2014-03-18 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob8038951e2263645c8a10274b0646f2cfbe04406c
1 2014-03-18  Richard Biener  <rguenther@suse.de>
3         * doc/invoke.texi (flto): Update for changes in 4.9.
5 2014-03-18  Richard Biener  <rguenther@suse.de>
7         * doc/loop.texi: Remove section on the removed lambda framework.
8         Update loop docs with recent changes in preserving loop structure.
10 2014-03-18  Richard Biener  <rguenther@suse.de>
12         * doc/lto.texi (-fresolution): Document.
14 2014-03-18  Richard Biener  <rguenther@suse.de>
16         * doc/contrib.texi: Adjust my name.
18 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
20         PR ipa/58721
21         * internal-fn.c: Include diagnostic-core.h.
22         (expand_BUILTIN_EXPECT): New function.
23         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
24         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
25         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
26         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
27         IFN_BUILTIN_EXPECT.
28         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
29         Revert 3 argument __builtin_expect code.
30         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
31         * gimple-fold.c (gimple_fold_call): Likewise.
32         * tree.h (fold_builtin_expect): New prototype.
33         * builtins.c (build_builtin_expect_predicate): Add predictor
34         argument, if non-NULL, create 3 argument __builtin_expect.
35         (fold_builtin_expect): No longer static.  Add ARG2 argument,
36         pass it through to build_builtin_expect_predicate.
37         (fold_builtin_2): Adjust caller.
38         (fold_builtin_3): Handle BUILT_IN_EXPECT.
39         * internal-fn.def (BUILTIN_EXPECT): New.
41 2014-03-18  Tobias Burnus  <burnus@net-b.de>
43         PR ipa/58721
44         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
45         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
46         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
48 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
50         PR ipa/58721
51         * predict.c (combine_predictions_for_bb): Fix up formatting.
52         (expr_expected_value_1, expr_expected_value): Add predictor argument,
53         fill what it points to if non-NULL.
54         (tree_predict_by_opcode): Adjust caller, use the predictor.
55         * predict.def (PRED_COMPARE_AND_SWAP): Add.
57 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
59         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
60         proper constant for the store mode.
62 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
64         * symtab.c (change_decl_assembler_name): Fix transparent alias
65         chain construction.
67 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
69         * config/aarch64/aarch64.c: Correct the comments about the
70         aarch64 stack layout.
72 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
74         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
75         check for GF_OMP_FOR_KIND_FOR.
77 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
79         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
80         ymm and zmm register names.
82 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
84         PR target/60516
85         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
86         note creation for the 2010-08-31 changes.
88 2014-03-17  Marek Polacek  <polacek@redhat.com>
90         PR middle-end/60534
91         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
92         as -fno-tree-loop-vectorize.
93         (expand_omp_simd): Likewise.
95 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
97         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
98         (eligible_for_call_delay): New prototype.
99         * config/sparc/sparc.c (tls_call_delay): Rename into...
100         (eligible_for_call_delay): ...this.  Return false if the instruction
101         cannot be put in the delay slot of a branch.
102         (eligible_for_restore_insn): Simplify.
103         (eligible_for_return_delay): Return false if the instruction cannot be
104         put in the delay slot of a branch and simplify.
105         (eligible_for_sibcall_delay): Return false if the instruction cannot be
106         put in the delay slot of a branch.
107         * config/sparc/sparc.md (fix_ut699): New attribute.
108         (tls_call_delay): Delete.
109         (in_call_delay): Reimplement.
110         (eligible_for_sibcall_delay): Rename into...
111         (in_sibcall_delay): ...this.
112         (eligible_for_return_delay): Rename into...
113         (in_return_delay): ...this.
114         (in_branch_delay): Reimplement.
115         (in_uncond_branch_delay): Delete.
116         (in_annul_branch_delay): Delete.
118 2014-03-14  Richard Henderson  <rth@redhat.com>
120         PR target/60525
121         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
122         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
123         (*floathi<X87MODEF>2_i387_with_temp): Remove.
124         (floathi splitters): Remove.
125         (float<SWI48x>xf2): New pattern.
126         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
127         code that tried to handle DImode for 32-bit, but which was excluded
128         by the pattern's condition.  Drop allocation of stack temporary.
129         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
130         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
131         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
132         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
133         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
134         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
135         (*float<SWI48><MODEF>2_sse_interunit): Remove.
136         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
137         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
138         (*float<SWI48x><X87MODEF>2_i387): Remove.
139         (all float _with_temp splitters): Remove.
140         (*float<SWI48x><MODEF>2_i387): New pattern.
141         (*float<SWI48><MODEF>2_sse): New pattern.
142         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
143         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
145 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
146             Marek Polacek  <polacek@redhat.com>
148         PR middle-end/60484
149         * common.opt (dump_base_name_prefixed): New Variable.
150         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
151         if x_dump_base_name_prefixed is already set, set it at the end.
153 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
155         PR rtl-optimization/60508
156         * lra-constraints.c (get_reload_reg): Add new parameter
157         in_subreg_p.
158         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
159         Pass the new parameter values.
161 2014-03-14  Richard Biener  <rguenther@suse.de>
163         * common.opt: Revert unintented changes from r205065.
164         * opts.c: Likewise.
166 2014-03-14  Richard Biener  <rguenther@suse.de>
168         PR middle-end/60518
169         * cfghooks.c (split_block): Properly adjust all loops the
170         block was a latch of.
172 2014-03-14  Martin Jambor  <mjambor@suse.cz>
174         PR lto/60461
175         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
176         and simplify it.
178 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
180         PR target/59396
181         * config/avr/avr.c (avr_set_current_function): Pass function name
182         through default_strip_name_encoding before sanity checking instead
183         of skipping the first char of the assembler name.
185 2014-03-13  Richard Henderson  <rth@redhat.com>
187         PR debug/60438
188         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
189         (ix86_force_to_memory, ix86_free_from_memory): Remove.
190         * config/i386/i386-protos.h: Likewise.
191         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
192         in the expander instead of a splitter.
193         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
194         any possibility of requiring a memory.
195         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
196         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
197         (fp branch splitters): Update for ix86_split_fp_branch.
198         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
199         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
200         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
201         (*fop_<MODEF>_2_i387): Remove f/r alternative.
202         (*fop_<MODEF>_3_i387): Likewise.
203         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
204         (splitters for the fop_* register patterns): Remove.
205         (fscalexf4_i387): Rename from *fscalexf4_i387.
206         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
207         
208 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
210         PR tree-optimization/59779
211         * tree-dfa.c (get_ref_base_and_extent): Use double_int
212         type for bitsize and maxsize instead of HOST_WIDE_INT.
214 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
216         PR rtl-optimization/57320
217         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
218         the CFG after thread_prologue_and_epilogue_insns.
220 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
222         PR rtl-optimization/57189
223         * lra-constraints.c (process_alt_operands): Disfavor spilling
224         vector pseudos.
226 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
228         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
229         messages.
231 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
233         PR tree-optimization/59025
234         PR middle-end/60418
235         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
236         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
238 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
240         PR target/60486
241         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
242         calls of avr_out_plus_1.
244 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
246         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
247         BB's single pred and update the father loop's latch info later.
249 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
251         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
252         (VEC_M): Likewise.
253         (VEC_N): Likewise.
254         (VEC_R): Likewise.
255         (VEC_base): Likewise.
256         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
257         registers, we need to swap double words in little endian mode.
259         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
260         to be a container mode for 128-bit integer operations added in ISA
261         2.07.  Unlike TImode and PTImode, the preferred register set is
262         the Altivec/VMX registers for the 128-bit operations.
264         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
265         declarations.
266         (rs6000_split_128bit_ok_p): Likewise.
268         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
269         macros for creating ISA 2.07 normal and overloaded builtin
270         functions with 3 arguments.
271         (BU_P8V_OVERLOAD_3): Likewise.
272         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
273         for use as overloaded functions.
274         (VPERM_1TI_UNS): Likewise.
275         (VSEL_1TI): Likewise.
276         (VSEL_1TI_UNS): Likewise.
277         (ST_INTERNAL_1ti): Likewise.
278         (LD_INTERNAL_1ti): Likewise.
279         (XXSEL_1TI): Likewise.
280         (XXSEL_1TI_UNS): Likewise.
281         (VPERM_1TI): Likewise.
282         (VPERM_1TI_UNS): Likewise.
283         (XXPERMDI_1TI): Likewise.
284         (SET_1TI): Likewise.
285         (LXVD2X_V1TI): Likewise.
286         (STXVD2X_V1TI): Likewise.
287         (VEC_INIT_V1TI): Likewise.
288         (VEC_SET_V1TI): Likewise.
289         (VEC_EXT_V1TI): Likewise.
290         (EQV_V1TI): Likewise.
291         (NAND_V1TI): Likewise.
292         (ORC_V1TI): Likewise.
293         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
294         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
295         overloaded builtin.
296         (VADDUQM): Likewise.
297         (VSUBCUQ): Likewise.
298         (VADDEUQM): Likewise.
299         (VADDECUQ): Likewise.
300         (VSUBEUQM): Likewise.
301         (VSUBECUQ): Likewise.
303         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
304         __int128_t and __uint128_t types.
305         (__uint128_type): Likewise.
306         (altivec_categorize_keyword): Add support for vector __int128_t,
307         vector __uint128_t, vector __int128, and vector unsigned __int128
308         as a container type for TImode operations that need to be done in
309         VSX/Altivec registers.
310         (rs6000_macro_to_expand): Likewise.
311         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
312         to support 128-bit integer instructions vaddcuq, vadduqm,
313         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
314         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
316         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
317         for V1TImode, and set up preferences to use VSX/Altivec registers.
318         Setup VSX reload handlers.
319         (rs6000_debug_reg_global): Likewise.
320         (rs6000_init_hard_regno_mode_ok): Likewise.
321         (rs6000_preferred_simd_mode): Likewise.
322         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
323         (easy_altivec_constant): Likewise.
324         (output_vec_const_move): Likewise.
325         (rs6000_expand_vector_set): Convert V1TImode set and extract to
326         simple move.
327         (rs6000_expand_vector_extract): Likewise.
328         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
329         addressing.
330         (rs6000_const_vec): Add support for V1TImode.
331         (rs6000_emit_le_vsx_load): Swap double words when loading or
332         storing TImode/V1TImode.
333         (rs6000_emit_le_vsx_store): Likewise.
334         (rs6000_emit_le_vsx_move): Likewise.
335         (rs6000_emit_move): Add support for V1TImode.
336         (altivec_expand_ld_builtin): Likewise.
337         (altivec_expand_st_builtin): Likewise.
338         (altivec_expand_vec_init_builtin): Likewise.
339         (altivec_expand_builtin): Likewise.
340         (rs6000_init_builtins): Add support for V1TImode type.  Add
341         support for ISA 2.07 128-bit integer builtins.  Define type names
342         for the VSX/Altivec vector types.
343         (altivec_init_builtins): Add support for overloaded vector
344         functions with V1TImode type.
345         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
346         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
347         external function.
348         (rs6000_split_128bit_ok_p): Likewise.
349         (rs6000_handle_altivec_attribute): Create V1TImode from vector
350         __int128_t and vector __uint128_t.
352         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
353         and mode attributes.
354         (VSX_M): Likewise.
355         (VSX_M2): Likewise.
356         (VSm): Likewise.
357         (VSs): Likewise.
358         (VSr): Likewise.
359         (VSv): Likewise.
360         (VS_scalar): Likewise.
361         (VS_double): Likewise.
362         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
364         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
365         we support the ISA 2.07 128-bit integer arithmetic instructions.
366         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
367         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
368         and TImode types for use with the builtin functions.
369         (V1TI_type_node): Likewise.
370         (unsigned_V1TI_type_node): Likewise.
371         (intTI_type_internal_node): Likewise.
372         (uintTI_type_internal_node): Likewise.
374         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
375         128-bit builtin functions.
376         (UNSPEC_VADDEUQM): Likewise.
377         (UNSPEC_VADDECUQ): Likewise.
378         (UNSPEC_VSUBCUQ): Likewise.
379         (UNSPEC_VSUBEUQM): Likewise.
380         (UNSPEC_VSUBECUQ): Likewise.
381         (VM): Add V1TImode to vector mode iterators.
382         (VM2): Likewise.
383         (VI_unit): Likewise.
384         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
385         (altivec_vaddcuq): Likewise.
386         (altivec_vsubuqm): Likewise.
387         (altivec_vsubcuq): Likewise.
388         (altivec_vaddeuqm): Likewise.
389         (altivec_vaddecuq): Likewise.
390         (altivec_vsubeuqm): Likewise.
391         (altivec_vsubecuq): Likewise.
393         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
394         mode iterators.
395         (BOOL_128): Likewise.
396         (BOOL_REGS_OUTPUT): Likewise.
397         (BOOL_REGS_OP1): Likewise.
398         (BOOL_REGS_OP2): Likewise.
399         (BOOL_REGS_UNARY): Likewise.
400         (BOOL_REGS_AND_CR0): Likewise.
402         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
403         128-bit integer builtin support.
404         (vec_vadduqm): Likewise.
405         (vec_vaddecuq): Likewise.
406         (vec_vaddeuqm): Likewise.
407         (vec_vsubecuq): Likewise.
408         (vec_vsubeuqm): Likewise.
409         (vec_vsubcuq): Likewise.
410         (vec_vsubuqm): Likewise.
412         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
413         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
414         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
415         128-bit integer add/subtract to ISA 2.07.
417 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
419         * config/arc/arc.c (arc_predicate_delay_insns):
420         Fix third argument passed to conditionalize_nonjump.
422 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
424         * config/aarch64/aarch64-builtins.c
425         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
426         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
427         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
428         instead of __builtin_lfloor.
429         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
431 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
433         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
434         (tree_ssa_ifcombine_bb_1): New function.
435         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
436         is an empty forwarder block to then_bb or vice versa and then_bb
437         and else_bb are effectively swapped.
439 2014-03-12  Christian Bruel  <christian.bruel@st.com>
441         PR target/60264
442         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
443         REG_CFA_DEF_CFA note.
444         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
445         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
447 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
449         PR tree-optimization/60454
450         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
452 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
454         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
455         Do not define target_cpu_default2 to generic.
456         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
457         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
458         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
460 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
461             Marc Glisse  <marc.glisse@inria.fr>
463         PR tree-optimization/60502
464         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
465         instead of build_low_bits_mask.
467 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
469         PR middle-end/60482
470         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
471         if there are multiple uses, but op doesn't live on E edge.
472         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
473         clobber stmts before __builtin_unreachable.
475 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
477         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
478         hard_frame_pointer_rtx.
479         * cse.c (cse_insn): Remove volatile check.
480         * cselib.c (cselib_process_insn): Likewise.
481         * dse.c (scan_insn): Likewise.
483 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
485         * config/arc/arc.c (conditionalize_nonjump): New function,
486         broken out of ...
487         (arc_ifcvt): ... this.
488         (arc_predicate_delay_insns): Use it.
490 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
492         * config/arc/predicates.md (extend_operand): During/after reload,
493         allow const_int_operand.
494         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
495         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
496         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
497         to "i".
498         (umulsi3_highpart_i): Likewise.
500 2014-03-11  Richard Biener  <rguenther@suse.de>
502         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
503         Add asserts to guard possible wrong-code bugs.
505 2014-03-11  Richard Biener  <rguenther@suse.de>
507         PR tree-optimization/60429
508         PR tree-optimization/60485
509         * tree-ssa-structalias.c (set_union_with_increment): Properly
510         take into account all fields that overlap the shifted vars.
511         (do_sd_constraint): Likewise.
512         (do_ds_constraint): Likewise.
513         (get_constraint_for_ptr_offset): Likewise.
515 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
517         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
518         (nios2_compute_frame_layout):
519         Add calculation of cfun->machine->fp_save_offset.
520         (nios2_expand_prologue): Correct setting of frame pointer register
521         in prologue.
522         (nios2_expand_epilogue): Update recovery of stack pointer from
523         frame pointer accordingly.
524         (nios2_initial_elimination_offset): Update calculation of offset
525         for eliminating to HARD_FRAME_POINTER_REGNUM.
527 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
529         PR ipa/60457
530         * ipa.c (symtab_remove_unreachable_nodes): Don't call
531         cgraph_get_create_node on VAR_DECLs.
533 2014-03-10  Richard Biener  <rguenther@suse.de>
535         PR middle-end/60474
536         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
538 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
540         * config/vms/vms.opt (vms_float_format): New variable.
542 2014-03-08  Tobias Burnus  <burnus@net-b.de>
544         * doc/invoke.texi (-fcilkplus): Update implementation status.
546 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
547             Richard Biener  <rguenther@suse.de>
549         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
550         consistently accross all TUs.
551         (run_gcc): Enable -fshort-double automatically at link at link-time
552         and disallow override.
554 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
556         PR target/58271
557         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
558         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
559         if they can't be used.
561 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
563         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
564         for Solaris 11/x86 ld.
565         * configure: Regenerate.
567 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
569         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
570         (LIB_TLS_SPEC): Save as ld_tls_libs.
571         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
572         (HAVE_AS_IX86_TLSLDM): New test.
573         * configure, config.in: Regenerate.
574         * config/i386/i386.c (legitimize_tls_address): Fall back to
575         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
576         cannot support TLS_MODEL_LOCAL_DYNAMIC.
577         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
578         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
580 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
582         * common.opt (fira-loop-pressure): Mark as optimization.
584 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
586         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
587         an OpenMP mappable type.
