2014-03-21 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blobbe0c8c484474c761502eb40838825a6f533b0801
1 2014-03-21  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/60577
4         * tree-core.h (struct tree_base): Document nothrow_flag use
5         in VAR_DECL_NONALIASED.
6         * tree.h (VAR_DECL_NONALIASED): New.
7         (may_be_aliased): Adjust.
8         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
10 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
12         * expr.c (expand_expr_real_1): Remove outdated comment.
14 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
16         PR middle-end/60597
17         * ira.c (adjust_cleared_regs): Call copy_rtx on
18         *reg_equiv[REGNO (loc)].src_p before passing it to
19         simplify_replace_fn_rtx.
21         PR target/60568
22         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
23         into CONST, put pic register as first operand of PLUS.  Use
24         gen_const_mem for both 32-bit and 64-bit PIC got loads.
26 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
30 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
32         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
33         around for store forwarding issue in the FPU on the UT699.
34         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
35         loads and operations if -mfix-ut699 is specified.
36         (divtf3_hq): Tweak attribute.
37         (sqrttf2_hq): Likewise.
39 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
41         * calls.c (store_one_arg): Remove incorrect const qualification on the
42         type of the temporary.
43         * cfgexpand.c (expand_return): Likewise.
44         * expr.c (expand_constructor): Likewise.
45         (expand_expr_real_1): Likewise.
47 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
49         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
50         of parts.
52 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
54         PR target/60039
55         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
57 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
59         * config/arm/aarch-common-protos.h
60         (alu_cost_table): Fix spelling of "extend".
61         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
63 2014-03-19  Richard Biener  <rguenther@suse.de>
65         PR middle-end/60553
66         * tree-core.h (tree_type_common): Re-order pointer members
67         to reduce recursion depth during GC walks.
69 2014-03-19  Marek Polacek  <polacek@redhat.com>
71         PR sanitizer/60569
72         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
73         before accessing it.
75 2014-03-19  Richard Biener  <rguenther@suse.de>
77         PR lto/59543
78         * lto-streamer-in.c (input_function): In WPA stage do not drop
79         debug stmts.
81 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
83         PR tree-optimization/60559
84         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
85         with build_zero_cst assignment.
87 2014-03-18  Kai Tietz  <ktietz@redhat.com>
89         PR rtl-optimization/56356
90         * sdbout.c (sdbout_parms): Verify that parms'
91         incoming argument is valid.
92         (sdbout_reg_parms): Likewise.
94 2014-03-18  Richard Henderson  <rth@redhat.com>
96         PR target/60562
97         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
98         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
99         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
101 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
103         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
104         * doc/plugins.texi (Plugin callbacks): Mention
105         PLUGIN_INCLUDE_FILE.
106         Italicize plugin event names in description.  Explain that
107         PLUGIN_PRAGMAS has no sense for lto1. Explain
108         PLUGIN_INCLUDE_FILE.
109         Remind that no GCC functions should be called after
110         PLUGIN_FINISH.
111         Explain what pragmas with expansion are.
113 2014-03-18  Martin Liska  <mliska@suse.cz>
115         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
116         gimple call statement is update.
117         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
118         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
120 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
122         PR sanitizer/60557
123         * ubsan.c (ubsan_instrument_unreachable): Call
124         initialize_sanitizer_builtins.
125         (ubsan_pass): Likewise.
127         PR sanitizer/60535
128         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
129         varpool_finalize_decl instead of rest_of_decl_compilation.
131 2014-03-18  Richard Biener  <rguenther@suse.de>
133         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
134         by using bitmap_and_compl instead of bitmap_and_compl_into.
135         (df_rd_transfer_function): Likewise.
137 2014-03-18  Richard Biener  <rguenther@suse.de>
139         * doc/lto.texi (fresolution): Fix typo.
141 2014-03-18  Richard Biener  <rguenther@suse.de>
143         * doc/invoke.texi (flto): Update for changes in 4.9.
