2014-03-18 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob23c0b194a2e19c4f9ecf86f3c9bbfb4ad4b4eacc
1 2014-03-18  Richard Biener  <rguenther@suse.de>
3         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
4         by using bitmap_and_compl instead of bitmap_and_compl_into.
5         (df_rd_transfer_function): Likewise.
7 2014-03-18  Richard Biener  <rguenther@suse.de>
9         * doc/lto.texi (fresolution): Fix typo.
11 2014-03-18  Richard Biener  <rguenther@suse.de>
13         * doc/invoke.texi (flto): Update for changes in 4.9.
15 2014-03-18  Richard Biener  <rguenther@suse.de>
17         * doc/loop.texi: Remove section on the removed lambda framework.
18         Update loop docs with recent changes in preserving loop structure.
20 2014-03-18  Richard Biener  <rguenther@suse.de>
22         * doc/lto.texi (-fresolution): Document.
24 2014-03-18  Richard Biener  <rguenther@suse.de>
26         * doc/contrib.texi: Adjust my name.
28 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
30         PR ipa/58721
31         * internal-fn.c: Include diagnostic-core.h.
32         (expand_BUILTIN_EXPECT): New function.
33         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
34         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
35         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
36         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
37         IFN_BUILTIN_EXPECT.
38         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
39         Revert 3 argument __builtin_expect code.
40         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
41         * gimple-fold.c (gimple_fold_call): Likewise.
42         * tree.h (fold_builtin_expect): New prototype.
43         * builtins.c (build_builtin_expect_predicate): Add predictor
44         argument, if non-NULL, create 3 argument __builtin_expect.
45         (fold_builtin_expect): No longer static.  Add ARG2 argument,
46         pass it through to build_builtin_expect_predicate.
47         (fold_builtin_2): Adjust caller.
48         (fold_builtin_3): Handle BUILT_IN_EXPECT.
49         * internal-fn.def (BUILTIN_EXPECT): New.
51 2014-03-18  Tobias Burnus  <burnus@net-b.de>
53         PR ipa/58721
54         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
55         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
56         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
58 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
60         PR ipa/58721
61         * predict.c (combine_predictions_for_bb): Fix up formatting.
62         (expr_expected_value_1, expr_expected_value): Add predictor argument,
63         fill what it points to if non-NULL.
64         (tree_predict_by_opcode): Adjust caller, use the predictor.
65         * predict.def (PRED_COMPARE_AND_SWAP): Add.
67 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
69         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
70         proper constant for the store mode.
72 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
74         * symtab.c (change_decl_assembler_name): Fix transparent alias
75         chain construction.
77 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
79         * config/aarch64/aarch64.c: Correct the comments about the
80         aarch64 stack layout.
82 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
84         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
85         check for GF_OMP_FOR_KIND_FOR.
87 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
89         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
90         ymm and zmm register names.
92 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
94         PR target/60516
95         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
96         note creation for the 2010-08-31 changes.
98 2014-03-17  Marek Polacek  <polacek@redhat.com>
100         PR middle-end/60534
101         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
102         as -fno-tree-loop-vectorize.
103         (expand_omp_simd): Likewise.
105 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
107         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
108         (eligible_for_call_delay): New prototype.
109         * config/sparc/sparc.c (tls_call_delay): Rename into...
110         (eligible_for_call_delay): ...this.  Return false if the instruction
111         cannot be put in the delay slot of a branch.
112         (eligible_for_restore_insn): Simplify.
113         (eligible_for_return_delay): Return false if the instruction cannot be
114         put in the delay slot of a branch and simplify.
115         (eligible_for_sibcall_delay): Return false if the instruction cannot be
116         put in the delay slot of a branch.
117         * config/sparc/sparc.md (fix_ut699): New attribute.
118         (tls_call_delay): Delete.
119         (in_call_delay): Reimplement.
120         (eligible_for_sibcall_delay): Rename into...
121         (in_sibcall_delay): ...this.
122         (eligible_for_return_delay): Rename into...
123         (in_return_delay): ...this.
124         (in_branch_delay): Reimplement.
125         (in_uncond_branch_delay): Delete.
126         (in_annul_branch_delay): Delete.
128 2014-03-14  Richard Henderson  <rth@redhat.com>
130         PR target/60525
131         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
132         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
133         (*floathi<X87MODEF>2_i387_with_temp): Remove.
134         (floathi splitters): Remove.
135         (float<SWI48x>xf2): New pattern.
136         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
137         code that tried to handle DImode for 32-bit, but which was excluded
138         by the pattern's condition.  Drop allocation of stack temporary.
139         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
140         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
141         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
142         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
143         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
144         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
145         (*float<SWI48><MODEF>2_sse_interunit): Remove.
146         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
147         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
148         (*float<SWI48x><X87MODEF>2_i387): Remove.
149         (all float _with_temp splitters): Remove.
150         (*float<SWI48x><MODEF>2_i387): New pattern.
151         (*float<SWI48><MODEF>2_sse): New pattern.
152         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
153         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
155 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
156             Marek Polacek  <polacek@redhat.com>
158         PR middle-end/60484
159         * common.opt (dump_base_name_prefixed): New Variable.
160         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
161         if x_dump_base_name_prefixed is already set, set it at the end.
163 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
165         PR rtl-optimization/60508
166         * lra-constraints.c (get_reload_reg): Add new parameter
167         in_subreg_p.
168         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
169         Pass the new parameter values.
171 2014-03-14  Richard Biener  <rguenther@suse.de>
173         * common.opt: Revert unintented changes from r205065.
174         * opts.c: Likewise.
176 2014-03-14  Richard Biener  <rguenther@suse.de>
178         PR middle-end/60518
179         * cfghooks.c (split_block): Properly adjust all loops the
180         block was a latch of.
182 2014-03-14  Martin Jambor  <mjambor@suse.cz>
184         PR lto/60461
185         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
186         and simplify it.
188 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
190         PR target/59396
191         * config/avr/avr.c (avr_set_current_function): Pass function name
192         through default_strip_name_encoding before sanity checking instead
193         of skipping the first char of the assembler name.
195 2014-03-13  Richard Henderson  <rth@redhat.com>
197         PR debug/60438
198         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
199         (ix86_force_to_memory, ix86_free_from_memory): Remove.
200         * config/i386/i386-protos.h: Likewise.
201         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
202         in the expander instead of a splitter.
203         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
204         any possibility of requiring a memory.
205         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
206         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
207         (fp branch splitters): Update for ix86_split_fp_branch.
208         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
209         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
210         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
211         (*fop_<MODEF>_2_i387): Remove f/r alternative.
212         (*fop_<MODEF>_3_i387): Likewise.
213         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
214         (splitters for the fop_* register patterns): Remove.
215         (fscalexf4_i387): Rename from *fscalexf4_i387.
216         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
217         
218 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
220         PR tree-optimization/59779
221         * tree-dfa.c (get_ref_base_and_extent): Use double_int
222         type for bitsize and maxsize instead of HOST_WIDE_INT.
224 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
226         PR rtl-optimization/57320
227         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
228         the CFG after thread_prologue_and_epilogue_insns.
230 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
232         PR rtl-optimization/57189
233         * lra-constraints.c (process_alt_operands): Disfavor spilling
234         vector pseudos.
236 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
238         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
239         messages.
241 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
243         PR tree-optimization/59025
244         PR middle-end/60418
245         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
246         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
248 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
250         PR target/60486
251         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
252         calls of avr_out_plus_1.
254 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
256         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
257         BB's single pred and update the father loop's latch info later.
259 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
261         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
262         (VEC_M): Likewise.
263         (VEC_N): Likewise.
264         (VEC_R): Likewise.
265         (VEC_base): Likewise.
266         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
267         registers, we need to swap double words in little endian mode.
269         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
270         to be a container mode for 128-bit integer operations added in ISA
271         2.07.  Unlike TImode and PTImode, the preferred register set is
272         the Altivec/VMX registers for the 128-bit operations.
274         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
275         declarations.
276         (rs6000_split_128bit_ok_p): Likewise.
278         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
279         macros for creating ISA 2.07 normal and overloaded builtin
280         functions with 3 arguments.
281         (BU_P8V_OVERLOAD_3): Likewise.
282         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
283         for use as overloaded functions.
284         (VPERM_1TI_UNS): Likewise.
285         (VSEL_1TI): Likewise.
286         (VSEL_1TI_UNS): Likewise.
287         (ST_INTERNAL_1ti): Likewise.
288         (LD_INTERNAL_1ti): Likewise.
289         (XXSEL_1TI): Likewise.
290         (XXSEL_1TI_UNS): Likewise.
291         (VPERM_1TI): Likewise.
292         (VPERM_1TI_UNS): Likewise.
293         (XXPERMDI_1TI): Likewise.
294         (SET_1TI): Likewise.
295         (LXVD2X_V1TI): Likewise.
296         (STXVD2X_V1TI): Likewise.
297         (VEC_INIT_V1TI): Likewise.
298         (VEC_SET_V1TI): Likewise.
299         (VEC_EXT_V1TI): Likewise.
300         (EQV_V1TI): Likewise.
301         (NAND_V1TI): Likewise.
302         (ORC_V1TI): Likewise.
303         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
304         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
305         overloaded builtin.
306         (VADDUQM): Likewise.
307         (VSUBCUQ): Likewise.
308         (VADDEUQM): Likewise.
309         (VADDECUQ): Likewise.
310         (VSUBEUQM): Likewise.
311         (VSUBECUQ): Likewise.
313         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
314         __int128_t and __uint128_t types.
315         (__uint128_type): Likewise.
316         (altivec_categorize_keyword): Add support for vector __int128_t,
317         vector __uint128_t, vector __int128, and vector unsigned __int128
318         as a container type for TImode operations that need to be done in
319         VSX/Altivec registers.
320         (rs6000_macro_to_expand): Likewise.
321         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
322         to support 128-bit integer instructions vaddcuq, vadduqm,
323         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
324         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
326         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
327         for V1TImode, and set up preferences to use VSX/Altivec registers.
328         Setup VSX reload handlers.
329         (rs6000_debug_reg_global): Likewise.
330         (rs6000_init_hard_regno_mode_ok): Likewise.
331         (rs6000_preferred_simd_mode): Likewise.
332         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
333         (easy_altivec_constant): Likewise.
334         (output_vec_const_move): Likewise.
335         (rs6000_expand_vector_set): Convert V1TImode set and extract to
336         simple move.
337         (rs6000_expand_vector_extract): Likewise.
338         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
339         addressing.
340         (rs6000_const_vec): Add support for V1TImode.
341         (rs6000_emit_le_vsx_load): Swap double words when loading or
342         storing TImode/V1TImode.
343         (rs6000_emit_le_vsx_store): Likewise.
344         (rs6000_emit_le_vsx_move): Likewise.
345         (rs6000_emit_move): Add support for V1TImode.
346         (altivec_expand_ld_builtin): Likewise.
347         (altivec_expand_st_builtin): Likewise.
348         (altivec_expand_vec_init_builtin): Likewise.
349         (altivec_expand_builtin): Likewise.
350         (rs6000_init_builtins): Add support for V1TImode type.  Add
351         support for ISA 2.07 128-bit integer builtins.  Define type names
352         for the VSX/Altivec vector types.
353         (altivec_init_builtins): Add support for overloaded vector
354         functions with V1TImode type.
355         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
356         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
357         external function.
358         (rs6000_split_128bit_ok_p): Likewise.
359         (rs6000_handle_altivec_attribute): Create V1TImode from vector
360         __int128_t and vector __uint128_t.
362         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
363         and mode attributes.
364         (VSX_M): Likewise.
365         (VSX_M2): Likewise.
366         (VSm): Likewise.
367         (VSs): Likewise.
368         (VSr): Likewise.
369         (VSv): Likewise.
370         (VS_scalar): Likewise.
371         (VS_double): Likewise.
372         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
374         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
375         we support the ISA 2.07 128-bit integer arithmetic instructions.
376         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
377         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
378         and TImode types for use with the builtin functions.
379         (V1TI_type_node): Likewise.
380         (unsigned_V1TI_type_node): Likewise.
381         (intTI_type_internal_node): Likewise.
382         (uintTI_type_internal_node): Likewise.
384         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
385         128-bit builtin functions.
386         (UNSPEC_VADDEUQM): Likewise.
387         (UNSPEC_VADDECUQ): Likewise.
388         (UNSPEC_VSUBCUQ): Likewise.
389         (UNSPEC_VSUBEUQM): Likewise.
390         (UNSPEC_VSUBECUQ): Likewise.
391         (VM): Add V1TImode to vector mode iterators.
392         (VM2): Likewise.
393         (VI_unit): Likewise.
394         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
395         (altivec_vaddcuq): Likewise.
396         (altivec_vsubuqm): Likewise.
397         (altivec_vsubcuq): Likewise.
398         (altivec_vaddeuqm): Likewise.
399         (altivec_vaddecuq): Likewise.
400         (altivec_vsubeuqm): Likewise.
401         (altivec_vsubecuq): Likewise.
403         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
404         mode iterators.
405         (BOOL_128): Likewise.
406         (BOOL_REGS_OUTPUT): Likewise.
407         (BOOL_REGS_OP1): Likewise.
408         (BOOL_REGS_OP2): Likewise.
409         (BOOL_REGS_UNARY): Likewise.
410         (BOOL_REGS_AND_CR0): Likewise.
412         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
413         128-bit integer builtin support.
414         (vec_vadduqm): Likewise.
415         (vec_vaddecuq): Likewise.
416         (vec_vaddeuqm): Likewise.
417         (vec_vsubecuq): Likewise.
418         (vec_vsubeuqm): Likewise.
419         (vec_vsubcuq): Likewise.
420         (vec_vsubuqm): Likewise.
422         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
423         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
424         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
425         128-bit integer add/subtract to ISA 2.07.
427 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
429         * config/arc/arc.c (arc_predicate_delay_insns):
430         Fix third argument passed to conditionalize_nonjump.
432 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
434         * config/aarch64/aarch64-builtins.c
435         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
436         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
437         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
438         instead of __builtin_lfloor.
439         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
441 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
443         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
444         (tree_ssa_ifcombine_bb_1): New function.
445         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
446         is an empty forwarder block to then_bb or vice versa and then_bb
447         and else_bb are effectively swapped.
449 2014-03-12  Christian Bruel  <christian.bruel@st.com>
451         PR target/60264
452         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
453         REG_CFA_DEF_CFA note.
454         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
455         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
457 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
459         PR tree-optimization/60454
460         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
462 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
464         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
465         Do not define target_cpu_default2 to generic.
466         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
467         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
468         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
470 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
471             Marc Glisse  <marc.glisse@inria.fr>
473         PR tree-optimization/60502
474         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
475         instead of build_low_bits_mask.
477 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
479         PR middle-end/60482
480         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
481         if there are multiple uses, but op doesn't live on E edge.
482         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
483         clobber stmts before __builtin_unreachable.
485 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
487         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
488         hard_frame_pointer_rtx.
489         * cse.c (cse_insn): Remove volatile check.
490         * cselib.c (cselib_process_insn): Likewise.
491         * dse.c (scan_insn): Likewise.
493 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
495         * config/arc/arc.c (conditionalize_nonjump): New function,
496         broken out of ...
497         (arc_ifcvt): ... this.
498         (arc_predicate_delay_insns): Use it.
500 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
502         * config/arc/predicates.md (extend_operand): During/after reload,
503         allow const_int_operand.
504         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
505         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
506         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
507         to "i".
508         (umulsi3_highpart_i): Likewise.
510 2014-03-11  Richard Biener  <rguenther@suse.de>
512         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
513         Add asserts to guard possible wrong-code bugs.
515 2014-03-11  Richard Biener  <rguenther@suse.de>
517         PR tree-optimization/60429
518         PR tree-optimization/60485
519         * tree-ssa-structalias.c (set_union_with_increment): Properly
520         take into account all fields that overlap the shifted vars.
521         (do_sd_constraint): Likewise.
522         (do_ds_constraint): Likewise.
523         (get_constraint_for_ptr_offset): Likewise.
525 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
527         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
528         (nios2_compute_frame_layout):
529         Add calculation of cfun->machine->fp_save_offset.
530         (nios2_expand_prologue): Correct setting of frame pointer register
531         in prologue.
532         (nios2_expand_epilogue): Update recovery of stack pointer from
533         frame pointer accordingly.
534         (nios2_initial_elimination_offset): Update calculation of offset
535         for eliminating to HARD_FRAME_POINTER_REGNUM.
537 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
539         PR ipa/60457
540         * ipa.c (symtab_remove_unreachable_nodes): Don't call
541         cgraph_get_create_node on VAR_DECLs.
543 2014-03-10  Richard Biener  <rguenther@suse.de>
545         PR middle-end/60474
546         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
548 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
550         * config/vms/vms.opt (vms_float_format): New variable.
552 2014-03-08  Tobias Burnus  <burnus@net-b.de>
554         * doc/invoke.texi (-fcilkplus): Update implementation status.
556 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
557             Richard Biener  <rguenther@suse.de>
559         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
560         consistently accross all TUs.
561         (run_gcc): Enable -fshort-double automatically at link at link-time
562         and disallow override.
564 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
566         PR target/58271
567         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
568         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
569         if they can't be used.
571 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
573         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
574         for Solaris 11/x86 ld.
575         * configure: Regenerate.
577 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
579         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
580         (LIB_TLS_SPEC): Save as ld_tls_libs.
581         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
582         (HAVE_AS_IX86_TLSLDM): New test.
583         * configure, config.in: Regenerate.
584         * config/i386/i386.c (legitimize_tls_address): Fall back to
585         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
586         cannot support TLS_MODEL_LOCAL_DYNAMIC.