589 2014-03-06  Matthias Klose  <doko@ubuntu.com>
591         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
592         MULTILIB_OSDIRNAMES is not defined.
594 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
595             Meador Inge  <meadori@codesourcery.com>
597         PR target/58595
598         * config/arm/arm.c (arm_tls_symbol_p): Remove.
599         (arm_legitimize_address): Call legitimize_tls_address for any
600         arm_tls_referenced_p expression, handle constant addend.  Call it
601         before testing for !TARGET_ARM.
602         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
604 2014-03-06  Richard Biener  <rguenther@suse.de>
606         PR middle-end/60445
607         PR lto/60424
608         PR lto/60427
609         Revert
610         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
612         * tree-streamer.c (record_common_node): Assert we don't record
613         nodes with type double.
614         (preload_common_node): Skip type double, complex double and double
615         pointer since it is now frontend dependent due to fshort-double option.
617 2014-03-06  Richard Biener  <rguenther@suse.de>
619         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
620         or -fno-lto is specified and the linker has full plugin support.
621         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
622         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
623         * lto-wrapper.c (merge_and_complain): Merge compile-time
624         optimization levels.
625         (run_gcc): And pass it through to the link options.
627 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
629         PR debug/60381
630         Revert:
631         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
632         PR debug/59992
633         * cselib.c (remove_useless_values): Skip to avoid quadratic
634         behavior if the condition moved from...
635         (cselib_process_insn): ... here holds.
637 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
639         PR plugins/59335
640         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
641         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
643         PR plugins/59335
644         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
645         (TM_H): Add x86-tune.def.
647 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
649         * config/aarch64/aarch64.c (generic_tunings):
650         Use cortexa57_extra_costs.
652 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
654         PR lto/60404
655         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
656         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
657         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
658         cost for in_lto_p.
660 2014-03-04  Heiher  <r@hev.cc>
662         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
663         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
665 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
667         * config/i386/predicates.md (const2356_operand): Change to ...
668         (const2367_operand): ... this.
669         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
670         const2367_operand.
671         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
672         (*avx512pf_scatterpf<mode>sf): Ditto.
673         (avx512pf_scatterpf<mode>df): Ditto.
674         (*avx512pf_scatterpf<mode>df_mask): Ditto.
675         (*avx512pf_scatterpf<mode>df): Ditto.
676         * config/i386/i386.c (ix86_expand_builtin): Update
677         incorrect hint operand error message.
679 2014-03-04  Richard Biener  <rguenther@suse.de>
681         * lto-section-in.c (lto_get_section_data): Fix const cast.
683 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
685         * tree-streamer.c (record_common_node): Assert we don't record
686         nodes with type double.
687         (preload_common_node): Skip type double, complex double and double
688         pointer since it is now frontend dependent due to fshort-double option.
690 2014-03-04  Richard Biener  <rguenther@suse.de>
692         PR lto/60405
693         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
694         (lto_input_toplevel_asms): Likewise.
695         * lto-section-in.c (lto_get_section_data): Instead do it here
696         for every section.
698 2014-03-04  Richard Biener  <rguenther@suse.de>
700         PR tree-optimization/60382
701         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
702         dead PHIs a reduction.
704 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
706         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
707         hint value.
708         (_mm_prefetch): Move out of GCC target("sse") pragma.
709         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
710         GCC target("prfchw") pragma.
711         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
712         for locality <= 2.
713         * config/i386/i386.c (ix86_option_override_internal): Enable
714         -mprfchw with -mprefetchwt1.
716 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
718         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
719         Mark as varying.
721 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
723         * opts.h (CL_PCH_IGNORE): Define.
724         * targhooks.c (option_affects_pch_p):
725         Return false for options that have CL_PCH_IGNORE set.
726         * opt-functions.awk: Process PchIgnore.
727         * doc/options.texi: Document PchIgnore.
729         * config/arc/arc.opt (misize): Add PchIgnore property.
731 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
733         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
734         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
735         constraint on constants to permit them being loaded into
736         GENERAL_REGS or BASE_REGS.
738 2014-03-03  Nick Clifton  <nickc@redhat.com>
740         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
741         anti-cacnonical alternatives.
742         (negandhi3_real): New pattern.
743         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
745 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
747         * config/avr/avr-mcus.def: Remove atxmega16x1.
748         * config/avr/avr-tables.opt: Regenerate.
749         * config/avr/t-multilib: Regenerate.
750         * doc/avr-mmcu.texi: Regenerate.
752 2014-03-03  Tobias Grosser  <tobias@grosser.es>
753             Mircea Namolaru  <mircea.namolaru@inria.fr>
755         PR tree-optimization/58028
756         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
757         scalar dimensions.
759 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
761         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
762         not handled by recognizers.
764 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
766         PR middle-end/60175
767         * function.c (expand_function_end): Don't emit
768         clobber_return_register sequence if clobber_after is a BARRIER.
769         * cfgexpand.c (construct_exit_block): Append instructions before
770         return_label to prev_bb.
772 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
774         * config/rs6000/constraints.md: Document reserved use of "wc".
776 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
778         PR ipa/60150
779         * ipa.c (function_and_variable_visibility): When dissolving comdat
780         group, also set all symbols to local.
782 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
784         PR ipa/60306
786         Revert:
787         2013-12-14   Jan Hubicka  <jh@suse.cz>
788         PR middle-end/58477
789         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
791 2014-03-02  Jon Beniston  <jon@beniston.com>
793         PR bootstrap/48230
794         PR bootstrap/50927
795         PR bootstrap/52466
796         PR target/46898
797         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
798           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
799         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
800         (simple_return, *simple_return): New patterns
801         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
802         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
804 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
806         * dwarf2out.c (gen_subprogram_die): Tidy.
808 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
810         PR target/60071
811         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
812         (*mov_t_msb_neg_negc): ... this new insn.
814 2014-02-28  Jason Merrill  <jason@redhat.com>
816         PR c++/58678
817         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
818         function.
820 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
822         PR c++/60314
823         * dwarf2out.c (decltype_auto_die): New static.
824         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
825         (gen_type_die_with_usage): Handle 'decltype(auto)'.
826         (is_cxx_auto): Likewise.
828 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
830         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
831         we are not using general regs only.
833 2014-02-28  Richard Biener  <rguenther@suse.de>
835         PR target/60280
836         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
837         previous fix and only allow to remove trivial pre-headers
838         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
839         (remove_forwarder_block): Properly update the latch of a loop.
841 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
843         PR debug/59992
844         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
845         (cselib_preserved_hash_table): New.
846         (preserve_constants_and_equivs): Move preserved vals to it.
847         (cselib_find_slot): Look it up first.
848         (cselib_init): Initialize it.
849         (cselib_finish): Release it.
850         (dump_cselib_table): Dump it.
852 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
854         PR debug/59992
855         * cselib.c (remove_useless_values): Skip to avoid quadratic
856         behavior if the condition moved from...
857         (cselib_process_insn): ... here holds.
859 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
861         PR debug/57232
862         * var-tracking.c (vt_initialize): Apply the same condition to
863         preserve the CFA base value.
865 2014-02-28  Joey Ye  <joey.ye@arm.com>
867         PR target/PR60169
868         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
869         if reload in progress or completed.
871 2014-02-28  Tobias Burnus  <burnus@net-b.de>
873         PR middle-end/60147
874         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
875         NAMELIST_DECL.
877 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
879         * doc/tm.texi.in (Condition Code Status): Update documention for
880         relative locations of cc0-setter and cc0-user.
882 2014-02-27  Jeff Law  <law@redhat.com>
884         PR rtl-optimization/52714
885         * combine.c (try_combine): When splitting an unrecognized PARALLEL
886         into two independent simple sets, if I3 is a jump, ensure the
887         pattern we place into I3 is a (set (pc) ...).
889 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
890             Jeff Law  <law@redhat.com>
892         PR rtl-optimization/49847
893         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
894         are in different blocks.
895         * doc/tm.texi (Condition Code Status): Update documention for
896         relative locations of cc0-setter and cc0-user.
898 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
900         PR target/59222
901         * lra.c (lra_emit_add): Check SUBREG too.
903 2014-02-27  Andreas Schwab  <schwab@suse.de>
905         * config/m68k/m68k.c (m68k_option_override): Disable
906         -flive-range-shrinkage for classic m68k.
907         (m68k_override_options_after_change): Likewise.
909 2014-02-27  Marek Polacek  <polacek@redhat.com>
911         PR middle-end/59223
912         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
913         -Wmaybe-uninitialized.
915 2014-02-27  Alan Modra  <amodra@gmail.com>
917         PR target/57936
918         * reload1.c (emit_input_reload_insns): When reload_override_in,
919         set old to rl->in_reg when rl->in_reg is a subreg.
921 2014-02-26  Richard Biener  <rguenther@suse.de>
923         PR bootstrap/60343
924         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
926 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
928         * common/config/i386/predicates.md (const1256_operand): Remove.
929         (const2356_operand): New.
930         (const_1_to_2_operand): Remove.
931         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
932         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
933         (*avx512pf_gatherpf<mode>sf): Ditto.
934         (avx512pf_gatherpf<mode>df): Ditto.
935         (*avx512pf_gatherpf<mode>df_mask): Ditto.
936         (*avx512pf_gatherpf<mode>df): Ditto.
937         (avx512pf_scatterpf<mode>sf): Ditto.
938         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
939         (*avx512pf_scatterpf<mode>sf): Ditto.
940         (avx512pf_scatterpf<mode>df): Ditto.
941         (*avx512pf_scatterpf<mode>df_mask): Ditto.
942         (*avx512pf_scatterpf<mode>df): Ditto.
943         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
945 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
947         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
948         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
949         (_mm512_mask_testn_epi64_mask): Move to ...
950         * config/i386/avx512cdintrin.h: Here.
951         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
952         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
953         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
954         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
955         TARGET_AVX512F from TARGET_AVX512CD.
957 2014-02-26  Richard Biener  <rguenther@suse.de>
959         PR ipa/60327
960         * ipa.c (walk_polymorphic_call_targets): Properly guard
961         call to inline_update_overall_summary.
963 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
965         PR target/60280
966         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
967         and latches only if requested.  Fix latch if it is removed.
968         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
969         LOOPS_HAVE_PREHEADERS.
971 2014-02-25  Andrew Pinski  <apinski@cavium.com>
973         * builtins.c (expand_builtin_thread_pointer): Create a new target
974         when the target is NULL.
976 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
978         PR rtl-optimization/60317
979         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
980         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
981         * lra-assigns.c: Include params.h.
982         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
983         other reload pseudos considerations.
985 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
987         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
988         to use canonical form for nor<mode>3.
990 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
992         PR target/55426
993         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
994         conversions.
996 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
998         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
999         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1000         (ix86_handle_option): Handle OPT_mprefetchwt1.
1001         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1002         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1003         PREFETCHWT1 CPUID.
1004         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1005         OPTION_MASK_ISA_PREFETCHWT1.
1006         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1007         (PTA_PREFETCHWT1): New.
1008         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1009         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1010         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1011         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1012         (*prefetch_avx512pf_<mode>_: Change into ...
1013         (*prefetch_prefetchwt1_<mode>: This.
1014         * config/i386/i386.opt (mprefetchwt1): New.
1015         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1016         (_mm_prefetch): Handle intent to write.
1017         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1019 2014-02-25  Richard Biener  <rguenther@suse.de>
1021         PR middle-end/60291
1022         * emit-rtl.c (mem_attrs_htab): Remove.
1023         (mem_attrs_htab_hash): Likewise.
1024         (mem_attrs_htab_eq): Likewise.
1025         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1026         (init_emit_once): Do not allocate mem_attrs_htab.
1028 2014-02-25  Richard Biener  <rguenther@suse.de>
1030         PR lto/60319
1031         * lto-opts.c (lto_write_options): Output non-explicit conservative
1032         -fwrapv, -fno-trapv and -fno-strict-overflow.
1033         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1034         (run_gcc): And pass them through.
1036 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1038         * sel-sched.c (calculate_new_fences): New parameter ptime.
1039         Calculate it as a maximum over all fence cycles.
1040         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1041         Print the final schedule timing when sched_verbose.
1043 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1045         PR rtl-optimization/60292
1046         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1047         bit fot the fence instruction.
1049 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1051         * calls.h: Fix typo in comment.
1053 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1055         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1056         adjusting offsetable addresses.
1058 2014-02-24  Guozhi Wei  <carrot@google.com>
1060         * sparseset.h (sparseset_pop): Fix the wrong index.
1062 2014-02-24  Walter Lee  <walt@tilera.com>
1064         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1065         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1066         triplet.
1067         * common/config/tilegx/tilegx-common.c
1068         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1069         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1070         (LINK_SPEC): Ditto.
1071         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1072         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1073         (tilegx_gimplify_va_arg_expr): Handle big endian.
1074         (tilegx_expand_unaligned_load): Ditto.
1075         (tilegx_expand_unaligned_store): Ditto.
1076         (TARGET_RETURN_IN_MSB): New.
1077         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1078         (TARGET_ENDIAN_DEFAULT): New.
1079         (TARGET_BIG_ENDIAN): Handle big endian.
1080         (BYTES_BIG_ENDIAN): Ditto.
1081         (WORDS_BIG_ENDIAN): Ditto.
1082         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1083         (ENDIAN_SPEC): New.
1084         (EXTRA_SPECS): New.
1085         * config/tilegx/tilegx.md (extv): Handle big endian.
1086         (extzv): Ditto.
1087         (insn_st<n>): Ditto.
1088         (insn_st<n>_add<bitsuffix>): Ditto.
1089         (insn_stnt<n>): Ditto.
1090         (insn_stnt<n>_add<bitsuffix>):Ditto.
1091         (vec_interleave_highv8qi): Handle big endian.
1092         (vec_interleave_highv8qi_be): New.
1093         (vec_interleave_highv8qi_le): New.
1094         (insn_v1int_h): Handle big endian.
1095         (vec_interleave_lowv8qi): Handle big endian.
1096         (vec_interleave_lowv8qi_be): New.
1097         (vec_interleave_lowv8qi_le): New.
1098         (insn_v1int_l): Handle big endian.
1099         (vec_interleave_highv4hi): Handle big endian.
1100         (vec_interleave_highv4hi_be): New.
1101         (vec_interleave_highv4hi_le): New.
1102         (insn_v2int_h): Handle big endian.
1103         (vec_interleave_lowv4hi): Handle big endian.
1104         (vec_interleave_lowv4hi_be): New.
1105         (vec_interleave_lowv4hi_le): New.
1106         (insn_v2int_l): Handle big endian.
1107         (vec_interleave_highv2si): Handle big endian.
1108         (vec_interleave_highv2si_be): New.
1109         (vec_interleave_highv2si_le): New.
1110         (insn_v4int_h): Handle big endian.
1111         (vec_interleave_lowv2si): Handle big endian.
1112         (vec_interleave_lowv2si_be): New.
1113         (vec_interleave_lowv2si_le): New.
1114         (insn_v4int_l): Handle big endian.
1115         * config/tilegx/tilegx.opt (mbig-endian): New option.
1116         (mlittle-endian): New option.
1117         * doc/install.texi: Document tilegxbe-linux.
1118         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1120 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1122         PR ipa/60266
1123         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1124         there are no parameter descriptors.
1126 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1128         PR rtl-optimization/60268
1129         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1130         initialization to ...
1131         (sched_rgn_init): ... here.
1132         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1134 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1136         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1137         names.
1139 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1141         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1142         definition.
1144 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1146         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1147         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1149 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1151         * config/microblaze/predicates.md: Add cmp_op predicate.
1152         * config/microblaze/microblaze.md: Add branch_compare instruction
1153         which uses cmp_op predicate and emits cmp insn before branch.
1154         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1155         to microblaze_expand_conditional_branch and consolidate logic.
1156         (microblaze_expand_conditional_branch): emit branch_compare
1157         insn instead of handling cmp op separate from branch insn.
1159 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1161         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1162         to permit subregs.
1164 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1166         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1167         define_insn with define_expand and new define_insn
1168         *altivec_lve<VI_char>x_internal.
1169         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1170         and new define_insn *altivec_stve<VI_char>x_internal.
1171         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1172         prototype.
1173         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1174         lve*x built-ins.
1175         (altivec_expand_stvex_be): New function.
1177 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1179         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1180         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1181         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1182         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1184 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1186         PR target/60298
1187         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1188         instead of emit_move_insn.
1190 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1192         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1193         vspltw with vsldoi.
1194         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1195         gen_altivec_vsumsws.
1197 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1199         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1200         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1201         (altivec_lvxl_<mode>): New define_expand incorporating
1202         -maltivec=be semantics where needed.
1203         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1204         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1205         semantics where needed.
1206         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1207         (altivec_stvx_<mode>): New define_expand incorporating
1208         -maltivec=be semantics where needed.
1209         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1210         VM2 iterator instead of V4SI.
1211         (altivec_stvxl_<mode>): New define_expand incorporating
1212         -maltivec=be semantics where needed.
1213         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1214         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1215         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1216         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1217         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1218         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1219         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1220         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1221         ALTIVEC_BUILTIN_STVXL.
1222         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1223         (altivec_expand_stvx_be): Likewise.
1224         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1225         (altivec_expand_lvx_be): Likewise.
1226         (altivec_expand_stvx_be): Likewise.
1227         (altivec_expand_builtin): Add cases for
1228         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1229         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1230         (altivec_init_builtins): Add definitions for
1231         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1232         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1234 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1236         * doc/invoke.texi (mvirt, mno-virt): Document.
1237         * config/mips/mips.opt (mvirt): New option.
1238         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1240 2014-02-21  Richard Biener  <rguenther@suse.de>
1242         PR tree-optimization/60276
1243         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1244         (STMT_VINFO_MIN_NEG_DIST): New macro.