145 2014-03-18  Richard Biener  <rguenther@suse.de>
147         * doc/loop.texi: Remove section on the removed lambda framework.
148         Update loop docs with recent changes in preserving loop structure.
150 2014-03-18  Richard Biener  <rguenther@suse.de>
152         * doc/lto.texi (-fresolution): Document.
154 2014-03-18  Richard Biener  <rguenther@suse.de>
156         * doc/contrib.texi: Adjust my name.
158 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
160         PR ipa/58721
161         * internal-fn.c: Include diagnostic-core.h.
162         (expand_BUILTIN_EXPECT): New function.
163         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
164         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
165         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
166         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
167         IFN_BUILTIN_EXPECT.
168         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
169         Revert 3 argument __builtin_expect code.
170         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
171         * gimple-fold.c (gimple_fold_call): Likewise.
172         * tree.h (fold_builtin_expect): New prototype.
173         * builtins.c (build_builtin_expect_predicate): Add predictor
174         argument, if non-NULL, create 3 argument __builtin_expect.
175         (fold_builtin_expect): No longer static.  Add ARG2 argument,
176         pass it through to build_builtin_expect_predicate.
177         (fold_builtin_2): Adjust caller.
178         (fold_builtin_3): Handle BUILT_IN_EXPECT.
179         * internal-fn.def (BUILTIN_EXPECT): New.
181 2014-03-18  Tobias Burnus  <burnus@net-b.de>
183         PR ipa/58721
184         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
185         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
186         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
188 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
190         PR ipa/58721
191         * predict.c (combine_predictions_for_bb): Fix up formatting.
192         (expr_expected_value_1, expr_expected_value): Add predictor argument,
193         fill what it points to if non-NULL.
194         (tree_predict_by_opcode): Adjust caller, use the predictor.
195         * predict.def (PRED_COMPARE_AND_SWAP): Add.
197 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
199         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
200         proper constant for the store mode.
202 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
204         * symtab.c (change_decl_assembler_name): Fix transparent alias
205         chain construction.
207 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
209         * config/aarch64/aarch64.c: Correct the comments about the
210         aarch64 stack layout.
212 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
214         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
215         check for GF_OMP_FOR_KIND_FOR.
217 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
219         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
220         ymm and zmm register names.
222 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
224         PR target/60516
225         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
226         note creation for the 2010-08-31 changes.
228 2014-03-17  Marek Polacek  <polacek@redhat.com>
230         PR middle-end/60534
231         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
232         as -fno-tree-loop-vectorize.
233         (expand_omp_simd): Likewise.
235 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
237         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
238         (eligible_for_call_delay): New prototype.
239         * config/sparc/sparc.c (tls_call_delay): Rename into...
240         (eligible_for_call_delay): ...this.  Return false if the instruction
241         cannot be put in the delay slot of a branch.
242         (eligible_for_restore_insn): Simplify.
243         (eligible_for_return_delay): Return false if the instruction cannot be
244         put in the delay slot of a branch and simplify.
245         (eligible_for_sibcall_delay): Return false if the instruction cannot be
246         put in the delay slot of a branch.
247         * config/sparc/sparc.md (fix_ut699): New attribute.
248         (tls_call_delay): Delete.
249         (in_call_delay): Reimplement.
250         (eligible_for_sibcall_delay): Rename into...
251         (in_sibcall_delay): ...this.
252         (eligible_for_return_delay): Rename into...
253         (in_return_delay): ...this.
254         (in_branch_delay): Reimplement.
255         (in_uncond_branch_delay): Delete.
256         (in_annul_branch_delay): Delete.
258 2014-03-14  Richard Henderson  <rth@redhat.com>
260         PR target/60525
261         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
262         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
263         (*floathi<X87MODEF>2_i387_with_temp): Remove.
264         (floathi splitters): Remove.
