if there
3073         are any, adjust make_gimple_omp_edges caller.  Call
3074         handle_abnormal_edges.
3075         (make_abnormal_goto_edges): Remove.
3076         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3077         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3078         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3079         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3080         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3081         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3082         filling *region also set *region_idx to (*region)->entry->index.
3084         PR other/58712
3085         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3086         For REGs set ORIGINAL_REGNO.
3088 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3090         * doc/md.texi: Mention that a target shouldn't implement
3091         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3092         than hi/lo pair.
3094 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3096         PR tree-optimization/59594
3097         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3098         a copy of the datarefs vector rather than the vector itself.
3100 2014-01-28  Jason Merrill  <jason@redhat.com>
3102         PR c++/53756
3103         * dwarf2out.c (auto_die): New static.
3104         (gen_type_die_with_usage): Handle C++1y 'auto'.
3105         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3106         on definition.
3108 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3110         PR target/59672
3111         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3112         (SPEC_X32): Likewise.
3113         (SPEC_64): Likewise.
3114         * config/i386/i386.c (ix86_option_override_internal): Turn off
3115         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3116         for TARGET_16BIT.
3117         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3118         * config/i386/i386.h (TARGET_16BIT): New macro.
3119         (TARGET_16BIT_P): Likewise.
3120         * config/i386/i386.opt: Add m16.
3121         * doc/invoke.texi: Document -m16.
3123 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3125         PR preprocessor/59935
3126         * input.c (location_get_source_line): Bail out on when line number
3127         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3129 2014-01-28  Richard Biener  <rguenther@suse.de>
3131         PR tree-optimization/58742
3132         * tree-ssa-forwprop.c (associate_plusminus): Handle
3133         pointer subtraction of the form (T)(P + A) - (T)P.
3135 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3137         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3138         at const_int_cost.
3140 2014-01-28  Richard Biener  <rguenther@suse.de>
3142         Revert
3143         2014-01-28  Richard Biener  <rguenther@suse.de>
3145         PR rtl-optimization/45364
3146         PR rtl-optimization/59890
3147         * var-tracking.c (local_get_addr_clear_given_value): Handle
3148         already cleared slot.
3149         (val_reset): Handle not allocated local_get_addr_cache.
3150         (vt_find_locations): Use post-order on the inverted CFG.
3152 2014-01-28  Richard Biener  <rguenther@suse.de>
3154         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3156 2014-01-28  Richard Biener  <rguenther@suse.de>
3158         PR rtl-optimization/45364
3159         PR rtl-optimization/59890
3160         * var-tracking.c (local_get_addr_clear_given_value): Handle
3161         already cleared slot.
3162         (val_reset): Handle not allocated local_get_addr_cache.
3163         (vt_find_locations): Use post-order on the inverted CFG.
3165 2014-01-28  Alan Modra  <amodra@gmail.com>
3167         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3168         * configure.ac <recursive call for build != host>: Define
3169         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3170         and LD_FOR_BUILD too.
3171         * configure: Regenerate.
3173 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3175         * config/i386/i386.c (get_builtin_code_for_version): Separate
3176         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3177         Broadwell from Haswell.
3179 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3181         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3182         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3183         * config/mips/mips.c (mips_option_override): Change setting
3184         of TARGET_DSP.
3185         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3186         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3187         Change from Mask to Var.
3189 2014-01-27  Jeff Law  <law@redhat.com>
3191         * ipa-inline.c (inline_small_functions): Fix typo.
3193 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3195         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3196         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3197         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3198         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3199         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3200         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3201         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3202         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3203         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3204         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3205         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3206         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3207         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3208         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3209         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3210         (_mm512_storeu_epi64): Ditto.
3211         (_mm512_cmpge_epi32_mask): Ditto.
3212         (_mm512_cmpge_epu32_mask): Ditto.
3213         (_mm512_cmpge_epi64_mask): Ditto.
3214         (_mm512_cmpge_epu64_mask): Ditto.
3215         (_mm512_cmple_epi32_mask): Ditto.
3216         (_mm512_cmple_epu32_mask): Ditto.
3217         (_mm512_cmple_epi64_mask): Ditto.
3218         (_mm512_cmple_epu64_mask): Ditto.
3219         (_mm512_cmplt_epi32_mask): Ditto.
3220         (_mm512_cmplt_epu32_mask): Ditto.
3221         (_mm512_cmplt_epi64_mask): Ditto.
3222         (_mm512_cmplt_epu64_mask): Ditto.
3223         (_mm512_cmpneq_epi32_mask): Ditto.
3224         (_mm512_cmpneq_epu32_mask): Ditto.
3225         (_mm512_cmpneq_epi64_mask): Ditto.
3226         (_mm512_cmpneq_epu64_mask): Ditto.
3227         (_mm512_expand_pd): Ditto.
3228         (_mm512_expand_ps): Ditto.
3229         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3230         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3231         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3232         * config/i386/i386.c (ix86_builtins): Add
3233         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3234         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3235         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3236         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3237         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3238         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3239         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3240         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3241         IX86_BUILTIN_PMOVUSQW512_MEM.
3242         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3243         __builtin_ia32_pmovsqd512mem_mask,
3244         __builtin_ia32_pmovqd512mem_mask,
3245         __builtin_ia32_pmovusqw512mem_mask,
3246         __builtin_ia32_pmovsqw512mem_mask,
3247         __builtin_ia32_pmovqw512mem_mask,
3248         __builtin_ia32_pmovusdw512mem_mask,
3249         __builtin_ia32_pmovsdw512mem_mask,
3250         __builtin_ia32_pmovdw512mem_mask,
3251         __builtin_ia32_pmovqb512mem_mask,
3252         __builtin_ia32_pmovusqb512mem_mask,
3253         __builtin_ia32_pmovsqb512mem_mask,
3254         __builtin_ia32_pmovusdb512mem_mask,
3255         __builtin_ia32_pmovsdb512mem_mask,
3256         __builtin_ia32_pmovdb512mem_mask.
3257         (bdesc_args): Add __builtin_ia32_expanddf512,
3258         __builtin_ia32_expandsf512.
3259         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3260         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3261         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3262         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3263         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3264         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3265         (avx512f_<code>v8div16qi2_mask_store): This.
3266         (avx512f_expand<mode>): New.
3268 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3270         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3271         New.
3272         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3273         (_mm512_prefetch_i32scatter_pd): Ditto.
3274         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3275         (_mm512_prefetch_i64scatter_pd): Ditto.
3276         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3277         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3278         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3279         (_mm512_prefetch_i32scatter_ps): Ditto.
3280         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3281         (_mm512_prefetch_i64scatter_ps): Ditto.
3282         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3283         * config/i386/i386-builtin-types.def: Define
3284         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3285         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3286         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3287         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3288         IX86_BUILTIN_SCATTERPFQPD.
3289         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3290         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3291         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3292         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3293         __builtin_ia32_scatterpfqps.
3294         (ix86_expand_builtin): Expand new built-ins.
3295         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3296         fix memory access data type.
3297         (*avx512pf_gatherpf<mode>_mask): Ditto.
3298         (*avx512pf_gatherpf<mode>): Ditto.
3299         (avx512pf_scatterpf<mode>): Ditto.
3300         (*avx512pf_scatterpf<mode>_mask): Ditto.