587         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
588         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
590 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
592         * common.opt (fira-loop-pressure): Mark as optimization.
594 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
596         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
597         an OpenMP mappable type.
599 2014-03-06  Matthias Klose  <doko@ubuntu.com>
601         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
602         MULTILIB_OSDIRNAMES is not defined.
604 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
605             Meador Inge  <meadori@codesourcery.com>
607         PR target/58595
608         * config/arm/arm.c (arm_tls_symbol_p): Remove.
609         (arm_legitimize_address): Call legitimize_tls_address for any
610         arm_tls_referenced_p expression, handle constant addend.  Call it
611         before testing for !TARGET_ARM.
612         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
614 2014-03-06  Richard Biener  <rguenther@suse.de>
616         PR middle-end/60445
617         PR lto/60424
618         PR lto/60427
619         Revert
620         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
622         * tree-streamer.c (record_common_node): Assert we don't record
623         nodes with type double.
624         (preload_common_node): Skip type double, complex double and double
625         pointer since it is now frontend dependent due to fshort-double option.
627 2014-03-06  Richard Biener  <rguenther@suse.de>
629         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
630         or -fno-lto is specified and the linker has full plugin support.
631         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
632         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
633         * lto-wrapper.c (merge_and_complain): Merge compile-time
634         optimization levels.
635         (run_gcc): And pass it through to the link options.
637 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
639         PR debug/60381
640         Revert:
641         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
642         PR debug/59992
643         * cselib.c (remove_useless_values): Skip to avoid quadratic
644         behavior if the condition moved from...
645         (cselib_process_insn): ... here holds.
647 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
649         PR plugins/59335
650         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
651         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
653         PR plugins/59335
654         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
655         (TM_H): Add x86-tune.def.
657 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
659         * config/aarch64/aarch64.c (generic_tunings):
660         Use cortexa57_extra_costs.
662 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
664         PR lto/60404
665         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
666         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
667         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
668         cost for in_lto_p.
670 2014-03-04  Heiher  <r@hev.cc>
672         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
673         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
675 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
677         * config/i386/predicates.md (const2356_operand): Change to ...
678         (const2367_operand): ... this.
679         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
680         const2367_operand.
681         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
682         (*avx512pf_scatterpf<mode>sf): Ditto.
683         (avx512pf_scatterpf<mode>df): Ditto.
684         (*avx512pf_scatterpf<mode>df_mask): Ditto.
685         (*avx512pf_scatterpf<mode>df): Ditto.
686         * config/i386/i386.c (ix86_expand_builtin): Update
687         incorrect hint operand error message.
689 2014-03-04  Richard Biener  <rguenther@suse.de>
691         * lto-section-in.c (lto_get_section_data): Fix const cast.
693 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
695         * tree-streamer.c (record_common_node): Assert we don't record
696         nodes with type double.
697         (preload_common_node): Skip type double, complex double and double
698         pointer since it is now frontend dependent due to fshort-double option.
700 2014-03-04  Richard Biener  <rguenther@suse.de>
702         PR lto/60405
703         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
704         (lto_input_toplevel_asms): Likewise.
705         * lto-section-in.c (lto_get_section_data): Instead do it here
706         for every section.
708 2014-03-04  Richard Biener  <rguenther@suse.de>
710         PR tree-optimization/60382
711         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
712         dead PHIs a reduction.
714 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
716         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
717         hint value.
718         (_mm_prefetch): Move out of GCC target("sse") pragma.
719         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
720         GCC target("prfchw") pragma.
721         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
722         for locality <= 2.
723         * config/i386/i386.c (ix86_option_override_internal): Enable
724         -mprfchw with -mprefetchwt1.
726 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
728         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
729         Mark as varying.
731 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
733         * opts.h (CL_PCH_IGNORE): Define.
734         * targhooks.c (option_affects_pch_p):
735         Return false for options that have CL_PCH_IGNORE set.
736         * opt-functions.awk: Process PchIgnore.
737         * doc/options.texi: Document PchIgnore.
739         * config/arc/arc.opt (misize): Add PchIgnore property.
741 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
743         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
744         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
745         constraint on constants to permit them being loaded into
746         GENERAL_REGS or BASE_REGS.
748 2014-03-03  Nick Clifton  <nickc@redhat.com>
750         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
751         anti-cacnonical alternatives.
752         (negandhi3_real): New pattern.
753         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
755 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
757         * config/avr/avr-mcus.def: Remove atxmega16x1.
758         * config/avr/avr-tables.opt: Regenerate.
759         * config/avr/t-multilib: Regenerate.
760         * doc/avr-mmcu.texi: Regenerate.
762 2014-03-03  Tobias Grosser  <tobias@grosser.es>
763             Mircea Namolaru  <mircea.namolaru@inria.fr>
765         PR tree-optimization/58028
766         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
767         scalar dimensions.
769 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
771         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
772         not handled by recognizers.
774 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
776         PR middle-end/60175
777         * function.c (expand_function_end): Don't emit
778         clobber_return_register sequence if clobber_after is a BARRIER.
779         * cfgexpand.c (construct_exit_block): Append instructions before
780         return_label to prev_bb.
782 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
784         * config/rs6000/constraints.md: Document reserved use of "wc".
786 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
788         PR ipa/60150
789         * ipa.c (function_and_variable_visibility): When dissolving comdat
790         group, also set all symbols to local.
792 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
794         PR ipa/60306
796         Revert:
797         2013-12-14   Jan Hubicka  <jh@suse.cz>
798         PR middle-end/58477
799         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
801 2014-03-02  Jon Beniston  <jon@beniston.com>
803         PR bootstrap/48230
804         PR bootstrap/50927
805         PR bootstrap/52466
806         PR target/46898
807         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
808           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
809         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
810         (simple_return, *simple_return): New patterns
811         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
812         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
814 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
816         * dwarf2out.c (gen_subprogram_die): Tidy.
818 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
820         PR target/60071
821         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
822         (*mov_t_msb_neg_negc): ... this new insn.
824 2014-02-28  Jason Merrill  <jason@redhat.com>
826         PR c++/58678
827         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
828         function.
830 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
832         PR c++/60314
833         * dwarf2out.c (decltype_auto_die): New static.
834         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
835         (gen_type_die_with_usage): Handle 'decltype(auto)'.
836         (is_cxx_auto): Likewise.
838 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
840         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
841         we are not using general regs only.
843 2014-02-28  Richard Biener  <rguenther@suse.de>
845         PR target/60280
846         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
847         previous fix and only allow to remove trivial pre-headers
848         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
849         (remove_forwarder_block): Properly update the latch of a loop.
851 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
853         PR debug/59992
854         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
855         (cselib_preserved_hash_table): New.
856         (preserve_constants_and_equivs): Move preserved vals to it.
857         (cselib_find_slot): Look it up first.
858         (cselib_init): Initialize it.
859         (cselib_finish): Release it.
860         (dump_cselib_table): Dump it.
862 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
864         PR debug/59992
865         * cselib.c (remove_useless_values): Skip to avoid quadratic
866         behavior if the condition moved from...
867         (cselib_process_insn): ... here holds.
869 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
871         PR debug/57232
872         * var-tracking.c (vt_initialize): Apply the same condition to
873         preserve the CFA base value.
875 2014-02-28  Joey Ye  <joey.ye@arm.com>
877         PR target/PR60169
878         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
879         if reload in progress or completed.
881 2014-02-28  Tobias Burnus  <burnus@net-b.de>
883         PR middle-end/60147
884         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
885         NAMELIST_DECL.
887 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
889         * doc/tm.texi.in (Condition Code Status): Update documention for
890         relative locations of cc0-setter and cc0-user.
892 2014-02-27  Jeff Law  <law@redhat.com>
894         PR rtl-optimization/52714
895         * combine.c (try_combine): When splitting an unrecognized PARALLEL
896         into two independent simple sets, if I3 is a jump, ensure the
897         pattern we place into I3 is a (set (pc) ...).
899 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
900             Jeff Law  <law@redhat.com>
902         PR rtl-optimization/49847
903         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
904         are in different blocks.
905         * doc/tm.texi (Condition Code Status): Update documention for
906         relative locations of cc0-setter and cc0-user.
908 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
910         PR target/59222
911         * lra.c (lra_emit_add): Check SUBREG too.
913 2014-02-27  Andreas Schwab  <schwab@suse.de>
915         * config/m68k/m68k.c (m68k_option_override): Disable
916         -flive-range-shrinkage for classic m68k.
917         (m68k_override_options_after_change): Likewise.
919 2014-02-27  Marek Polacek  <polacek@redhat.com>
921         PR middle-end/59223
922         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
923         -Wmaybe-uninitialized.
925 2014-02-27  Alan Modra  <amodra@gmail.com>
927         PR target/57936
928         * reload1.c (emit_input_reload_insns): When reload_override_in,
929         set old to rl->in_reg when rl->in_reg is a subreg.
931 2014-02-26  Richard Biener  <rguenther@suse.de>
933         PR bootstrap/60343
934         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
936 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
938         * common/config/i386/predicates.md (const1256_operand): Remove.
939         (const2356_operand): New.
940         (const_1_to_2_operand): Remove.
941         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
942         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
943         (*avx512pf_gatherpf<mode>sf): Ditto.
944         (avx512pf_gatherpf<mode>df): Ditto.
945         (*avx512pf_gatherpf<mode>df_mask): Ditto.
946         (*avx512pf_gatherpf<mode>df): Ditto.
947         (avx512pf_scatterpf<mode>sf): Ditto.
948         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
949         (*avx512pf_scatterpf<mode>sf): Ditto.
950         (avx512pf_scatterpf<mode>df): Ditto.
951         (*avx512pf_scatterpf<mode>df_mask): Ditto.
952         (*avx512pf_scatterpf<mode>df): Ditto.
953         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
955 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
957         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
958         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
959         (_mm512_mask_testn_epi64_mask): Move to ...
960         * config/i386/avx512cdintrin.h: Here.
961         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
962         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
963         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
964         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
965         TARGET_AVX512F from TARGET_AVX512CD.
967 2014-02-26  Richard Biener  <rguenther@suse.de>
969         PR ipa/60327
970         * ipa.c (walk_polymorphic_call_targets): Properly guard
971         call to inline_update_overall_summary.
973 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
975         PR target/60280
976         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
977         and latches only if requested.  Fix latch if it is removed.
978         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
979         LOOPS_HAVE_PREHEADERS.
981 2014-02-25  Andrew Pinski  <apinski@cavium.com>
983         * builtins.c (expand_builtin_thread_pointer): Create a new target
984         when the target is NULL.
986 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
988         PR rtl-optimization/60317
989         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
990         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
991         * lra-assigns.c: Include params.h.
992         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
993         other reload pseudos considerations.
995 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
997         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
998         to use canonical form for nor<mode>3.
1000 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1002         PR target/55426
1003         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1004         conversions.
1006 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1008         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1009         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1010         (ix86_handle_option): Handle OPT_mprefetchwt1.
1011         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1012         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1013         PREFETCHWT1 CPUID.
1014         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1015         OPTION_MASK_ISA_PREFETCHWT1.
1016         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1017         (PTA_PREFETCHWT1): New.
1018         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1019         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1020         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1021         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1022         (*prefetch_avx512pf_<mode>_: Change into ...
1023         (*prefetch_prefetchwt1_<mode>: This.
1024         * config/i386/i386.opt (mprefetchwt1): New.
1025         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1026         (_mm_prefetch): Handle intent to write.
1027         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1029 2014-02-25  Richard Biener  <rguenther@suse.de>
1031         PR middle-end/60291
1032         * emit-rtl.c (mem_attrs_htab): Remove.
1033         (mem_attrs_htab_hash): Likewise.
1034         (mem_attrs_htab_eq): Likewise.
1035         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1036         (init_emit_once): Do not allocate mem_attrs_htab.
1038 2014-02-25  Richard Biener  <rguenther@suse.de>
1040         PR lto/60319
1041         * lto-opts.c (lto_write_options): Output non-explicit conservative
1042         -fwrapv, -fno-trapv and -fno-strict-overflow.
1043         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1044         (run_gcc): And pass them through.
1046 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1048         * sel-sched.c (calculate_new_fences): New parameter ptime.
1049         Calculate it as a maximum over all fence cycles.
1050         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1051         Print the final schedule timing when sched_verbose.
1053 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1055         PR rtl-optimization/60292
1056         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1057         bit fot the fence instruction.
1059 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1061         * calls.h: Fix typo in comment.
1063 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1065         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1066         adjusting offsetable addresses.
1068 2014-02-24  Guozhi Wei  <carrot@google.com>
1070         * sparseset.h (sparseset_pop): Fix the wrong index.
1072 2014-02-24  Walter Lee  <walt@tilera.com>
1074         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1075         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1076         triplet.
1077         * common/config/tilegx/tilegx-common.c
1078         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1079         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1080         (LINK_SPEC): Ditto.
1081         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1082         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1083         (tilegx_gimplify_va_arg_expr): Handle big endian.
1084         (tilegx_expand_unaligned_load): Ditto.
1085         (tilegx_expand_unaligned_store): Ditto.
1086         (TARGET_RETURN_IN_MSB): New.
1087         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1088         (TARGET_ENDIAN_DEFAULT): New.
1089         (TARGET_BIG_ENDIAN): Handle big endian.
1090         (BYTES_BIG_ENDIAN): Ditto.
1091         (WORDS_BIG_ENDIAN): Ditto.
1092         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1093         (ENDIAN_SPEC): New.
1094         (EXTRA_SPECS): New.
1095         * config/tilegx/tilegx.md (extv): Handle big endian.
1096         (extzv): Ditto.
1097         (insn_st<n>): Ditto.
1098         (insn_st<n>_add<bitsuffix>): Ditto.
1099         (insn_stnt<n>): Ditto.
1100         (insn_stnt<n>_add<bitsuffix>):Ditto.
1101         (vec_interleave_highv8qi): Handle big endian.
1102         (vec_interleave_highv8qi_be): New.
1103         (vec_interleave_highv8qi_le): New.
1104         (insn_v1int_h): Handle big endian.
1105         (vec_interleave_lowv8qi): Handle big endian.
1106         (vec_interleave_lowv8qi_be): New.
1107         (vec_interleave_lowv8qi_le): New.
1108         (insn_v1int_l): Handle big endian.
1109         (vec_interleave_highv4hi): Handle big endian.
1110         (vec_interleave_highv4hi_be): New.
1111         (vec_interleave_highv4hi_le): New.
1112         (insn_v2int_h): Handle big endian.
1113         (vec_interleave_lowv4hi): Handle big endian.
1114         (vec_interleave_lowv4hi_be): New.
1115         (vec_interleave_lowv4hi_le): New.
1116         (insn_v2int_l): Handle big endian.
1117         (vec_interleave_highv2si): Handle big endian.
1118         (vec_interleave_highv2si_be): New.
1119         (vec_interleave_highv2si_le): New.
1120         (insn_v4int_h): Handle big endian.
1121         (vec_interleave_lowv2si): Handle big endian.
1122         (vec_interleave_lowv2si_be): New.
1123         (vec_interleave_lowv2si_le): New.
1124         (insn_v4int_l): Handle big endian.
1125         * config/tilegx/tilegx.opt (mbig-endian): New option.
1126         (mlittle-endian): New option.
1127         * doc/install.texi: Document tilegxbe-linux.
1128         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1130 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1132         PR ipa/60266
1133         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1134         there are no parameter descriptors.
1136 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1138         PR rtl-optimization/60268
1139         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1140         initialization to ...
1141         (sched_rgn_init): ... here.
1142         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1144 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1146         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1147         names.
1149 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1151         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1152         definition.
1154 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1156         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1157         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1159 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1161         * config/microblaze/predicates.md: Add cmp_op predicate.
1162         * config/microblaze/microblaze.md: Add branch_compare instruction
1163         which uses cmp_op predicate and emits cmp insn before branch.
1164         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1165         to microblaze_expand_conditional_branch and consolidate logic.
1166         (microblaze_expand_conditional_branch): emit branch_compare
1167         insn instead of handling cmp op separate from branch insn.
1169 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1171         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1172         to permit subregs.
1174 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1176         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1177         define_insn with define_expand and new define_insn
1178         *altivec_lve<VI_char>x_internal.
1179         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1180         and new define_insn *altivec_stve<VI_char>x_internal.
1181         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1182         prototype.
1183         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1184         lve*x built-ins.
1185         (altivec_expand_stvex_be): New function.
1187 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1189         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1190         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1191         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1192         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1194 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1196         PR target/60298
1197         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1198         instead of emit_move_insn.
1200 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1202         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1203         vspltw with vsldoi.
1204         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1205         gen_altivec_vsumsws.
1207 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1209         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1210         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1211         (altivec_lvxl_<mode>): New define_expand incorporating
1212         -maltivec=be semantics where needed.
1213         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1214         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1215         semantics where needed.
1216         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1217         (altivec_stvx_<mode>): New define_expand incorporating
1218         -maltivec=be semantics where needed.
1219         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1220         VM2 iterator instead of V4SI.
1221         (altivec_stvxl_<mode>): New define_expand incorporating
1222         -maltivec=be semantics where needed.
1223         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1224         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1225         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1226         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1227         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1228         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1229         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1230         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1231         ALTIVEC_BUILTIN_STVXL.
1232         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1233         (altivec_expand_stvx_be): Likewise.
1234         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1235         (altivec_expand_lvx_be): Likewise.
1236         (altivec_expand_stvx_be): Likewise.
1237         (altivec_expand_builtin): Add cases for
1238         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1239         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1240         (altivec_init_builtins): Add definitions for
1241         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1242         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1244 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1246         * doc/invoke.texi (mvirt, mno-virt): Document.