265         (float<SWI48x>xf2): New pattern.
266         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
267         code that tried to handle DImode for 32-bit, but which was excluded
268         by the pattern's condition.  Drop allocation of stack temporary.
269         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
270         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
271         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
272         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
273         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
274         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
275         (*float<SWI48><MODEF>2_sse_interunit): Remove.
276         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
277         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
278         (*float<SWI48x><X87MODEF>2_i387): Remove.
279         (all float _with_temp splitters): Remove.
280         (*float<SWI48x><MODEF>2_i387): New pattern.
281         (*float<SWI48><MODEF>2_sse): New pattern.
282         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
283         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
285 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
286             Marek Polacek  <polacek@redhat.com>
288         PR middle-end/60484
289         * common.opt (dump_base_name_prefixed): New Variable.
290         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
291         if x_dump_base_name_prefixed is already set, set it at the end.
293 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
295         PR rtl-optimization/60508
296         * lra-constraints.c (get_reload_reg): Add new parameter
297         in_subreg_p.
298         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
299         Pass the new parameter values.
301 2014-03-14  Richard Biener  <rguenther@suse.de>
303         * common.opt: Revert unintented changes from r205065.
304         * opts.c: Likewise.
306 2014-03-14  Richard Biener  <rguenther@suse.de>
308         PR middle-end/60518
309         * cfghooks.c (split_block): Properly adjust all loops the
310         block was a latch of.
312 2014-03-14  Martin Jambor  <mjambor@suse.cz>
314         PR lto/60461
315         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
316         and simplify it.
318 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
320         PR target/59396
321         * config/avr/avr.c (avr_set_current_function): Pass function name
322         through default_strip_name_encoding before sanity checking instead
323         of skipping the first char of the assembler name.
325 2014-03-13  Richard Henderson  <rth@redhat.com>
327         PR debug/60438
328         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
329         (ix86_force_to_memory, ix86_free_from_memory): Remove.
330         * config/i386/i386-protos.h: Likewise.
331         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
332         in the expander instead of a splitter.
333         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
334         any possibility of requiring a memory.
335         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
336         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
337         (fp branch splitters): Update for ix86_split_fp_branch.
338         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
339         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
340         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
341         (*fop_<MODEF>_2_i387): Remove f/r alternative.
342         (*fop_<MODEF>_3_i387): Likewise.
343         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
344         (splitters for the fop_* register patterns): Remove.
345         (fscalexf4_i387): Rename from *fscalexf4_i387.
346         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
347         
348 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
350         PR tree-optimization/59779
351         * tree-dfa.c (get_ref_base_and_extent): Use double_int
352         type for bitsize and maxsize instead of HOST_WIDE_INT.
354 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
356         PR rtl-optimization/57320
357         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
358         the CFG after thread_prologue_and_epilogue_insns.
360 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
362         PR rtl-optimization/57189
363         * lra-constraints.c (process_alt_operands): Disfavor spilling
364         vector pseudos.
366 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
368         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
369         messages.
371 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
373         PR tree-optimization/59025
374         PR middle-end/60418
375         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
376         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
378 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
380         PR target/60486
381         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
382         calls of avr_out_plus_1.
384 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
386         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
387         BB's single pred and update the father loop's latch info later.
389 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
391         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
392         (VEC_M): Likewise.
393         (VEC_N): Likewise.
394         (VEC_R): Likewise.
395         (VEC_base): Likewise.
396         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
397         registers, we need to swap double words in little endian mode.
399         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
400         to be a container mode for 128-bit integer operations added in ISA
401         2.07.  Unlike TImode and PTImode, the preferred register set is
402         the Altivec/VMX registers for the 128-bit operations.
404         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
405         declarations.
406         (rs6000_split_128bit_ok_p): Likewise.
408         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
409         macros for creating ISA 2.07 normal and overloaded builtin
410         functions with 3 arguments.