3301         (*avx512pf_scatterpf<mode>): Ditto.
3302         (GATHER_SCATTER_SF_MEM_MODE): New.
3303         (avx512pf_gatherpf<mode>df): Ditto.
3304         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3305         (*avx512pf_scatterpf<mode>df): Ditto.
3307 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3309         PR bootstrap/59934
3310         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3311         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3312         reached.
3314 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3316         * common/config/arm/arm-common.c
3317         (arm_rewrite_mcpu): Handle multiple names.
3318         * config/arm/arm.h
3319         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3321 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3323         * gimple-builder.h (create_gimple_tmp): Delete.
3325 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3327         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3328         words comparisons.
3330 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3332         * config/pa/pa.md (call): Generate indirect long calls to non-local
3333         functions when outputing 32-bit code.
3334         (call_value): Likewise except for special call to buggy powf function.
3336         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3337         portable runtime and PIC indirect calls.
3338         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3339         and PIC call sequences.  Use ldo instead of blr to set return register
3340         in PIC call sequence.
3342 2014-01-25  Walter Lee  <walt@tilera.com>
3344         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3345         avoid clobbering a live register.
3347 2014-01-25  Walter Lee  <walt@tilera.com>
3349         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3350         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3351         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3352         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3354 2014-01-25  Walter Lee  <walt@tilera.com>
3356         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3357         arguments on even registers.
3358         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3359         STACK_BOUNDARY.
3360         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3361         (BIGGEST_ALIGNMENT): Ditto.
3362         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3364 2014-01-25  Walter Lee  <walt@tilera.com>
3366         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3367         insns before bundling.
3368         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3370 2014-01-25  Walter Lee  <walt@tilera.com>
3372         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3373         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3374         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3376 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3378         * config/mips/constraints.md (kl): Delete.
3379         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3380         define expands, using...
3381         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3382         instructions for MIPS16.
3383         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3384         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3386 2014-01-25  Walter Lee  <walt@tilera.com>
3388         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3389         (clzdi2): Ditto.
3390         (ffsdi2): Ditto.
3392 2014-01-25  Walter Lee  <walt@tilera.com>
3394         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3395         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3397 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3399         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3400         Handle XOR.
3402 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3404         * print-rtl.c (in_call_function_usage): New var.
3405         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3406         EXPR_LIST mode as mode and not as reg note name.
3408         PR middle-end/59561
3409         * cfgloopmanip.c (copy_loop_info): If
3410         loop->warned_aggressive_loop_optimizations, make sure
3411         the flag is set in target loop too.
3413 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3415         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3416         flag_cilkplus.
3417         * builtins.def: Likewise.
3418         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3419         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3420         * ira.c (ira_setup_eliminable_regset): Likewise.
3421         * omp-low.c (gate_expand_omp): Likewise.
3422         (execute_lower_omp): Likewise.
3423         (diagnose_sb_0): Likewise.
3424         (gate_diagnose_omp_blocks): Likewise.
3425         (simd_clone_clauses_extract): Likewise.
3426         (gate): Likewise.
3428 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3430         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3431         correction for little endian...
3432         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3433         here.
3435 2014-01-24  Jeff Law  <law@redhat.com>
3437         PR tree-optimization/59919
3438         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3439         for non-returning calls.
3441 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3443         * common/config/aarch64/aarch64-common.c
3444         (aarch64_rewrite_mcpu): Handle multiple names.
3445         * config/aarch64/aarch64.h
3446         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3448 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3450         * input.c (add_file_to_cache_tab): Handle the case where fopen
3451         returns NULL.
3453 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3455         PR target/59929
3456         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3457         from push operand if code of push isn't PRE_DEC.
3459 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3461         PR target/59909
3462         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3463         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3464         it is only used for non-atomic loads/stores.
3466         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3467         -mquad-memory or -mquad-memory-atomic switches.
3469         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3470         -mquad-memory-atomic to ISA 2.07 support.
3472         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3473         to separate support of normal quad word memory operations (ldq, stq)
3474         from the atomic quad word memory operations.
3476         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3477         support to separate non-atomic quad word operations from atomic
3478         quad word operations.  Disable non-atomic quad word operations in
3479         little endian mode so that we don't have to swap words after the
3480         load and before the store.
3481         (quad_load_store_p): Add comment about atomic quad word support.
3482         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3483         options printed with -mdebug=reg.
3485         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3486         -mquad-memory-atomic as the test for whether we have quad word
3487         atomic instructions.
3488         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3489         or -mp8-vector are used, allow byte/half-word atomic operations.
3491         * config/rs6000/sync.md (load_lockedti): Insure that the address
3492         is a proper indexed or indirect address for the lqarx instruction.
3493         On little endian systems, swap the hi/lo registers after the lqarx
3494         instruction.
3495         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3496         insure the address is valid for the lqarx instruction.
3497         (store_conditionalti): Insure that the address is a proper indexed
3498         or indirect address for the stqcrx. instruction.  On little endian
3499         systems, swap the hi/lo registers before doing the stqcrx.
3500         instruction.
3501         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3502         insure the address is valid for the stqcrx. instruction.
3504         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3505         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3506         type of quad memory support is available.
3508 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3510         PR regression/59915
3511         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3512         there is a danger of looping.
3514 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3516         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3517         force flag_ira_loop_pressure if set via command line.
3519 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3521         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3522         (ashr_simd): New builtin handling DI mode.
3523         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3524         (aarch64_sshr_simddi): New match pattern.
3525         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3526         (vshrd_n_s64): Likewise.
3527         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3529 2014-01-23  Nick Clifton  <nickc@redhat.com>
3531         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3532         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3533         favour of mcu specific scripts.
3534         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3535         430x multilibs.
3537 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3538             Alex Velenko  <Alex.Velenko@arm.com>
3540         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3541         (vaddv_s16): Likewise.
3542         (vaddv_s32): Likewise.
3543         (vaddv_u8): Likewise.
3544         (vaddv_u16): Likewise.
3545         (vaddv_u32): Likewise.
3546         (vaddvq_s8): Likewise.
3547         (vaddvq_s16): Likewise.
3548         (vaddvq_s32): Likewise.
3549         (vaddvq_s64): Likewise.
3550         (vaddvq_u8): Likewise.
3551         (vaddvq_u16): Likewise.
3552         (vaddvq_u32): Likewise.
3553         (vaddvq_u64): Likewise.
3554         (vaddv_f32): Likewise.
3555         (vaddvq_f32): Likewise.
3556         (vaddvq_f64): Likewise.
3557         (vmaxv_f32): Likewise.
3558         (vmaxv_s8): Likewise.
3559         (vmaxv_s16): Likewise.
3560         (vmaxv_s32): Likewise.
3561         (vmaxv_u8): Likewise.
3562         (vmaxv_u16): Likewise.
3563         (vmaxv_u32): Likewise.
3564         (vmaxvq_f32): Likewise.
3565         (vmaxvq_f64): Likewise.
3566         (vmaxvq_s8): Likewise.
3567         (vmaxvq_s16): Likewise.
3568         (vmaxvq_s32): Likewise.
3569         (vmaxvq_u8): Likewise.
3570         (vmaxvq_u16): Likewise.
3571         (vmaxvq_u32): Likewise.
3572         (vmaxnmv_f32): Likewise.