1247         * config/mips/mips.opt (mvirt): New option.
1248         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1250 2014-02-21  Richard Biener  <rguenther@suse.de>
1252         PR tree-optimization/60276
1253         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1254         (STMT_VINFO_MIN_NEG_DIST): New macro.
1255         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1256         STMT_VINFO_MIN_NEG_DIST.
1257         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1258         made for negative dependence distances still hold.
1260 2014-02-21  Richard Biener  <rguenther@suse.de>
1262         PR middle-end/60291
1263         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1264         DECL_INITIAL for globals not in the current function context.
1266 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1268         PR tree-optimization/56490
1269         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1270         * tree-ssa-uninit.c: Include params.h.
1271         (compute_control_dep_chain): Add num_calls argument, return false
1272         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1273         num_calls to recursive call.
1274         (find_predicates): Change dep_chain into normal array,
1275         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1276         variable and adjust compute_control_dep_chain caller.
1277         (find_def_preds): Likewise.
1279 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1281         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1282         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1284 2014-02-21  Nick Clifton  <nickc@redhat.com>
1286         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1287         (pushhi1): Likewise.
1288         (popqi1): Add mode to pre_dec.
1289         (pophi1): Likewise.
1291 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1293         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1294         mode for mask of V8SFmode permutation.
1296 2014-02-20  Richard Henderson  <rth@redhat.com>
1298         PR c++/60272
1299         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1300         a new pseudo for OLDVAL.
1302 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1304         PR target/57896
1305         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1306         gen_reg_rtx if d->testing_p.
1307         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1308         if d->testing_p and we will certainly return true.
1309         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1310         if d->testing_p.
1312 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1314         * emit-rtl.c (gen_reg_rtx): Assert that
1315         crtl->emit.regno_pointer_align_length is non-zero.
1317 2014-02-20  Richard Henderson  <rth@redhat.com>
1319         PR c++/60272
1320         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1321         on failure the store back into EXPECT.
1323 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1324             Sandra Loosemore  <sandra@codesourcery.com>
1326         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1327         * config/nios2/nios2.c (nios2_function_profiler): Add
1328         -fPIC (flag_pic == 2) support.
1329         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1330         (nios2_large_offset_p): New function.
1331         (nios2_unspec_reloc_p): Move up position, update to use
1332         nios2_large_offset_p.
1333         (nios2_unspec_address): Remove function.
1334         (nios2_unspec_offset): New function.
1335         (nios2_large_got_address): New function.
1336         (nios2_got_address): Add large offset support.
1337         (nios2_legitimize_tls_address): Update usage of removed and new
1338         functions.
1339         (nios2_symbol_binds_local_p): New function.
1340         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1341         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1342         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1343         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1344         processing for (const (unspec ...)).
1345         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1347 2014-02-20  Richard Biener  <rguenther@suse.de>
1349         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1350         doing the substitution.
1351         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1353 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1355         PR ipa/55260
1356         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1357         info when checking whether lattices are bottom.
1359 2014-02-20  Richard Biener  <rguenther@suse.de>
1361         PR middle-end/60221
1362         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1363         regions at -O0.
1365 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1367         PR ipa/58555
1368         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1369         parameter specifying the scaling.
1370         (inline_call): Update.
1371         (want_inline_recursively): Guard division by zero.
1372         (recursive_inlining): Update.
1373         * ipa-inline.h (clone_inlined_nodes): Update.
1375 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1377         PR target/60204
1378         * config/i386/i386.c (classify_argument): Pass structures of size
1379         64 bytes or less in register.
1381 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1382             Kirill Yukhin  <kirill.yukhin@intel.com>
1384         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1385         (_mm_rcp28_round_ss): Ditto.
1386         (_mm_rsqrt28_round_sd): Ditto.
1387         (_mm_rsqrt28_round_ss): Ditto.
1388         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1389         (_mm_rcp14_round_ss): Ditto.
1390         (_mm_rsqrt14_round_sd): Ditto.
1391         (_mm_rsqrt14_round_ss): Ditto.
1392         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1393         the first input operand, get rid of match_dup.
1394         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1395         attribute to sse.
1396         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1397         Ditto.
1398         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1399         operand as the first input operand, set type attribute.
1400         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1401         Set type attribute.
1402         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1403         operand as the first input operand, set type attribute.
1405 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1407         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1408         bit of zero.
1410 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1412         PR target/60207
1413         * config/i386/i386.c (construct_container): Remove TFmode check
1414         for X86_64_INTEGER_CLASS.
1416 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1418         PR target/59794
1419         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1420         only when -Wpsabi is enabled.
1422 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1424          PR target/59799
1425         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1426         passing arrays in registers are the same as for structs, so remove the
1427         special case for them.
1429 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1431         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1432         destination type, extract only the valid bits if the source type is not
1433         integral and has a different mode.
1435 2014-02-19  Richard Biener  <rguenther@suse.de>
1437         PR ipa/60243
1438         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1439         for all calls.
1441 2014-02-19  Richard Biener  <rguenther@suse.de>
1443         PR ipa/60243
1444         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1445         (ipa_modify_call_arguments): Emit an argument load explicitely and
1446         preserve virtual SSA form there and for the replacement call.
1447         Do not update SSA form nor free dominance info.
1449 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1451         * ipa.c (function_and_variable_visibility): Also clear WEAK
1452         flag when disolving COMDAT_GROUP.
1454 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1456         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1457         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1458         not devirtualizing.
1459         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1460         do more sanity checks.
1461         (detect_type_change): Return true when giving up early.
1462         (compute_complex_assign_jump_func): Fix type parameter of
1463         ipa_set_ancestor_jf.
1464         (compute_complex_ancestor_jump_func): Likewise.
1465         (update_jump_functions_after_inlining): Fix updating of
1466         ancestor function.
1467         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1469 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1471         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1472         inline clones when edge disappears.
1474 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1476         PR target/60203
1477         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1478         Split 64-bit moves into 2 patterns.  Do not allow the use of
1479         direct move for TDmode in little endian, since the decimal value
1480         has little endian bytes within a word, but the 64-bit pieces are
1481         ordered in a big endian fashion, and normal subreg's of TDmode are
1482         not allowed.
1483         (mov<mode>_64bit_dm): Likewise.
1484         (movtd_64bit_nodm): Likewise.
1486 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1488         PR tree-optimization/60174
1489         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1490         statement of an SSA_NAME that occurs in an abnormal PHI node.
1492 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1494         PR sanitizer/60142
1495         * final.c (SEEN_BB): Remove.
1496         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1497         (final_scan_insn): Don't force_source_line on second
1498         NOTE_INSN_BASIC_BLOCK.
1500 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1502         PR target/60205
1503         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1504         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1505         (type_natural_mode): Warn ABI change when %zmm register is not
1506         available for AVX512F vector value passing.
1508 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1510         PR target/60193
1511         * config/i386/i386.c (ix86_expand_prologue): Use value in
1512         rax register as displacement when restoring %r10 or %rax.
1513         Fix wrong offset when restoring both registers.
1515 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1517         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1518         assertion with conditional return.
1520 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1521             Uros Bizjak  <ubizjak@gmail.com>
1523         PR driver/60233
1524         * config/i386/driver-i386.c (host_detect_local_cpu): If
1525         YMM state is not saved by the OS, also clear has_f16c.  Move
1526         CPUID 0x80000001 handling before YMM state saving checking.
1528 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1530         PR rtl-optimization/58960
1531         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1532         factored out from ...
1533         (sched_init): ... here.
1534         (free_global_sched_pressure_data): New, factored out from ...
1535         (sched_finish): ... here.
1536         * sched-int.h (free_global_sched_pressure_data): Declare.
1537         * sched-rgn.c (nr_regions_initial): New static global.
1538         (haifa_find_rgns): Initialize it.
1539         (schedule_region): Disable sched-pressure for the newly
1540         generated regions.
1542 2014-02-17  Richard Biener  <rguenther@suse.de>
1544         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1545         release SSA defs of pattern stmts.
1547 2014-02-17  Richard Biener  <rguenther@suse.de>
1549         * tree-inline.c (expand_call_inline): Release the virtual
1550         operand defined by the call we are about to inline.
1552 2014-02-17  Richard Biener  <rguenther@suse.de>
1554         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1556 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1557             Ilya Tocar  <ilya.tocar@intel.com>
1559         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1560         arguments order in builtin.
1561         (_mm512_permutexvar_epi64): Ditto.
1562         (_mm512_mask_permutexvar_epi64): Ditto
1563         (_mm512_maskz_permutexvar_epi32): Ditto
1564         (_mm512_permutexvar_epi32): Ditto
1565         (_mm512_mask_permutexvar_epi32): Ditto
1567 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1569         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1570         (p8_vmrgow): Likewise.
1572 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1574         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1575         endian targets.
1577 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1579         PR target/60203
1580         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1581         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1582         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1583         using direct move instructions on ISA 2.07.  Also adjust
1584         instruction length for 64-bit.
1585         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1586         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1588 2014-02-15  Alan Modra  <amodra@gmail.com>
1590         PR target/58675
1591         PR target/57935
1592         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1593         find_replacement on parts of insn rtl that might be reloaded.
1595 2014-02-15  Richard Biener  <rguenther@suse.de>
1597         PR tree-optimization/60183
1598         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1599         (tree_ssa_phiprop): Calculate and free post-dominators.
1601 2014-02-14  Jeff Law  <law@redhat.com>
1603         PR rtl-optimization/60131
1604         * ree.c (get_extended_src_reg): New function.
1605         (combine_reaching_defs): Use it rather than assuming location of REG.
1606         (find_and_remove_re): Verify first operand of extension is
1607         a REG before adding the insns to the copy list.
1609 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1611         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1612         * configure: Regenerated.
1613         * config.in: Regenerated.
1614         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1615         instead of ASM_SHORT.
1617 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1618             Richard Earnshaw  <rearnsha@arm.com>
1620         PR rtl-optimization/59535
1621         * lra-constraints.c (process_alt_operands): Encourage alternative
1622         when unassigned pseudo class is superset of the alternative class.
1623         (inherit_reload_reg): Don't inherit when optimizing for code size.
1624         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1625         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1626         modes not less than 4 for Thumb1.
1628 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1630         PR pch/60010
1631         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1633 2014-02-14  Richard Biener  <rguenther@suse.de>
1635         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1636         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1637         Do not strip INDIRECT_REFs.
1639 2014-02-14  Richard Biener  <rguenther@suse.de>
1641         PR lto/60179
1642         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1643         DECL_FUNCTION_SPECIFIC_TARGET.
1644         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1645         * tree-streamer-out.c (pack_ts_target_option): Remove.
1646         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1647         (write_ts_function_decl_tree_pointers): Do not stream
1648         DECL_FUNCTION_SPECIFIC_TARGET.
1649         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1650         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1651         (lto_input_ts_function_decl_tree_pointers): Do not stream
1652         DECL_FUNCTION_SPECIFIC_TARGET.
1654 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1656         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1657         (get_initial_def_for_induction, vectorizable_induction): Ignore
1658         debug stmts when looking for exit_phi.
1659         (vectorizable_live_operation): Fix up condition.
1661 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1663         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1664         nreverse() because it changes the content of original tree list.
1666 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1668         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1669         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1671 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1673         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1674         GNU coding standards.
1676 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1678         PR debug/60152
1679         * dwarf2out.c (gen_subprogram_die): Don't call
1680         add_calling_convention_attribute if subr_die is old_die.
1682 2014-02-13  Sharad Singhai  <singhai@google.com>
1684         * doc/optinfo.texi: Fix order of nodes.
1686 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1688         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1689         operands[2], not operands[3].
1691 2014-02-13  Richard Biener  <rguenther@suse.de>
1693         PR bootstrap/59878
1694         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1695         mention the possibility of an in-tree build.
1696         (CLooG): Update recommended version to 0.18.1, mention the
1697         possibility of an in-tree build and clarify that the ISL
1698         bundled with CLooG does not work.
1700 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1702         PR target/43546
1703         * expr.c (compress_float_constant): If x is a hard register,
1704         extend into a pseudo and then move to x.
1706 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1708         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1709         caused by bad second argument to warning_at() with -mhotpatch and
1710         nested functions (e.g. with gfortran).
1712 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1714         * opts.c (option_name): Remove "enabled by default" rider.
1716 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1718         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1720 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1721             Uros Bizjak  <ubizjak@gmail.com>
1723         PR target/60151
1724         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1725         * configure: Regenerated.
1727 2014-02-12  Richard Biener  <rguenther@suse.de>
1729         * vec.c (vec_prefix::calculate_allocation): Move as
1730         inline variant to vec.h.
1731         (vec_prefix::calculate_allocation_1): New out-of-line version.
1732         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1733         (vec_prefix::m_has_auto_buf): Rename to ...
1734         (vec_prefix::m_using_auto_storage): ... this.
1735         (vec_prefix::calculate_allocation): Inline the easy cases
1736         and dispatch to calculate_allocation_1 which doesn't need the
1737         prefix address.
1738         (va_heap::reserve): Use gcc_checking_assert.
1739         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1740         m_using_auto_storage.
1741         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1742         member and adjust.
1743         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1744         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1745         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1747 2014-02-12  Richard Biener  <rguenther@suse.de>
1749         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1750         when we found a dependence.
1752 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1754         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1755         common code...
1756         (maybe_fold_stmt): ... into this new function.
1757         * omp-low.c (lower_omp): Update comment.
1759         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1760         last use.
1762         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1763         dereference.
1765 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1767         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1768         identifiers in comments.
1769         (cortexa53_extra_costs): Likewise.
1770         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1771         (cortexa7_extra_costs): Likewise.
1772         (cortexa12_extra_costs): Likewise.
1773         (cortexa15_extra_costs): Likewise.
1774         (v7m_extra_costs): Likewise.
1776 2014-02-12  Richard Biener  <rguenther@suse.de>
1778         PR middle-end/60092
1779         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1780         of posix_memalign being successful.
1781         (lower_stmt): Restrict lowering of posix_memalign to when
1782         -ftree-bit-ccp is enabled.
1784 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1786         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1787         arg_loc.
1788         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1790 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1792         PR rtl-optimization/60116
1793         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1794         other_insn once the combination has been validated.
1796 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1798         PR lto/59468
1799         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1800         and wrapper.
1801         * ipa-devirt.c: Include demangle.h
1802         (odr_violation_reported): New static variable.
1803         (add_type_duplicate): Update odr_violations.
1804         (maybe_record_node): Add completep parameter; update it.
1805         (record_target_from_binfo): Add COMPLETEP parameter;
1806         update it as needed.
1807         (possible_polymorphic_call_targets_1): Likewise.
1808         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1809         rename FINAL to COMPLETE.
1810         (record_targets_from_bases): Sanity check we found the binfo;
1811         fix COMPLETEP updating.
1812         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1813         parameter, fix computing of COMPLETEP.
1814         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1815         at LTO time do demangling.
1816         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1817         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1818         parameter.
1819         (gimple_get_virt_method_for_binfo): Likewise.
1820         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1821         gimple_get_virt_method_for_vtable): Update prototypes.
1823 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1825         PR target/49008
1826         * genautomata.c (add_presence_absence): Fix typo with
1827         {final_}presence_list.
1829 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1831         PR target/60137
1832         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1833         for VSX/Altivec vectors that land in GPR registers.
1835 2014-02-11  Richard Henderson  <rth@redhat.com>
1836             Jakub Jelinek  <jakub@redhat.com>
1838         PR debug/59776
1839         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1840         around drhs if type conversion to lacc->type is not useless.
1842 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1844         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1845         tuning struct.
1846         (cortex-a57.cortex-a53): Likewise.
1847         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1849 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1851         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1852         arm_restrict_it.
1854 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1856         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1857         add_options_for_arm_vfp3.
1859 2014-02-11  Jeff Law  <law@redhat.com>
1861         PR middle-end/54041
1862         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1863         object with an undesirable mode.
1865 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1867         PR libgomp/60107
1868         * config/i386/sol2-9.h: New file.
1869         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1870         *-*-solaris2.9*): Use it.
1872 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1874         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1875         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1877 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1879         * config/microblaze/microblaze.c: Extend mcpu version format
1881 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
1883         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1885 2014-02-10  Richard Henderson  <rth@redhat.com>
1887         PR target/59927
1888         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1889         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1890         ms-abi vs -mno-accumulate-outgoing-args.
1891         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1892         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1893         respect to ms-abi.
1895 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1897         PR middle-end/60080
1898         * cfgexpand.c (expand_asm_operands): Attach source location to
1899         ASM_INPUT rtx objects.
1900         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1902 2014-02-10  Nick Clifton  <nickc@redhat.com>
1904         * config/mn10300/mn10300.c (popcount): New function.
1905         (mn10300_expand_prologue): Include saved registers in stack usage
1906         count.
1908 2014-02-10  Jeff Law  <law@redhat.com>
1910         PR middle-end/52306
1911         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1912         when changing the SET_DEST of a prior insn to avoid an input reload.
1914 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
1916         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1917         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1918         -mcall-openbsd, or -mcall-linux.
1919         (CC1_ENDIAN_BIG_SPEC): Remove.
1920         (CC1_ENDIAN_LITTLE_SPEC): Remove.
1921         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1922         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1923         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1924         and %cc1_endian_default.
1925         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1927 2014-02-10  Richard Biener  <rguenther@suse.de>
1929         PR tree-optimization/60115
1930         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1931         MEM_REF handling.  Properly verify that the accesses are not
1932         out of the objects bound.
1934 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1936         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1937         coretex to cortex.
1939 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
1941         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1942         proper constants and fix formatting.
1943         (possible_polymorphic_call_targets): Fix formatting.