1245         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1246         STMT_VINFO_MIN_NEG_DIST.
1247         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1248         made for negative dependence distances still hold.
1250 2014-02-21  Richard Biener  <rguenther@suse.de>
1252         PR middle-end/60291
1253         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1254         DECL_INITIAL for globals not in the current function context.
1256 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1258         PR tree-optimization/56490
1259         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1260         * tree-ssa-uninit.c: Include params.h.
1261         (compute_control_dep_chain): Add num_calls argument, return false
1262         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1263         num_calls to recursive call.
1264         (find_predicates): Change dep_chain into normal array,
1265         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1266         variable and adjust compute_control_dep_chain caller.
1267         (find_def_preds): Likewise.
1269 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1271         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1272         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1274 2014-02-21  Nick Clifton  <nickc@redhat.com>
1276         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1277         (pushhi1): Likewise.
1278         (popqi1): Add mode to pre_dec.
1279         (pophi1): Likewise.
1281 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1283         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1284         mode for mask of V8SFmode permutation.
1286 2014-02-20  Richard Henderson  <rth@redhat.com>
1288         PR c++/60272
1289         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1290         a new pseudo for OLDVAL.
1292 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1294         PR target/57896
1295         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1296         gen_reg_rtx if d->testing_p.
1297         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1298         if d->testing_p and we will certainly return true.
1299         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1300         if d->testing_p.
1302 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1304         * emit-rtl.c (gen_reg_rtx): Assert that
1305         crtl->emit.regno_pointer_align_length is non-zero.
1307 2014-02-20  Richard Henderson  <rth@redhat.com>
1309         PR c++/60272
1310         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1311         on failure the store back into EXPECT.
1313 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1314             Sandra Loosemore  <sandra@codesourcery.com>
1316         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1317         * config/nios2/nios2.c (nios2_function_profiler): Add
1318         -fPIC (flag_pic == 2) support.
1319         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1320         (nios2_large_offset_p): New function.
1321         (nios2_unspec_reloc_p): Move up position, update to use
1322         nios2_large_offset_p.
1323         (nios2_unspec_address): Remove function.
1324         (nios2_unspec_offset): New function.
1325         (nios2_large_got_address): New function.
1326         (nios2_got_address): Add large offset support.
1327         (nios2_legitimize_tls_address): Update usage of removed and new
1328         functions.
1329         (nios2_symbol_binds_local_p): New function.
1330         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1331         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1332         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1333         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1334         processing for (const (unspec ...)).
1335         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1337 2014-02-20  Richard Biener  <rguenther@suse.de>
1339         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1340         doing the substitution.
1341         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1343 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1345         PR ipa/55260
1346         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1347         info when checking whether lattices are bottom.
1349 2014-02-20  Richard Biener  <rguenther@suse.de>
1351         PR middle-end/60221
1352         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1353         regions at -O0.
1355 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1357         PR ipa/58555
1358         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1359         parameter specifying the scaling.
1360         (inline_call): Update.
1361         (want_inline_recursively): Guard division by zero.
1362         (recursive_inlining): Update.
1363         * ipa-inline.h (clone_inlined_nodes): Update.
1365 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1367         PR target/60204
1368         * config/i386/i386.c (classify_argument): Pass structures of size
1369         64 bytes or less in register.
1371 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1372             Kirill Yukhin  <kirill.yukhin@intel.com>
1374         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1375         (_mm_rcp28_round_ss): Ditto.
1376         (_mm_rsqrt28_round_sd): Ditto.
1377         (_mm_rsqrt28_round_ss): Ditto.
1378         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1379         (_mm_rcp14_round_ss): Ditto.
1380         (_mm_rsqrt14_round_sd): Ditto.
1381         (_mm_rsqrt14_round_ss): Ditto.
1382         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1383         the first input operand, get rid of match_dup.
1384         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1385         attribute to sse.
1386         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1387         Ditto.
1388         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1389         operand as the first input operand, set type attribute.
1390         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1391         Set type attribute.
1392         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1393         operand as the first input operand, set type attribute.
1395 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1397         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1398         bit of zero.
1400 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1402         PR target/60207
1403         * config/i386/i386.c (construct_container): Remove TFmode check
1404         for X86_64_INTEGER_CLASS.
1406 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1408         PR target/59794
1409         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1410         only when -Wpsabi is enabled.
1412 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1414          PR target/59799
1415         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1416         passing arrays in registers are the same as for structs, so remove the
1417         special case for them.
1419 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1421         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1422         destination type, extract only the valid bits if the source type is not
1423         integral and has a different mode.
1425 2014-02-19  Richard Biener  <rguenther@suse.de>
1427         PR ipa/60243
1428         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1429         for all calls.
1431 2014-02-19  Richard Biener  <rguenther@suse.de>
1433         PR ipa/60243
1434         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1435         (ipa_modify_call_arguments): Emit an argument load explicitely and
1436         preserve virtual SSA form there and for the replacement call.
1437         Do not update SSA form nor free dominance info.
1439 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1441         * ipa.c (function_and_variable_visibility): Also clear WEAK
1442         flag when disolving COMDAT_GROUP.
1444 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1446         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1447         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1448         not devirtualizing.
1449         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1450         do more sanity checks.
1451         (detect_type_change): Return true when giving up early.
1452         (compute_complex_assign_jump_func): Fix type parameter of
1453         ipa_set_ancestor_jf.
1454         (compute_complex_ancestor_jump_func): Likewise.
1455         (update_jump_functions_after_inlining): Fix updating of
1456         ancestor function.
1457         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1459 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1461         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1462         inline clones when edge disappears.
1464 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1466         PR target/60203
1467         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1468         Split 64-bit moves into 2 patterns.  Do not allow the use of
1469         direct move for TDmode in little endian, since the decimal value
1470         has little endian bytes within a word, but the 64-bit pieces are
1471         ordered in a big endian fashion, and normal subreg's of TDmode are
1472         not allowed.
1473         (mov<mode>_64bit_dm): Likewise.
1474         (movtd_64bit_nodm): Likewise.
1476 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1478         PR tree-optimization/60174
1479         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1480         statement of an SSA_NAME that occurs in an abnormal PHI node.
1482 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1484         PR sanitizer/60142
1485         * final.c (SEEN_BB): Remove.
1486         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1487         (final_scan_insn): Don't force_source_line on second
1488         NOTE_INSN_BASIC_BLOCK.
1490 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1492         PR target/60205
1493         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1494         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1495         (type_natural_mode): Warn ABI change when %zmm register is not
1496         available for AVX512F vector value passing.
1498 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1500         PR target/60193
1501         * config/i386/i386.c (ix86_expand_prologue): Use value in
1502         rax register as displacement when restoring %r10 or %rax.
1503         Fix wrong offset when restoring both registers.
1505 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1507         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1508         assertion with conditional return.
1510 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1511             Uros Bizjak  <ubizjak@gmail.com>
1513         PR driver/60233
1514         * config/i386/driver-i386.c (host_detect_local_cpu): If
1515         YMM state is not saved by the OS, also clear has_f16c.  Move
1516         CPUID 0x80000001 handling before YMM state saving checking.
1518 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1520         PR rtl-optimization/58960
1521         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1522         factored out from ...
1523         (sched_init): ... here.
1524         (free_global_sched_pressure_data): New, factored out from ...
1525         (sched_finish): ... here.
1526         * sched-int.h (free_global_sched_pressure_data): Declare.
1527         * sched-rgn.c (nr_regions_initial): New static global.
1528         (haifa_find_rgns): Initialize it.
1529         (schedule_region): Disable sched-pressure for the newly
1530         generated regions.
1532 2014-02-17  Richard Biener  <rguenther@suse.de>
1534         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1535         release SSA defs of pattern stmts.
1537 2014-02-17  Richard Biener  <rguenther@suse.de>
1539         * tree-inline.c (expand_call_inline): Release the virtual
1540         operand defined by the call we are about to inline.
1542 2014-02-17  Richard Biener  <rguenther@suse.de>
1544         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1546 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1547             Ilya Tocar  <ilya.tocar@intel.com>
1549         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1550         arguments order in builtin.
1551         (_mm512_permutexvar_epi64): Ditto.
1552         (_mm512_mask_permutexvar_epi64): Ditto
1553         (_mm512_maskz_permutexvar_epi32): Ditto
1554         (_mm512_permutexvar_epi32): Ditto
1555         (_mm512_mask_permutexvar_epi32): Ditto
1557 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1559         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1560         (p8_vmrgow): Likewise.
1562 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1564         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1565         endian targets.
1567 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1569         PR target/60203
1570         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1571         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1572         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1573         using direct move instructions on ISA 2.07.  Also adjust
1574         instruction length for 64-bit.
1575         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1576         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1578 2014-02-15  Alan Modra  <amodra@gmail.com>
1580         PR target/58675
1581         PR target/57935
1582         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1583         find_replacement on parts of insn rtl that might be reloaded.
1585 2014-02-15  Richard Biener  <rguenther@suse.de>
1587         PR tree-optimization/60183
1588         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1589         (tree_ssa_phiprop): Calculate and free post-dominators.
1591 2014-02-14  Jeff Law  <law@redhat.com>
1593         PR rtl-optimization/60131
1594         * ree.c (get_extended_src_reg): New function.
1595         (combine_reaching_defs): Use it rather than assuming location of REG.
1596         (find_and_remove_re): Verify first operand of extension is
1597         a REG before adding the insns to the copy list.
1599 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1601         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1602         * configure: Regenerated.
1603         * config.in: Regenerated.
1604         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1605         instead of ASM_SHORT.
1607 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1608             Richard Earnshaw  <rearnsha@arm.com>
1610         PR rtl-optimization/59535
1611         * lra-constraints.c (process_alt_operands): Encourage alternative
1612         when unassigned pseudo class is superset of the alternative class.
1613         (inherit_reload_reg): Don't inherit when optimizing for code size.
1614         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1615         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1616         modes not less than 4 for Thumb1.
1618 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1620         PR pch/60010
1621         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1623 2014-02-14  Richard Biener  <rguenther@suse.de>
1625         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1626         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1627         Do not strip INDIRECT_REFs.
1629 2014-02-14  Richard Biener  <rguenther@suse.de>
1631         PR lto/60179
1632         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1633         DECL_FUNCTION_SPECIFIC_TARGET.
1634         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1635         * tree-streamer-out.c (pack_ts_target_option): Remove.
1636         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1637         (write_ts_function_decl_tree_pointers): Do not stream
1638         DECL_FUNCTION_SPECIFIC_TARGET.
1639         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1640         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1641         (lto_input_ts_function_decl_tree_pointers): Do not stream
1642         DECL_FUNCTION_SPECIFIC_TARGET.
1644 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1646         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1647         (get_initial_def_for_induction, vectorizable_induction): Ignore
1648         debug stmts when looking for exit_phi.
1649         (vectorizable_live_operation): Fix up condition.
1651 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1653         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1654         nreverse() because it changes the content of original tree list.
1656 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1658         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1659         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1661 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1663         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1664         GNU coding standards.
1666 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1668         PR debug/60152
1669         * dwarf2out.c (gen_subprogram_die): Don't call
1670         add_calling_convention_attribute if subr_die is old_die.
1672 2014-02-13  Sharad Singhai  <singhai@google.com>
1674         * doc/optinfo.texi: Fix order of nodes.
1676 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1678         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1679         operands[2], not operands[3].
1681 2014-02-13  Richard Biener  <rguenther@suse.de>
1683         PR bootstrap/59878
1684         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1685         mention the possibility of an in-tree build.
1686         (CLooG): Update recommended version to 0.18.1, mention the
1687         possibility of an in-tree build and clarify that the ISL
1688         bundled with CLooG does not work.
1690 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1692         PR target/43546
1693         * expr.c (compress_float_constant): If x is a hard register,
1694         extend into a pseudo and then move to x.
1696 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1698         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1699         caused by bad second argument to warning_at() with -mhotpatch and
1700         nested functions (e.g. with gfortran).
1702 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1704         * opts.c (option_name): Remove "enabled by default" rider.
1706 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1708         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1710 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1711             Uros Bizjak  <ubizjak@gmail.com>
1713         PR target/60151
1714         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1715         * configure: Regenerated.
1717 2014-02-12  Richard Biener  <rguenther@suse.de>
1719         * vec.c (vec_prefix::calculate_allocation): Move as
1720         inline variant to vec.h.
1721         (vec_prefix::calculate_allocation_1): New out-of-line version.
1722         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1723         (vec_prefix::m_has_auto_buf): Rename to ...
1724         (vec_prefix::m_using_auto_storage): ... this.
1725         (vec_prefix::calculate_allocation): Inline the easy cases
1726         and dispatch to calculate_allocation_1 which doesn't need the
1727         prefix address.
1728         (va_heap::reserve): Use gcc_checking_assert.
1729         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1730         m_using_auto_storage.
1731         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1732         member and adjust.
1733         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1734         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1735         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1737 2014-02-12  Richard Biener  <rguenther@suse.de>
1739         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1740         when we found a dependence.
1742 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1744         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1745         common code...
1746         (maybe_fold_stmt): ... into this new function.
1747         * omp-low.c (lower_omp): Update comment.
1749         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1750         last use.
1752         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1753         dereference.
1755 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1757         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1758         identifiers in comments.
1759         (cortexa53_extra_costs): Likewise.
1760         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1761         (cortexa7_extra_costs): Likewise.
1762         (cortexa12_extra_costs): Likewise.
1763         (cortexa15_extra_costs): Likewise.
1764         (v7m_extra_costs): Likewise.
1766 2014-02-12  Richard Biener  <rguenther@suse.de>
1768         PR middle-end/60092
1769         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1770         of posix_memalign being successful.
1771         (lower_stmt): Restrict lowering of posix_memalign to when
1772         -ftree-bit-ccp is enabled.
1774 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1776         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1777         arg_loc.
1778         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1780 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1782         PR rtl-optimization/60116
1783         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1784         other_insn once the combination has been validated.
1786 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1788         PR lto/59468
1789         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1790         and wrapper.
1791         * ipa-devirt.c: Include demangle.h
1792         (odr_violation_reported): New static variable.
1793         (add_type_duplicate): Update odr_violations.
1794         (maybe_record_node): Add completep parameter; update it.
1795         (record_target_from_binfo): Add COMPLETEP parameter;
1796         update it as needed.
1797         (possible_polymorphic_call_targets_1): Likewise.
1798         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1799         rename FINAL to COMPLETE.
1800         (record_targets_from_bases): Sanity check we found the binfo;
1801         fix COMPLETEP updating.
1802         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1803         parameter, fix computing of COMPLETEP.
1804         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1805         at LTO time do demangling.
1806         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1807         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1808         parameter.
1809         (gimple_get_virt_method_for_binfo): Likewise.
1810         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1811         gimple_get_virt_method_for_vtable): Update prototypes.
1813 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1815         PR target/49008
1816         * genautomata.c (add_presence_absence): Fix typo with
1817         {final_}presence_list.
1819 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1821         PR target/60137
1822         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1823         for VSX/Altivec vectors that land in GPR registers.
1825 2014-02-11  Richard Henderson  <rth@redhat.com>
1826             Jakub Jelinek  <jakub@redhat.com>
1828         PR debug/59776
1829         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1830         around drhs if type conversion to lacc->type is not useless.
1832 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1834         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1835         tuning struct.
1836         (cortex-a57.cortex-a53): Likewise.
1837         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1839 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1841         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1842         arm_restrict_it.
1844 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1846         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1847         add_options_for_arm_vfp3.
1849 2014-02-11  Jeff Law  <law@redhat.com>
1851         PR middle-end/54041
1852         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1853         object with an undesirable mode.
1855 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1857         PR libgomp/60107
1858         * config/i386/sol2-9.h: New file.
1859         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1860         *-*-solaris2.9*): Use it.
1862 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1864         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1865         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1867 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1869         * config/microblaze/microblaze.c: Extend mcpu version format
1871 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
1873         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1875 2014-02-10  Richard Henderson  <rth@redhat.com>
1877         PR target/59927
1878         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1879         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1880         ms-abi vs -mno-accumulate-outgoing-args.
1881         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1882         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1883         respect to ms-abi.
1885 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1887         PR middle-end/60080
1888         * cfgexpand.c (expand_asm_operands): Attach source location to
1889         ASM_INPUT rtx objects.
1890         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1892 2014-02-10  Nick Clifton  <nickc@redhat.com>
1894         * config/mn10300/mn10300.c (popcount): New function.
1895         (mn10300_expand_prologue): Include saved registers in stack usage
1896         count.
1898 2014-02-10  Jeff Law  <law@redhat.com>
1900         PR middle-end/52306
1901         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1902         when changing the SET_DEST of a prior insn to avoid an input reload.
1904 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
1906         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1907         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1908         -mcall-openbsd, or -mcall-linux.
1909         (CC1_ENDIAN_BIG_SPEC): Remove.
1910         (CC1_ENDIAN_LITTLE_SPEC): Remove.
1911         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1912         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1913         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1914         and %cc1_endian_default.
1915         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1917 2014-02-10  Richard Biener  <rguenther@suse.de>
1919         PR tree-optimization/60115
1920         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1921         MEM_REF handling.  Properly verify that the accesses are not
1922         out of the objects bound.
1924 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1926         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1927         coretex to cortex.
1929 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
1931         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1932         proper constants and fix formatting.
1933         (possible_polymorphic_call_targets): Fix formatting.
1935 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
1936             Ilya Tocar  <ilya.tocar@intel.com>
1938         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1939         (_mm512_loadu_epi32): Renamed into...
1940         (_mm512_loadu_si512): This.
1941         (_mm512_storeu_epi32): Renamed into...