3573         (vmaxnmvq_f32): Likewise.
3574         (vmaxnmvq_f64): Likewise.
3575         (vminv_f32): Likewise.
3576         (vminv_s8): Likewise.
3577         (vminv_s16): Likewise.
3578         (vminv_s32): Likewise.
3579         (vminv_u8): Likewise.
3580         (vminv_u16): Likewise.
3581         (vminv_u32): Likewise.
3582         (vminvq_f32): Likewise.
3583         (vminvq_f64): Likewise.
3584         (vminvq_s8): Likewise.
3585         (vminvq_s16): Likewise.
3586         (vminvq_s32): Likewise.
3587         (vminvq_u8): Likewise.
3588         (vminvq_u16): Likewise.
3589         (vminvq_u32): Likewise.
3590         (vminnmv_f32): Likewise.
3591         (vminnmvq_f32): Likewise.
3592         (vminnmvq_f64): Likewise.
3594 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3596         * config/aarch64/aarch64-simd.md
3597         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3598         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3599         (*aarch64_mul3_elt<mode>): Likewise.
3600         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3601         (*aarch64_mul3_elt_to_64v2df): Likewise.
3602         (*aarch64_mla_elt<mode>): Likewise.
3603         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3604         (*aarch64_mls_elt<mode>): Likewise.
3605         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3606         (*aarch64_fma4_elt<mode>): Likewise.
3607         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3608         (*aarch64_fma4_elt_to_64v2df): Likewise.
3609         (*aarch64_fnma4_elt<mode>): Likewise.
3610         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3611         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3612         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3613         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3614         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3615         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3616         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3617         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3618         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3620 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3622         * config/aarch64/aarch64-simd.md
3623         (aarch64_be_checked_get_lane<mode>): New define_expand.
3624         * config/aarch64/aarch64-simd-builtins.def
3625         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3626         New builtin definition.
3627         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3628         Use new safe be builtin.
3630 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3632         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3633         New define_insn.
3634         (aarch64_be_st1<mode>): Likewise.
3635         (aarch_ld1<VALL:mode>): Define_expand modified.
3636         (aarch_st1<VALL:mode>): Likewise.
3637         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3638         (UNSPEC_ST1): Likewise.
3640 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3642         * config/microblaze/microblaze.md: Add trap insn and attribute
3644 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3646         PR preprocessor/58580
3647         * input.h (location_get_source_line): Take an additional line_size
3648         parameter.
3649         (void diagnostics_file_cache_fini): Declare new function.
3650         * input.c (struct fcache): New type.
3651         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3652         New static constants.
3653         (diagnostic_file_cache_init, total_lines_num)
3654         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3655         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3656         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3657         (get_next_line, read_next_line, goto_next_line, read_line_num):
3658         New static function definitions.
3659         (diagnostic_file_cache_fini): New function.
3660         (location_get_source_line): Take an additional output line_len
3661         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3662         read_line_num.
3663         * diagnostic.c (diagnostic_finish): Call
3664         diagnostic_file_cache_fini.
3665         (adjust_line): Take an additional input parameter for the length
3666         of the line, rather than calculating it with strlen.
3667         (diagnostic_show_locus): Adjust the use of
3668         location_get_source_line and adjust_line with respect to their new
3669         signature.  While displaying a line now, do not stop at the first
3670         null byte.  Rather, display the zero byte as a space and keep
3671         going until we reach the size of the line.
3672         * Makefile.in: Add vec.o to OBJS-libcommon
3674 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3675             Ilya Tocar     <ilya.tocar@intel.com>
3677         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3678         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3679         (__builtin_ia32_kmov16): Ditto.
3680         * config/i386/i386.md (UNSPEC_KMOV): New.
3681         (kmovw): Ditto.
3683 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3685         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3686         (_mm512_storeu_si512): Ditto.
3688 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3690         PR target/52125
3691         * rtl.h (get_referenced_operands): Declare.
3692         * recog.c (get_referenced_operands): New function.
3693         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3694         operands have been referenced when recording LO_SUM references.
3696 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3698         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3700 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3702         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3703         Enable for generic and recent AMD targets.
3705 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3707         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3708         ARG_SIZE note when adjustment was eliminated.
3710 2014-01-22  Jeff Law  <law@redhat.com>
3712         PR tree-optimization/59597
3713         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3714         in file.  Accept new argument REGISTERING and use it to modify
3715         dump output appropriately.
3716         (register_jump_thread): Corresponding changes.
3717         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3718         thread paths involving joiner blocks.  Add code to dump cancelled
3719         jump threading paths.
3721 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3723         PR rtl-optimization/59477
3724         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3725         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3727 2014-01-22  Tom Tromey  <tromey@redhat.com>
3729         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3730         PARAMS.
3731         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3733 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3735         PR rtl-optimization/59896
3736         * lra-constraints.c (process_alt_operands): Check unused note for
3737         matched operands of insn with no output reloads.
3739 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3741         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3742         (mips_move_from_gpr_cost): Likewise.
3744 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3746         PR rtl-optimization/59858
3747         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3748         ira_class_hard_regs_num.
3749         (process_alt_operands): Increase reject for dying matched operand.
3751 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3753         PR target/59003
3754         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3755         smaller than size, perform several stores or loads and stores
3756         at dst + count - size to store or copy all of size bytes, rather
3757         than just last modesize bytes.
3759 2014-01-20  DJ Delorie  <dj@redhat.com>
3761         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3762         that CLOBBERs are REGs before propogating their values.
3764 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3766         PR middle-end/59789
3767         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3768         (cgraph_inline_failed_type): New function.
3769         * cgraph.h (DEFCIFCODE): Add type.
3770         (cgraph_inline_failed_type_t): New enum.
3771         (cgraph_inline_failed_type): New prototype.
3772         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3773         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3774         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3775         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3776         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3777         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3778         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3779         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3780         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3781         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3782         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3783         OPTIMIZATION_MISMATCH.
3784         * tree-inline.c (expand_call_inline): Emit errors during
3785         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3787 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3789         PR target/59685
3790         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3791         mode attribute in insn output.
3793 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3795         * output.h (output_constant): Delete.
3796         * varasm.c (output_constant): Make private.
3798 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3800         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3802 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3804         PR middle-end/59860
3805         * tree.h (fold_builtin_strcat): New prototype.
3806         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3807         argument, if non-NULL, don't call c_strlen.  Optimize
3808         directly into __builtin_memcpy instead of __builtin_strcpy.
3809         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3810         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3812 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3814         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3815         for SImode_address_operand operands, having only a REG argument.
3817 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3819         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3820         loader name using mbig-endian.
3821         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3823 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3825         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3826         (-mtune): Likewise.
3827         (-mcpu): Likewise.
3829 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3831         * config/aarch64/aarch64-protos.h
3832         (aarch64_cannot_change_mode_class_ptr): Declare.
3833         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3834         aarch64_cannot_change_mode_class_ptr): New.
3835         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3836         backend hook aarch64_cannot_change_mode_class.
3838 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3840         * common/config/aarch64/aarch64-common.c
3841         (aarch64_handle_option): Don't handle any option order logic here.
3842         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3843         selected_cpu, warn on architecture version mismatch.
3844         (aarch64_override_options): Fix parsing order for option strings.