1945 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
1946             Ilya Tocar  <ilya.tocar@intel.com>
1948         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1949         (_mm512_loadu_epi32): Renamed into...
1950         (_mm512_loadu_si512): This.
1951         (_mm512_storeu_epi32): Renamed into...
1952         (_mm512_storeu_si512): This.
1953         (_mm512_maskz_ceil_ps): Removed.
1954         (_mm512_maskz_ceil_pd): Ditto.
1955         (_mm512_maskz_floor_ps): Ditto.
1956         (_mm512_maskz_floor_pd): Ditto.
1957         (_mm512_floor_round_ps): Ditto.
1958         (_mm512_floor_round_pd): Ditto.
1959         (_mm512_ceil_round_ps): Ditto.
1960         (_mm512_ceil_round_pd): Ditto.
1961         (_mm512_mask_floor_round_ps): Ditto.
1962         (_mm512_mask_floor_round_pd): Ditto.
1963         (_mm512_mask_ceil_round_ps): Ditto.
1964         (_mm512_mask_ceil_round_pd): Ditto.
1965         (_mm512_maskz_floor_round_ps): Ditto.
1966         (_mm512_maskz_floor_round_pd): Ditto.
1967         (_mm512_maskz_ceil_round_ps): Ditto.
1968         (_mm512_maskz_ceil_round_pd): Ditto.
1969         (_mm512_expand_pd): Ditto.
1970         (_mm512_expand_ps): Ditto.
1971         * config/i386/i386.c (ix86_builtins): Remove
1972         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1973         (bdesc_args): Ditto.
1974         * config/i386/predicates.md (const1256_operand): New.
1975         (const_1_to_2_operand): Ditto.
1976         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1977         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1978         (*avx512pf_gatherpf<mode>sf): Ditto.
1979         (avx512pf_gatherpf<mode>df): Ditto.
1980         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1981         (*avx512pf_gatherpf<mode>df): Ditto.
1982         (avx512pf_scatterpf<mode>sf): Ditto.
1983         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1984         (*avx512pf_scatterpf<mode>sf): Ditto.
1985         (avx512pf_scatterpf<mode>df): Ditto.
1986         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1987         (*avx512pf_scatterpf<mode>df): Ditto.
1988         (avx512f_expand<mode>): Removed.
1989         (<shift_insn><mode>3<mask_name>): Change predicate type.
1991 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1993         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1994         not at the end of datarefs vector use ordered_remove to avoid
1995         reordering datarefs vector.
1997         PR c/59984
1998         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1999         mark local addressable non-static vars as GOVD_PRIVATE
2000         instead of GOVD_LOCAL.
2001         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2002         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2003         than copying them.
2005         PR middle-end/60092
2006         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2007         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2008         assume_aligned or alloc_align attributes.
2009         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2010         arguments.  Handle also assume_aligned and alloc_align attributes.
2011         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2012         calls to functions with assume_aligned or alloc_align attributes.
2013         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2015 2014-02-08  Terry Guo  <terry.guo@arm.com>
2017         * doc/invoke.texi: Document ARM -march=armv7e-m.
2019 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2021         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2022         flag on __cilkrts_rethrow builtin.
2024         PR ipa/60026
2025         * ipa-cp.c (determine_versionability): Fail at -O0
2026         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2027         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2029         Revert:
2030         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2032         PR ipa/60026
2033         * tree-inline.c (copy_forbidden): Fail for
2034         __attribute__((optimize (0))) functions.
2036 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2038         * varpool.c: Include pointer-set.h.
2039         (varpool_remove_unreferenced_decls): Variables in other partitions
2040         will not be output; be however careful to not lose information
2041         about partitioning.
2043 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2045         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2046         lookup in the vtable constructor.
2048 2014-02-07  Jeff Law  <law@redhat.com>
2050         PR target/40977
2051         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2052         define_insn_and_split.
2054         * ipa-inline.c (inline_small_functions): Fix typos.
2056 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2058         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2059         (s390_can_use_return_insn): Declare.
2060         * config/s390/s390.h (EPILOGUE_USES): Define.
2061         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2062         instructions.
2063         (s390_chunkify_start): Handle return JUMP_LABELs.
2064         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2065         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2066         (s390_can_use_return_insn): New functions.
2067         (s390_fix_long_loop_prediction): Handle conditional returns.
2068         (TARGET_SET_UP_BY_PROLOGUE): Define.
2069         * config/s390/s390.md (ANY_RETURN): New code iterator.
2070         (*creturn, *csimple_return, return, simple_return): New patterns.
2072 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2074         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2075         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2076         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2077         REG_CFA_RESTORE list when deciding not to restore a register.
2079 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2081         * config/s390/s390.c: Include tree-pass.h and context.h.
2082         (s390_early_mach): New function, split out from...
2083         (s390_emit_prologue): ...here.
2084         (pass_data_s390_early_mach): New pass structure.
2085         (pass_s390_early_mach): New class.
2086         (s390_option_override): Create and register early_mach pass.
2087         Move to end of file.
2089 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2091         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2092         to match for the exit block.
2094 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2096         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2097         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2098         Reject misaligned operands.
2100 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2102         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2104 2014-02-07  Richard Biener  <rguenther@suse.de>
2106         PR middle-end/60092
2107         * gimple-low.c (lower_builtin_posix_memalign): New function.
2108         (lower_stmt): Call it to lower posix_memalign in a way
2109         to make alignment info accessible.
2111 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2113         PR c++/60082
2114         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2115         __builtin_setjmp_receiver.
2117 2014-02-07  Richard Biener  <rguenther@suse.de>
2119         PR middle-end/60092
2120         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2121         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2122         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2123         Handle BUILT_IN_POSIX_MEMALIGN.
2124         (find_func_clobbers): Likewise.
2125         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2126         (call_may_clobber_ref_p_1): Likewise.
2128 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2130         PR ipa/59918
2131         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2132         sanity check.
2134 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2136         PR ipa/59469
2137         * lto-cgraph.c (lto_output_node): Use
2138         symtab_get_symbol_partitioning_class.
2139         (lto_output_varpool_node): likewise.
2140         (symtab_get_symbol_partitioning_class): Move here from
2141         lto/lto-partition.c
2142         * cgraph.h (symbol_partitioning_class): Likewise.
2143         (symtab_get_symbol_partitioning_class): Declare.
2145 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2147         * ggc.h (ggc_internal_cleared_alloc): New macro.
2148         * vec.h (vec_safe_copy): Handle memory stats.
2149         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2150         * target-globals.c (save_target_globals): Likewise.
2152 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2154         PR target/60077
2155         * expr.c (emit_move_resolve_push): Export; be bit more selective
2156         on when to clear alias set.
2157         * expr.h (emit_move_resolve_push): Declare.
2158         * function.h (struct function): Add tail_call_marked.
2159         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2160         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2161         * config/i386/i386.md (TImode move expander): De not call
2162         ix86_expand_push.
2163         (FP push expanders): Preserve memory attributes.
2164         * config/i386/sse.md (push<mode>1): Remove.
2165         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2166         (ix86_expand_push): Remove.
2167         * config/i386/mmx.md (push<mode>1): Remove.
2169 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2171         PR rtl-optimization/60030
2172         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2173         lopart with paradoxical subreg before shifting it up by hprec.
2175 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2177         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2178         Remove extra newline at end of file.
2179         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2180         (arm_issue_rate): Handle cortexa57.
2181         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2182         (cortex-a57.cortex-a53): Likewise.
2184 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2186         PR target/59575
2187         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2188         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2189         bitmask.
2190         (arm_expand_prologue): Adjust all callers.
2191         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2192         info, registers also at the lowest numbered registers side.  Use
2193         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2194         XEXP.
2196         PR debug/59992
2197         * var-tracking.c (adjust_mems): Before adding a SET to
2198         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2200 2014-02-06  Alan Modra  <amodra@gmail.com>
2202         PR target/60032
2203         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2204         change SDmode to DDmode when lra_in_progress.
2206 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2208         PR middle-end/59150
2209         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2210         free_data_ref on the dr first, and before goto again also set dr
2211         to the next dr.  For simd_lane_access, free old datarefs[i] before
2212         overwriting it.  For get_vectype_for_scalar_type failure, don't
2213         free_data_ref if simd_lane_access.
2215         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2217         PR target/60062
2218         * tree.h (opts_for_fn): New inline function.
2219         (opt_for_fn): Define.
2220         * config/i386/i386.c (ix86_function_regparm): Use
2221         opt_for_fn (decl, optimize) instead of optimize.
2223 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2225         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2226         for SYMBOL_REF in large memory model.
2228 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2230         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2231         and crypto support.
2232         (cortex-a57): Likewise.
2233         (cortex-a57.cortex-a53): Likewise.
2235 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2236             Kugan Vivekanandarajah  <kuganv@linaro.org>
2238         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2239         unaligned_access.
2240         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2242 2014-02-06  Richard Biener  <rguenther@suse.de>
2244         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2245         set_loop_copy and initialize_original_copy_tables.
2247 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2249         * config/aarch64/aarch64-simd.md
2250         (aarch64_ashr_simddi): Change QI to SI.
2252 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2253             Jakub Jelinek  <jakub@redhat.com>
2255         PR middle-end/60013
2256         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2257         of the dataflow.
2259 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2261         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2262         CODE_FOR_altivec_vpku[hw]um to
2263         CODE_FOR_altivec_vpku[hw]um_direct.
2264         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2265         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2266         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2267         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2269 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2271         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2272         generation for -maltivec=be.
2273         (altivec_vsumsws): Simplify redundant test.
2275 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2277         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2278         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2279         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2280         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2281         gen_altivec_vpkuwum.
2282         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2283         BYTES_BIG_ENDIAN.
2284         (altivec_vpks<VI_char>ss): Likewise.
2285         (altivec_vpks<VI_char>us): Likewise.
2286         (altivec_vpku<VI_char>us): Likewise.
2287         (altivec_vpku<VI_char>um): Likewise.
2288         (altivec_vpku<VI_char>um_direct): New (copy of
2289         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2290         internal use).
2291         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2292         target is little endian and -maltivec=be is not specified.
2293         (*altivec_vupkhs<VU_char>_direct): New (copy of
2294         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2295         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2296         target is little endian and -maltivec=be is not specified.
2297         (*altivec_vupkls<VU_char>_direct): New (copy of
2298         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2299         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2300         little endian and -maltivec=be is not specified.
2301         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2302         little endian and -maltivec=be is not specified.
2304 2014-02-05  Richard Henderson  <rth@redhat.com>
2306         PR debug/52727
2307         * combine-stack-adj.c: Revert r206943.
2308         * sched-int.h (struct deps_desc): Add last_args_size.
2309         * sched-deps.c (init_deps): Initialize it.
2310         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2311         contain REG_ARGS_SIZE notes.
2313 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2315         * lto-cgraph.c (asm_nodes_output): Make global.
2316         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2317         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2318         (driver_handle_option): Handle OPT_fwpa.
2320 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2322         PR ipa/59947
2323         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2324         a comment typo and formatting issue.  If odr_hash hasn't been
2325         created, return vNULL and set *completep to false.
2327         PR middle-end/57499
2328         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2329         bb with no successors.
2331 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2333         PR target/59718
2334         * doc/invoke.texi (-march): Clarify documentation for ARM.
2335         (-mtune): Likewise.
2336         (-mcpu): Likewise.
2338 2014-02-05  Richard Biener  <rguenther@suse.de>
2340         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2341         when not vectorizing because of too many alias checks.
2342         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2343         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2345 2014-02-05  Nick Clifton  <nickc@redhat.com>
2347         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2348         accept extended registers in any mode when compiling for the MN10300.
2350 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2352         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2353         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2354         sanitization attributes.
2355         (can_inline_edge_p): Likewise.
2356         (sanitize_attrs_match_for_inline_p): New function.
2358 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2360         * ipa-prop.c (detect_type_change): Shor circuit testing of
2361         type changes on THIS pointer.
2363 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2365         PR target/59777
2366         * config/pa/pa.c (legitimize_tls_address): Return original address
2367         if not passed a SYMBOL_REF rtx.
2368         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2369         addresses.
2370         (pa_emit_move_sequence): Simplify TLS source operands.
2371         (pa_legitimate_constant_p): Reject all TLS constants.
2372         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2373         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2375 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2377         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2378         groups when we know they are controlled by LTO.
2379         * varasm.c (default_binds_local_p_1): If object is in other partition,
2380         it will be resolved locally.
2382 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2384         * config/host-linux.c (linux_gt_pch_use_address): Don't
2385         use SSIZE_MAX because it is not always defined.
2387 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2389         PR bootstrap/59913
2390         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2391         threshold for pseudo splitting.
2392         (update_ebb_live_info): Process call argument hard registers and
2393         hard registers from insn definition too.
2394         (max_small_class_regs_num): New constant.
2395         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2396         reloads_num only for small register classes.  Don't split for
2397         outputs of jumps.
2399 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2401         PR ipa/60058
2402         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2403         is non-null.
2405 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2407         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2408         visibility is safe.
2410 2014-02-04  Marek Polacek  <polacek@redhat.com>
2412         * gdbinit.in (pel): Define.
2414 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2416         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2417         behavior.
2419 2014-02-04  Richard Biener  <rguenther@suse.de>
2421         PR lto/59723
2422         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2423         in function context local.
2424         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2425         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2426         similar to LTO_imported_decl_ref.
2428 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2430         PR tree-optimization/60002
2431         * cgraphclones.c (build_function_decl_skip_args): Clear
2432         DECL_LANG_SPECIFIC.
2434         PR tree-optimization/60023
2435         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2436         false to gsi_replace.
2437         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2438         has been in some EH region and vec_stmt could throw, add
2439         vec_stmt into the same EH region.
2440         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2441         has no lhs, ignore it.
2442         * internal-fn.c (expand_MASK_LOAD): Likewise.
2444         PR ipa/60026
2445         * tree-inline.c (copy_forbidden): Fail for
2446         __attribute__((optimize (0))) functions.
2448         PR other/58712
2449         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2450         is set, copy one less argument.
2451         (expand_simd_clones): Don't subtract clone_info->inbranch
2452         from simd_clone_struct_alloc argument.
2454         PR rtl-optimization/57915
2455         * recog.c (simplify_while_replacing): If all unary/binary/relational
2456         operation arguments are constant, attempt to simplify those.
2458         PR middle-end/59261
2459         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2460         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2462 2014-02-04  Richard Biener  <rguenther@suse.de>
2464         PR tree-optimization/60012
2465         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2466         TBAA disambiguation to all DDRs.
2468 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2470         PR target/59788
2471         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2472         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2474 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2476         PR ipa/59882
2477         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2479 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2481         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2482         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2484 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2486         PR ipa/59831
2487         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2488         to figure out targets of polymorphic calls with known decl.
2489         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2490         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2491         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2492         (get_polymorphic_call_info): ... here.
2493         (get_polymorphic_call_info_from_invariant): New function.
2495 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2497         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2498         lookup via vtable pointer; check for type consistency
2499         and turn inconsitent facts into UNREACHABLE.
2500         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2501         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2502         type inconsistent querries; return UNREACHABLE instead.
2504 2014-02-03  Richard Henderson  <rth@twiddle.net>
2506         PR tree-opt/59924
2507         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2508         already processed this node.
2509         (normalize_one_pred_1): Pass along mark_set.
2510         (normalize_one_pred): Create and destroy a pointer_set_t.
2511         (normalize_one_pred_chain): Likewise.
2513 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2515         PR gcov-profile/58602
2516         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2518 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2520         PR ipa/59831
2521         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2522         -fno-devirtualize; try to devirtualize by the knowledge of
2523         virtual table pointer given by aggregate propagation.
2524         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2525         (ipa_print_node_jump_functions): Dump also offset that
2526         is relevant for polymorphic calls.
2527         (determine_known_aggregate_parts): Add arg_type parameter; use it
2528         instead of determining the type from pointer type.
2529         (ipa_compute_jump_functions_for_edge): Update call of
2530         determine_known_aggregate_parts.
2531         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2532         (gimple_get_virt_method_for_binfo): ... here; simplify using
2533         vtable_pointer_value_to_vtable.
2534         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2535         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2536         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2537         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2538         POINTER_PLUS_EXPR.
2539         (vtable_pointer_value_to_binfo): ... here.
2540         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2542 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2544         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2545         redef of outer loop index variable.
2547 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2549         PR c++/53017
2550         PR c++/59211
2551         * doc/extend.texi (Function Attributes): Typo.
2553 2014-02-03  Cong Hou  <congh@google.com>
2555         PR tree-optimization/60000
2556         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2557         if the vectorized statement is a store.  A store statement can only
2558         appear at the end of pattern statements.
2560 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2562         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2563         (ix86_option_override_internal): Default long double to 64-bit for
2564         32-bit Bionic and to 128-bit for 64-bit Bionic.
2566         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2567         TARGET_LONG_DOUBLE_128 is true.
2568         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2570         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2571         (mlong-double-64): Negate -mlong-double-128.
2572         (mlong-double-128): New option.
2574         * config/i386/i386-c.c (ix86_target_macros): Define
2575         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2577         * doc/invoke.texi: Document -mlong-double-128.
2579 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2581         PR rtl-optimization/60024
2582         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2584 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2586         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2588 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2590         PR rtl-optimization/57662
2591         * sel-sched.c (code_motion_path_driver): Do not mark already not
2592         existing blocks in the visiting bitmap.
2594 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2596         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2597         on the insn being emitted.
2599 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2600             Will Deacon  <will.deacon@arm.com>
2602         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2604 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2606         * config/arm/arm-tables.opt: Regenerate.
2608 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2610         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2611         for vector types other than V16QImode.