1942         (_mm512_storeu_si512): This.
1943         (_mm512_maskz_ceil_ps): Removed.
1944         (_mm512_maskz_ceil_pd): Ditto.
1945         (_mm512_maskz_floor_ps): Ditto.
1946         (_mm512_maskz_floor_pd): Ditto.
1947         (_mm512_floor_round_ps): Ditto.
1948         (_mm512_floor_round_pd): Ditto.
1949         (_mm512_ceil_round_ps): Ditto.
1950         (_mm512_ceil_round_pd): Ditto.
1951         (_mm512_mask_floor_round_ps): Ditto.
1952         (_mm512_mask_floor_round_pd): Ditto.
1953         (_mm512_mask_ceil_round_ps): Ditto.
1954         (_mm512_mask_ceil_round_pd): Ditto.
1955         (_mm512_maskz_floor_round_ps): Ditto.
1956         (_mm512_maskz_floor_round_pd): Ditto.
1957         (_mm512_maskz_ceil_round_ps): Ditto.
1958         (_mm512_maskz_ceil_round_pd): Ditto.
1959         (_mm512_expand_pd): Ditto.
1960         (_mm512_expand_ps): Ditto.
1961         * config/i386/i386.c (ix86_builtins): Remove
1962         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1963         (bdesc_args): Ditto.
1964         * config/i386/predicates.md (const1256_operand): New.
1965         (const_1_to_2_operand): Ditto.
1966         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1967         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1968         (*avx512pf_gatherpf<mode>sf): Ditto.
1969         (avx512pf_gatherpf<mode>df): Ditto.
1970         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1971         (*avx512pf_gatherpf<mode>df): Ditto.
1972         (avx512pf_scatterpf<mode>sf): Ditto.
1973         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1974         (*avx512pf_scatterpf<mode>sf): Ditto.
1975         (avx512pf_scatterpf<mode>df): Ditto.
1976         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1977         (*avx512pf_scatterpf<mode>df): Ditto.
1978         (avx512f_expand<mode>): Removed.
1979         (<shift_insn><mode>3<mask_name>): Change predicate type.
1981 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1983         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1984         not at the end of datarefs vector use ordered_remove to avoid
1985         reordering datarefs vector.
1987         PR c/59984
1988         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1989         mark local addressable non-static vars as GOVD_PRIVATE
1990         instead of GOVD_LOCAL.
1991         * omp-low.c (lower_omp_for): Move gimple_bind_vars
1992         and BLOCK_VARS of gimple_bind_block to new_stmt rather
1993         than copying them.
1995         PR middle-end/60092
1996         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
1997         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
1998         assume_aligned or alloc_align attributes.
1999         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2000         arguments.  Handle also assume_aligned and alloc_align attributes.
2001         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2002         calls to functions with assume_aligned or alloc_align attributes.
2003         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2005 2014-02-08  Terry Guo  <terry.guo@arm.com>
2007         * doc/invoke.texi: Document ARM -march=armv7e-m.
2009 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2011         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2012         flag on __cilkrts_rethrow builtin.
2014         PR ipa/60026
2015         * ipa-cp.c (determine_versionability): Fail at -O0
2016         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2017         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2019         Revert:
2020         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2022         PR ipa/60026
2023         * tree-inline.c (copy_forbidden): Fail for
2024         __attribute__((optimize (0))) functions.
2026 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2028         * varpool.c: Include pointer-set.h.
2029         (varpool_remove_unreferenced_decls): Variables in other partitions
2030         will not be output; be however careful to not lose information
2031         about partitioning.
2033 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2035         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2036         lookup in the vtable constructor.
2038 2014-02-07  Jeff Law  <law@redhat.com>
2040         PR target/40977
2041         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2042         define_insn_and_split.
2044         * ipa-inline.c (inline_small_functions): Fix typos.
2046 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2048         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2049         (s390_can_use_return_insn): Declare.
2050         * config/s390/s390.h (EPILOGUE_USES): Define.
2051         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2052         instructions.
2053         (s390_chunkify_start): Handle return JUMP_LABELs.
2054         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2055         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2056         (s390_can_use_return_insn): New functions.
2057         (s390_fix_long_loop_prediction): Handle conditional returns.
2058         (TARGET_SET_UP_BY_PROLOGUE): Define.
2059         * config/s390/s390.md (ANY_RETURN): New code iterator.
2060         (*creturn, *csimple_return, return, simple_return): New patterns.
2062 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2064         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2065         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2066         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2067         REG_CFA_RESTORE list when deciding not to restore a register.
2069 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2071         * config/s390/s390.c: Include tree-pass.h and context.h.
2072         (s390_early_mach): New function, split out from...
2073         (s390_emit_prologue): ...here.
2074         (pass_data_s390_early_mach): New pass structure.
2075         (pass_s390_early_mach): New class.
2076         (s390_option_override): Create and register early_mach pass.
2077         Move to end of file.
2079 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2081         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2082         to match for the exit block.
2084 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2086         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2087         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2088         Reject misaligned operands.
2090 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2092         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2094 2014-02-07  Richard Biener  <rguenther@suse.de>
2096         PR middle-end/60092
2097         * gimple-low.c (lower_builtin_posix_memalign): New function.
2098         (lower_stmt): Call it to lower posix_memalign in a way
2099         to make alignment info accessible.
2101 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2103         PR c++/60082
2104         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2105         __builtin_setjmp_receiver.
2107 2014-02-07  Richard Biener  <rguenther@suse.de>
2109         PR middle-end/60092
2110         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2111         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2112         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2113         Handle BUILT_IN_POSIX_MEMALIGN.
2114         (find_func_clobbers): Likewise.
2115         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2116         (call_may_clobber_ref_p_1): Likewise.
2118 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2120         PR ipa/59918
2121         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2122         sanity check.
2124 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2126         PR ipa/59469
2127         * lto-cgraph.c (lto_output_node): Use
2128         symtab_get_symbol_partitioning_class.
2129         (lto_output_varpool_node): likewise.
2130         (symtab_get_symbol_partitioning_class): Move here from
2131         lto/lto-partition.c
2132         * cgraph.h (symbol_partitioning_class): Likewise.
2133         (symtab_get_symbol_partitioning_class): Declare.
2135 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2137         * ggc.h (ggc_internal_cleared_alloc): New macro.
2138         * vec.h (vec_safe_copy): Handle memory stats.
2139         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2140         * target-globals.c (save_target_globals): Likewise.
2142 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2144         PR target/60077
2145         * expr.c (emit_move_resolve_push): Export; be bit more selective
2146         on when to clear alias set.
2147         * expr.h (emit_move_resolve_push): Declare.
2148         * function.h (struct function): Add tail_call_marked.
2149         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2150         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2151         * config/i386/i386.md (TImode move expander): De not call
2152         ix86_expand_push.
2153         (FP push expanders): Preserve memory attributes.
2154         * config/i386/sse.md (push<mode>1): Remove.
2155         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2156         (ix86_expand_push): Remove.
2157         * config/i386/mmx.md (push<mode>1): Remove.
2159 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2161         PR rtl-optimization/60030
2162         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2163         lopart with paradoxical subreg before shifting it up by hprec.
2165 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2167         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2168         Remove extra newline at end of file.
2169         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2170         (arm_issue_rate): Handle cortexa57.
2171         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2172         (cortex-a57.cortex-a53): Likewise.
2174 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2176         PR target/59575
2177         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2178         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2179         bitmask.
2180         (arm_expand_prologue): Adjust all callers.
2181         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2182         info, registers also at the lowest numbered registers side.  Use
2183         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2184         XEXP.
2186         PR debug/59992
2187         * var-tracking.c (adjust_mems): Before adding a SET to
2188         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2190 2014-02-06  Alan Modra  <amodra@gmail.com>
2192         PR target/60032
2193         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2194         change SDmode to DDmode when lra_in_progress.
2196 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2198         PR middle-end/59150
2199         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2200         free_data_ref on the dr first, and before goto again also set dr
2201         to the next dr.  For simd_lane_access, free old datarefs[i] before
2202         overwriting it.  For get_vectype_for_scalar_type failure, don't
2203         free_data_ref if simd_lane_access.
2205         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2207         PR target/60062
2208         * tree.h (opts_for_fn): New inline function.
2209         (opt_for_fn): Define.
2210         * config/i386/i386.c (ix86_function_regparm): Use
2211         opt_for_fn (decl, optimize) instead of optimize.
2213 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2215         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2216         for SYMBOL_REF in large memory model.
2218 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2220         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2221         and crypto support.
2222         (cortex-a57): Likewise.
2223         (cortex-a57.cortex-a53): Likewise.
2225 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2226             Kugan Vivekanandarajah  <kuganv@linaro.org>
2228         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2229         unaligned_access.
2230         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2232 2014-02-06  Richard Biener  <rguenther@suse.de>
2234         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2235         set_loop_copy and initialize_original_copy_tables.
2237 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2239         * config/aarch64/aarch64-simd.md
2240         (aarch64_ashr_simddi): Change QI to SI.
2242 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2243             Jakub Jelinek  <jakub@redhat.com>
2245         PR middle-end/60013
2246         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2247         of the dataflow.
2249 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2251         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2252         CODE_FOR_altivec_vpku[hw]um to
2253         CODE_FOR_altivec_vpku[hw]um_direct.
2254         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2255         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2256         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2257         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2259 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2261         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2262         generation for -maltivec=be.
2263         (altivec_vsumsws): Simplify redundant test.
2265 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2267         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2268         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2269         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2270         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2271         gen_altivec_vpkuwum.
2272         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2273         BYTES_BIG_ENDIAN.
2274         (altivec_vpks<VI_char>ss): Likewise.
2275         (altivec_vpks<VI_char>us): Likewise.
2276         (altivec_vpku<VI_char>us): Likewise.
2277         (altivec_vpku<VI_char>um): Likewise.
2278         (altivec_vpku<VI_char>um_direct): New (copy of
2279         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2280         internal use).
2281         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2282         target is little endian and -maltivec=be is not specified.
2283         (*altivec_vupkhs<VU_char>_direct): New (copy of
2284         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2285         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2286         target is little endian and -maltivec=be is not specified.
2287         (*altivec_vupkls<VU_char>_direct): New (copy of
2288         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2289         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2290         little endian and -maltivec=be is not specified.
2291         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2292         little endian and -maltivec=be is not specified.
2294 2014-02-05  Richard Henderson  <rth@redhat.com>
2296         PR debug/52727
2297         * combine-stack-adj.c: Revert r206943.
2298         * sched-int.h (struct deps_desc): Add last_args_size.
2299         * sched-deps.c (init_deps): Initialize it.
2300         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2301         contain REG_ARGS_SIZE notes.
2303 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2305         * lto-cgraph.c (asm_nodes_output): Make global.
2306         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2307         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2308         (driver_handle_option): Handle OPT_fwpa.
2310 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2312         PR ipa/59947
2313         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2314         a comment typo and formatting issue.  If odr_hash hasn't been
2315         created, return vNULL and set *completep to false.
2317         PR middle-end/57499
2318         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2319         bb with no successors.
2321 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2323         PR target/59718
2324         * doc/invoke.texi (-march): Clarify documentation for ARM.
2325         (-mtune): Likewise.
2326         (-mcpu): Likewise.
2328 2014-02-05  Richard Biener  <rguenther@suse.de>
2330         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2331         when not vectorizing because of too many alias checks.
2332         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2333         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2335 2014-02-05  Nick Clifton  <nickc@redhat.com>
2337         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2338         accept extended registers in any mode when compiling for the MN10300.
2340 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2342         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2343         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2344         sanitization attributes.
2345         (can_inline_edge_p): Likewise.
2346         (sanitize_attrs_match_for_inline_p): New function.
2348 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2350         * ipa-prop.c (detect_type_change): Shor circuit testing of
2351         type changes on THIS pointer.
2353 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2355         PR target/59777
2356         * config/pa/pa.c (legitimize_tls_address): Return original address
2357         if not passed a SYMBOL_REF rtx.
2358         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2359         addresses.
2360         (pa_emit_move_sequence): Simplify TLS source operands.
2361         (pa_legitimate_constant_p): Reject all TLS constants.
2362         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2363         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2365 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2367         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2368         groups when we know they are controlled by LTO.
2369         * varasm.c (default_binds_local_p_1): If object is in other partition,
2370         it will be resolved locally.
2372 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2374         * config/host-linux.c (linux_gt_pch_use_address): Don't
2375         use SSIZE_MAX because it is not always defined.
2377 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2379         PR bootstrap/59913
2380         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2381         threshold for pseudo splitting.
2382         (update_ebb_live_info): Process call argument hard registers and
2383         hard registers from insn definition too.
2384         (max_small_class_regs_num): New constant.
2385         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2386         reloads_num only for small register classes.  Don't split for
2387         outputs of jumps.
2389 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2391         PR ipa/60058
2392         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2393         is non-null.
2395 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2397         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2398         visibility is safe.
2400 2014-02-04  Marek Polacek  <polacek@redhat.com>
2402         * gdbinit.in (pel): Define.
2404 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2406         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2407         behavior.
2409 2014-02-04  Richard Biener  <rguenther@suse.de>
2411         PR lto/59723
2412         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2413         in function context local.
2414         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2415         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2416         similar to LTO_imported_decl_ref.
2418 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2420         PR tree-optimization/60002
2421         * cgraphclones.c (build_function_decl_skip_args): Clear
2422         DECL_LANG_SPECIFIC.
2424         PR tree-optimization/60023
2425         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2426         false to gsi_replace.
2427         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2428         has been in some EH region and vec_stmt could throw, add
2429         vec_stmt into the same EH region.
2430         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2431         has no lhs, ignore it.
2432         * internal-fn.c (expand_MASK_LOAD): Likewise.
2434         PR ipa/60026
2435         * tree-inline.c (copy_forbidden): Fail for
2436         __attribute__((optimize (0))) functions.
2438         PR other/58712
2439         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2440         is set, copy one less argument.
2441         (expand_simd_clones): Don't subtract clone_info->inbranch
2442         from simd_clone_struct_alloc argument.
2444         PR rtl-optimization/57915
2445         * recog.c (simplify_while_replacing): If all unary/binary/relational
2446         operation arguments are constant, attempt to simplify those.
2448         PR middle-end/59261
2449         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2450         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2452 2014-02-04  Richard Biener  <rguenther@suse.de>
2454         PR tree-optimization/60012
2455         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2456         TBAA disambiguation to all DDRs.
2458 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2460         PR target/59788
2461         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2462         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2464 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2466         PR ipa/59882
2467         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2469 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2471         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2472         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2474 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2476         PR ipa/59831
2477         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2478         to figure out targets of polymorphic calls with known decl.
2479         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2480         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2481         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2482         (get_polymorphic_call_info): ... here.
2483         (get_polymorphic_call_info_from_invariant): New function.
2485 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2487         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2488         lookup via vtable pointer; check for type consistency
2489         and turn inconsitent facts into UNREACHABLE.
2490         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2491         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2492         type inconsistent querries; return UNREACHABLE instead.
2494 2014-02-03  Richard Henderson  <rth@twiddle.net>
2496         PR tree-opt/59924
2497         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2498         already processed this node.
2499         (normalize_one_pred_1): Pass along mark_set.
2500         (normalize_one_pred): Create and destroy a pointer_set_t.
2501         (normalize_one_pred_chain): Likewise.
2503 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2505         PR gcov-profile/58602
2506         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2508 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2510         PR ipa/59831
2511         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2512         -fno-devirtualize; try to devirtualize by the knowledge of
2513         virtual table pointer given by aggregate propagation.
2514         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2515         (ipa_print_node_jump_functions): Dump also offset that
2516         is relevant for polymorphic calls.
2517         (determine_known_aggregate_parts): Add arg_type parameter; use it
2518         instead of determining the type from pointer type.
2519         (ipa_compute_jump_functions_for_edge): Update call of
2520         determine_known_aggregate_parts.
2521         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2522         (gimple_get_virt_method_for_binfo): ... here; simplify using
2523         vtable_pointer_value_to_vtable.
2524         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2525         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2526         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2527         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2528         POINTER_PLUS_EXPR.
2529         (vtable_pointer_value_to_binfo): ... here.
2530         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2532 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2534         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2535         redef of outer loop index variable.
2537 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2539         PR c++/53017
2540         PR c++/59211
2541         * doc/extend.texi (Function Attributes): Typo.
2543 2014-02-03  Cong Hou  <congh@google.com>
2545         PR tree-optimization/60000
2546         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2547         if the vectorized statement is a store.  A store statement can only
2548         appear at the end of pattern statements.
2550 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2552         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2553         (ix86_option_override_internal): Default long double to 64-bit for
2554         32-bit Bionic and to 128-bit for 64-bit Bionic.
2556         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2557         TARGET_LONG_DOUBLE_128 is true.
2558         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2560         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2561         (mlong-double-64): Negate -mlong-double-128.
2562         (mlong-double-128): New option.
2564         * config/i386/i386-c.c (ix86_target_macros): Define
2565         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2567         * doc/invoke.texi: Document -mlong-double-128.
2569 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2571         PR rtl-optimization/60024
2572         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2574 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2576         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2578 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2580         PR rtl-optimization/57662
2581         * sel-sched.c (code_motion_path_driver): Do not mark already not
2582         existing blocks in the visiting bitmap.
2584 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2586         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2587         on the insn being emitted.
2589 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2590             Will Deacon  <will.deacon@arm.com>
2592         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2594 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2596         * config/arm/arm-tables.opt: Regenerate.
2598 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2600         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2601         for vector types other than V16QImode.
2602         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2603         define_expand, and call altivec_expand_vec_perm_le when producing
2604         code with little endian element order.