3846 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3847             Iain Sandoe  <iain@codesourcery.com>
3849         PR bootstrap/59496
3850         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3851         warning.  Amend comment to reflect current functionality.
3853 2014-01-20  Richard Biener  <rguenther@suse.de>
3855         PR middle-end/59860
3856         * builtins.c (fold_builtin_strcat): Remove case better handled
3857         by tree-ssa-strlen.c.
3859 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
3861         * config/aarch64/aarch64.opt
3862         (mcpu, march, mtune): Make case-insensitive.
3864 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3866         PR target/59880
3867         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3868         if operands[1] is a REG or ZERO_EXTEND of a REG.
3870 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
3872         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3874 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
3876         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3877         long non-pic millicode calls.
3879 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3881         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3883 2014-01-19  Kito Cheng  <kito@0xlab.org>
3885         * builtins.c (expand_movstr): Check movstr expand done or fail.
3887 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3888             H.J. Lu  <hongjiu.lu@intel.com>
3890         PR target/59379
3891         * config/i386/i386.md (*lea<mode>): Zero-extend return register
3892         to DImode for zero-extended addresses.
3894 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
3896         PR rtl-optimization/57763
3897         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3898         on the new indirect jump_insn and increment LABEL_NUSES (label).
3900 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
3902         PR bootstrap/59580
3903         PR bootstrap/59583
3904         * config.gcc (x86_archs): New variable.
3905         (x86_64_archs): Likewise.
3906         (x86_cpus): Likewise.
3907         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3908         --with-arch/--with-cpu= options.
3909         Support --with-arch=/--with-cpu={nehalem,westmere,
3910         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3912 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3914         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3915         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
3917 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3919         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3921 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
3923         PR target/58944
3924         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3925         clear cpp_get_options (parse_in)->warn_unused_macros for
3926         ix86_target_macros_internal with cpp_define.
3928 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
3930         * jump.c (delete_related_insns): Keep (use (insn))s.
3931         * reorg.c (redundant_insn): Check for barriers too.
3933 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3935         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3937 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
3939         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3940         call to $$dyncall when TARGET_LONG_CALLS is true.
3942 2014-01-17  Jeff Law  <law@redhat.com>
3944         * ree.c (combine_set_extension): Temporarily disable test for
3945         changing number of hard registers.
3947 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3949         PR middle-end/58125
3950         * ipa-inline-analysis.c (inline_free_summary):
3951         Do not free summary of aliases.
3953 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3955         PR middle-end/59706
3956         * gimplify.c (gimplify_expr): Use create_tmp_var
3957         instead of create_tmp_var_raw.  If cond doesn't have
3958         integral type, don't add the IFN_ANNOTATE builtin at all.
3960 2014-01-17  Martin Jambor  <mjambor@suse.cz>
3962         PR ipa/59736
3963         * ipa-cp.c (prev_edge_clone): New variable.
3964         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3965         Also resize prev_edge_clone vector.
3966         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3967         (ipcp_edge_removal_hook): New function.
3968         (ipcp_driver): Register ipcp_edge_removal_hook.
3970 2014-01-17  Andrew Pinski <apinski@cavium.com>
3971             Steve Ellcey  <sellcey@mips.com>
3973         PR target/59462
3974         * config/mips/mips.c (mips_print_operand): Check operand mode instead
3975         of operator mode.
3977 2014-01-17  Jeff Law  <law@redhat.com>
3979         PR middle-end/57904
3980         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3981         so that pass_ccp runs first.
3983 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3985         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3986         (ix86_adjust_cost): Use !TARGET_XXX.
3987         (do_reorder_for_imul): Likewise.
3988         (swap_top_of_ready_list): Likewise.
3989         (ix86_sched_reorder): Likewise.
3991 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3993         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3994         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
3995         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
3996         (intel_memset): New.  Duplicate slm_memset.
3997         (intel_cost): New.  Duplicate slm_cost.
3998         (m_INTEL): New macro.
3999         (processor_target_table): Add "intel".
4000         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4001         with PROCESSOR_INTEL for "intel".
4002         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4003         PROCESSOR_SILVERMONT.
4004         (ix86_issue_rate): Likewise.
4005         (ix86_adjust_cost): Likewise.
4006         (ia32_multipass_dfa_lookahead): Likewise.
4007         (swap_top_of_ready_list): Likewise.
4008         (ix86_sched_reorder): Likewise.
4009         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4010         instead of TARGET_OPT_AGU.
4011         * config/i386/i386.h (TARGET_INTEL): New.
4012         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4013         (processor_type): Add PROCESSOR_INTEL.
4014         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4015         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4017 2014-01-17  Marek Polacek  <polacek@redhat.com>
4019         PR c/58346
4020         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4021         size is zero.
4023 2014-01-17  Richard Biener  <rguenther@suse.de>
4025         PR tree-optimization/46590
4026         * opts.c (default_options_table): Add entries for
4027         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4028         all enabled at -O1 but not for -Og.
4029         * common.opt (fbranch-count-reg): Remove Init(1).
4030         (fmove-loop-invariants): Likewise.
4031         (ftree-pta): Likewise.
4033 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4035         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4036         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4037         decls to at least the GCC 4.8 used alignments.
4039         PR fortran/59440
4040         * tree-nested.c (convert_nonlocal_reference_stmt,
4041         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4042         of GIMPLE_BIND stmts, adjust associated decls.
4044 2014-01-17  Richard Biener  <rguenther@suse.de>
4046         PR tree-optimization/46590
4047         * vec.h (vec<>::bseach): New member function implementing
4048         binary search according to C89 bsearch.
4049         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4050         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4051         bitmap pointer again.  Make accesses_in_loop a flat array.
4052         (mem_ref_obstack): New global.
4053         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4054         (mark_ref_stored): Likewise.
4055         (ref_indep_loop_p_2): Likewise.
4056         (set_ref_stored_in_loop): New helper function.
4057         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4058         (memref_free): Adjust.
4059         (record_mem_ref_loc): Simplify.
4060         (gather_mem_refs_stmt): Adjust.
4061         (sort_locs_in_loop_postorder_cmp): New function.
4062         (analyze_memory_references): Sort accesses_in_loop after
4063         loop postorder number.
4064         (find_ref_loc_in_loop_cmp): New function.
4065         (for_all_locs_in_loop): Find relevant cluster of locs in
4066         accesses_in_loop and iterate without recursion.
4067         (execute_sm): Avoid uninit warning.
4068         (struct ref_always_accessed): Simplify.
4069         (ref_always_accessed::operator ()): Likewise.
4070         (ref_always_accessed_p): Likewise.
4071         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4072         loop postorder numbers here.
4073         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4074         numbers.
4076 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4078         PR c++/57945
4079         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4080         on decls for which assemble_alias has been called.
4082 2014-01-17  Nick Clifton  <nickc@redhat.com>
4084         * config/msp430/msp430.opt: (mcpu): New option.
4085         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4086         (msp430_option_override): Parse target_cpu.  If the MCU name
4087         matches a generic string, clear target_mcu.
4088         (msp430_attr): Allow numeric interrupt values up to 63.
4089         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4090         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4091         option.
4092         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4093         Add mcpu matches.
4094         * config/msp430/msp430.md (popm): Use %J rather than %I.