2612         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2613         define_expand, and call altivec_expand_vec_perm_le when producing
2614         code with little endian element order.
2615         (*altivec_vperm_<mode>_internal): New insn having previous
2616         behavior of altivec_vperm_<mode>.
2617         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2618         altivec_expand_vec_perm_le when producing code with little endian
2619         element order.
2620         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2621         behavior of altivec_vperm_<mode>_uns.
2623 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2625         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2626         (altivec_vsumsws): Add handling for -maltivec=be with a little
2627         endian target.
2628         (altivec_vsumsws_direct): New.
2629         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2630         gen_altivec_vsumsws.
2632 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2634         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2635         vtable_pointer_value_to_binfo): New functions.
2636         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2637         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2639 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2641         * config/nios2/nios2.md (load_got_register): Initialize GOT
2642         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2643         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2645 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2647         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2648         preserverd by passthrough, do not propagate the type.
2650 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2652         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2653         (mips_atomic_assign_expand_fenv): New function.
2654         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2656 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2658         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2659         (__builtin_mips_set_fcsr): Likewise.
2660         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2661         MIPS_USI_FTYPE_VOID.
2662         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2663         (mips16_expand_set_fcsr): Likewise.
2664         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2665         (mips16_set_fcsr_stub): Likewise.
2666         (mips16_get_fcsr_one_only_stub): New class.
2667         (mips16_set_fcsr_one_only_stub): Likewise.
2668         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2669         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2670         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2671         (hard_float): New availability predicate.
2672         (mips_builtins): Add get_fcsr and set_fcsr.
2673         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2674         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2675         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2676         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2677         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2678         patterns.
2680 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2682         * config/mips/mips.c (mips_one_only_stub): New class.
2683         (mips_need_mips16_rdhwr_p): Replace with...
2684         (mips16_rdhwr_stub): ...this new variable.
2685         (mips16_stub_call_address): New function.
2686         (mips16_rdhwr_one_only_stub): New class.
2687         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2688         (mips_output_mips16_rdhwr): Delete.
2689         (mips_finish_stub): New function.
2690         (mips_code_end): Use it to handle rdhwr stubs.
2692 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2694         PR target/60017
2695         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2696         when calculating size of integer atomic types.
2698 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2700         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2702 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2704         PR tree-optimization/60003
2705         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2706         * profile.c (branch_prob): Use gimple_call_builtin_p
2707         to check for BUILT_IN_SETJMP_RECEIVER.
2708         * tree-inline.c (copy_bb): Call notice_special_calls.
2710 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2712         PR bootstrap/59985
2713         * lra-constraints.c (process_alt_operands): Update reload_sum only
2714         on the first pass.
2716 2014-01-31  Richard Henderson  <rth@redhat.com>
2718         PR middle-end/60004
2719         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2720         until after else_eh is processed.
2722 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2724         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2725         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2726         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2727         in smmintrin.h, remove them.
2728         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2729         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2730         * config/i386/i386.md (ROUND_SAE): Fix value.
2731         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2732         (const48_operand): New.
2733         * config/i386/subst.md (round), (round_expand): Use
2734         const_4_or_8_to_11_operand.
2735         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2737 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2739         * config/i386/constraints.md (Yk): Swap meaning with k.
2740         * config/i386/i386.md (movhi_internal): Change Yk to k.
2741         (movqi_internal): Ditto.
2742         (*k<logic><mode>): Ditto.
2743         (*andhi_1): Ditto.
2744         (*andqi_1): Ditto.
2745         (kandn<mode>): Ditto.
2746         (*<code>hi_1): Ditto.
2747         (*<code>qi_1): Ditto.
2748         (kxnor<mode>): Ditto.
2749         (kortestzhi): Ditto.
2750         (kortestchi): Ditto.
2751         (kunpckhi): Ditto.
2752         (*one_cmplhi2_1): Ditto.
2753         (*one_cmplqi2_1): Ditto.
2754         * config/i386/sse.md (): Change k to Yk.
2755         (avx512f_load<mode>_mask): Ditto.
2756         (avx512f_blendm<mode>): Ditto.
2757         (avx512f_store<mode>_mask): Ditto.
2758         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2759         (avx512f_storedqu<mode>_mask): Ditto.
2760         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2761         Ditto.
2762         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2763         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2764         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2765         (avx512f_maskcmp<mode>3): Ditto.
2766         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2767         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2768         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2769         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2770         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2771         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2772         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2773         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2774         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2775         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2776         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2777         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2778         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2779         (vec_extract_lo_<mode>_maskm): Ditto.
2780         (vec_extract_hi_<mode>_maskm): Ditto.
2781         (avx512f_vternlog<mode>_mask): Ditto.
2782         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2783         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2784         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2785         (avx512f_<code>v8div16qi2_mask): Ditto.
2786         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2787         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2788         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2789         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2790         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2791         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2792         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2793         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2794         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2795         (avx512cd_maskb_vec_dupv8di): Ditto.
2796         (avx512cd_maskw_vec_dupv16si): Ditto.
2797         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2798         (avx512f_vpermi2var<mode>3_mask): Ditto.
2799         (avx512f_vpermi2var<mode>3_mask): Ditto.
2800         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2801         (*avx512f_gathersi<mode>): Ditto.
2802         (*avx512f_gathersi<mode>_2): Ditto.
2803         (*avx512f_gatherdi<mode>): Ditto.
2804         (*avx512f_gatherdi<mode>_2): Ditto.
2805         (*avx512f_scattersi<mode>): Ditto.
2806         (*avx512f_scatterdi<mode>): Ditto.
2807         (avx512f_compress<mode>_mask): Ditto.
2808         (avx512f_compressstore<mode>_mask): Ditto.
2809         (avx512f_expand<mode>_mask): Ditto.
2810         * config/i386/subst.md (mask): Change k to Yk.
2811         (mask_scalar_merge): Ditto.
2812         (sd): Ditto.
2814 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2816         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2818 2014-01-31  Richard Biener  <rguenther@suse.de>
2820         PR middle-end/59990
2821         * builtins.c (fold_builtin_memory_op): Make sure to not
2822         use a floating-point mode or a boolean or enumeral type for
2823         the copy operation.
2825 2014-01-30  DJ Delorie  <dj@redhat.com>
2827         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2828         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2829         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2830         whenever main() has an epilogue.
2832 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2834         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2835         unused variable "field".
2836         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2837         (vsx_mergeh_<mode>): Likewise.
2838         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2839         (altivec_vmrghh): Likewise.
2840         (altivec_vmrghw): Likewise.
2841         (altivec_vmrglb): Likewise.
2842         (altivec_vmrglh): Likewise.
2843         (altivec_vmrglw): Likewise.
2844         (altivec_vspltb): Add missing uses.
2845         (altivec_vsplth): Likewise.
2846         (altivec_vspltw): Likewise.
2847         (altivec_vspltsf): Likewise.
2849 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2851         PR target/59923
2852         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2853         frame related instructions.
2855 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2857         PR rtl-optimization/59959
2858         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2859         any reload of register whose subreg is invalid.
2861 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2863         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2864         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2865         Add missing return type - void.
2867 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2869         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2870         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2871         remove element index adjustment for endian (now handled in vsx.md
2872         and altivec.md).
2873         (altivec_expand_vec_perm_const): Use
2874         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2875         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2876         (vsx_xxspltw_<mode>): Adjust element index for little endian.
2877         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2878         define_expand and a new define_insn *altivec_vspltb_internal;
2879         adjust for -maltivec=be on a little endian target.
2880         (altivec_vspltb_direct): New.
2881         (altivec_vsplth): Divide into a define_expand and a new
2882         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2883         little endian target.
2884         (altivec_vsplth_direct): New.
2885         (altivec_vspltw): Divide into a define_expand and a new
2886         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2887         little endian target.
2888         (altivec_vspltw_direct): New.
2889         (altivec_vspltsf): Divide into a define_expand and a new
2890         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2891         a little endian target.
2893 2014-01-30  Richard Biener  <rguenther@suse.de>
2895         PR tree-optimization/59993
2896         * tree-ssa-forwprop.c (associate_pointerplus): Check we
2897         can propagate form the earlier stmt and avoid the transform
2898         when the intermediate result is needed.
2900 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
2902         * README.Portability: Fix typo.
2904 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
2906         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2907         comparison_operator with ordered_comparison_operator.
2909 2014-01-30  Nick Clifton  <nickc@redhat.com>
2911         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2912         Rename to mn10300_store_multiple_regs.
2913         * config/mn10300/mn10300.c: Likewise.
2914         * config/mn10300/mn10300.md (store_movm): Fix typo: call
2915         store_multiple_regs.
2916         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2917         Call mn10300_store_multiple_regs.
2919 2014-01-30  Nick Clifton  <nickc@redhat.com>
2920             DJ Delorie  <dj@redhat.com>
2922         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2923         %fp 2 to keep registers after it properly word-aligned.
2924         (rl78_alloc_physical_registers_umul): Handle the case where both
2925         input operands are the same.
2927 2014-01-30  Richard Biener  <rguenther@suse.de>
2929         PR tree-optimization/59903
2930         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2931         check properly.
2933 2014-01-30  Jason Merrill  <jason@redhat.com>
2935         PR c++/59633
2936         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2938         PR c++/59645
2939         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2941 2014-01-30  Richard Biener  <rguenther@suse.de>
2943         PR tree-optimization/59951
2944         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2946 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
2948         PR target/59784
2949         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2950         SFmode to DFmode case.
2952 2014-01-29  DJ Delorie  <dj@redhat.com>
2954         * config/msp430/msp430.opt (-minrt): New.
2955         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2956         if -minrt given.
2957         (ENDFILE_SPEC): Likewise.
2959 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
2961         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2962         (estimate_function_body_sizes): Use it.
2964 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
2966         PR c++/58561
2967         * dwarf2out.c (is_cxx_auto): New.
2968         (is_base_type): Use it.
2969         (gen_type_die_with_usage): Likewise.
2971 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2973         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
2974         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2975         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2976         -maltivec=be with LE targets.
2977         (vsx_mergeh_<mode>): Likewise.
2978         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2979         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2980         (altivec_vmrghb): Replace with define_expand and new
2981         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2982         (altivec_vmrghb_direct): New define_insn.
2983         (altivec_vmrghh): Replace with define_expand and new
2984         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2985         (altivec_vmrghh_direct): New define_insn.
2986         (altivec_vmrghw): Replace with define_expand and new
2987         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2988         (altivec_vmrghw_direct): New define_insn.
2989         (*altivec_vmrghsf): Adjust for endianness.
2990         (altivec_vmrglb): Replace with define_expand and new
2991         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2992         (altivec_vmrglb_direct): New define_insn.
2993         (altivec_vmrglh): Replace with define_expand and new
2994         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2995         (altivec_vmrglh_direct): New define_insn.
2996         (altivec_vmrglw): Replace with define_expand and new
2997         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2998         (altivec_vmrglw_direct): New define_insn.
2999         (*altivec_vmrglsf): Adjust for endianness.
3000         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3001         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3002         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3003         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3004         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3005         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3006         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3007         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3009 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3011         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3012         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3013         whitespace.
3015 2014-01-29  Richard Biener  <rguenther@suse.de>
3017         PR tree-optimization/58742
3018         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3019         associate_pointerplus_align.
3020         (associate_pointerplus_diff): New function.
3021         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3022         and associate_pointerplus_diff.
3024 2014-01-29  Richard Biener  <rguenther@suse.de>
3026         * lto-streamer.h (LTO_major_version): Bump to 3.
3027         (LTO_minor_version): Reset to 0.
3029 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3031         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3032         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3033         (arm_file_start): Generate correct asm header for armv7ve.
3034         * config/arm/bpabi.h: Add multilib support for armv7ve.
3035         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3036         and cortex-a15 to armv7ve.
3037         * config/arm/t-aprofile: Add multilib support for armv7ve.
3038         * doc/invoke.texi: Document -march=armv7ve.
3040 2014-01-29  Richard Biener  <rguenther@suse.de>
3042         PR tree-optimization/58742
3043         * tree-ssa-forwprop.c (associate_plusminus): Return true
3044         if we changed sth, defer EH cleanup to ...
3045         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3046         (simplify_mult): New function.
3048 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3050         PR middle-end/59917
3051         PR tree-optimization/59920
3052         * tree.c (build_common_builtin_nodes): Remove
3053         __builtin_setjmp_dispatcher initialization.
3054         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3055         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3056         instead of gsi_after_labels + manually skipping debug stmts.
3057         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3058         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3059         * tree-inline.c (copy_edges_for_bb): Remove
3060         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3061         argument.  Ignore computed_goto_p stmts.  Don't call
3062         make_abnormal_goto_edges.  If a call might need abnormal edges
3063         for non-local gotos, see if it already has an edge to
3064         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3065         with true argument, don't do anything then, otherwise add
3066         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3067         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3068         caller.
3069         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3070         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3071         (lower_stmt): Don't set data->calls_builtin_setjmp.
3072         (lower_builtin_setjmp): Adjust comment.
3073         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3074         * tree-cfg.c (found_computed_goto): Remove.
3075         (factor_computed_gotos): Remove.
3076         (make_goto_expr_edges): Return bool, true for computed gotos.
3077         Don't call make_abnormal_goto_edges.
3078         (build_gimple_cfg): Don't set found_computed_goto, don't call
3079         factor_computed_gotos.
3080         (computed_goto_p): No longer static.
3081         (make_blocks): Don't set found_computed_goto.
3082         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3083         (make_edges): If make_goto_expr_edges returns true, push bb
3084         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3085         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3086         vector.  Record mapping between bbs and OpenMP regions if there
3087         are any, adjust make_gimple_omp_edges caller.  Call
3088         handle_abnormal_edges.
3089         (make_abnormal_goto_edges): Remove.
3090         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3091         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3092         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3093         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3094         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3095         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3096         filling *region also set *region_idx to (*region)->entry->index.
3098         PR other/58712
3099         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3100         For REGs set ORIGINAL_REGNO.
3102 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3104         * doc/md.texi: Mention that a target shouldn't implement
3105         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3106         than hi/lo pair.
3108 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3110         PR tree-optimization/59594
3111         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3112         a copy of the datarefs vector rather than the vector itself.
3114 2014-01-28  Jason Merrill  <jason@redhat.com>
3116         PR c++/53756
3117         * dwarf2out.c (auto_die): New static.
3118         (gen_type_die_with_usage): Handle C++1y 'auto'.
3119         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3120         on definition.
3122 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3124         PR target/59672
3125         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3126         (SPEC_X32): Likewise.
3127         (SPEC_64): Likewise.
3128         * config/i386/i386.c (ix86_option_override_internal): Turn off
3129         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3130         for TARGET_16BIT.
3131         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3132         * config/i386/i386.h (TARGET_16BIT): New macro.
3133         (TARGET_16BIT_P): Likewise.
3134         * config/i386/i386.opt: Add m16.
3135         * doc/invoke.texi: Document -m16.
3137 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3139         PR preprocessor/59935
3140         * input.c (location_get_source_line): Bail out on when line number
3141         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3143 2014-01-28  Richard Biener  <rguenther@suse.de>
3145         PR tree-optimization/58742
3146         * tree-ssa-forwprop.c (associate_plusminus): Handle
3147         pointer subtraction of the form (T)(P + A) - (T)P.
3149 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3151         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3152         at const_int_cost.
3154 2014-01-28  Richard Biener  <rguenther@suse.de>
3156         Revert
3157         2014-01-28  Richard Biener  <rguenther@suse.de>
3159         PR rtl-optimization/45364
3160         PR rtl-optimization/59890
3161         * var-tracking.c (local_get_addr_clear_given_value): Handle
3162         already cleared slot.
3163         (val_reset): Handle not allocated local_get_addr_cache.
3164         (vt_find_locations): Use post-order on the inverted CFG.
3166 2014-01-28  Richard Biener  <rguenther@suse.de>
3168         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3170 2014-01-28  Richard Biener  <rguenther@suse.de>
3172         PR rtl-optimization/45364
3173         PR rtl-optimization/59890
3174         * var-tracking.c (local_get_addr_clear_given_value): Handle
3175         already cleared slot.
3176         (val_reset): Handle not allocated local_get_addr_cache.
3177         (vt_find_locations): Use post-order on the inverted CFG.
3179 2014-01-28  Alan Modra  <amodra@gmail.com>
3181         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3182         * configure.ac <recursive call for build != host>: Define
3183         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3184         and LD_FOR_BUILD too.
3185         * configure: Regenerate.
3187 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3189         * config/i386/i386.c (get_builtin_code_for_version): Separate
3190         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3191         Broadwell from Haswell.
3193 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3195         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3196         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3197         * config/mips/mips.c (mips_option_override): Change setting
3198         of TARGET_DSP.
3199         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3200         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3201         Change from Mask to Var.
3203 2014-01-27  Jeff Law  <law@redhat.com>
3205         * ipa-inline.c (inline_small_functions): Fix typo.
3207 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3209         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3210         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3211         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3212         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3213         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3214         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3215         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3216         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3217         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3218         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3219         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3220         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3221         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3222         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3223         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3224         (_mm512_storeu_epi64): Ditto.
3225         (_mm512_cmpge_epi32_mask): Ditto.
3226         (_mm512_cmpge_epu32_mask): Ditto.
3227         (_mm512_cmpge_epi64_mask): Ditto.
3228         (_mm512_cmpge_epu64_mask): Ditto.
3229         (_mm512_cmple_epi32_mask): Ditto.
3230         (_mm512_cmple_epu32_mask): Ditto.
3231         (_mm512_cmple_epi64_mask): Ditto.
3232         (_mm512_cmple_epu64_mask): Ditto.