2605         (*altivec_vperm_<mode>_internal): New insn having previous
2606         behavior of altivec_vperm_<mode>.
2607         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2608         altivec_expand_vec_perm_le when producing code with little endian
2609         element order.
2610         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2611         behavior of altivec_vperm_<mode>_uns.
2613 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2615         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2616         (altivec_vsumsws): Add handling for -maltivec=be with a little
2617         endian target.
2618         (altivec_vsumsws_direct): New.
2619         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2620         gen_altivec_vsumsws.
2622 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2624         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2625         vtable_pointer_value_to_binfo): New functions.
2626         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2627         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2629 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2631         * config/nios2/nios2.md (load_got_register): Initialize GOT
2632         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2633         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2635 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2637         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2638         preserverd by passthrough, do not propagate the type.
2640 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2642         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2643         (mips_atomic_assign_expand_fenv): New function.
2644         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2646 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2648         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2649         (__builtin_mips_set_fcsr): Likewise.
2650         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2651         MIPS_USI_FTYPE_VOID.
2652         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2653         (mips16_expand_set_fcsr): Likewise.
2654         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2655         (mips16_set_fcsr_stub): Likewise.
2656         (mips16_get_fcsr_one_only_stub): New class.
2657         (mips16_set_fcsr_one_only_stub): Likewise.
2658         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2659         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2660         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2661         (hard_float): New availability predicate.
2662         (mips_builtins): Add get_fcsr and set_fcsr.
2663         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2664         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2665         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2666         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2667         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2668         patterns.
2670 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2672         * config/mips/mips.c (mips_one_only_stub): New class.
2673         (mips_need_mips16_rdhwr_p): Replace with...
2674         (mips16_rdhwr_stub): ...this new variable.
2675         (mips16_stub_call_address): New function.
2676         (mips16_rdhwr_one_only_stub): New class.
2677         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2678         (mips_output_mips16_rdhwr): Delete.
2679         (mips_finish_stub): New function.
2680         (mips_code_end): Use it to handle rdhwr stubs.
2682 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2684         PR target/60017
2685         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2686         when calculating size of integer atomic types.
2688 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2690         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2692 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2694         PR tree-optimization/60003
2695         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2696         * profile.c (branch_prob): Use gimple_call_builtin_p
2697         to check for BUILT_IN_SETJMP_RECEIVER.
2698         * tree-inline.c (copy_bb): Call notice_special_calls.
2700 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2702         PR bootstrap/59985
2703         * lra-constraints.c (process_alt_operands): Update reload_sum only
2704         on the first pass.
2706 2014-01-31  Richard Henderson  <rth@redhat.com>
2708         PR middle-end/60004
2709         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2710         until after else_eh is processed.
2712 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2714         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2715         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2716         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2717         in smmintrin.h, remove them.
2718         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2719         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2720         * config/i386/i386.md (ROUND_SAE): Fix value.
2721         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2722         (const48_operand): New.
2723         * config/i386/subst.md (round), (round_expand): Use
2724         const_4_or_8_to_11_operand.
2725         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2727 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2729         * config/i386/constraints.md (Yk): Swap meaning with k.
2730         * config/i386/i386.md (movhi_internal): Change Yk to k.
2731         (movqi_internal): Ditto.
2732         (*k<logic><mode>): Ditto.
2733         (*andhi_1): Ditto.
2734         (*andqi_1): Ditto.
2735         (kandn<mode>): Ditto.
2736         (*<code>hi_1): Ditto.
2737         (*<code>qi_1): Ditto.
2738         (kxnor<mode>): Ditto.
2739         (kortestzhi): Ditto.
2740         (kortestchi): Ditto.
2741         (kunpckhi): Ditto.
2742         (*one_cmplhi2_1): Ditto.
2743         (*one_cmplqi2_1): Ditto.
2744         * config/i386/sse.md (): Change k to Yk.
2745         (avx512f_load<mode>_mask): Ditto.
2746         (avx512f_blendm<mode>): Ditto.
2747         (avx512f_store<mode>_mask): Ditto.
2748         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2749         (avx512f_storedqu<mode>_mask): Ditto.
2750         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2751         Ditto.
2752         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2753         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2754         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2755         (avx512f_maskcmp<mode>3): Ditto.
2756         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2757         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2758         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2759         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2760         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2761         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2762         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2763         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2764         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2765         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2766         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2767         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2768         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2769         (vec_extract_lo_<mode>_maskm): Ditto.
2770         (vec_extract_hi_<mode>_maskm): Ditto.
2771         (avx512f_vternlog<mode>_mask): Ditto.
2772         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2773         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2774         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2775         (avx512f_<code>v8div16qi2_mask): Ditto.
2776         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2777         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2778         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2779         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2780         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2781         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2782         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2783         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2784         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2785         (avx512cd_maskb_vec_dupv8di): Ditto.
2786         (avx512cd_maskw_vec_dupv16si): Ditto.
2787         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2788         (avx512f_vpermi2var<mode>3_mask): Ditto.
2789         (avx512f_vpermi2var<mode>3_mask): Ditto.
2790         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2791         (*avx512f_gathersi<mode>): Ditto.
2792         (*avx512f_gathersi<mode>_2): Ditto.
2793         (*avx512f_gatherdi<mode>): Ditto.
2794         (*avx512f_gatherdi<mode>_2): Ditto.
2795         (*avx512f_scattersi<mode>): Ditto.
2796         (*avx512f_scatterdi<mode>): Ditto.
2797         (avx512f_compress<mode>_mask): Ditto.
2798         (avx512f_compressstore<mode>_mask): Ditto.
2799         (avx512f_expand<mode>_mask): Ditto.
2800         * config/i386/subst.md (mask): Change k to Yk.
2801         (mask_scalar_merge): Ditto.
2802         (sd): Ditto.
2804 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2806         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2808 2014-01-31  Richard Biener  <rguenther@suse.de>
2810         PR middle-end/59990
2811         * builtins.c (fold_builtin_memory_op): Make sure to not
2812         use a floating-point mode or a boolean or enumeral type for
2813         the copy operation.
2815 2014-01-30  DJ Delorie  <dj@redhat.com>
2817         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2818         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2819         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2820         whenever main() has an epilogue.
2822 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2824         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2825         unused variable "field".
2826         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2827         (vsx_mergeh_<mode>): Likewise.
2828         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2829         (altivec_vmrghh): Likewise.
2830         (altivec_vmrghw): Likewise.
2831         (altivec_vmrglb): Likewise.
2832         (altivec_vmrglh): Likewise.
2833         (altivec_vmrglw): Likewise.
2834         (altivec_vspltb): Add missing uses.
2835         (altivec_vsplth): Likewise.
2836         (altivec_vspltw): Likewise.
2837         (altivec_vspltsf): Likewise.
2839 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2841         PR target/59923
2842         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2843         frame related instructions.
2845 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2847         PR rtl-optimization/59959
2848         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2849         any reload of register whose subreg is invalid.
2851 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2853         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2854         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2855         Add missing return type - void.
2857 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2859         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2860         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2861         remove element index adjustment for endian (now handled in vsx.md
2862         and altivec.md).
2863         (altivec_expand_vec_perm_const): Use
2864         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2865         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2866         (vsx_xxspltw_<mode>): Adjust element index for little endian.
2867         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2868         define_expand and a new define_insn *altivec_vspltb_internal;
2869         adjust for -maltivec=be on a little endian target.
2870         (altivec_vspltb_direct): New.
2871         (altivec_vsplth): Divide into a define_expand and a new
2872         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2873         little endian target.
2874         (altivec_vsplth_direct): New.
2875         (altivec_vspltw): Divide into a define_expand and a new
2876         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2877         little endian target.
2878         (altivec_vspltw_direct): New.
2879         (altivec_vspltsf): Divide into a define_expand and a new
2880         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2881         a little endian target.
2883 2014-01-30  Richard Biener  <rguenther@suse.de>
2885         PR tree-optimization/59993
2886         * tree-ssa-forwprop.c (associate_pointerplus): Check we
2887         can propagate form the earlier stmt and avoid the transform
2888         when the intermediate result is needed.
2890 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
2892         * README.Portability: Fix typo.
2894 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
2896         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2897         comparison_operator with ordered_comparison_operator.
2899 2014-01-30  Nick Clifton  <nickc@redhat.com>
2901         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2902         Rename to mn10300_store_multiple_regs.
2903         * config/mn10300/mn10300.c: Likewise.
2904         * config/mn10300/mn10300.md (store_movm): Fix typo: call
2905         store_multiple_regs.
2906         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2907         Call mn10300_store_multiple_regs.
2909 2014-01-30  Nick Clifton  <nickc@redhat.com>
2910             DJ Delorie  <dj@redhat.com>
2912         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2913         %fp 2 to keep registers after it properly word-aligned.
2914         (rl78_alloc_physical_registers_umul): Handle the case where both
2915         input operands are the same.
2917 2014-01-30  Richard Biener  <rguenther@suse.de>
2919         PR tree-optimization/59903
2920         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2921         check properly.
2923 2014-01-30  Jason Merrill  <jason@redhat.com>
2925         PR c++/59633
2926         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2928         PR c++/59645
2929         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2931 2014-01-30  Richard Biener  <rguenther@suse.de>
2933         PR tree-optimization/59951
2934         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2936 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
2938         PR target/59784
2939         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2940         SFmode to DFmode case.
2942 2014-01-29  DJ Delorie  <dj@redhat.com>
2944         * config/msp430/msp430.opt (-minrt): New.
2945         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2946         if -minrt given.
2947         (ENDFILE_SPEC): Likewise.
2949 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
2951         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2952         (estimate_function_body_sizes): Use it.
2954 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
2956         PR c++/58561
2957         * dwarf2out.c (is_cxx_auto): New.
2958         (is_base_type): Use it.
2959         (gen_type_die_with_usage): Likewise.
2961 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2963         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
2964         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2965         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2966         -maltivec=be with LE targets.
2967         (vsx_mergeh_<mode>): Likewise.
2968         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2969         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2970         (altivec_vmrghb): Replace with define_expand and new
2971         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2972         (altivec_vmrghb_direct): New define_insn.
2973         (altivec_vmrghh): Replace with define_expand and new
2974         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2975         (altivec_vmrghh_direct): New define_insn.
2976         (altivec_vmrghw): Replace with define_expand and new
2977         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2978         (altivec_vmrghw_direct): New define_insn.
2979         (*altivec_vmrghsf): Adjust for endianness.
2980         (altivec_vmrglb): Replace with define_expand and new
2981         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2982         (altivec_vmrglb_direct): New define_insn.
2983         (altivec_vmrglh): Replace with define_expand and new
2984         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2985         (altivec_vmrglh_direct): New define_insn.
2986         (altivec_vmrglw): Replace with define_expand and new
2987         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2988         (altivec_vmrglw_direct): New define_insn.
2989         (*altivec_vmrglsf): Adjust for endianness.
2990         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2991         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2992         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2993         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2994         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2995         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2996         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2997         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2999 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3001         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3002         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3003         whitespace.
3005 2014-01-29  Richard Biener  <rguenther@suse.de>
3007         PR tree-optimization/58742
3008         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3009         associate_pointerplus_align.
3010         (associate_pointerplus_diff): New function.
3011         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3012         and associate_pointerplus_diff.
3014 2014-01-29  Richard Biener  <rguenther@suse.de>
3016         * lto-streamer.h (LTO_major_version): Bump to 3.
3017         (LTO_minor_version): Reset to 0.
3019 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3021         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3022         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3023         (arm_file_start): Generate correct asm header for armv7ve.
3024         * config/arm/bpabi.h: Add multilib support for armv7ve.
3025         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3026         and cortex-a15 to armv7ve.
3027         * config/arm/t-aprofile: Add multilib support for armv7ve.
3028         * doc/invoke.texi: Document -march=armv7ve.
3030 2014-01-29  Richard Biener  <rguenther@suse.de>
3032         PR tree-optimization/58742
3033         * tree-ssa-forwprop.c (associate_plusminus): Return true
3034         if we changed sth, defer EH cleanup to ...
3035         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3036         (simplify_mult): New function.
3038 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3040         PR middle-end/59917
3041         PR tree-optimization/59920
3042         * tree.c (build_common_builtin_nodes): Remove
3043         __builtin_setjmp_dispatcher initialization.
3044         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3045         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3046         instead of gsi_after_labels + manually skipping debug stmts.
3047         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3048         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3049         * tree-inline.c (copy_edges_for_bb): Remove
3050         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3051         argument.  Ignore computed_goto_p stmts.  Don't call
3052         make_abnormal_goto_edges.  If a call might need abnormal edges
3053         for non-local gotos, see if it already has an edge to
3054         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3055         with true argument, don't do anything then, otherwise add
3056         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3057         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3058         caller.
3059         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3060         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3061         (lower_stmt): Don't set data->calls_builtin_setjmp.
3062         (lower_builtin_setjmp): Adjust comment.
3063         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3064         * tree-cfg.c (found_computed_goto): Remove.
3065         (factor_computed_gotos): Remove.
3066         (make_goto_expr_edges): Return bool, true for computed gotos.
3067         Don't call make_abnormal_goto_edges.
3068         (build_gimple_cfg): Don't set found_computed_goto, don't call
3069         factor_computed_gotos.
3070         (computed_goto_p): No longer static.
3071         (make_blocks): Don't set found_computed_goto.
3072         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3073         (make_edges): If make_goto_expr_edges returns true, push bb
3074         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3075         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3076         vector.  Record mapping between bbs and OpenMP regions if there
3077         are any, adjust make_gimple_omp_edges caller.  Call
3078         handle_abnormal_edges.
3079         (make_abnormal_goto_edges): Remove.
3080         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3081         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3082         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3083         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3084         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3085         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3086         filling *region also set *region_idx to (*region)->entry->index.
3088         PR other/58712
3089         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3090         For REGs set ORIGINAL_REGNO.
3092 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3094         * doc/md.texi: Mention that a target shouldn't implement
3095         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3096         than hi/lo pair.
3098 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3100         PR tree-optimization/59594
3101         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3102         a copy of the datarefs vector rather than the vector itself.
3104 2014-01-28  Jason Merrill  <jason@redhat.com>
3106         PR c++/53756
3107         * dwarf2out.c (auto_die): New static.
3108         (gen_type_die_with_usage): Handle C++1y 'auto'.
3109         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3110         on definition.
3112 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3114         PR target/59672
3115         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3116         (SPEC_X32): Likewise.
3117         (SPEC_64): Likewise.
3118         * config/i386/i386.c (ix86_option_override_internal): Turn off
3119         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3120         for TARGET_16BIT.
3121         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3122         * config/i386/i386.h (TARGET_16BIT): New macro.
3123         (TARGET_16BIT_P): Likewise.
3124         * config/i386/i386.opt: Add m16.
3125         * doc/invoke.texi: Document -m16.
3127 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3129         PR preprocessor/59935
3130         * input.c (location_get_source_line): Bail out on when line number
3131         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3133 2014-01-28  Richard Biener  <rguenther@suse.de>
3135         PR tree-optimization/58742
3136         * tree-ssa-forwprop.c (associate_plusminus): Handle
3137         pointer subtraction of the form (T)(P + A) - (T)P.
3139 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3141         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3142         at const_int_cost.
3144 2014-01-28  Richard Biener  <rguenther@suse.de>
3146         Revert
3147         2014-01-28  Richard Biener  <rguenther@suse.de>
3149         PR rtl-optimization/45364
3150         PR rtl-optimization/59890
3151         * var-tracking.c (local_get_addr_clear_given_value): Handle
3152         already cleared slot.
3153         (val_reset): Handle not allocated local_get_addr_cache.
3154         (vt_find_locations): Use post-order on the inverted CFG.
3156 2014-01-28  Richard Biener  <rguenther@suse.de>
3158         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3160 2014-01-28  Richard Biener  <rguenther@suse.de>
3162         PR rtl-optimization/45364
3163         PR rtl-optimization/59890
3164         * var-tracking.c (local_get_addr_clear_given_value): Handle
3165         already cleared slot.
3166         (val_reset): Handle not allocated local_get_addr_cache.
3167         (vt_find_locations): Use post-order on the inverted CFG.
3169 2014-01-28  Alan Modra  <amodra@gmail.com>
3171         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3172         * configure.ac <recursive call for build != host>: Define
3173         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3174         and LD_FOR_BUILD too.
3175         * configure: Regenerate.
3177 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3179         * config/i386/i386.c (get_builtin_code_for_version): Separate
3180         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3181         Broadwell from Haswell.
3183 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3185         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3186         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3187         * config/mips/mips.c (mips_option_override): Change setting
3188         of TARGET_DSP.
3189         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3190         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3191         Change from Mask to Var.
3193 2014-01-27  Jeff Law  <law@redhat.com>
3195         * ipa-inline.c (inline_small_functions): Fix typo.
3197 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3199         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3200         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3201         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3202         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3203         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3204         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3205         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3206         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3207         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3208         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3209         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3210         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3211         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3212         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3213         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3214         (_mm512_storeu_epi64): Ditto.
3215         (_mm512_cmpge_epi32_mask): Ditto.
3216         (_mm512_cmpge_epu32_mask): Ditto.
3217         (_mm512_cmpge_epi64_mask): Ditto.
3218         (_mm512_cmpge_epu64_mask): Ditto.
3219         (_mm512_cmple_epi32_mask): Ditto.
3220         (_mm512_cmple_epu32_mask): Ditto.
3221         (_mm512_cmple_epi64_mask): Ditto.
3222         (_mm512_cmple_epu64_mask): Ditto.
3223         (_mm512_cmplt_epi32_mask): Ditto.