4095         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4096         (addhi_cy_i): Use immediate_operand for operand 2.
4097         * doc/invoke.texi: Document -mcpu option.
4099 2014-01-17  Richard Biener  <rguenther@suse.de>
4101         PR rtl-optimization/38518
4102         * df.h (df_analyze_loop): Declare.
4103         * df-core.c: Include cfgloop.h.
4104         (df_analyze_1): Split out main part of df_analyze.
4105         (df_analyze): Adjust.
4106         (loop_inverted_post_order_compute): New function.
4107         (loop_post_order_compute): Likewise.
4108         (df_analyze_loop): New function avoiding whole-function
4109         postorder computes.
4110         * loop-invariant.c (find_defs): Use df_analyze_loop.
4111         (find_invariants): Adjust.
4112         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4114 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4116         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4117         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4119 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4121         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4122         traversal when removing references.
4124 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4126         PR ipa/59775
4127         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4129 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4131         PR middle-end/56791
4132         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4133         pushing a reload for an autoinc when we had previously reloaded an
4134         inner part of the address.
4136 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4138         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4139         field.
4140         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4141         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4142         when not giving up or versioning for alias only because of
4143         loop->safelen.
4144         (vect_analyze_data_ref_dependences): Set to true.
4145         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4146         is a GIMPLE_PHI.
4147         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4148         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4149         to the condition.
4151         PR middle-end/58344
4152         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4154         PR target/59839
4155         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4156         operand 0 predicate for gathers, use a new pseudo as subtarget.
4158 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4160         PR middle-end/59609
4161         * lra-constraints.c (process_alt_operands): Add printing debug info.
4162         Check absence of input/output reloads for matched operands too.
4164 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4166         PR rtl-optimization/59835
4167         * ira.c (ira_init_register_move_cost): Increase cost for
4168         impossible modes.
4170 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4172         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4174 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4176         PR target/59780
4177         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4178         non-register objects.  Use gen_(high/low)part more consistently.
4179         Fix assertions.
4181 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4183         PR target/59844
4184         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4185         endian support, remove tests for WORDS_BIG_ENDIAN.
4186         (p8_mfvsrd_3_<mode>): Likewise.
4187         (reload_gpr_from_vsx<mode>): Likewise.
4188         (reload_gpr_from_vsxsf): Likewise.
4189         (p8_mfvsrd_4_disf): Likewise.
4191 2014-01-16  Richard Biener  <rguenther@suse.de>
4193         PR rtl-optimization/46590
4194         * lcm.c (compute_antinout_edge): Use postorder iteration.
4195         (compute_laterin): Use inverted postorder iteration.
4197 2014-01-16  Nick Clifton  <nickc@redhat.com>
4199         PR middle-end/28865
4200         * varasm.c (output_constant): Return the number of bytes actually
4201         emitted.
4202         (output_constructor_array_range): Update the field size with the
4203         number of bytes emitted by output_constant.
4204         (output_constructor_regular_field): Likewise.  Also do not
4205         complain if the total number of bytes emitted is now greater
4206         than the expected fieldpos.
4207         * output.h (output_constant): Update prototype and descriptive comment.
4209 2014-01-16  Marek Polacek  <polacek@redhat.com>
4211         PR middle-end/59827
4212         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4213         it is error_mark_node.
4215 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4217         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4218         VALID_AVX256_REG_OR_OI_MODE.
4220 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4222         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4223         current procedure should be profiled.
4225 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4227         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4228         of moving from/to the STACK_REG register class.
4230 2014-01-15  Richard Henderson  <rth@redhat.com>
4232         PR debug/54694
4233         * reginfo.c (global_regs_decl): Globalize.
4234         * rtl.h (global_regs_decl): Declare.
4235         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4236         reserved via global_regs.
4238 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4240         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4242 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4244         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4245         and vmulosh rather than call gen_vec_widen_smult_*.
4246         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4247         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4248         (vec_widen_smult_even_v16qi): Likewise.
4249         (vec_widen_umult_even_v8hi): Likewise.
4250         (vec_widen_smult_even_v8hi): Likewise.
4251         (vec_widen_umult_odd_v16qi): Likewise.
4252         (vec_widen_smult_odd_v16qi): Likewise.
4253         (vec_widen_umult_odd_v8hi): Likewise.
4254         (vec_widen_smult_odd_v8hi): Likewise.
4255         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4256         vmuloub rather than call gen_vec_widen_umult_*.
4257         (vec_widen_umult_lo_v16qi): Likewise.
4258         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4259         vmulosb rather than call gen_vec_widen_smult_*.
4260         (vec_widen_smult_lo_v16qi): Likewise.
4261         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4262         rather than call gen_vec_widen_umult_*.
4263         (vec_widen_umult_lo_v8hi): Likewise.
4264         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4265         rather than call gen_vec_widen_smult_*.
4266         (vec_widen_smult_lo_v8hi): Likewise.
4268 2014-01-15  Jeff Law  <law@redhat.com>
4270         PR tree-optimization/59747
4271         * ree.c (find_and_remove_re): Properly handle case where a second
4272         eliminated extension requires widening a copy created for elimination
4273         of a prior extension.
4274         (combine_set_extension): Ensure that the number of hard regs needed
4275         for a destination register does not change when we widen it.
4277 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4279         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4280         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4281         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4282         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4283         (avr-*-rtems*): Likewise.
4284         (bfin*-rtems*): Likewise.
4285         (moxie-*-rtems*): Likewise.
4286         (h8300-*-rtems*): Likewise.
4287         (i[34567]86-*-rtems*): Likewise.
4288         (lm32-*-rtems*): Likewise.
4289         (m32r-*-rtems*): Likewise.
4290         (m68k-*-rtems*): Likewise.
4291         (microblaze*-*-rtems*): Likewise.
4292         (mips*-*-rtems*): Likewise.
4293         (powerpc-*-rtems*): Likewise.
4294         (sh-*-rtems*): Likewise.
4295         (sparc-*-rtems*): Likewise.
4296         (sparc64-*-rtems*): Likewise.
4297         (v850-*-rtems*): Likewise.
4298         (m32c-*-rtems*): Likewise.
4300 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4302         PR rtl-optimization/59511
4303         * ira.c (ira_init_register_move_cost): Use memory costs for some
4304         cases of register move cost calculations.
4305         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4306         instead of BB frequency.
4307         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4308         * lra-assigns.c (find_hard_regno_for): Ditto.
4310 2014-01-15  Richard Biener  <rguenther@suse.de>
4312         PR tree-optimization/59822
4313         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4314         (vectorizable_load): Use it to hoist defs of uses of invariant
4315         loads out of the loop.
4317 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4318             Kugan Vivekanandarajah  <kuganv@linaro.org>
4320         PR target/59695
4321         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4322         truncation.
4324 2014-01-15  Richard Biener  <rguenther@suse.de>
4326         PR rtl-optimization/59802
4327         * lcm.c (compute_available): Use inverted postorder to seed
4328         the initial worklist.
4330 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4332         PR target/59803
4333         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4334         ADDR_REGS for invalid symrefs in non-PIC code.
4336 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4338         PR other/58712
4339         * builtins.c (determine_block_size): Initialize *probable_max_size
4340         even if len_rtx is CONST_INT.
4342 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4344         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4345         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4346         (cortexa53_tunings): Likewise.