3233         (_mm512_cmplt_epi32_mask): Ditto.
3234         (_mm512_cmplt_epu32_mask): Ditto.
3235         (_mm512_cmplt_epi64_mask): Ditto.
3236         (_mm512_cmplt_epu64_mask): Ditto.
3237         (_mm512_cmpneq_epi32_mask): Ditto.
3238         (_mm512_cmpneq_epu32_mask): Ditto.
3239         (_mm512_cmpneq_epi64_mask): Ditto.
3240         (_mm512_cmpneq_epu64_mask): Ditto.
3241         (_mm512_expand_pd): Ditto.
3242         (_mm512_expand_ps): Ditto.
3243         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3244         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3245         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3246         * config/i386/i386.c (ix86_builtins): Add
3247         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3248         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3249         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3250         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3251         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3252         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3253         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3254         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3255         IX86_BUILTIN_PMOVUSQW512_MEM.
3256         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3257         __builtin_ia32_pmovsqd512mem_mask,
3258         __builtin_ia32_pmovqd512mem_mask,
3259         __builtin_ia32_pmovusqw512mem_mask,
3260         __builtin_ia32_pmovsqw512mem_mask,
3261         __builtin_ia32_pmovqw512mem_mask,
3262         __builtin_ia32_pmovusdw512mem_mask,
3263         __builtin_ia32_pmovsdw512mem_mask,
3264         __builtin_ia32_pmovdw512mem_mask,
3265         __builtin_ia32_pmovqb512mem_mask,
3266         __builtin_ia32_pmovusqb512mem_mask,
3267         __builtin_ia32_pmovsqb512mem_mask,
3268         __builtin_ia32_pmovusdb512mem_mask,
3269         __builtin_ia32_pmovsdb512mem_mask,
3270         __builtin_ia32_pmovdb512mem_mask.
3271         (bdesc_args): Add __builtin_ia32_expanddf512,
3272         __builtin_ia32_expandsf512.
3273         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3274         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3275         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3276         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3277         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3278         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3279         (avx512f_<code>v8div16qi2_mask_store): This.
3280         (avx512f_expand<mode>): New.
3282 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3284         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3285         New.
3286         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3287         (_mm512_prefetch_i32scatter_pd): Ditto.
3288         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3289         (_mm512_prefetch_i64scatter_pd): Ditto.
3290         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3291         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3292         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3293         (_mm512_prefetch_i32scatter_ps): Ditto.
3294         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3295         (_mm512_prefetch_i64scatter_ps): Ditto.
3296         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3297         * config/i386/i386-builtin-types.def: Define
3298         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3299         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3300         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3301         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3302         IX86_BUILTIN_SCATTERPFQPD.
3303         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3304         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3305         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3306         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3307         __builtin_ia32_scatterpfqps.
3308         (ix86_expand_builtin): Expand new built-ins.
3309         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3310         fix memory access data type.
3311         (*avx512pf_gatherpf<mode>_mask): Ditto.
3312         (*avx512pf_gatherpf<mode>): Ditto.
3313         (avx512pf_scatterpf<mode>): Ditto.
3314         (*avx512pf_scatterpf<mode>_mask): Ditto.
3315         (*avx512pf_scatterpf<mode>): Ditto.
3316         (GATHER_SCATTER_SF_MEM_MODE): New.
3317         (avx512pf_gatherpf<mode>df): Ditto.
3318         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3319         (*avx512pf_scatterpf<mode>df): Ditto.
3321 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3323         PR bootstrap/59934
3324         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3325         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3326         reached.
3328 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3330         * common/config/arm/arm-common.c
3331         (arm_rewrite_mcpu): Handle multiple names.
3332         * config/arm/arm.h
3333         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3335 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3337         * gimple-builder.h (create_gimple_tmp): Delete.
3339 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3341         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3342         words comparisons.
3344 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3346         * config/pa/pa.md (call): Generate indirect long calls to non-local
3347         functions when outputing 32-bit code.
3348         (call_value): Likewise except for special call to buggy powf function.
3350         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3351         portable runtime and PIC indirect calls.
3352         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3353         and PIC call sequences.  Use ldo instead of blr to set return register
3354         in PIC call sequence.
3356 2014-01-25  Walter Lee  <walt@tilera.com>
3358         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3359         avoid clobbering a live register.
3361 2014-01-25  Walter Lee  <walt@tilera.com>
3363         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3364         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3365         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3366         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3368 2014-01-25  Walter Lee  <walt@tilera.com>
3370         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3371         arguments on even registers.
3372         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3373         STACK_BOUNDARY.
3374         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3375         (BIGGEST_ALIGNMENT): Ditto.
3376         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3378 2014-01-25  Walter Lee  <walt@tilera.com>
3380         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3381         insns before bundling.
3382         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3384 2014-01-25  Walter Lee  <walt@tilera.com>
3386         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3387         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3388         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3390 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3392         * config/mips/constraints.md (kl): Delete.
3393         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3394         define expands, using...
3395         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3396         instructions for MIPS16.
3397         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3398         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3400 2014-01-25  Walter Lee  <walt@tilera.com>
3402         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3403         (clzdi2): Ditto.
3404         (ffsdi2): Ditto.
3406 2014-01-25  Walter Lee  <walt@tilera.com>
3408         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3409         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3411 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3413         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3414         Handle XOR.
3416 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3418         * print-rtl.c (in_call_function_usage): New var.
3419         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3420         EXPR_LIST mode as mode and not as reg note name.
3422         PR middle-end/59561
3423         * cfgloopmanip.c (copy_loop_info): If
3424         loop->warned_aggressive_loop_optimizations, make sure
3425         the flag is set in target loop too.
3427 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3429         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3430         flag_cilkplus.
3431         * builtins.def: Likewise.
3432         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3433         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3434         * ira.c (ira_setup_eliminable_regset): Likewise.
3435         * omp-low.c (gate_expand_omp): Likewise.
3436         (execute_lower_omp): Likewise.
3437         (diagnose_sb_0): Likewise.
3438         (gate_diagnose_omp_blocks): Likewise.
3439         (simd_clone_clauses_extract): Likewise.
3440         (gate): Likewise.
3442 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3444         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3445         correction for little endian...
3446         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3447         here.
3449 2014-01-24  Jeff Law  <law@redhat.com>
3451         PR tree-optimization/59919
3452         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3453         for non-returning calls.
3455 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3457         * common/config/aarch64/aarch64-common.c
3458         (aarch64_rewrite_mcpu): Handle multiple names.
3459         * config/aarch64/aarch64.h
3460         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3462 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3464         * input.c (add_file_to_cache_tab): Handle the case where fopen
3465         returns NULL.
3467 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3469         PR target/59929
3470         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3471         from push operand if code of push isn't PRE_DEC.
3473 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3475         PR target/59909
3476         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3477         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3478         it is only used for non-atomic loads/stores.
3480         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3481         -mquad-memory or -mquad-memory-atomic switches.
3483         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3484         -mquad-memory-atomic to ISA 2.07 support.
3486         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3487         to separate support of normal quad word memory operations (ldq, stq)
3488         from the atomic quad word memory operations.
3490         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3491         support to separate non-atomic quad word operations from atomic
3492         quad word operations.  Disable non-atomic quad word operations in
3493         little endian mode so that we don't have to swap words after the
3494         load and before the store.
3495         (quad_load_store_p): Add comment about atomic quad word support.
3496         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3497         options printed with -mdebug=reg.
3499         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3500         -mquad-memory-atomic as the test for whether we have quad word
3501         atomic instructions.
3502         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3503         or -mp8-vector are used, allow byte/half-word atomic operations.
3505         * config/rs6000/sync.md (load_lockedti): Insure that the address
3506         is a proper indexed or indirect address for the lqarx instruction.
3507         On little endian systems, swap the hi/lo registers after the lqarx
3508         instruction.
3509         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3510         insure the address is valid for the lqarx instruction.
3511         (store_conditionalti): Insure that the address is a proper indexed
3512         or indirect address for the stqcrx. instruction.  On little endian
3513         systems, swap the hi/lo registers before doing the stqcrx.
3514         instruction.
3515         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3516         insure the address is valid for the stqcrx. instruction.
3518         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3519         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3520         type of quad memory support is available.
3522 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3524         PR regression/59915
3525         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3526         there is a danger of looping.
3528 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3530         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3531         force flag_ira_loop_pressure if set via command line.
3533 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3535         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3536         (ashr_simd): New builtin handling DI mode.
3537         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3538         (aarch64_sshr_simddi): New match pattern.
3539         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3540         (vshrd_n_s64): Likewise.
3541         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3543 2014-01-23  Nick Clifton  <nickc@redhat.com>
3545         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3546         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3547         favour of mcu specific scripts.
3548         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3549         430x multilibs.
3551 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3552             Alex Velenko  <Alex.Velenko@arm.com>
3554         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3555         (vaddv_s16): Likewise.
3556         (vaddv_s32): Likewise.
3557         (vaddv_u8): Likewise.
3558         (vaddv_u16): Likewise.
3559         (vaddv_u32): Likewise.
3560         (vaddvq_s8): Likewise.
3561         (vaddvq_s16): Likewise.
3562         (vaddvq_s32): Likewise.
3563         (vaddvq_s64): Likewise.
3564         (vaddvq_u8): Likewise.
3565         (vaddvq_u16): Likewise.
3566         (vaddvq_u32): Likewise.
3567         (vaddvq_u64): Likewise.
3568         (vaddv_f32): Likewise.
3569         (vaddvq_f32): Likewise.
3570         (vaddvq_f64): Likewise.
3571         (vmaxv_f32): Likewise.
3572         (vmaxv_s8): Likewise.
3573         (vmaxv_s16): Likewise.
3574         (vmaxv_s32): Likewise.
3575         (vmaxv_u8): Likewise.
3576         (vmaxv_u16): Likewise.
3577         (vmaxv_u32): Likewise.
3578         (vmaxvq_f32): Likewise.
3579         (vmaxvq_f64): Likewise.
3580         (vmaxvq_s8): Likewise.
3581         (vmaxvq_s16): Likewise.
3582         (vmaxvq_s32): Likewise.
3583         (vmaxvq_u8): Likewise.
3584         (vmaxvq_u16): Likewise.
3585         (vmaxvq_u32): Likewise.
3586         (vmaxnmv_f32): Likewise.
3587         (vmaxnmvq_f32): Likewise.
3588         (vmaxnmvq_f64): Likewise.
3589         (vminv_f32): Likewise.
3590         (vminv_s8): Likewise.
3591         (vminv_s16): Likewise.
3592         (vminv_s32): Likewise.
3593         (vminv_u8): Likewise.
3594         (vminv_u16): Likewise.
3595         (vminv_u32): Likewise.
3596         (vminvq_f32): Likewise.
3597         (vminvq_f64): Likewise.
3598         (vminvq_s8): Likewise.
3599         (vminvq_s16): Likewise.
3600         (vminvq_s32): Likewise.
3601         (vminvq_u8): Likewise.
3602         (vminvq_u16): Likewise.
3603         (vminvq_u32): Likewise.
3604         (vminnmv_f32): Likewise.
3605         (vminnmvq_f32): Likewise.
3606         (vminnmvq_f64): Likewise.
3608 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3610         * config/aarch64/aarch64-simd.md
3611         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3612         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3613         (*aarch64_mul3_elt<mode>): Likewise.
3614         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3615         (*aarch64_mul3_elt_to_64v2df): Likewise.
3616         (*aarch64_mla_elt<mode>): Likewise.
3617         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3618         (*aarch64_mls_elt<mode>): Likewise.
3619         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3620         (*aarch64_fma4_elt<mode>): Likewise.
3621         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3622         (*aarch64_fma4_elt_to_64v2df): Likewise.
3623         (*aarch64_fnma4_elt<mode>): Likewise.
3624         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3625         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3626         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3627         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3628         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3629         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3630         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3631         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3632         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3634 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3636         * config/aarch64/aarch64-simd.md
3637         (aarch64_be_checked_get_lane<mode>): New define_expand.
3638         * config/aarch64/aarch64-simd-builtins.def
3639         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3640         New builtin definition.
3641         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3642         Use new safe be builtin.
3644 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3646         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3647         New define_insn.
3648         (aarch64_be_st1<mode>): Likewise.
3649         (aarch_ld1<VALL:mode>): Define_expand modified.
3650         (aarch_st1<VALL:mode>): Likewise.
3651         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3652         (UNSPEC_ST1): Likewise.
3654 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3656         * config/microblaze/microblaze.md: Add trap insn and attribute
3658 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3660         PR preprocessor/58580
3661         * input.h (location_get_source_line): Take an additional line_size
3662         parameter.
3663         (void diagnostics_file_cache_fini): Declare new function.
3664         * input.c (struct fcache): New type.
3665         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3666         New static constants.
3667         (diagnostic_file_cache_init, total_lines_num)
3668         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3669         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3670         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3671         (get_next_line, read_next_line, goto_next_line, read_line_num):
3672         New static function definitions.
3673         (diagnostic_file_cache_fini): New function.
3674         (location_get_source_line): Take an additional output line_len
3675         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3676         read_line_num.
3677         * diagnostic.c (diagnostic_finish): Call
3678         diagnostic_file_cache_fini.
3679         (adjust_line): Take an additional input parameter for the length
3680         of the line, rather than calculating it with strlen.
3681         (diagnostic_show_locus): Adjust the use of
3682         location_get_source_line and adjust_line with respect to their new
3683         signature.  While displaying a line now, do not stop at the first
3684         null byte.  Rather, display the zero byte as a space and keep
3685         going until we reach the size of the line.
3686         * Makefile.in: Add vec.o to OBJS-libcommon
3688 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3689             Ilya Tocar     <ilya.tocar@intel.com>
3691         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3692         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3693         (__builtin_ia32_kmov16): Ditto.
3694         * config/i386/i386.md (UNSPEC_KMOV): New.
3695         (kmovw): Ditto.
3697 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3699         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3700         (_mm512_storeu_si512): Ditto.
3702 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3704         PR target/52125
3705         * rtl.h (get_referenced_operands): Declare.
3706         * recog.c (get_referenced_operands): New function.
3707         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3708         operands have been referenced when recording LO_SUM references.
3710 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3712         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3714 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3716         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3717         Enable for generic and recent AMD targets.
3719 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3721         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3722         ARG_SIZE note when adjustment was eliminated.
3724 2014-01-22  Jeff Law  <law@redhat.com>
3726         PR tree-optimization/59597
3727         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3728         in file.  Accept new argument REGISTERING and use it to modify
3729         dump output appropriately.
3730         (register_jump_thread): Corresponding changes.
3731         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3732         thread paths involving joiner blocks.  Add code to dump cancelled
3733         jump threading paths.
3735 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3737         PR rtl-optimization/59477
3738         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3739         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3741 2014-01-22  Tom Tromey  <tromey@redhat.com>
3743         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3744         PARAMS.
3745         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3747 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3749         PR rtl-optimization/59896
3750         * lra-constraints.c (process_alt_operands): Check unused note for
3751         matched operands of insn with no output reloads.
3753 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3755         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3756         (mips_move_from_gpr_cost): Likewise.
3758 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3760         PR rtl-optimization/59858
3761         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3762         ira_class_hard_regs_num.
3763         (process_alt_operands): Increase reject for dying matched operand.
3765 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3767         PR target/59003
3768         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3769         smaller than size, perform several stores or loads and stores
3770         at dst + count - size to store or copy all of size bytes, rather
3771         than just last modesize bytes.
3773 2014-01-20  DJ Delorie  <dj@redhat.com>
3775         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3776         that CLOBBERs are REGs before propogating their values.
3778 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3780         PR middle-end/59789
3781         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3782         (cgraph_inline_failed_type): New function.
3783         * cgraph.h (DEFCIFCODE): Add type.
3784         (cgraph_inline_failed_type_t): New enum.
3785         (cgraph_inline_failed_type): New prototype.
3786         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3787         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3788         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3789         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3790         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3791         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3792         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3793         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3794         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3795         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3796         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3797         OPTIMIZATION_MISMATCH.
3798         * tree-inline.c (expand_call_inline): Emit errors during
3799         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3801 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3803         PR target/59685
3804         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3805         mode attribute in insn output.
3807 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3809         * output.h (output_constant): Delete.
3810         * varasm.c (output_constant): Make private.
3812 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3814         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3816 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3818         PR middle-end/59860
3819         * tree.h (fold_builtin_strcat): New prototype.
3820         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3821         argument, if non-NULL, don't call c_strlen.  Optimize
3822         directly into __builtin_memcpy instead of __builtin_strcpy.
3823         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3824         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3826 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3828         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3829         for SImode_address_operand operands, having only a REG argument.
3831 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3833         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3834         loader name using mbig-endian.
3835         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3837 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3839         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3840         (-mtune): Likewise.
3841         (-mcpu): Likewise.
3843 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3845         * config/aarch64/aarch64-protos.h
3846         (aarch64_cannot_change_mode_class_ptr): Declare.
3847         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3848         aarch64_cannot_change_mode_class_ptr): New.
3849         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3850         backend hook aarch64_cannot_change_mode_class.
3852 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3854         * common/config/aarch64/aarch64-common.c
3855         (aarch64_handle_option): Don't handle any option order logic here.
3856         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3857         selected_cpu, warn on architecture version mismatch.
3858         (aarch64_override_options): Fix parsing order for option strings.
3860 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3861             Iain Sandoe  <iain@codesourcery.com>
3863         PR bootstrap/59496
3864         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3865         warning.  Amend comment to reflect current functionality.
3867 2014-01-20  Richard Biener  <rguenther@suse.de>
3869         PR middle-end/59860
3870         * builtins.c (fold_builtin_strcat): Remove case better handled
3871         by tree-ssa-strlen.c.