3224         (_mm512_cmplt_epu32_mask): Ditto.
3225         (_mm512_cmplt_epi64_mask): Ditto.
3226         (_mm512_cmplt_epu64_mask): Ditto.
3227         (_mm512_cmpneq_epi32_mask): Ditto.
3228         (_mm512_cmpneq_epu32_mask): Ditto.
3229         (_mm512_cmpneq_epi64_mask): Ditto.
3230         (_mm512_cmpneq_epu64_mask): Ditto.
3231         (_mm512_expand_pd): Ditto.
3232         (_mm512_expand_ps): Ditto.
3233         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3234         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3235         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3236         * config/i386/i386.c (ix86_builtins): Add
3237         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3238         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3239         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3240         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3241         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3242         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3243         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3244         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3245         IX86_BUILTIN_PMOVUSQW512_MEM.
3246         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3247         __builtin_ia32_pmovsqd512mem_mask,
3248         __builtin_ia32_pmovqd512mem_mask,
3249         __builtin_ia32_pmovusqw512mem_mask,
3250         __builtin_ia32_pmovsqw512mem_mask,
3251         __builtin_ia32_pmovqw512mem_mask,
3252         __builtin_ia32_pmovusdw512mem_mask,
3253         __builtin_ia32_pmovsdw512mem_mask,
3254         __builtin_ia32_pmovdw512mem_mask,
3255         __builtin_ia32_pmovqb512mem_mask,
3256         __builtin_ia32_pmovusqb512mem_mask,
3257         __builtin_ia32_pmovsqb512mem_mask,
3258         __builtin_ia32_pmovusdb512mem_mask,
3259         __builtin_ia32_pmovsdb512mem_mask,
3260         __builtin_ia32_pmovdb512mem_mask.
3261         (bdesc_args): Add __builtin_ia32_expanddf512,
3262         __builtin_ia32_expandsf512.
3263         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3264         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3265         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3266         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3267         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3268         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3269         (avx512f_<code>v8div16qi2_mask_store): This.
3270         (avx512f_expand<mode>): New.
3272 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3274         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3275         New.
3276         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3277         (_mm512_prefetch_i32scatter_pd): Ditto.
3278         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3279         (_mm512_prefetch_i64scatter_pd): Ditto.
3280         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3281         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3282         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3283         (_mm512_prefetch_i32scatter_ps): Ditto.
3284         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3285         (_mm512_prefetch_i64scatter_ps): Ditto.
3286         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3287         * config/i386/i386-builtin-types.def: Define
3288         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3289         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3290         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3291         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3292         IX86_BUILTIN_SCATTERPFQPD.
3293         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3294         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3295         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3296         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3297         __builtin_ia32_scatterpfqps.
3298         (ix86_expand_builtin): Expand new built-ins.
3299         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3300         fix memory access data type.
3301         (*avx512pf_gatherpf<mode>_mask): Ditto.
3302         (*avx512pf_gatherpf<mode>): Ditto.
3303         (avx512pf_scatterpf<mode>): Ditto.
3304         (*avx512pf_scatterpf<mode>_mask): Ditto.
3305         (*avx512pf_scatterpf<mode>): Ditto.
3306         (GATHER_SCATTER_SF_MEM_MODE): New.
3307         (avx512pf_gatherpf<mode>df): Ditto.
3308         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3309         (*avx512pf_scatterpf<mode>df): Ditto.
3311 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3313         PR bootstrap/59934
3314         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3315         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3316         reached.
3318 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3320         * common/config/arm/arm-common.c
3321         (arm_rewrite_mcpu): Handle multiple names.
3322         * config/arm/arm.h
3323         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3325 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3327         * gimple-builder.h (create_gimple_tmp): Delete.
3329 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3331         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3332         words comparisons.
3334 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3336         * config/pa/pa.md (call): Generate indirect long calls to non-local
3337         functions when outputing 32-bit code.
3338         (call_value): Likewise except for special call to buggy powf function.
3340         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3341         portable runtime and PIC indirect calls.
3342         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3343         and PIC call sequences.  Use ldo instead of blr to set return register
3344         in PIC call sequence.
3346 2014-01-25  Walter Lee  <walt@tilera.com>
3348         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3349         avoid clobbering a live register.
3351 2014-01-25  Walter Lee  <walt@tilera.com>
3353         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3354         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3355         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3356         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3358 2014-01-25  Walter Lee  <walt@tilera.com>
3360         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3361         arguments on even registers.
3362         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3363         STACK_BOUNDARY.
3364         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3365         (BIGGEST_ALIGNMENT): Ditto.
3366         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3368 2014-01-25  Walter Lee  <walt@tilera.com>
3370         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3371         insns before bundling.
3372         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3374 2014-01-25  Walter Lee  <walt@tilera.com>
3376         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3377         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3378         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3380 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3382         * config/mips/constraints.md (kl): Delete.
3383         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3384         define expands, using...
3385         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3386         instructions for MIPS16.
3387         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3388         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3390 2014-01-25  Walter Lee  <walt@tilera.com>
3392         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3393         (clzdi2): Ditto.
3394         (ffsdi2): Ditto.
3396 2014-01-25  Walter Lee  <walt@tilera.com>
3398         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3399         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3401 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3403         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3404         Handle XOR.
3406 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3408         * print-rtl.c (in_call_function_usage): New var.
3409         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3410         EXPR_LIST mode as mode and not as reg note name.
3412         PR middle-end/59561
3413         * cfgloopmanip.c (copy_loop_info): If
3414         loop->warned_aggressive_loop_optimizations, make sure
3415         the flag is set in target loop too.
3417 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3419         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3420         flag_cilkplus.
3421         * builtins.def: Likewise.
3422         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3423         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3424         * ira.c (ira_setup_eliminable_regset): Likewise.
3425         * omp-low.c (gate_expand_omp): Likewise.
3426         (execute_lower_omp): Likewise.
3427         (diagnose_sb_0): Likewise.
3428         (gate_diagnose_omp_blocks): Likewise.
3429         (simd_clone_clauses_extract): Likewise.
3430         (gate): Likewise.
3432 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3434         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3435         correction for little endian...
3436         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3437         here.
3439 2014-01-24  Jeff Law  <law@redhat.com>
3441         PR tree-optimization/59919
3442         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3443         for non-returning calls.
3445 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3447         * common/config/aarch64/aarch64-common.c
3448         (aarch64_rewrite_mcpu): Handle multiple names.
3449         * config/aarch64/aarch64.h
3450         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3452 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3454         * input.c (add_file_to_cache_tab): Handle the case where fopen
3455         returns NULL.
3457 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3459         PR target/59929
3460         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3461         from push operand if code of push isn't PRE_DEC.
3463 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3465         PR target/59909
3466         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3467         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3468         it is only used for non-atomic loads/stores.
3470         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3471         -mquad-memory or -mquad-memory-atomic switches.
3473         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3474         -mquad-memory-atomic to ISA 2.07 support.
3476         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3477         to separate support of normal quad word memory operations (ldq, stq)
3478         from the atomic quad word memory operations.
3480         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3481         support to separate non-atomic quad word operations from atomic
3482         quad word operations.  Disable non-atomic quad word operations in
3483         little endian mode so that we don't have to swap words after the
3484         load and before the store.
3485         (quad_load_store_p): Add comment about atomic quad word support.
3486         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3487         options printed with -mdebug=reg.
3489         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3490         -mquad-memory-atomic as the test for whether we have quad word
3491         atomic instructions.
3492         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3493         or -mp8-vector are used, allow byte/half-word atomic operations.
3495         * config/rs6000/sync.md (load_lockedti): Insure that the address
3496         is a proper indexed or indirect address for the lqarx instruction.
3497         On little endian systems, swap the hi/lo registers after the lqarx
3498         instruction.
3499         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3500         insure the address is valid for the lqarx instruction.
3501         (store_conditionalti): Insure that the address is a proper indexed
3502         or indirect address for the stqcrx. instruction.  On little endian
3503         systems, swap the hi/lo registers before doing the stqcrx.
3504         instruction.
3505         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3506         insure the address is valid for the stqcrx. instruction.
3508         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3509         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3510         type of quad memory support is available.
3512 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3514         PR regression/59915
3515         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3516         there is a danger of looping.
3518 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3520         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3521         force flag_ira_loop_pressure if set via command line.
3523 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3525         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3526         (ashr_simd): New builtin handling DI mode.
3527         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3528         (aarch64_sshr_simddi): New match pattern.
3529         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3530         (vshrd_n_s64): Likewise.
3531         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3533 2014-01-23  Nick Clifton  <nickc@redhat.com>
3535         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3536         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3537         favour of mcu specific scripts.
3538         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3539         430x multilibs.
3541 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3542             Alex Velenko  <Alex.Velenko@arm.com>
3544         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3545         (vaddv_s16): Likewise.
3546         (vaddv_s32): Likewise.
3547         (vaddv_u8): Likewise.
3548         (vaddv_u16): Likewise.
3549         (vaddv_u32): Likewise.
3550         (vaddvq_s8): Likewise.
3551         (vaddvq_s16): Likewise.
3552         (vaddvq_s32): Likewise.
3553         (vaddvq_s64): Likewise.
3554         (vaddvq_u8): Likewise.
3555         (vaddvq_u16): Likewise.
3556         (vaddvq_u32): Likewise.
3557         (vaddvq_u64): Likewise.
3558         (vaddv_f32): Likewise.
3559         (vaddvq_f32): Likewise.
3560         (vaddvq_f64): Likewise.
3561         (vmaxv_f32): Likewise.
3562         (vmaxv_s8): Likewise.
3563         (vmaxv_s16): Likewise.
3564         (vmaxv_s32): Likewise.
3565         (vmaxv_u8): Likewise.
3566         (vmaxv_u16): Likewise.
3567         (vmaxv_u32): Likewise.
3568         (vmaxvq_f32): Likewise.
3569         (vmaxvq_f64): Likewise.
3570         (vmaxvq_s8): Likewise.
3571         (vmaxvq_s16): Likewise.
3572         (vmaxvq_s32): Likewise.
3573         (vmaxvq_u8): Likewise.
3574         (vmaxvq_u16): Likewise.
3575         (vmaxvq_u32): Likewise.
3576         (vmaxnmv_f32): Likewise.
3577         (vmaxnmvq_f32): Likewise.
3578         (vmaxnmvq_f64): Likewise.
3579         (vminv_f32): Likewise.
3580         (vminv_s8): Likewise.
3581         (vminv_s16): Likewise.
3582         (vminv_s32): Likewise.
3583         (vminv_u8): Likewise.
3584         (vminv_u16): Likewise.
3585         (vminv_u32): Likewise.
3586         (vminvq_f32): Likewise.
3587         (vminvq_f64): Likewise.
3588         (vminvq_s8): Likewise.
3589         (vminvq_s16): Likewise.
3590         (vminvq_s32): Likewise.
3591         (vminvq_u8): Likewise.
3592         (vminvq_u16): Likewise.
3593         (vminvq_u32): Likewise.
3594         (vminnmv_f32): Likewise.
3595         (vminnmvq_f32): Likewise.
3596         (vminnmvq_f64): Likewise.
3598 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3600         * config/aarch64/aarch64-simd.md
3601         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3602         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3603         (*aarch64_mul3_elt<mode>): Likewise.
3604         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3605         (*aarch64_mul3_elt_to_64v2df): Likewise.
3606         (*aarch64_mla_elt<mode>): Likewise.
3607         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3608         (*aarch64_mls_elt<mode>): Likewise.
3609         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3610         (*aarch64_fma4_elt<mode>): Likewise.
3611         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3612         (*aarch64_fma4_elt_to_64v2df): Likewise.
3613         (*aarch64_fnma4_elt<mode>): Likewise.
3614         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3615         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3616         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3617         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3618         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3619         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3620         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3621         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3622         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3624 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3626         * config/aarch64/aarch64-simd.md
3627         (aarch64_be_checked_get_lane<mode>): New define_expand.
3628         * config/aarch64/aarch64-simd-builtins.def
3629         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3630         New builtin definition.
3631         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3632         Use new safe be builtin.
3634 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3636         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3637         New define_insn.
3638         (aarch64_be_st1<mode>): Likewise.
3639         (aarch_ld1<VALL:mode>): Define_expand modified.
3640         (aarch_st1<VALL:mode>): Likewise.
3641         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3642         (UNSPEC_ST1): Likewise.
3644 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3646         * config/microblaze/microblaze.md: Add trap insn and attribute
3648 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3650         PR preprocessor/58580
3651         * input.h (location_get_source_line): Take an additional line_size
3652         parameter.
3653         (void diagnostics_file_cache_fini): Declare new function.
3654         * input.c (struct fcache): New type.
3655         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3656         New static constants.
3657         (diagnostic_file_cache_init, total_lines_num)
3658         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3659         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3660         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3661         (get_next_line, read_next_line, goto_next_line, read_line_num):
3662         New static function definitions.
3663         (diagnostic_file_cache_fini): New function.
3664         (location_get_source_line): Take an additional output line_len
3665         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3666         read_line_num.
3667         * diagnostic.c (diagnostic_finish): Call
3668         diagnostic_file_cache_fini.
3669         (adjust_line): Take an additional input parameter for the length
3670         of the line, rather than calculating it with strlen.
3671         (diagnostic_show_locus): Adjust the use of
3672         location_get_source_line and adjust_line with respect to their new
3673         signature.  While displaying a line now, do not stop at the first
3674         null byte.  Rather, display the zero byte as a space and keep
3675         going until we reach the size of the line.
3676         * Makefile.in: Add vec.o to OBJS-libcommon
3678 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3679             Ilya Tocar     <ilya.tocar@intel.com>
3681         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3682         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3683         (__builtin_ia32_kmov16): Ditto.
3684         * config/i386/i386.md (UNSPEC_KMOV): New.
3685         (kmovw): Ditto.
3687 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3689         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3690         (_mm512_storeu_si512): Ditto.
3692 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3694         PR target/52125
3695         * rtl.h (get_referenced_operands): Declare.
3696         * recog.c (get_referenced_operands): New function.
3697         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3698         operands have been referenced when recording LO_SUM references.
3700 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3702         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3704 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3706         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3707         Enable for generic and recent AMD targets.
3709 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3711         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3712         ARG_SIZE note when adjustment was eliminated.
3714 2014-01-22  Jeff Law  <law@redhat.com>
3716         PR tree-optimization/59597
3717         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3718         in file.  Accept new argument REGISTERING and use it to modify
3719         dump output appropriately.
3720         (register_jump_thread): Corresponding changes.
3721         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3722         thread paths involving joiner blocks.  Add code to dump cancelled
3723         jump threading paths.
3725 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3727         PR rtl-optimization/59477
3728         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3729         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3731 2014-01-22  Tom Tromey  <tromey@redhat.com>
3733         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3734         PARAMS.
3735         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3737 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3739         PR rtl-optimization/59896
3740         * lra-constraints.c (process_alt_operands): Check unused note for
3741         matched operands of insn with no output reloads.
3743 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3745         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3746         (mips_move_from_gpr_cost): Likewise.
3748 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3750         PR rtl-optimization/59858
3751         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3752         ira_class_hard_regs_num.
3753         (process_alt_operands): Increase reject for dying matched operand.
3755 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3757         PR target/59003
3758         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3759         smaller than size, perform several stores or loads and stores
3760         at dst + count - size to store or copy all of size bytes, rather
3761         than just last modesize bytes.
3763 2014-01-20  DJ Delorie  <dj@redhat.com>
3765         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3766         that CLOBBERs are REGs before propogating their values.
3768 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3770         PR middle-end/59789
3771         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3772         (cgraph_inline_failed_type): New function.
3773         * cgraph.h (DEFCIFCODE): Add type.
3774         (cgraph_inline_failed_type_t): New enum.
3775         (cgraph_inline_failed_type): New prototype.
3776         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3777         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3778         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3779         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3780         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3781         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3782         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3783         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3784         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3785         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3786         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3787         OPTIMIZATION_MISMATCH.
3788         * tree-inline.c (expand_call_inline): Emit errors during
3789         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3791 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3793         PR target/59685
3794         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3795         mode attribute in insn output.
3797 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3799         * output.h (output_constant): Delete.
3800         * varasm.c (output_constant): Make private.
3802 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3804         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3806 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3808         PR middle-end/59860
3809         * tree.h (fold_builtin_strcat): New prototype.
3810         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3811         argument, if non-NULL, don't call c_strlen.  Optimize
3812         directly into __builtin_memcpy instead of __builtin_strcpy.
3813         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3814         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3816 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3818         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3819         for SImode_address_operand operands, having only a REG argument.
3821 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3823         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3824         loader name using mbig-endian.
3825         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3827 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3829         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3830         (-mtune): Likewise.
3831         (-mcpu): Likewise.
3833 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3835         * config/aarch64/aarch64-protos.h
3836         (aarch64_cannot_change_mode_class_ptr): Declare.
3837         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3838         aarch64_cannot_change_mode_class_ptr): New.
3839         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3840         backend hook aarch64_cannot_change_mode_class.
3842 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3844         * common/config/aarch64/aarch64-common.c
3845         (aarch64_handle_option): Don't handle any option order logic here.
3846         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3847         selected_cpu, warn on architecture version mismatch.
3848         (aarch64_override_options): Fix parsing order for option strings.
3850 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3851             Iain Sandoe  <iain@codesourcery.com>
3853         PR bootstrap/59496
3854         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3855         warning.  Amend comment to reflect current functionality.
3857 2014-01-20  Richard Biener  <rguenther@suse.de>
3859         PR middle-end/59860
3860         * builtins.c (fold_builtin_strcat): Remove case better handled
3861         by tree-ssa-strlen.c.