4347         (aarch64_sched_issue_rate): New function.
4348         (TARGET_SCHED_ISSUE_RATE): Define.
4350 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4352         * ira-costs.c (find_costs_and_classes): Add missed
4353         ira_init_register_move_cost_if_necessary.
4355 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4357         PR target/59787
4358         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4360 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4362         PR target/59794
4363         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4364         to indicate if type is used for function return value.  Warn ABI
4365         change if the vector mode isn't available for function return value.
4366         (ix86_function_arg_advance): Pass false to type_natural_mode.
4367         (ix86_function_arg): Likewise.
4368         (ix86_gimplify_va_arg): Likewise.
4369         (function_arg_32): Don't warn ABI change.
4370         (ix86_function_value): Pass true to type_natural_mode.
4371         (ix86_return_in_memory): Likewise.
4372         (ix86_struct_value_rtx): Removed.
4373         (TARGET_STRUCT_VALUE_RTX): Likewise.
4375 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4377         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4378         converting a conditional jump into a conditional return.
4380 2014-01-14  Richard Biener  <rguenther@suse.de>
4382         PR tree-optimization/58921
4383         PR tree-optimization/59006
4384         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4385         hoisting invariant stmts.
4386         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4387         invariant loads on the preheader edge if possible.
4389 2014-01-14  Joey Ye  <joey.ye@arm.com>
4391         * doc/plugin.texi (Building GCC plugins): Update to C++.
4393 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4395         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4396         (_mm_rcp28_round_ss): Ditto.
4397         (_mm_rsqrt28_round_sd): Ditto.
4398         (_mm_rsqrt28_round_ss): Ditto.
4399         (_mm_rcp28_sd): Ditto.
4400         (_mm_rcp28_ss): Ditto.
4401         (_mm_rsqrt28_sd): Ditto.
4402         (_mm_rsqrt28_ss): Ditto.
4403         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4404         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4405         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4406         (IX86_BUILTIN_RCP28SD): Ditto.
4407         (IX86_BUILTIN_RCP28SS): Ditto.
4408         (IX86_BUILTIN_RSQRT28SD): Ditto.
4409         (IX86_BUILTIN_RSQRT28SS): Ditto.
4410         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4411         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4412         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4413         (ix86_expand_special_args_builtin): Expand new FTYPE.
4414         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4415         (srcp14<mode>): Make insn unary.
4416         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4417         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4418         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4419         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4420         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4421         Fix rounding: make it SAE only.
4422         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4423         Ditto.
4424         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4425         Ditto.
4426         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4427         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4428         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4429         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4430         (round_saeonly_mask_scalar_operand4): Ditto.
4431         (round_saeonly_mask_scalar_op3): Ditto.
4432         (round_saeonly_mask_scalar_op4): Ditto.
4434 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4436         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4437         Implement -maltivec=be for vec_insert and vec_extract.
4439 2014-01-10  DJ Delorie  <dj@redhat.com>
4441         * config/msp430/msp430.md (call_internal): Don't allow memory
4442         references with SP as the base register.
4443         (call_value_internal): Likewise.
4444         * config/msp430/constraints.md (Yc): New.  For memory references
4445         that don't use SP as a base register.
4447         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4448         "an integer without a # prefix"
4449         * config/msp430/msp430.md (epilogue_helper): Use it.
4451 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4453         PR target/59617
4454         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4455         AVX512F gather builtins.
4456         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4457         on gather decls with INTEGER_TYPE masktype.
4458         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4459         directly into the builtin rather than hoisting it before loop.
4461         PR tree-optimization/59387
4462         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4463         (scev_const_prop): If folded_casts and type has undefined overflow,
4464         use force_gimple_operand instead of force_gimple_operand_gsi and
4465         for each added stmt if it is assign with
4466         arith_code_with_undefined_signed_overflow, call
4467         rewrite_to_defined_overflow.
4468         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4469         gimple-fold.h instead.
4470         (arith_code_with_undefined_signed_overflow,
4471         rewrite_to_defined_overflow): Moved to ...
4472         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4473         rewrite_to_defined_overflow): ... here.  No longer static.
4474         Include gimplify-me.h.
4475         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4476         rewrite_to_defined_overflow): New prototypes.
4478 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4480         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4482 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4484         * builtins.c (get_object_alignment_2): Minor tweak.
4485         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4487 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4489         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4490         optimized non constant lengths.
4492 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4494         PR libgomp/59194
4495         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4496         load as __atomic_load_N if possible.
4498 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4500         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4501         target parameter.
4502         (rs6000_expand_builtin): Adjust call.
4504 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4506         PR target/58115
4507         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4508         * config/rs6000/rs6000.c: Include target-globals.h.
4509         (rs6000_set_current_function): Instead of doing target_reinit
4510         unconditionally, use save_target_globals_default_opts and
4511         restore_target_globals.
4513         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4514         FPSCR.
4515         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4516         (rs6000_expand_builtin): Handle mffs and mtfsf.
4517         (rs6000_init_builtins): Define mffs and mtfsf.
4518         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4519         (rs6000_mffs): New pattern.
4520         (rs6000_mtfsf): New pattern.
4522 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4524         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4525         Start narrowing with START.  Apply candidate-use pair
4526         and check overall cost in narrowing.
4527         (iv_ca_prune): Pass new argument.
4529 2014-01-10  Jeff Law  <law@redhat.com>
4531         PR middle-end/59743
4532         * ree.c (combine_reaching_defs): Ensure the defining statement
4533         occurs before the extension when optimizing extensions with
4534         different source and destination hard registers.
4536 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4538         PR ipa/58585
4539         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4540         vtables into the type inheritance graph.
4542 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4544         PR rtl-optimization/59754
4545         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4546         modes in the REGNO != REGNO case.
4548 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4550         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4552 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4554         PR tree-optimization/59745
4555         * tree-predcom.c (tree_predictive_commoning_loop): Call
4556         free_affine_expand_cache if giving up because components is NULL.
4558         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4559         GC in payload of target_globals struct instead of allocating them on
4560         the heap and the larger structs separately using GC.
4561         * target-globals.h (struct target_globals): Make regs, hard_regs,
4562         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4563         of GTY((skip)) and change type to void *.
4564         (reset_target_globals): Cast loads from those fields to corresponding
4565         types.
4567 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4569         PR plugins/59335
4570         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4571         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4572         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4574 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4576         PR target/59744
4577         * aarch64-modes.def (CC_Zmode): New flags mode.
4578         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4579         represents an equality.
4580         (aarch64_get_condition_code): Handle CC_Zmode.
4581         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4583 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4585         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4586         extraction in good case.
4588 2014-01-10  Richard Biener  <rguenther@suse.de>
4590         PR tree-optimization/59374
4591         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4592         checking after SLP discovery.  Mark stmts not participating
4593         in any SLP instance properly.
4595 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4597         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4598         when handling a SET rtx.
4600 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4602         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4603         (cortex-a57): Likewise.
4604         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4606 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4608         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4609         non-iwmmxt builtins.
4611 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4613         PR ipa/58252
4614         PR ipa/59226
4615         * ipa-devirt.c record_target_from_binfo): Take as argument
4616         stack of binfos and lookup matching one for virtual inheritance.
4617         (possible_polymorphic_call_targets_1): Update.