3873 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
3875         * config/aarch64/aarch64.opt
3876         (mcpu, march, mtune): Make case-insensitive.
3878 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3880         PR target/59880
3881         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3882         if operands[1] is a REG or ZERO_EXTEND of a REG.
3884 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
3886         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3888 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
3890         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3891         long non-pic millicode calls.
3893 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3895         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3897 2014-01-19  Kito Cheng  <kito@0xlab.org>
3899         * builtins.c (expand_movstr): Check movstr expand done or fail.
3901 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3902             H.J. Lu  <hongjiu.lu@intel.com>
3904         PR target/59379
3905         * config/i386/i386.md (*lea<mode>): Zero-extend return register
3906         to DImode for zero-extended addresses.
3908 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
3910         PR rtl-optimization/57763
3911         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3912         on the new indirect jump_insn and increment LABEL_NUSES (label).
3914 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
3916         PR bootstrap/59580
3917         PR bootstrap/59583
3918         * config.gcc (x86_archs): New variable.
3919         (x86_64_archs): Likewise.
3920         (x86_cpus): Likewise.
3921         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3922         --with-arch/--with-cpu= options.
3923         Support --with-arch=/--with-cpu={nehalem,westmere,
3924         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3926 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3928         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3929         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
3931 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3933         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3935 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
3937         PR target/58944
3938         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3939         clear cpp_get_options (parse_in)->warn_unused_macros for
3940         ix86_target_macros_internal with cpp_define.
3942 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
3944         * jump.c (delete_related_insns): Keep (use (insn))s.
3945         * reorg.c (redundant_insn): Check for barriers too.
3947 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3949         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3951 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
3953         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3954         call to $$dyncall when TARGET_LONG_CALLS is true.
3956 2014-01-17  Jeff Law  <law@redhat.com>
3958         * ree.c (combine_set_extension): Temporarily disable test for
3959         changing number of hard registers.
3961 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3963         PR middle-end/58125
3964         * ipa-inline-analysis.c (inline_free_summary):
3965         Do not free summary of aliases.
3967 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3969         PR middle-end/59706
3970         * gimplify.c (gimplify_expr): Use create_tmp_var
3971         instead of create_tmp_var_raw.  If cond doesn't have
3972         integral type, don't add the IFN_ANNOTATE builtin at all.
3974 2014-01-17  Martin Jambor  <mjambor@suse.cz>
3976         PR ipa/59736
3977         * ipa-cp.c (prev_edge_clone): New variable.
3978         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3979         Also resize prev_edge_clone vector.
3980         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3981         (ipcp_edge_removal_hook): New function.
3982         (ipcp_driver): Register ipcp_edge_removal_hook.
3984 2014-01-17  Andrew Pinski <apinski@cavium.com>
3985             Steve Ellcey  <sellcey@mips.com>
3987         PR target/59462
3988         * config/mips/mips.c (mips_print_operand): Check operand mode instead
3989         of operator mode.
3991 2014-01-17  Jeff Law  <law@redhat.com>
3993         PR middle-end/57904
3994         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3995         so that pass_ccp runs first.
3997 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3999         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4000         (ix86_adjust_cost): Use !TARGET_XXX.
4001         (do_reorder_for_imul): Likewise.
4002         (swap_top_of_ready_list): Likewise.
4003         (ix86_sched_reorder): Likewise.
4005 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4007         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4008         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4009         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4010         (intel_memset): New.  Duplicate slm_memset.
4011         (intel_cost): New.  Duplicate slm_cost.
4012         (m_INTEL): New macro.
4013         (processor_target_table): Add "intel".
4014         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4015         with PROCESSOR_INTEL for "intel".
4016         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4017         PROCESSOR_SILVERMONT.
4018         (ix86_issue_rate): Likewise.
4019         (ix86_adjust_cost): Likewise.
4020         (ia32_multipass_dfa_lookahead): Likewise.
4021         (swap_top_of_ready_list): Likewise.
4022         (ix86_sched_reorder): Likewise.
4023         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4024         instead of TARGET_OPT_AGU.
4025         * config/i386/i386.h (TARGET_INTEL): New.
4026         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4027         (processor_type): Add PROCESSOR_INTEL.
4028         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4029         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4031 2014-01-17  Marek Polacek  <polacek@redhat.com>
4033         PR c/58346
4034         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4035         size is zero.
4037 2014-01-17  Richard Biener  <rguenther@suse.de>
4039         PR tree-optimization/46590
4040         * opts.c (default_options_table): Add entries for
4041         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4042         all enabled at -O1 but not for -Og.
4043         * common.opt (fbranch-count-reg): Remove Init(1).
4044         (fmove-loop-invariants): Likewise.
4045         (ftree-pta): Likewise.
4047 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4049         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4050         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4051         decls to at least the GCC 4.8 used alignments.
4053         PR fortran/59440
4054         * tree-nested.c (convert_nonlocal_reference_stmt,
4055         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4056         of GIMPLE_BIND stmts, adjust associated decls.
4058 2014-01-17  Richard Biener  <rguenther@suse.de>
4060         PR tree-optimization/46590
4061         * vec.h (vec<>::bseach): New member function implementing
4062         binary search according to C89 bsearch.
4063         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4064         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4065         bitmap pointer again.  Make accesses_in_loop a flat array.
4066         (mem_ref_obstack): New global.
4067         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4068         (mark_ref_stored): Likewise.
4069         (ref_indep_loop_p_2): Likewise.
4070         (set_ref_stored_in_loop): New helper function.
4071         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4072         (memref_free): Adjust.
4073         (record_mem_ref_loc): Simplify.
4074         (gather_mem_refs_stmt): Adjust.
4075         (sort_locs_in_loop_postorder_cmp): New function.
4076         (analyze_memory_references): Sort accesses_in_loop after
4077         loop postorder number.
4078         (find_ref_loc_in_loop_cmp): New function.
4079         (for_all_locs_in_loop): Find relevant cluster of locs in
4080         accesses_in_loop and iterate without recursion.
4081         (execute_sm): Avoid uninit warning.
4082         (struct ref_always_accessed): Simplify.
4083         (ref_always_accessed::operator ()): Likewise.
4084         (ref_always_accessed_p): Likewise.
4085         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4086         loop postorder numbers here.
4087         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4088         numbers.
4090 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4092         PR c++/57945
4093         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4094         on decls for which assemble_alias has been called.
4096 2014-01-17  Nick Clifton  <nickc@redhat.com>
4098         * config/msp430/msp430.opt: (mcpu): New option.
4099         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4100         (msp430_option_override): Parse target_cpu.  If the MCU name
4101         matches a generic string, clear target_mcu.
4102         (msp430_attr): Allow numeric interrupt values up to 63.
4103         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4104         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4105         option.
4106         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4107         Add mcpu matches.
4108         * config/msp430/msp430.md (popm): Use %J rather than %I.
4109         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4110         (addhi_cy_i): Use immediate_operand for operand 2.
4111         * doc/invoke.texi: Document -mcpu option.
4113 2014-01-17  Richard Biener  <rguenther@suse.de>
4115         PR rtl-optimization/38518
4116         * df.h (df_analyze_loop): Declare.
4117         * df-core.c: Include cfgloop.h.
4118         (df_analyze_1): Split out main part of df_analyze.
4119         (df_analyze): Adjust.
4120         (loop_inverted_post_order_compute): New function.
4121         (loop_post_order_compute): Likewise.
4122         (df_analyze_loop): New function avoiding whole-function
4123         postorder computes.
4124         * loop-invariant.c (find_defs): Use df_analyze_loop.
4125         (find_invariants): Adjust.
4126         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4128 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4130         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4131         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4133 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4135         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4136         traversal when removing references.
4138 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4140         PR ipa/59775
4141         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4143 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4145         PR middle-end/56791
4146         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4147         pushing a reload for an autoinc when we had previously reloaded an
4148         inner part of the address.
4150 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4152         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4153         field.
4154         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4155         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4156         when not giving up or versioning for alias only because of
4157         loop->safelen.
4158         (vect_analyze_data_ref_dependences): Set to true.
4159         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4160         is a GIMPLE_PHI.
4161         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4162         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4163         to the condition.
4165         PR middle-end/58344
4166         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4168         PR target/59839
4169         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4170         operand 0 predicate for gathers, use a new pseudo as subtarget.
4172 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4174         PR middle-end/59609
4175         * lra-constraints.c (process_alt_operands): Add printing debug info.
4176         Check absence of input/output reloads for matched operands too.
4178 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4180         PR rtl-optimization/59835
4181         * ira.c (ira_init_register_move_cost): Increase cost for
4182         impossible modes.
4184 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4186         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4188 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4190         PR target/59780
4191         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4192         non-register objects.  Use gen_(high/low)part more consistently.
4193         Fix assertions.
4195 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4197         PR target/59844
4198         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4199         endian support, remove tests for WORDS_BIG_ENDIAN.
4200         (p8_mfvsrd_3_<mode>): Likewise.
4201         (reload_gpr_from_vsx<mode>): Likewise.
4202         (reload_gpr_from_vsxsf): Likewise.
4203         (p8_mfvsrd_4_disf): Likewise.
4205 2014-01-16  Richard Biener  <rguenther@suse.de>
4207         PR rtl-optimization/46590
4208         * lcm.c (compute_antinout_edge): Use postorder iteration.
4209         (compute_laterin): Use inverted postorder iteration.
4211 2014-01-16  Nick Clifton  <nickc@redhat.com>
4213         PR middle-end/28865
4214         * varasm.c (output_constant): Return the number of bytes actually
4215         emitted.
4216         (output_constructor_array_range): Update the field size with the
4217         number of bytes emitted by output_constant.
4218         (output_constructor_regular_field): Likewise.  Also do not
4219         complain if the total number of bytes emitted is now greater
4220         than the expected fieldpos.
4221         * output.h (output_constant): Update prototype and descriptive comment.
4223 2014-01-16  Marek Polacek  <polacek@redhat.com>
4225         PR middle-end/59827
4226         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4227         it is error_mark_node.
4229 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4231         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4232         VALID_AVX256_REG_OR_OI_MODE.
4234 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4236         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4237         current procedure should be profiled.
4239 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4241         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4242         of moving from/to the STACK_REG register class.
4244 2014-01-15  Richard Henderson  <rth@redhat.com>
4246         PR debug/54694
4247         * reginfo.c (global_regs_decl): Globalize.
4248         * rtl.h (global_regs_decl): Declare.
4249         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4250         reserved via global_regs.
4252 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4254         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4256 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4258         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4259         and vmulosh rather than call gen_vec_widen_smult_*.
4260         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4261         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4262         (vec_widen_smult_even_v16qi): Likewise.
4263         (vec_widen_umult_even_v8hi): Likewise.
4264         (vec_widen_smult_even_v8hi): Likewise.
4265         (vec_widen_umult_odd_v16qi): Likewise.
4266         (vec_widen_smult_odd_v16qi): Likewise.
4267         (vec_widen_umult_odd_v8hi): Likewise.
4268         (vec_widen_smult_odd_v8hi): Likewise.
4269         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4270         vmuloub rather than call gen_vec_widen_umult_*.
4271         (vec_widen_umult_lo_v16qi): Likewise.
4272         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4273         vmulosb rather than call gen_vec_widen_smult_*.
4274         (vec_widen_smult_lo_v16qi): Likewise.
4275         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4276         rather than call gen_vec_widen_umult_*.
4277         (vec_widen_umult_lo_v8hi): Likewise.
4278         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4279         rather than call gen_vec_widen_smult_*.
4280         (vec_widen_smult_lo_v8hi): Likewise.
4282 2014-01-15  Jeff Law  <law@redhat.com>
4284         PR tree-optimization/59747
4285         * ree.c (find_and_remove_re): Properly handle case where a second
4286         eliminated extension requires widening a copy created for elimination
4287         of a prior extension.
4288         (combine_set_extension): Ensure that the number of hard regs needed
4289         for a destination register does not change when we widen it.
4291 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4293         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4294         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4295         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4296         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4297         (avr-*-rtems*): Likewise.
4298         (bfin*-rtems*): Likewise.
4299         (moxie-*-rtems*): Likewise.
4300         (h8300-*-rtems*): Likewise.
4301         (i[34567]86-*-rtems*): Likewise.
4302         (lm32-*-rtems*): Likewise.
4303         (m32r-*-rtems*): Likewise.
4304         (m68k-*-rtems*): Likewise.
4305         (microblaze*-*-rtems*): Likewise.
4306         (mips*-*-rtems*): Likewise.
4307         (powerpc-*-rtems*): Likewise.
4308         (sh-*-rtems*): Likewise.
4309         (sparc-*-rtems*): Likewise.
4310         (sparc64-*-rtems*): Likewise.
4311         (v850-*-rtems*): Likewise.
4312         (m32c-*-rtems*): Likewise.
4314 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4316         PR rtl-optimization/59511
4317         * ira.c (ira_init_register_move_cost): Use memory costs for some
4318         cases of register move cost calculations.
4319         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4320         instead of BB frequency.
4321         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4322         * lra-assigns.c (find_hard_regno_for): Ditto.
4324 2014-01-15  Richard Biener  <rguenther@suse.de>
4326         PR tree-optimization/59822
4327         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4328         (vectorizable_load): Use it to hoist defs of uses of invariant
4329         loads out of the loop.
4331 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4332             Kugan Vivekanandarajah  <kuganv@linaro.org>
4334         PR target/59695
4335         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4336         truncation.
4338 2014-01-15  Richard Biener  <rguenther@suse.de>
4340         PR rtl-optimization/59802
4341         * lcm.c (compute_available): Use inverted postorder to seed
4342         the initial worklist.
4344 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4346         PR target/59803
4347         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4348         ADDR_REGS for invalid symrefs in non-PIC code.
4350 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4352         PR other/58712
4353         * builtins.c (determine_block_size): Initialize *probable_max_size
4354         even if len_rtx is CONST_INT.
4356 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4358         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4359         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4360         (cortexa53_tunings): Likewise.
4361         (aarch64_sched_issue_rate): New function.
4362         (TARGET_SCHED_ISSUE_RATE): Define.
4364 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4366         * ira-costs.c (find_costs_and_classes): Add missed
4367         ira_init_register_move_cost_if_necessary.
4369 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4371         PR target/59787
4372         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4374 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4376         PR target/59794
4377         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4378         to indicate if type is used for function return value.  Warn ABI
4379         change if the vector mode isn't available for function return value.
4380         (ix86_function_arg_advance): Pass false to type_natural_mode.
4381         (ix86_function_arg): Likewise.
4382         (ix86_gimplify_va_arg): Likewise.
4383         (function_arg_32): Don't warn ABI change.
4384         (ix86_function_value): Pass true to type_natural_mode.
4385         (ix86_return_in_memory): Likewise.
4386         (ix86_struct_value_rtx): Removed.
4387         (TARGET_STRUCT_VALUE_RTX): Likewise.
4389 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4391         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4392         converting a conditional jump into a conditional return.
4394 2014-01-14  Richard Biener  <rguenther@suse.de>
4396         PR tree-optimization/58921
4397         PR tree-optimization/59006
4398         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4399         hoisting invariant stmts.
4400         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4401         invariant loads on the preheader edge if possible.
4403 2014-01-14  Joey Ye  <joey.ye@arm.com>
4405         * doc/plugin.texi (Building GCC plugins): Update to C++.
4407 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4409         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4410         (_mm_rcp28_round_ss): Ditto.
4411         (_mm_rsqrt28_round_sd): Ditto.
4412         (_mm_rsqrt28_round_ss): Ditto.
4413         (_mm_rcp28_sd): Ditto.
4414         (_mm_rcp28_ss): Ditto.
4415         (_mm_rsqrt28_sd): Ditto.
4416         (_mm_rsqrt28_ss): Ditto.
4417         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4418         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4419         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4420         (IX86_BUILTIN_RCP28SD): Ditto.
4421         (IX86_BUILTIN_RCP28SS): Ditto.
4422         (IX86_BUILTIN_RSQRT28SD): Ditto.
4423         (IX86_BUILTIN_RSQRT28SS): Ditto.
4424         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4425         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4426         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4427         (ix86_expand_special_args_builtin): Expand new FTYPE.
4428         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4429         (srcp14<mode>): Make insn unary.
4430         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4431         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4432         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4433         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4434         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4435         Fix rounding: make it SAE only.
4436         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4437         Ditto.
4438         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4439         Ditto.
4440         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4441         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4442         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4443         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4444         (round_saeonly_mask_scalar_operand4): Ditto.
4445         (round_saeonly_mask_scalar_op3): Ditto.
4446         (round_saeonly_mask_scalar_op4): Ditto.
4448 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4450         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4451         Implement -maltivec=be for vec_insert and vec_extract.
4453 2014-01-10  DJ Delorie  <dj@redhat.com>
4455         * config/msp430/msp430.md (call_internal): Don't allow memory
4456         references with SP as the base register.
4457         (call_value_internal): Likewise.
4458         * config/msp430/constraints.md (Yc): New.  For memory references
4459         that don't use SP as a base register.
4461         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4462         "an integer without a # prefix"
4463         * config/msp430/msp430.md (epilogue_helper): Use it.
4465 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4467         PR target/59617
4468         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4469         AVX512F gather builtins.
4470         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4471         on gather decls with INTEGER_TYPE masktype.
4472         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4473         directly into the builtin rather than hoisting it before loop.
4475         PR tree-optimization/59387
4476         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4477         (scev_const_prop): If folded_casts and type has undefined overflow,
4478         use force_gimple_operand instead of force_gimple_operand_gsi and
4479         for each added stmt if it is assign with
4480         arith_code_with_undefined_signed_overflow, call
4481         rewrite_to_defined_overflow.