3863 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
3865         * config/aarch64/aarch64.opt
3866         (mcpu, march, mtune): Make case-insensitive.
3868 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3870         PR target/59880
3871         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3872         if operands[1] is a REG or ZERO_EXTEND of a REG.
3874 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
3876         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3878 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
3880         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3881         long non-pic millicode calls.
3883 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3885         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3887 2014-01-19  Kito Cheng  <kito@0xlab.org>
3889         * builtins.c (expand_movstr): Check movstr expand done or fail.
3891 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3892             H.J. Lu  <hongjiu.lu@intel.com>
3894         PR target/59379
3895         * config/i386/i386.md (*lea<mode>): Zero-extend return register
3896         to DImode for zero-extended addresses.
3898 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
3900         PR rtl-optimization/57763
3901         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3902         on the new indirect jump_insn and increment LABEL_NUSES (label).
3904 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
3906         PR bootstrap/59580
3907         PR bootstrap/59583
3908         * config.gcc (x86_archs): New variable.
3909         (x86_64_archs): Likewise.
3910         (x86_cpus): Likewise.
3911         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3912         --with-arch/--with-cpu= options.
3913         Support --with-arch=/--with-cpu={nehalem,westmere,
3914         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3916 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3918         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3919         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
3921 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3923         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3925 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
3927         PR target/58944
3928         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3929         clear cpp_get_options (parse_in)->warn_unused_macros for
3930         ix86_target_macros_internal with cpp_define.
3932 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
3934         * jump.c (delete_related_insns): Keep (use (insn))s.
3935         * reorg.c (redundant_insn): Check for barriers too.
3937 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3939         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3941 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
3943         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3944         call to $$dyncall when TARGET_LONG_CALLS is true.
3946 2014-01-17  Jeff Law  <law@redhat.com>
3948         * ree.c (combine_set_extension): Temporarily disable test for
3949         changing number of hard registers.
3951 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3953         PR middle-end/58125
3954         * ipa-inline-analysis.c (inline_free_summary):
3955         Do not free summary of aliases.
3957 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3959         PR middle-end/59706
3960         * gimplify.c (gimplify_expr): Use create_tmp_var
3961         instead of create_tmp_var_raw.  If cond doesn't have
3962         integral type, don't add the IFN_ANNOTATE builtin at all.
3964 2014-01-17  Martin Jambor  <mjambor@suse.cz>
3966         PR ipa/59736
3967         * ipa-cp.c (prev_edge_clone): New variable.
3968         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3969         Also resize prev_edge_clone vector.
3970         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3971         (ipcp_edge_removal_hook): New function.
3972         (ipcp_driver): Register ipcp_edge_removal_hook.
3974 2014-01-17  Andrew Pinski <apinski@cavium.com>
3975             Steve Ellcey  <sellcey@mips.com>
3977         PR target/59462
3978         * config/mips/mips.c (mips_print_operand): Check operand mode instead
3979         of operator mode.
3981 2014-01-17  Jeff Law  <law@redhat.com>
3983         PR middle-end/57904
3984         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3985         so that pass_ccp runs first.
3987 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3989         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3990         (ix86_adjust_cost): Use !TARGET_XXX.
3991         (do_reorder_for_imul): Likewise.
3992         (swap_top_of_ready_list): Likewise.
3993         (ix86_sched_reorder): Likewise.
3995 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3997         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3998         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
3999         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4000         (intel_memset): New.  Duplicate slm_memset.
4001         (intel_cost): New.  Duplicate slm_cost.
4002         (m_INTEL): New macro.
4003         (processor_target_table): Add "intel".
4004         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4005         with PROCESSOR_INTEL for "intel".
4006         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4007         PROCESSOR_SILVERMONT.
4008         (ix86_issue_rate): Likewise.
4009         (ix86_adjust_cost): Likewise.
4010         (ia32_multipass_dfa_lookahead): Likewise.
4011         (swap_top_of_ready_list): Likewise.
4012         (ix86_sched_reorder): Likewise.
4013         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4014         instead of TARGET_OPT_AGU.
4015         * config/i386/i386.h (TARGET_INTEL): New.
4016         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4017         (processor_type): Add PROCESSOR_INTEL.
4018         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4019         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4021 2014-01-17  Marek Polacek  <polacek@redhat.com>
4023         PR c/58346
4024         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4025         size is zero.
4027 2014-01-17  Richard Biener  <rguenther@suse.de>
4029         PR tree-optimization/46590
4030         * opts.c (default_options_table): Add entries for
4031         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4032         all enabled at -O1 but not for -Og.
4033         * common.opt (fbranch-count-reg): Remove Init(1).
4034         (fmove-loop-invariants): Likewise.
4035         (ftree-pta): Likewise.
4037 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4039         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4040         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4041         decls to at least the GCC 4.8 used alignments.
4043         PR fortran/59440
4044         * tree-nested.c (convert_nonlocal_reference_stmt,
4045         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4046         of GIMPLE_BIND stmts, adjust associated decls.
4048 2014-01-17  Richard Biener  <rguenther@suse.de>
4050         PR tree-optimization/46590
4051         * vec.h (vec<>::bseach): New member function implementing
4052         binary search according to C89 bsearch.
4053         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4054         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4055         bitmap pointer again.  Make accesses_in_loop a flat array.
4056         (mem_ref_obstack): New global.
4057         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4058         (mark_ref_stored): Likewise.
4059         (ref_indep_loop_p_2): Likewise.
4060         (set_ref_stored_in_loop): New helper function.
4061         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4062         (memref_free): Adjust.
4063         (record_mem_ref_loc): Simplify.
4064         (gather_mem_refs_stmt): Adjust.
4065         (sort_locs_in_loop_postorder_cmp): New function.
4066         (analyze_memory_references): Sort accesses_in_loop after
4067         loop postorder number.
4068         (find_ref_loc_in_loop_cmp): New function.
4069         (for_all_locs_in_loop): Find relevant cluster of locs in
4070         accesses_in_loop and iterate without recursion.
4071         (execute_sm): Avoid uninit warning.
4072         (struct ref_always_accessed): Simplify.
4073         (ref_always_accessed::operator ()): Likewise.
4074         (ref_always_accessed_p): Likewise.
4075         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4076         loop postorder numbers here.
4077         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4078         numbers.
4080 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4082         PR c++/57945
4083         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4084         on decls for which assemble_alias has been called.
4086 2014-01-17  Nick Clifton  <nickc@redhat.com>
4088         * config/msp430/msp430.opt: (mcpu): New option.
4089         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4090         (msp430_option_override): Parse target_cpu.  If the MCU name
4091         matches a generic string, clear target_mcu.
4092         (msp430_attr): Allow numeric interrupt values up to 63.
4093         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4094         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4095         option.
4096         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4097         Add mcpu matches.
4098         * config/msp430/msp430.md (popm): Use %J rather than %I.
4099         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4100         (addhi_cy_i): Use immediate_operand for operand 2.
4101         * doc/invoke.texi: Document -mcpu option.
4103 2014-01-17  Richard Biener  <rguenther@suse.de>
4105         PR rtl-optimization/38518
4106         * df.h (df_analyze_loop): Declare.
4107         * df-core.c: Include cfgloop.h.
4108         (df_analyze_1): Split out main part of df_analyze.
4109         (df_analyze): Adjust.
4110         (loop_inverted_post_order_compute): New function.
4111         (loop_post_order_compute): Likewise.
4112         (df_analyze_loop): New function avoiding whole-function
4113         postorder computes.
4114         * loop-invariant.c (find_defs): Use df_analyze_loop.
4115         (find_invariants): Adjust.
4116         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4118 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4120         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4121         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4123 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4125         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4126         traversal when removing references.
4128 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4130         PR ipa/59775
4131         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4133 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4135         PR middle-end/56791
4136         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4137         pushing a reload for an autoinc when we had previously reloaded an
4138         inner part of the address.
4140 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4142         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4143         field.
4144         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4145         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4146         when not giving up or versioning for alias only because of
4147         loop->safelen.
4148         (vect_analyze_data_ref_dependences): Set to true.
4149         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4150         is a GIMPLE_PHI.
4151         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4152         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4153         to the condition.
4155         PR middle-end/58344
4156         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4158         PR target/59839
4159         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4160         operand 0 predicate for gathers, use a new pseudo as subtarget.
4162 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4164         PR middle-end/59609
4165         * lra-constraints.c (process_alt_operands): Add printing debug info.
4166         Check absence of input/output reloads for matched operands too.
4168 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4170         PR rtl-optimization/59835
4171         * ira.c (ira_init_register_move_cost): Increase cost for
4172         impossible modes.
4174 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4176         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4178 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4180         PR target/59780
4181         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4182         non-register objects.  Use gen_(high/low)part more consistently.
4183         Fix assertions.
4185 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4187         PR target/59844
4188         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4189         endian support, remove tests for WORDS_BIG_ENDIAN.
4190         (p8_mfvsrd_3_<mode>): Likewise.
4191         (reload_gpr_from_vsx<mode>): Likewise.
4192         (reload_gpr_from_vsxsf): Likewise.
4193         (p8_mfvsrd_4_disf): Likewise.
4195 2014-01-16  Richard Biener  <rguenther@suse.de>
4197         PR rtl-optimization/46590
4198         * lcm.c (compute_antinout_edge): Use postorder iteration.
4199         (compute_laterin): Use inverted postorder iteration.
4201 2014-01-16  Nick Clifton  <nickc@redhat.com>
4203         PR middle-end/28865
4204         * varasm.c (output_constant): Return the number of bytes actually
4205         emitted.
4206         (output_constructor_array_range): Update the field size with the
4207         number of bytes emitted by output_constant.
4208         (output_constructor_regular_field): Likewise.  Also do not
4209         complain if the total number of bytes emitted is now greater
4210         than the expected fieldpos.
4211         * output.h (output_constant): Update prototype and descriptive comment.
4213 2014-01-16  Marek Polacek  <polacek@redhat.com>
4215         PR middle-end/59827
4216         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4217         it is error_mark_node.
4219 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4221         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4222         VALID_AVX256_REG_OR_OI_MODE.
4224 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4226         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4227         current procedure should be profiled.
4229 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4231         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4232         of moving from/to the STACK_REG register class.
4234 2014-01-15  Richard Henderson  <rth@redhat.com>
4236         PR debug/54694
4237         * reginfo.c (global_regs_decl): Globalize.
4238         * rtl.h (global_regs_decl): Declare.
4239         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4240         reserved via global_regs.
4242 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4244         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4246 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4248         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4249         and vmulosh rather than call gen_vec_widen_smult_*.
4250         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4251         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4252         (vec_widen_smult_even_v16qi): Likewise.
4253         (vec_widen_umult_even_v8hi): Likewise.
4254         (vec_widen_smult_even_v8hi): Likewise.
4255         (vec_widen_umult_odd_v16qi): Likewise.
4256         (vec_widen_smult_odd_v16qi): Likewise.
4257         (vec_widen_umult_odd_v8hi): Likewise.
4258         (vec_widen_smult_odd_v8hi): Likewise.
4259         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4260         vmuloub rather than call gen_vec_widen_umult_*.
4261         (vec_widen_umult_lo_v16qi): Likewise.
4262         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4263         vmulosb rather than call gen_vec_widen_smult_*.
4264         (vec_widen_smult_lo_v16qi): Likewise.
4265         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4266         rather than call gen_vec_widen_umult_*.
4267         (vec_widen_umult_lo_v8hi): Likewise.
4268         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4269         rather than call gen_vec_widen_smult_*.
4270         (vec_widen_smult_lo_v8hi): Likewise.
4272 2014-01-15  Jeff Law  <law@redhat.com>
4274         PR tree-optimization/59747
4275         * ree.c (find_and_remove_re): Properly handle case where a second
4276         eliminated extension requires widening a copy created for elimination
4277         of a prior extension.
4278         (combine_set_extension): Ensure that the number of hard regs needed
4279         for a destination register does not change when we widen it.
4281 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4283         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4284         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4285         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4286         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4287         (avr-*-rtems*): Likewise.
4288         (bfin*-rtems*): Likewise.
4289         (moxie-*-rtems*): Likewise.
4290         (h8300-*-rtems*): Likewise.
4291         (i[34567]86-*-rtems*): Likewise.
4292         (lm32-*-rtems*): Likewise.
4293         (m32r-*-rtems*): Likewise.
4294         (m68k-*-rtems*): Likewise.
4295         (microblaze*-*-rtems*): Likewise.
4296         (mips*-*-rtems*): Likewise.
4297         (powerpc-*-rtems*): Likewise.
4298         (sh-*-rtems*): Likewise.
4299         (sparc-*-rtems*): Likewise.
4300         (sparc64-*-rtems*): Likewise.
4301         (v850-*-rtems*): Likewise.
4302         (m32c-*-rtems*): Likewise.
4304 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4306         PR rtl-optimization/59511
4307         * ira.c (ira_init_register_move_cost): Use memory costs for some
4308         cases of register move cost calculations.
4309         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4310         instead of BB frequency.
4311         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4312         * lra-assigns.c (find_hard_regno_for): Ditto.
4314 2014-01-15  Richard Biener  <rguenther@suse.de>
4316         PR tree-optimization/59822
4317         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4318         (vectorizable_load): Use it to hoist defs of uses of invariant
4319         loads out of the loop.
4321 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4322             Kugan Vivekanandarajah  <kuganv@linaro.org>
4324         PR target/59695
4325         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4326         truncation.
4328 2014-01-15  Richard Biener  <rguenther@suse.de>
4330         PR rtl-optimization/59802
4331         * lcm.c (compute_available): Use inverted postorder to seed
4332         the initial worklist.
4334 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4336         PR target/59803
4337         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4338         ADDR_REGS for invalid symrefs in non-PIC code.
4340 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4342         PR other/58712
4343         * builtins.c (determine_block_size): Initialize *probable_max_size
4344         even if len_rtx is CONST_INT.
4346 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4348         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4349         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4350         (cortexa53_tunings): Likewise.
4351         (aarch64_sched_issue_rate): New function.
4352         (TARGET_SCHED_ISSUE_RATE): Define.
4354 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4356         * ira-costs.c (find_costs_and_classes): Add missed
4357         ira_init_register_move_cost_if_necessary.
4359 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4361         PR target/59787
4362         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4364 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4366         PR target/59794
4367         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4368         to indicate if type is used for function return value.  Warn ABI
4369         change if the vector mode isn't available for function return value.
4370         (ix86_function_arg_advance): Pass false to type_natural_mode.
4371         (ix86_function_arg): Likewise.
4372         (ix86_gimplify_va_arg): Likewise.
4373         (function_arg_32): Don't warn ABI change.
4374         (ix86_function_value): Pass true to type_natural_mode.
4375         (ix86_return_in_memory): Likewise.
4376         (ix86_struct_value_rtx): Removed.
4377         (TARGET_STRUCT_VALUE_RTX): Likewise.
4379 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4381         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4382         converting a conditional jump into a conditional return.
4384 2014-01-14  Richard Biener  <rguenther@suse.de>
4386         PR tree-optimization/58921
4387         PR tree-optimization/59006
4388         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4389         hoisting invariant stmts.
4390         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4391         invariant loads on the preheader edge if possible.
4393 2014-01-14  Joey Ye  <joey.ye@arm.com>
4395         * doc/plugin.texi (Building GCC plugins): Update to C++.
4397 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4399         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4400         (_mm_rcp28_round_ss): Ditto.
4401         (_mm_rsqrt28_round_sd): Ditto.
4402         (_mm_rsqrt28_round_ss): Ditto.
4403         (_mm_rcp28_sd): Ditto.
4404         (_mm_rcp28_ss): Ditto.
4405         (_mm_rsqrt28_sd): Ditto.
4406         (_mm_rsqrt28_ss): Ditto.
4407         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4408         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4409         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4410         (IX86_BUILTIN_RCP28SD): Ditto.
4411         (IX86_BUILTIN_RCP28SS): Ditto.
4412         (IX86_BUILTIN_RSQRT28SD): Ditto.
4413         (IX86_BUILTIN_RSQRT28SS): Ditto.
4414         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4415         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4416         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4417         (ix86_expand_special_args_builtin): Expand new FTYPE.
4418         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4419         (srcp14<mode>): Make insn unary.
4420         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4421         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4422         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4423         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4424         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4425         Fix rounding: make it SAE only.
4426         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4427         Ditto.
4428         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4429         Ditto.
4430         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4431         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4432         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4433         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4434         (round_saeonly_mask_scalar_operand4): Ditto.
4435         (round_saeonly_mask_scalar_op3): Ditto.
4436         (round_saeonly_mask_scalar_op4): Ditto.
4438 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4440         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4441         Implement -maltivec=be for vec_insert and vec_extract.
4443 2014-01-10  DJ Delorie  <dj@redhat.com>
4445         * config/msp430/msp430.md (call_internal): Don't allow memory
4446         references with SP as the base register.
4447         (call_value_internal): Likewise.
4448         * config/msp430/constraints.md (Yc): New.  For memory references
4449         that don't use SP as a base register.
4451         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4452         "an integer without a # prefix"
4453         * config/msp430/msp430.md (epilogue_helper): Use it.
4455 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4457         PR target/59617
4458         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4459         AVX512F gather builtins.
4460         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4461         on gather decls with INTEGER_TYPE masktype.
4462         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4463         directly into the builtin rather than hoisting it before loop.
4465         PR tree-optimization/59387
4466         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4467         (scev_const_prop): If folded_casts and type has undefined overflow,
4468         use force_gimple_operand instead of force_gimple_operand_gsi and
4469         for each added stmt if it is assign with
4470         arith_code_with_undefined_signed_overflow, call
4471         rewrite_to_defined_overflow.