411         (BU_P8V_OVERLOAD_3): Likewise.
412         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
413         for use as overloaded functions.
414         (VPERM_1TI_UNS): Likewise.
415         (VSEL_1TI): Likewise.
416         (VSEL_1TI_UNS): Likewise.
417         (ST_INTERNAL_1ti): Likewise.
418         (LD_INTERNAL_1ti): Likewise.
419         (XXSEL_1TI): Likewise.
420         (XXSEL_1TI_UNS): Likewise.
421         (VPERM_1TI): Likewise.
422         (VPERM_1TI_UNS): Likewise.
423         (XXPERMDI_1TI): Likewise.
424         (SET_1TI): Likewise.
425         (LXVD2X_V1TI): Likewise.
426         (STXVD2X_V1TI): Likewise.
427         (VEC_INIT_V1TI): Likewise.
428         (VEC_SET_V1TI): Likewise.
429         (VEC_EXT_V1TI): Likewise.
430         (EQV_V1TI): Likewise.
431         (NAND_V1TI): Likewise.
432         (ORC_V1TI): Likewise.
433         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
434         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
435         overloaded builtin.
436         (VADDUQM): Likewise.
437         (VSUBCUQ): Likewise.
438         (VADDEUQM): Likewise.
439         (VADDECUQ): Likewise.
440         (VSUBEUQM): Likewise.
441         (VSUBECUQ): Likewise.
443         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
444         __int128_t and __uint128_t types.
445         (__uint128_type): Likewise.
446         (altivec_categorize_keyword): Add support for vector __int128_t,
447         vector __uint128_t, vector __int128, and vector unsigned __int128
448         as a container type for TImode operations that need to be done in
449         VSX/Altivec registers.
450         (rs6000_macro_to_expand): Likewise.
451         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
452         to support 128-bit integer instructions vaddcuq, vadduqm,
453         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
454         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
456         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
457         for V1TImode, and set up preferences to use VSX/Altivec registers.
458         Setup VSX reload handlers.
459         (rs6000_debug_reg_global): Likewise.
460         (rs6000_init_hard_regno_mode_ok): Likewise.
461         (rs6000_preferred_simd_mode): Likewise.
462         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
463         (easy_altivec_constant): Likewise.
464         (output_vec_const_move): Likewise.
465         (rs6000_expand_vector_set): Convert V1TImode set and extract to
466         simple move.
467         (rs6000_expand_vector_extract): Likewise.
468         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
469         addressing.
470         (rs6000_const_vec): Add support for V1TImode.
471         (rs6000_emit_le_vsx_load): Swap double words when loading or
472         storing TImode/V1TImode.
473         (rs6000_emit_le_vsx_store): Likewise.
474         (rs6000_emit_le_vsx_move): Likewise.
475         (rs6000_emit_move): Add support for V1TImode.
476         (altivec_expand_ld_builtin): Likewise.
477         (altivec_expand_st_builtin): Likewise.
478         (altivec_expand_vec_init_builtin): Likewise.
479         (altivec_expand_builtin): Likewise.
480         (rs6000_init_builtins): Add support for V1TImode type.  Add
481         support for ISA 2.07 128-bit integer builtins.  Define type names
482         for the VSX/Altivec vector types.
483         (altivec_init_builtins): Add support for overloaded vector
484         functions with V1TImode type.
485         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
486         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
487         external function.
488         (rs6000_split_128bit_ok_p): Likewise.
489         (rs6000_handle_altivec_attribute): Create V1TImode from vector
490         __int128_t and vector __uint128_t.
492         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
493         and mode attributes.
494         (VSX_M): Likewise.
495         (VSX_M2): Likewise.
496         (VSm): Likewise.
497         (VSs): Likewise.
498         (VSr): Likewise.
499         (VSv): Likewise.
500         (VS_scalar): Likewise.
501         (VS_double): Likewise.