4482         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4483         gimple-fold.h instead.
4484         (arith_code_with_undefined_signed_overflow,
4485         rewrite_to_defined_overflow): Moved to ...
4486         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4487         rewrite_to_defined_overflow): ... here.  No longer static.
4488         Include gimplify-me.h.
4489         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4490         rewrite_to_defined_overflow): New prototypes.
4492 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4494         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4496 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4498         * builtins.c (get_object_alignment_2): Minor tweak.
4499         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4501 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4503         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4504         optimized non constant lengths.
4506 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4508         PR libgomp/59194
4509         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4510         load as __atomic_load_N if possible.
4512 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4514         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4515         target parameter.
4516         (rs6000_expand_builtin): Adjust call.
4518 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4520         PR target/58115
4521         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4522         * config/rs6000/rs6000.c: Include target-globals.h.
4523         (rs6000_set_current_function): Instead of doing target_reinit
4524         unconditionally, use save_target_globals_default_opts and
4525         restore_target_globals.
4527         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4528         FPSCR.
4529         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4530         (rs6000_expand_builtin): Handle mffs and mtfsf.
4531         (rs6000_init_builtins): Define mffs and mtfsf.
4532         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4533         (rs6000_mffs): New pattern.
4534         (rs6000_mtfsf): New pattern.
4536 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4538         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4539         Start narrowing with START.  Apply candidate-use pair
4540         and check overall cost in narrowing.
4541         (iv_ca_prune): Pass new argument.
4543 2014-01-10  Jeff Law  <law@redhat.com>
4545         PR middle-end/59743
4546         * ree.c (combine_reaching_defs): Ensure the defining statement
4547         occurs before the extension when optimizing extensions with
4548         different source and destination hard registers.
4550 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4552         PR ipa/58585
4553         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4554         vtables into the type inheritance graph.
4556 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4558         PR rtl-optimization/59754
4559         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4560         modes in the REGNO != REGNO case.
4562 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4564         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4566 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4568         PR tree-optimization/59745
4569         * tree-predcom.c (tree_predictive_commoning_loop): Call
4570         free_affine_expand_cache if giving up because components is NULL.
4572         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4573         GC in payload of target_globals struct instead of allocating them on
4574         the heap and the larger structs separately using GC.
4575         * target-globals.h (struct target_globals): Make regs, hard_regs,
4576         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4577         of GTY((skip)) and change type to void *.
4578         (reset_target_globals): Cast loads from those fields to corresponding
4579         types.
4581 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4583         PR plugins/59335
4584         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4585         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4586         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4588 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4590         PR target/59744
4591         * aarch64-modes.def (CC_Zmode): New flags mode.
4592         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4593         represents an equality.
4594         (aarch64_get_condition_code): Handle CC_Zmode.
4595         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4597 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4599         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4600         extraction in good case.
4602 2014-01-10  Richard Biener  <rguenther@suse.de>
4604         PR tree-optimization/59374
4605         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4606         checking after SLP discovery.  Mark stmts not participating
4607         in any SLP instance properly.
4609 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4611         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4612         when handling a SET rtx.
4614 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4616         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4617         (cortex-a57): Likewise.
4618         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4620 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4622         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4623         non-iwmmxt builtins.
4625 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4627         PR ipa/58252
4628         PR ipa/59226
4629         * ipa-devirt.c record_target_from_binfo): Take as argument
4630         stack of binfos and lookup matching one for virtual inheritance.
4631         (possible_polymorphic_call_targets_1): Update.
4633 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4635         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4636         kernel strings for Loongson-2E/2F/3A.
4638 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4640         PR middle-end/59670
4641         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4642         is_gimple_call before calling gimple_call_internal_p.
4644 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4646         * Makefile.in (TREE_FLOW_H): Remove.
4647         (TREE_SSA_H): Add file names from tree-flow.h.
4648         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4649         * tree.h: Remove tree-flow.h reference.
4650         * hash-table.h: Remove tree-flow.h reference.
4651         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4652         reference with tree-ssa-loop.h.
4654 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4656         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4657         default element-order behavior for -maltivec.
4658         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4659         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4660         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4661         when targeting big endian, at least for now.
4662         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4664 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4666         PR middle-end/47735
4667         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4668         var satisfies use_register_for_decl, just take into account type
4669         alignment, rather than decl alignment.
4671         PR tree-optimization/59622
4672         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4673         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4674         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4675         Don't devirtualize for inplace at all.  For targets.length () == 1,
4676         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4678 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4680         * config/i386/i386.md (cpu): Remove the unused btver1.
4682 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4684         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4686 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4688         PR target/58115
4689         * tree-core.h (struct target_globals): New forward declaration.
4690         (struct tree_target_option): Add globals field.
4691         * tree.h (TREE_TARGET_GLOBALS): Define.
4692         (prepare_target_option_nodes_for_pch): New prototype.
4693         * target-globals.h (struct target_globals): Define even if
4694         !SWITCHABLE_TARGET.
4695         * tree.c (prepare_target_option_node_for_pch,
4696         prepare_target_option_nodes_for_pch): New functions.
4697         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4698         * config/i386/i386.c: Include target-globals.h.
4699         (ix86_set_current_function): Instead of doing target_reinit
4700         unconditionally, use save_target_globals_default_opts and
4701         restore_target_globals.
4703 2014-01-09  Richard Biener  <rguenther@suse.de>
4705         PR tree-optimization/59715
4706         * tree-cfg.h (split_critical_edges): Declare.
4707         * tree-cfg.c (split_critical_edges): Export.
4708         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4710 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4712         * cfgexpand.c (expand_stack_vars): Optionally disable
4713         asan stack protection.
4714         (expand_used_vars): Likewise.
4715         (partition_stack_vars): Likewise.
4716         * asan.c (asan_emit_stack_protection): Optionally disable
4717         after return stack usage.
4718         (instrument_derefs): Optionally disable memory access instrumentation.
4719         (instrument_builtin_call): Likewise.
4720         (instrument_strlen_call): Likewise.
4721         (asan_protect_global): Optionally disable global variables protection.
4722         * doc/invoke.texi: Added doc for new options.
4723         * params.def: Added new options.
4724         * params.h: Likewise.
4726 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4728         PR rtl-optimization/59724
4729         * ifcvt.c (cond_exec_process_if_block): Don't call
4730         flow_find_head_matching_sequence with 0 longest_match.
4731         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4732         non-active insns if !stop_after.
4733         (try_head_merge_bb): Revert 2014-01-07 changes.
4735 2014-01-08  Jeff Law  <law@redhat.com>
4737         * ree.c (get_sub_rtx): New function, extracted from...
4738         (merge_def_and_ext): Here.
4739         (combine_reaching_defs): Use get_sub_rtx.
4741 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4743         * cgraph.h (varpool_variable_node): Do not choke on null node.
4745 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4747         * config/mips/mips.md (simple_return): Attempt to use JRC
4748         for microMIPS.
4749         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4751 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4753         PR rtl-optimization/59137
4754         * reorg.c (steal_delay_list_from_target): Call update_block for
4755         elided insns.
4756         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4758 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4760         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4761         two duplicate entries.
4763 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4765         Revert:
4766         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4768         * config/mips/mips.c (mips_truncated_op_cost): New function.
4769         (mips_rtx_costs): Adjust test for BADDU.
4770         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4772         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4774         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4775         (*baddu_si): ...this new pattern.
4777 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4779         PR ipa/59722
4780         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4782 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4784         PR middle-end/57748
4785         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4786         inner_reference_p.
4787         (expand_expr, expand_normal): Adjust.
4788         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4789         inner_reference_p. Use inner_reference_p to expand inner references.
4790         (store_expr): Adjust.
4791         * cfgexpand.c (expand_call_stmt): Adjust.
4793 2014-01-08  Rong Xu  <xur@google.com>
4795         * gcov-io.c (gcov_var): Move from gcov-io.h.
4796         (gcov_position): Ditto.
4797         (gcov_is_error): Ditto.
4798         (gcov_rewrite): Ditto.
4799         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4800         only part to libgcc/libgcov.h.
4802 2014-01-08  Marek Polacek  <polacek@redhat.com>
4804         PR middle-end/59669
4805         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4807 2014-01-08  Marek Polacek  <polacek@redhat.com>
4809         PR sanitizer/59667
4810         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4812 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4814         PR rtl-optimization/59649
4815         * stor-layout.c (get_mode_bounds): For BImode return
4816         0 and STORE_FLAG_VALUE.
4818 2014-01-08  Richard Biener  <rguenther@suse.de>
4820         PR middle-end/59630
4821         * gimple.h (is_gimple_builtin_call): Remove.
4822         (gimple_builtin_call_types_compatible_p): New.
4823         (gimple_call_builtin_p): New overload.
4824         * gimple.c (is_gimple_builtin_call): Remove.
4825         (validate_call): Rename to ...
4826         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4827         check return types.
4828         (validate_type): New static function.
4829         (gimple_call_builtin_p): New overload and adjust.
4830         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4831         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4832         (gimple_fold_stmt_to_constant_1): Likewise.
4833         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4835 2014-01-08  Richard Biener  <rguenther@suse.de>
4837         PR middle-end/59471
4838         * gimplify.c (gimplify_expr): Gimplify register-register type
4839         VIEW_CONVERT_EXPRs to separate stmts.
4841 2014-01-07  Jeff Law  <law@redhat.com>
4843         PR middle-end/53623
4844         * ree.c (combine_set_extension): Handle case where source
4845         and destination registers in an extension insn are different.
4846         (combine_reaching_defs): Allow source and destination registers
4847         in extension to be different under limited circumstances.
4848         (add_removable_extension): Remove restriction that the
4849         source and destination registers in the extension are the same.
4850         (find_and_remove_re): Emit a copy from the extension's
4851         destination to its source after the defining insn if
4852         the source and destination registers are different.
4854         PR middle-end/59285
4855         * ifcvt.c (merge_if_block): If we are merging a block with more than
4856         one successor with a block with no successors, remove any BARRIER
4857         after the second block.
4859 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4861         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4863 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4865         PR target/59652
4866         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4867         for 14-bit register offsets when INT14_OK_STRICT is false.
4869 2014-01-07  Roland Stigge  <stigge@antcom.de>
4870             Michael Meissner  <meissner@linux.vnet.ibm.com>
4872         PR 57386/target
4873         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4874         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
4876 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
4878         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4879         -mcpu.
4881 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
4883         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4884         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4885         rtx is const0_rtx or not.
4887 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
4889         PR target/58115
4890         * target-globals.c (save_target_globals): Remove this_fn_optab
4891         handling.
4892         * toplev.c: Include optabs.h.
4893         (target_reinit): Temporarily restore the global options if another
4894         set of options are in force.
4896 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
4898         PR rtl-optimization/58668
4899         * cfgcleanup.c (flow_find_cross_jump): Don't count
4900         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
4901         to determine what is counted.
4902         (flow_find_head_matching_sequence): Use active_insn_p to determine
4903         what is counted.
4904         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4905         counting change.
4906         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4907         determine what is counted.
4909         PR tree-optimization/59643
4910         * tree-predcom.c (split_data_refs_to_components): If one dr is
4911         read and one write, determine_offset fails and the write isn't
4912         in the bad component, just put the read into the bad component.
4914 2014-01-07  Mike Stump  <mikestump@comcast.net>
4915             Jakub Jelinek  <jakub@redhat.com>
4917         PR pch/59436
4918         * tree-core.h (struct tree_optimization_option): Change optabs
4919         type from unsigned char * to void *.
4920         * optabs.c (init_tree_optimization_optabs): Adjust
4921         TREE_OPTIMIZATION_OPTABS initialization.
4923 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
4925         PR target/59644
4926         * config/i386/i386.h (struct machine_function): Add
4927         no_drap_save_restore field.
4928         * config/i386/i386.c (ix86_save_reg): Use
4929         !cfun->machine->no_drap_save_restore instead of
4930         crtl->stack_realign_needed.
4931         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4932         this function clears frame_pointer_needed.  Set
4933         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4934         and DRAP reg is needed.
4936 2014-01-06  Marek Polacek  <polacek@redhat.com>
4938         PR c/57773
4939         * doc/implement-c.texi: Mention that other integer types are
4940         permitted as bit-field types in strictly conforming mode.
4942 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
4944         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4945         is newly allocated.
4947 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
4949         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4951 2014-01-06  Martin Jambor  <mjambor@suse.cz>
4953         PR ipa/59008
4954         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4955         to int.
4956         * ipa-prop.c (ipa_print_node_params): Fix indentation.
4958 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
4960         PR debug/59350
4961         PR debug/59510
4962         * var-tracking.c (add_stores): Preserve the value of the source even if
4963         we don't record the store.
4965 2014-01-06  Terry Guo  <terry.guo@arm.com>
4967         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4969 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
4971         PR bootstrap/59541
4972         * config/darwin.c (darwin_function_section): Adjust return values to
4973         correspond to optimisation changes made in r206070.
4975 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
4977         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4978         from prefetch_block tune setting.
4979         (nocona_cost): Correct size of prefetch block to 64.
4981 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
4983         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4984         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4985         used to save the static chain register in the computation of the offset
4986         from which the FP registers need to be restored.
4988 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
4990         PR tree-optimization/59519
4991         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4992         ICE if get_current_def (current_new_name) is already non-NULL, as long
4993         as it is a phi result of some other phi in *new_exit_bb that has
4994         the same argument.
4996         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4997         or vmovdqu* for misaligned_operand.
4998         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4999         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5000         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5001         aligned_mem for AVX512F masked aligned load and store builtins and for
5002         non-temporal moves.
5004 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5006         PR tree-optimization/59651
5007         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5008         Address range for negative step should be added by TYPE_SIZE_UNIT.
5010 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5012         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5013         overlapping registers also for registers other than the stack pointer.
5015 2014-01-03  Marek Polacek  <polacek@redhat.com>
5017         PR other/59661
5018         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5019         __builtin_FILE.
5021 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5023         PR target/59625
5024         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5025         asm goto as jump.
5027         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5028         (push splitter): Use <P:MODE_SIZE> instead of
5029         GET_MODE_SIZE (<P:MODE>mode).
5030         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5031         (mov -1, reg peephole2): Likewise.
5032         * config/i386/sse.md (*mov<mode>_internal,
5033         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5034         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5035         *<code><mode>3, *andnot<mode>3<mask_name>,
5036         <mask_codefor><code><mode>3<mask_name>): Likewise.
5037         * config/i386/subst.md (mask_mode512bit_condition,
5038         sd_mask_mode512bit_condition): Likewise.
5040 2014-01-02  Xinliang David Li  <davidxl@google.com>
5042         PR tree-optimization/59303
5043         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5044         (dump_predicates): Better output format.
5045         (pred_equal_p): New function.
5046         (is_neq_relop_p): Ditto.
5047         (is_neq_zero_form_p): Ditto.
5048         (pred_expr_equal_p): Ditto.
5049         (pred_neg_p): Ditto.
5050         (simplify_pred): Ditto.
5051         (simplify_preds_2): Ditto.
5052         (simplify_preds_3): Ditto.
5053         (simplify_preds_4): Ditto.
5054         (simplify_preds): Ditto.
5055         (push_pred): Ditto.
5056         (push_to_worklist): Ditto.
5057         (get_pred_info_from_cmp): Ditto.
5058         (is_degenerated_phi): Ditto.
5059         (normalize_one_pred_1): Ditto.
5060         (normalize_one_pred): Ditto.
5061         (normalize_one_pred_chain): Ditto.
5062         (normalize_preds): Ditto.
5063         (normalize_cond_1): Remove function.
5064         (normalize_cond): Ditto.
5065         (is_gcond_subset_of): Ditto.
5066         (is_subset_of_any): Ditto.
5067         (is_or_set_subset_of): Ditto.
5068         (is_and_set_subset_of): Ditto.
5069         (is_norm_cond_subset_of): Ditto.
5070         (pred_chain_length_cmp): Ditto.
5071         (convert_control_dep_chain_into_preds): Type change.
5072         (find_predicates): Ditto.
5073         (find_def_preds): Ditto.
5074         (destroy_predicates_vecs): Ditto.
5075         (find_matching_predicates_in_rest_chains): Ditto.
5076         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5077         (is_pred_expr_subset): Ditto.
5078         (is_pred_chain_subset_of): Ditto.
5079         (is_included_in): Ditto.
5080         (is_superset_of): Ditto.
5082 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5084         Update copyright years.
5086 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5088         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5089         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5090         config/arc/arc.md, config/arc/arc.opt,
5091         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5092         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5093         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5094         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5095         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5096         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5097         vtable-verify.c, vtable-verify.h: Use the standard form for the
5098         copyright notice.
5100 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5102         * gcc.c (process_command): Update copyright notice dates.
5103         * gcov-dump.c: Ditto.
5104         * gcov.c: Ditto.
5105         * doc/cpp.texi: Bump @copying's copyright year.
5106         * doc/cppinternals.texi: Ditto.
5107         * doc/gcc.texi: Ditto.
5108         * doc/gccint.texi: Ditto.
5109         * doc/gcov.texi: Ditto.
5110         * doc/install.texi: Ditto.
5111         * doc/invoke.texi: Ditto.
5113 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5115         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5117 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5119         * config/i386/sse.md (*mov<mode>_internal): Guard
5120         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5122         PR rtl-optimization/59647
5123         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5124         new_rtx into UNSIGNED_FLOAT rtxes.
5126 Copyright (C) 2014 Free Software Foundation, Inc.
5128 Copying and distribution of this file, with or without modification,
5129 are permitted in any medium without royalty provided the copyright
5130 notice and this notice are preserved.