4619 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4621         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4622         kernel strings for Loongson-2E/2F/3A.
4624 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4626         PR middle-end/59670
4627         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4628         is_gimple_call before calling gimple_call_internal_p.
4630 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4632         * Makefile.in (TREE_FLOW_H): Remove.
4633         (TREE_SSA_H): Add file names from tree-flow.h.
4634         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4635         * tree.h: Remove tree-flow.h reference.
4636         * hash-table.h: Remove tree-flow.h reference.
4637         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4638         reference with tree-ssa-loop.h.
4640 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4642         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4643         default element-order behavior for -maltivec.
4644         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4645         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4646         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4647         when targeting big endian, at least for now.
4648         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4650 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4652         PR middle-end/47735
4653         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4654         var satisfies use_register_for_decl, just take into account type
4655         alignment, rather than decl alignment.
4657         PR tree-optimization/59622
4658         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4659         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4660         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4661         Don't devirtualize for inplace at all.  For targets.length () == 1,
4662         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4664 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4666         * config/i386/i386.md (cpu): Remove the unused btver1.
4668 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4670         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4672 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4674         PR target/58115
4675         * tree-core.h (struct target_globals): New forward declaration.
4676         (struct tree_target_option): Add globals field.
4677         * tree.h (TREE_TARGET_GLOBALS): Define.
4678         (prepare_target_option_nodes_for_pch): New prototype.
4679         * target-globals.h (struct target_globals): Define even if
4680         !SWITCHABLE_TARGET.
4681         * tree.c (prepare_target_option_node_for_pch,
4682         prepare_target_option_nodes_for_pch): New functions.
4683         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4684         * config/i386/i386.c: Include target-globals.h.
4685         (ix86_set_current_function): Instead of doing target_reinit
4686         unconditionally, use save_target_globals_default_opts and
4687         restore_target_globals.
4689 2014-01-09  Richard Biener  <rguenther@suse.de>
4691         PR tree-optimization/59715
4692         * tree-cfg.h (split_critical_edges): Declare.
4693         * tree-cfg.c (split_critical_edges): Export.
4694         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4696 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4698         * cfgexpand.c (expand_stack_vars): Optionally disable
4699         asan stack protection.
4700         (expand_used_vars): Likewise.
4701         (partition_stack_vars): Likewise.
4702         * asan.c (asan_emit_stack_protection): Optionally disable
4703         after return stack usage.
4704         (instrument_derefs): Optionally disable memory access instrumentation.
4705         (instrument_builtin_call): Likewise.
4706         (instrument_strlen_call): Likewise.
4707         (asan_protect_global): Optionally disable global variables protection.
4708         * doc/invoke.texi: Added doc for new options.
4709         * params.def: Added new options.
4710         * params.h: Likewise.
4712 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4714         PR rtl-optimization/59724
4715         * ifcvt.c (cond_exec_process_if_block): Don't call
4716         flow_find_head_matching_sequence with 0 longest_match.
4717         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4718         non-active insns if !stop_after.
4719         (try_head_merge_bb): Revert 2014-01-07 changes.
4721 2014-01-08  Jeff Law  <law@redhat.com>
4723         * ree.c (get_sub_rtx): New function, extracted from...
4724         (merge_def_and_ext): Here.
4725         (combine_reaching_defs): Use get_sub_rtx.
4727 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4729         * cgraph.h (varpool_variable_node): Do not choke on null node.
4731 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4733         * config/mips/mips.md (simple_return): Attempt to use JRC
4734         for microMIPS.
4735         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4737 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4739         PR rtl-optimization/59137
4740         * reorg.c (steal_delay_list_from_target): Call update_block for
4741         elided insns.
4742         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4744 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4746         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4747         two duplicate entries.
4749 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4751         Revert:
4752         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4754         * config/mips/mips.c (mips_truncated_op_cost): New function.
4755         (mips_rtx_costs): Adjust test for BADDU.
4756         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4758         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4760         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4761         (*baddu_si): ...this new pattern.
4763 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4765         PR ipa/59722
4766         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4768 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4770         PR middle-end/57748
4771         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4772         inner_reference_p.
4773         (expand_expr, expand_normal): Adjust.
4774         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4775         inner_reference_p. Use inner_reference_p to expand inner references.
4776         (store_expr): Adjust.
4777         * cfgexpand.c (expand_call_stmt): Adjust.
4779 2014-01-08  Rong Xu  <xur@google.com>
4781         * gcov-io.c (gcov_var): Move from gcov-io.h.
4782         (gcov_position): Ditto.
4783         (gcov_is_error): Ditto.
4784         (gcov_rewrite): Ditto.
4785         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4786         only part to libgcc/libgcov.h.
4788 2014-01-08  Marek Polacek  <polacek@redhat.com>
4790         PR middle-end/59669
4791         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4793 2014-01-08  Marek Polacek  <polacek@redhat.com>
4795         PR sanitizer/59667
4796         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4798 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4800         PR rtl-optimization/59649
4801         * stor-layout.c (get_mode_bounds): For BImode return
4802         0 and STORE_FLAG_VALUE.
4804 2014-01-08  Richard Biener  <rguenther@suse.de>
4806         PR middle-end/59630
4807         * gimple.h (is_gimple_builtin_call): Remove.
4808         (gimple_builtin_call_types_compatible_p): New.
4809         (gimple_call_builtin_p): New overload.
4810         * gimple.c (is_gimple_builtin_call): Remove.
4811         (validate_call): Rename to ...
4812         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4813         check return types.
4814         (validate_type): New static function.
4815         (gimple_call_builtin_p): New overload and adjust.
4816         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4817         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4818         (gimple_fold_stmt_to_constant_1): Likewise.
4819         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4821 2014-01-08  Richard Biener  <rguenther@suse.de>
4823         PR middle-end/59471
4824         * gimplify.c (gimplify_expr): Gimplify register-register type
4825         VIEW_CONVERT_EXPRs to separate stmts.
4827 2014-01-07  Jeff Law  <law@redhat.com>
4829         PR middle-end/53623
4830         * ree.c (combine_set_extension): Handle case where source
4831         and destination registers in an extension insn are different.
4832         (combine_reaching_defs): Allow source and destination registers
4833         in extension to be different under limited circumstances.
4834         (add_removable_extension): Remove restriction that the
4835         source and destination registers in the extension are the same.
4836         (find_and_remove_re): Emit a copy from the extension's
4837         destination to its source after the defining insn if
4838         the source and destination registers are different.
4840         PR middle-end/59285
4841         * ifcvt.c (merge_if_block): If we are merging a block with more than
4842         one successor with a block with no successors, remove any BARRIER
4843         after the second block.
4845 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4847         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4849 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4851         PR target/59652
4852         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4853         for 14-bit register offsets when INT14_OK_STRICT is false.
4855 2014-01-07  Roland Stigge  <stigge@antcom.de>
4856             Michael Meissner  <meissner@linux.vnet.ibm.com>
4858         PR 57386/target
4859         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4860         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
4862 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
4864         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4865         -mcpu.
4867 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
4869         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4870         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4871         rtx is const0_rtx or not.
4873 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
4875         PR target/58115
4876         * target-globals.c (save_target_globals): Remove this_fn_optab
4877         handling.
4878         * toplev.c: Include optabs.h.