4472         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4473         gimple-fold.h instead.
4474         (arith_code_with_undefined_signed_overflow,
4475         rewrite_to_defined_overflow): Moved to ...
4476         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4477         rewrite_to_defined_overflow): ... here.  No longer static.
4478         Include gimplify-me.h.
4479         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4480         rewrite_to_defined_overflow): New prototypes.
4482 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4484         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4486 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4488         * builtins.c (get_object_alignment_2): Minor tweak.
4489         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4491 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4493         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4494         optimized non constant lengths.
4496 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4498         PR libgomp/59194
4499         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4500         load as __atomic_load_N if possible.
4502 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4504         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4505         target parameter.
4506         (rs6000_expand_builtin): Adjust call.
4508 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4510         PR target/58115
4511         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4512         * config/rs6000/rs6000.c: Include target-globals.h.
4513         (rs6000_set_current_function): Instead of doing target_reinit
4514         unconditionally, use save_target_globals_default_opts and
4515         restore_target_globals.
4517         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4518         FPSCR.
4519         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4520         (rs6000_expand_builtin): Handle mffs and mtfsf.
4521         (rs6000_init_builtins): Define mffs and mtfsf.
4522         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4523         (rs6000_mffs): New pattern.
4524         (rs6000_mtfsf): New pattern.
4526 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4528         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4529         Start narrowing with START.  Apply candidate-use pair
4530         and check overall cost in narrowing.
4531         (iv_ca_prune): Pass new argument.
4533 2014-01-10  Jeff Law  <law@redhat.com>
4535         PR middle-end/59743
4536         * ree.c (combine_reaching_defs): Ensure the defining statement
4537         occurs before the extension when optimizing extensions with
4538         different source and destination hard registers.
4540 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4542         PR ipa/58585
4543         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4544         vtables into the type inheritance graph.
4546 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4548         PR rtl-optimization/59754
4549         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4550         modes in the REGNO != REGNO case.
4552 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4554         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4556 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4558         PR tree-optimization/59745
4559         * tree-predcom.c (tree_predictive_commoning_loop): Call
4560         free_affine_expand_cache if giving up because components is NULL.
4562         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4563         GC in payload of target_globals struct instead of allocating them on
4564         the heap and the larger structs separately using GC.
4565         * target-globals.h (struct target_globals): Make regs, hard_regs,
4566         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4567         of GTY((skip)) and change type to void *.
4568         (reset_target_globals): Cast loads from those fields to corresponding
4569         types.
4571 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4573         PR plugins/59335
4574         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4575         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4576         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4578 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4580         PR target/59744
4581         * aarch64-modes.def (CC_Zmode): New flags mode.
4582         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4583         represents an equality.
4584         (aarch64_get_condition_code): Handle CC_Zmode.
4585         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4587 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4589         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4590         extraction in good case.
4592 2014-01-10  Richard Biener  <rguenther@suse.de>
4594         PR tree-optimization/59374
4595         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4596         checking after SLP discovery.  Mark stmts not participating
4597         in any SLP instance properly.
4599 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4601         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4602         when handling a SET rtx.
4604 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4606         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4607         (cortex-a57): Likewise.
4608         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4610 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4612         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4613         non-iwmmxt builtins.
4615 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4617         PR ipa/58252
4618         PR ipa/59226
4619         * ipa-devirt.c record_target_from_binfo): Take as argument
4620         stack of binfos and lookup matching one for virtual inheritance.
4621         (possible_polymorphic_call_targets_1): Update.
4623 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4625         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4626         kernel strings for Loongson-2E/2F/3A.
4628 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4630         PR middle-end/59670
4631         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4632         is_gimple_call before calling gimple_call_internal_p.
4634 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4636         * Makefile.in (TREE_FLOW_H): Remove.
4637         (TREE_SSA_H): Add file names from tree-flow.h.
4638         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4639         * tree.h: Remove tree-flow.h reference.
4640         * hash-table.h: Remove tree-flow.h reference.
4641         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4642         reference with tree-ssa-loop.h.
4644 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4646         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4647         default element-order behavior for -maltivec.
4648         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4649         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4650         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4651         when targeting big endian, at least for now.
4652         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4654 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4656         PR middle-end/47735
4657         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4658         var satisfies use_register_for_decl, just take into account type
4659         alignment, rather than decl alignment.
4661         PR tree-optimization/59622
4662         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4663         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4664         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4665         Don't devirtualize for inplace at all.  For targets.length () == 1,
4666         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4668 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4670         * config/i386/i386.md (cpu): Remove the unused btver1.
4672 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4674         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4676 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4678         PR target/58115
4679         * tree-core.h (struct target_globals): New forward declaration.
4680         (struct tree_target_option): Add globals field.
4681         * tree.h (TREE_TARGET_GLOBALS): Define.
4682         (prepare_target_option_nodes_for_pch): New prototype.
4683         * target-globals.h (struct target_globals): Define even if
4684         !SWITCHABLE_TARGET.
4685         * tree.c (prepare_target_option_node_for_pch,
4686         prepare_target_option_nodes_for_pch): New functions.
4687         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4688         * config/i386/i386.c: Include target-globals.h.
4689         (ix86_set_current_function): Instead of doing target_reinit
4690         unconditionally, use save_target_globals_default_opts and
4691         restore_target_globals.
4693 2014-01-09  Richard Biener  <rguenther@suse.de>
4695         PR tree-optimization/59715
4696         * tree-cfg.h (split_critical_edges): Declare.
4697         * tree-cfg.c (split_critical_edges): Export.
4698         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4700 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4702         * cfgexpand.c (expand_stack_vars): Optionally disable
4703         asan stack protection.
4704         (expand_used_vars): Likewise.
4705         (partition_stack_vars): Likewise.
4706         * asan.c (asan_emit_stack_protection): Optionally disable
4707         after return stack usage.
4708         (instrument_derefs): Optionally disable memory access instrumentation.
4709         (instrument_builtin_call): Likewise.
4710         (instrument_strlen_call): Likewise.
4711         (asan_protect_global): Optionally disable global variables protection.
4712         * doc/invoke.texi: Added doc for new options.
4713         * params.def: Added new options.
4714         * params.h: Likewise.
4716 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4718         PR rtl-optimization/59724
4719         * ifcvt.c (cond_exec_process_if_block): Don't call
4720         flow_find_head_matching_sequence with 0 longest_match.
4721         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4722         non-active insns if !stop_after.
4723         (try_head_merge_bb): Revert 2014-01-07 changes.
4725 2014-01-08  Jeff Law  <law@redhat.com>
4727         * ree.c (get_sub_rtx): New function, extracted from...
4728         (merge_def_and_ext): Here.
4729         (combine_reaching_defs): Use get_sub_rtx.
4731 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4733         * cgraph.h (varpool_variable_node): Do not choke on null node.
4735 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4737         * config/mips/mips.md (simple_return): Attempt to use JRC
4738         for microMIPS.
4739         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4741 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4743         PR rtl-optimization/59137
4744         * reorg.c (steal_delay_list_from_target): Call update_block for
4745         elided insns.
4746         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4748 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4750         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4751         two duplicate entries.
4753 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4755         Revert:
4756         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4758         * config/mips/mips.c (mips_truncated_op_cost): New function.
4759         (mips_rtx_costs): Adjust test for BADDU.
4760         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4762         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4764         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4765         (*baddu_si): ...this new pattern.
4767 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4769         PR ipa/59722
4770         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4772 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4774         PR middle-end/57748
4775         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4776         inner_reference_p.
4777         (expand_expr, expand_normal): Adjust.
4778         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4779         inner_reference_p. Use inner_reference_p to expand inner references.
4780         (store_expr): Adjust.
4781         * cfgexpand.c (expand_call_stmt): Adjust.
4783 2014-01-08  Rong Xu  <xur@google.com>
4785         * gcov-io.c (gcov_var): Move from gcov-io.h.
4786         (gcov_position): Ditto.
4787         (gcov_is_error): Ditto.
4788         (gcov_rewrite): Ditto.
4789         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4790         only part to libgcc/libgcov.h.
4792 2014-01-08  Marek Polacek  <polacek@redhat.com>
4794         PR middle-end/59669
4795         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4797 2014-01-08  Marek Polacek  <polacek@redhat.com>
4799         PR sanitizer/59667
4800         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4802 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4804         PR rtl-optimization/59649
4805         * stor-layout.c (get_mode_bounds): For BImode return
4806         0 and STORE_FLAG_VALUE.
4808 2014-01-08  Richard Biener  <rguenther@suse.de>
4810         PR middle-end/59630
4811         * gimple.h (is_gimple_builtin_call): Remove.
4812         (gimple_builtin_call_types_compatible_p): New.
4813         (gimple_call_builtin_p): New overload.
4814         * gimple.c (is_gimple_builtin_call): Remove.
4815         (validate_call): Rename to ...
4816         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4817         check return types.
4818         (validate_type): New static function.
4819         (gimple_call_builtin_p): New overload and adjust.
4820         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4821         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4822         (gimple_fold_stmt_to_constant_1): Likewise.
4823         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4825 2014-01-08  Richard Biener  <rguenther@suse.de>
4827         PR middle-end/59471
4828         * gimplify.c (gimplify_expr): Gimplify register-register type
4829         VIEW_CONVERT_EXPRs to separate stmts.
4831 2014-01-07  Jeff Law  <law@redhat.com>
4833         PR middle-end/53623
4834         * ree.c (combine_set_extension): Handle case where source
4835         and destination registers in an extension insn are different.
4836         (combine_reaching_defs): Allow source and destination registers
4837         in extension to be different under limited circumstances.
4838         (add_removable_extension): Remove restriction that the
4839         source and destination registers in the extension are the same.
4840         (find_and_remove_re): Emit a copy from the extension's
4841         destination to its source after the defining insn if
4842         the source and destination registers are different.
4844         PR middle-end/59285
4845         * ifcvt.c (merge_if_block): If we are merging a block with more than
4846         one successor with a block with no successors, remove any BARRIER
4847         after the second block.
4849 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4851         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4853 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4855         PR target/59652
4856         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4857         for 14-bit register offsets when INT14_OK_STRICT is false.
4859 2014-01-07  Roland Stigge  <stigge@antcom.de>
4860             Michael Meissner  <meissner@linux.vnet.ibm.com>
4862         PR 57386/target
4863         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4864         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
4866 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
4868         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4869         -mcpu.
4871 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
4873         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4874         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4875         rtx is const0_rtx or not.
4877 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
4879         PR target/58115
4880         * target-globals.c (save_target_globals): Remove this_fn_optab
4881         handling.
4882         * toplev.c: Include optabs.h.
4883         (target_reinit): Temporarily restore the global options if another
4884         set of options are in force.
4886 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
4888         PR rtl-optimization/58668
4889         * cfgcleanup.c (flow_find_cross_jump): Don't count
4890         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
4891         to determine what is counted.
4892         (flow_find_head_matching_sequence): Use active_insn_p to determine
4893         what is counted.
4894         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4895         counting change.
4896         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4897         determine what is counted.
4899         PR tree-optimization/59643
4900         * tree-predcom.c (split_data_refs_to_components): If one dr is
4901         read and one write, determine_offset fails and the write isn't
4902         in the bad component, just put the read into the bad component.
4904 2014-01-07  Mike Stump  <mikestump@comcast.net>
4905             Jakub Jelinek  <jakub@redhat.com>
4907         PR pch/59436
4908         * tree-core.h (struct tree_optimization_option): Change optabs
4909         type from unsigned char * to void *.
4910         * optabs.c (init_tree_optimization_optabs): Adjust
4911         TREE_OPTIMIZATION_OPTABS initialization.
4913 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
4915         PR target/59644
4916         * config/i386/i386.h (struct machine_function): Add
4917         no_drap_save_restore field.
4918         * config/i386/i386.c (ix86_save_reg): Use
4919         !cfun->machine->no_drap_save_restore instead of
4920         crtl->stack_realign_needed.
4921         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4922         this function clears frame_pointer_needed.  Set
4923         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4924         and DRAP reg is needed.
4926 2014-01-06  Marek Polacek  <polacek@redhat.com>
4928         PR c/57773
4929         * doc/implement-c.texi: Mention that other integer types are
4930         permitted as bit-field types in strictly conforming mode.
4932 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
4934         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4935         is newly allocated.
4937 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
4939         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4941 2014-01-06  Martin Jambor  <mjambor@suse.cz>
4943         PR ipa/59008
4944         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4945         to int.
4946         * ipa-prop.c (ipa_print_node_params): Fix indentation.
4948 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
4950         PR debug/59350
4951         PR debug/59510
4952         * var-tracking.c (add_stores): Preserve the value of the source even if
4953         we don't record the store.
4955 2014-01-06  Terry Guo  <terry.guo@arm.com>
4957         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4959 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
4961         PR bootstrap/59541
4962         * config/darwin.c (darwin_function_section): Adjust return values to
4963         correspond to optimisation changes made in r206070.
4965 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
4967         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4968         from prefetch_block tune setting.
4969         (nocona_cost): Correct size of prefetch block to 64.
4971 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
4973         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4974         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4975         used to save the static chain register in the computation of the offset
4976         from which the FP registers need to be restored.
4978 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
4980         PR tree-optimization/59519
4981         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4982         ICE if get_current_def (current_new_name) is already non-NULL, as long
4983         as it is a phi result of some other phi in *new_exit_bb that has
4984         the same argument.
4986         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4987         or vmovdqu* for misaligned_operand.
4988         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4989         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4990         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4991         aligned_mem for AVX512F masked aligned load and store builtins and for
4992         non-temporal moves.
4994 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
4996         PR tree-optimization/59651
4997         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
4998         Address range for negative step should be added by TYPE_SIZE_UNIT.
5000 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5002         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5003         overlapping registers also for registers other than the stack pointer.
5005 2014-01-03  Marek Polacek  <polacek@redhat.com>
5007         PR other/59661
5008         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5009         __builtin_FILE.
5011 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5013         PR target/59625
5014         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5015         asm goto as jump.
5017         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5018         (push splitter): Use <P:MODE_SIZE> instead of
5019         GET_MODE_SIZE (<P:MODE>mode).
5020         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5021         (mov -1, reg peephole2): Likewise.
5022         * config/i386/sse.md (*mov<mode>_internal,
5023         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5024         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5025         *<code><mode>3, *andnot<mode>3<mask_name>,
5026         <mask_codefor><code><mode>3<mask_name>): Likewise.
5027         * config/i386/subst.md (mask_mode512bit_condition,
5028         sd_mask_mode512bit_condition): Likewise.
5030 2014-01-02  Xinliang David Li  <davidxl@google.com>
5032         PR tree-optimization/59303
5033         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5034         (dump_predicates): Better output format.
5035         (pred_equal_p): New function.
5036         (is_neq_relop_p): Ditto.
5037         (is_neq_zero_form_p): Ditto.
5038         (pred_expr_equal_p): Ditto.
5039         (pred_neg_p): Ditto.
5040         (simplify_pred): Ditto.
5041         (simplify_preds_2): Ditto.
5042         (simplify_preds_3): Ditto.
5043         (simplify_preds_4): Ditto.
5044         (simplify_preds): Ditto.
5045         (push_pred): Ditto.
5046         (push_to_worklist): Ditto.
5047         (get_pred_info_from_cmp): Ditto.
5048         (is_degenerated_phi): Ditto.
5049         (normalize_one_pred_1): Ditto.
5050         (normalize_one_pred): Ditto.
5051         (normalize_one_pred_chain): Ditto.
5052         (normalize_preds): Ditto.
5053         (normalize_cond_1): Remove function.
5054         (normalize_cond): Ditto.
5055         (is_gcond_subset_of): Ditto.
5056         (is_subset_of_any): Ditto.
5057         (is_or_set_subset_of): Ditto.
5058         (is_and_set_subset_of): Ditto.
5059         (is_norm_cond_subset_of): Ditto.
5060         (pred_chain_length_cmp): Ditto.
5061         (convert_control_dep_chain_into_preds): Type change.
5062         (find_predicates): Ditto.
5063         (find_def_preds): Ditto.
5064         (destroy_predicates_vecs): Ditto.
5065         (find_matching_predicates_in_rest_chains): Ditto.
5066         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5067         (is_pred_expr_subset): Ditto.
5068         (is_pred_chain_subset_of): Ditto.
5069         (is_included_in): Ditto.
5070         (is_superset_of): Ditto.
5072 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5074         Update copyright years.
5076 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5078         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5079         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5080         config/arc/arc.md, config/arc/arc.opt,
5081         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5082         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5083         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5084         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5085         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5086         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5087         vtable-verify.c, vtable-verify.h: Use the standard form for the
5088         copyright notice.
5090 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5092         * gcc.c (process_command): Update copyright notice dates.
5093         * gcov-dump.c: Ditto.
5094         * gcov.c: Ditto.
5095         * doc/cpp.texi: Bump @copying's copyright year.
5096         * doc/cppinternals.texi: Ditto.
5097         * doc/gcc.texi: Ditto.
5098         * doc/gccint.texi: Ditto.
5099         * doc/gcov.texi: Ditto.
5100         * doc/install.texi: Ditto.
5101         * doc/invoke.texi: Ditto.
5103 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5105         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5107 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5109         * config/i386/sse.md (*mov<mode>_internal): Guard
5110         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5112         PR rtl-optimization/59647
5113         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5114         new_rtx into UNSIGNED_FLOAT rtxes.
5116 Copyright (C) 2014 Free Software Foundation, Inc.
5118 Copying and distribution of this file, with or without modification,
5119 are permitted in any medium without royalty provided the copyright
5120 notice and this notice are preserved.