4879         (target_reinit): Temporarily restore the global options if another
4880         set of options are in force.
4882 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
4884         PR rtl-optimization/58668
4885         * cfgcleanup.c (flow_find_cross_jump): Don't count
4886         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
4887         to determine what is counted.
4888         (flow_find_head_matching_sequence): Use active_insn_p to determine
4889         what is counted.
4890         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4891         counting change.
4892         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4893         determine what is counted.
4895         PR tree-optimization/59643
4896         * tree-predcom.c (split_data_refs_to_components): If one dr is
4897         read and one write, determine_offset fails and the write isn't
4898         in the bad component, just put the read into the bad component.
4900 2014-01-07  Mike Stump  <mikestump@comcast.net>
4901             Jakub Jelinek  <jakub@redhat.com>
4903         PR pch/59436
4904         * tree-core.h (struct tree_optimization_option): Change optabs
4905         type from unsigned char * to void *.
4906         * optabs.c (init_tree_optimization_optabs): Adjust
4907         TREE_OPTIMIZATION_OPTABS initialization.
4909 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
4911         PR target/59644
4912         * config/i386/i386.h (struct machine_function): Add
4913         no_drap_save_restore field.
4914         * config/i386/i386.c (ix86_save_reg): Use
4915         !cfun->machine->no_drap_save_restore instead of
4916         crtl->stack_realign_needed.
4917         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4918         this function clears frame_pointer_needed.  Set
4919         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4920         and DRAP reg is needed.
4922 2014-01-06  Marek Polacek  <polacek@redhat.com>
4924         PR c/57773
4925         * doc/implement-c.texi: Mention that other integer types are
4926         permitted as bit-field types in strictly conforming mode.
4928 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
4930         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4931         is newly allocated.
4933 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
4935         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4937 2014-01-06  Martin Jambor  <mjambor@suse.cz>
4939         PR ipa/59008
4940         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4941         to int.
4942         * ipa-prop.c (ipa_print_node_params): Fix indentation.
4944 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
4946         PR debug/59350
4947         PR debug/59510
4948         * var-tracking.c (add_stores): Preserve the value of the source even if
4949         we don't record the store.
4951 2014-01-06  Terry Guo  <terry.guo@arm.com>
4953         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4955 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
4957         PR bootstrap/59541
4958         * config/darwin.c (darwin_function_section): Adjust return values to
4959         correspond to optimisation changes made in r206070.
4961 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
4963         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4964         from prefetch_block tune setting.
4965         (nocona_cost): Correct size of prefetch block to 64.
4967 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
4969         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4970         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4971         used to save the static chain register in the computation of the offset
4972         from which the FP registers need to be restored.
4974 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
4976         PR tree-optimization/59519
4977         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4978         ICE if get_current_def (current_new_name) is already non-NULL, as long
4979         as it is a phi result of some other phi in *new_exit_bb that has
4980         the same argument.
4982         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4983         or vmovdqu* for misaligned_operand.
4984         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4985         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4986         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4987         aligned_mem for AVX512F masked aligned load and store builtins and for
4988         non-temporal moves.
4990 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
4992         PR tree-optimization/59651
4993         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
4994         Address range for negative step should be added by TYPE_SIZE_UNIT.
4996 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
4998         * config/m68k/m68k.c (handle_move_double): Handle pushes with
4999         overlapping registers also for registers other than the stack pointer.
5001 2014-01-03  Marek Polacek  <polacek@redhat.com>
5003         PR other/59661
5004         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5005         __builtin_FILE.
5007 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5009         PR target/59625
5010         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5011         asm goto as jump.
5013         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5014         (push splitter): Use <P:MODE_SIZE> instead of
5015         GET_MODE_SIZE (<P:MODE>mode).
5016         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5017         (mov -1, reg peephole2): Likewise.
5018         * config/i386/sse.md (*mov<mode>_internal,
5019         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5020         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5021         *<code><mode>3, *andnot<mode>3<mask_name>,
5022         <mask_codefor><code><mode>3<mask_name>): Likewise.
5023         * config/i386/subst.md (mask_mode512bit_condition,
5024         sd_mask_mode512bit_condition): Likewise.
5026 2014-01-02  Xinliang David Li  <davidxl@google.com>
5028         PR tree-optimization/59303
5029         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5030         (dump_predicates): Better output format.
5031         (pred_equal_p): New function.
5032         (is_neq_relop_p): Ditto.
5033         (is_neq_zero_form_p): Ditto.
5034         (pred_expr_equal_p): Ditto.
5035         (pred_neg_p): Ditto.
5036         (simplify_pred): Ditto.
5037         (simplify_preds_2): Ditto.
5038         (simplify_preds_3): Ditto.
5039         (simplify_preds_4): Ditto.
5040         (simplify_preds): Ditto.
5041         (push_pred): Ditto.
5042         (push_to_worklist): Ditto.
5043         (get_pred_info_from_cmp): Ditto.
5044         (is_degenerated_phi): Ditto.
5045         (normalize_one_pred_1): Ditto.
5046         (normalize_one_pred): Ditto.
5047         (normalize_one_pred_chain): Ditto.
5048         (normalize_preds): Ditto.
5049         (normalize_cond_1): Remove function.
5050         (normalize_cond): Ditto.
5051         (is_gcond_subset_of): Ditto.
5052         (is_subset_of_any): Ditto.
5053         (is_or_set_subset_of): Ditto.
5054         (is_and_set_subset_of): Ditto.
5055         (is_norm_cond_subset_of): Ditto.
5056         (pred_chain_length_cmp): Ditto.
5057         (convert_control_dep_chain_into_preds): Type change.
5058         (find_predicates): Ditto.
5059         (find_def_preds): Ditto.
5060         (destroy_predicates_vecs): Ditto.
5061         (find_matching_predicates_in_rest_chains): Ditto.
5062         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5063         (is_pred_expr_subset): Ditto.
5064         (is_pred_chain_subset_of): Ditto.
5065         (is_included_in): Ditto.
5066         (is_superset_of): Ditto.
5068 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5070         Update copyright years.
5072 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5074         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5075         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5076         config/arc/arc.md, config/arc/arc.opt,
5077         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5078         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5079         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5080         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5081         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5082         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5083         vtable-verify.c, vtable-verify.h: Use the standard form for the
5084         copyright notice.
5086 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5088         * gcc.c (process_command): Update copyright notice dates.
5089         * gcov-dump.c: Ditto.
5090         * gcov.c: Ditto.
5091         * doc/cpp.texi: Bump @copying's copyright year.
5092         * doc/cppinternals.texi: Ditto.
5093         * doc/gcc.texi: Ditto.
5094         * doc/gccint.texi: Ditto.
5095         * doc/gcov.texi: Ditto.
5096         * doc/install.texi: Ditto.
5097         * doc/invoke.texi: Ditto.
5099 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5101         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5103 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5105         * config/i386/sse.md (*mov<mode>_internal): Guard
5106         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5108         PR rtl-optimization/59647
5109         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5110         new_rtx into UNSIGNED_FLOAT rtxes.
5112 Copyright (C) 2014 Free Software Foundation, Inc.
5114 Copying and distribution of this file, with or without modification,
5115 are permitted in any medium without royalty provided the copyright
5116 notice and this notice are preserved.