502         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
504         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
505         we support the ISA 2.07 128-bit integer arithmetic instructions.
506         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
507         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
508         and TImode types for use with the builtin functions.
509         (V1TI_type_node): Likewise.
510         (unsigned_V1TI_type_node): Likewise.
511         (intTI_type_internal_node): Likewise.
512         (uintTI_type_internal_node): Likewise.
514         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
515         128-bit builtin functions.
516         (UNSPEC_VADDEUQM): Likewise.
517         (UNSPEC_VADDECUQ): Likewise.
518         (UNSPEC_VSUBCUQ): Likewise.
519         (UNSPEC_VSUBEUQM): Likewise.
520         (UNSPEC_VSUBECUQ): Likewise.
521         (VM): Add V1TImode to vector mode iterators.
522         (VM2): Likewise.
523         (VI_unit): Likewise.
524         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
525         (altivec_vaddcuq): Likewise.
526         (altivec_vsubuqm): Likewise.
527         (altivec_vsubcuq): Likewise.
528         (altivec_vaddeuqm): Likewise.
529         (altivec_vaddecuq): Likewise.
530         (altivec_vsubeuqm): Likewise.
531         (altivec_vsubecuq): Likewise.
533         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
534         mode iterators.
535         (BOOL_128): Likewise.
536         (BOOL_REGS_OUTPUT): Likewise.
537         (BOOL_REGS_OP1): Likewise.
538         (BOOL_REGS_OP2): Likewise.
539         (BOOL_REGS_UNARY): Likewise.
540         (BOOL_REGS_AND_CR0): Likewise.
542         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
543         128-bit integer builtin support.
544         (vec_vadduqm): Likewise.
545         (vec_vaddecuq): Likewise.
546         (vec_vaddeuqm): Likewise.
547         (vec_vsubecuq): Likewise.
548         (vec_vsubeuqm): Likewise.
549         (vec_vsubcuq): Likewise.
550         (vec_vsubuqm): Likewise.
552         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
553         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
554         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
555         128-bit integer add/subtract to ISA 2.07.
557 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
559         * config/arc/arc.c (arc_predicate_delay_insns):
560         Fix third argument passed to conditionalize_nonjump.
562 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
564         * config/aarch64/aarch64-builtins.c
565         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
566         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
567         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
568         instead of __builtin_lfloor.
569         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
571 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
573         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
574         (tree_ssa_ifcombine_bb_1): New function.
575         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
576         is an empty forwarder block to then_bb or vice versa and then_bb
577         and else_bb are effectively swapped.
579 2014-03-12  Christian Bruel  <christian.bruel@st.com>
581         PR target/60264
582         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
583         REG_CFA_DEF_CFA note.
584         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
585         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
587 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
589         PR tree-optimization/60454
590         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
592 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
594         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
595         Do not define target_cpu_default2 to generic.
596         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
597         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
598         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
600 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
601             Marc Glisse  <marc.glisse@inria.fr>
603         PR tree-optimization/60502
604         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
605         instead of build_low_bits_mask.
607 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
609         PR middle-end/60482
610         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
611         if there are multiple uses, but op doesn't live on E edge.
612         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
613         clobber stmts before __builtin_unreachable.
615 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
617         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
618         hard_frame_pointer_rtx.
619         * cse.c (cse_insn): Remove volatile check.
620         * cselib.c (cselib_process_insn): Likewise.
621         * dse.c (scan_insn): Likewise.
623 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
625         * config/arc/arc.c (conditionalize_nonjump): New function,
626         broken out of ...
627         (arc_ifcvt): ... this.
628         (arc_predicate_delay_insns): Use it.
630 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
632         * config/arc/predicates.md (extend_operand): During/after reload,
633         allow const_int_operand.
634         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
635         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
636         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
637         to "i".
638         (umulsi3_highpart_i): Likewise.
640 2014-03-11  Richard Biener  <rguenther@suse.de>
642         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
643         Add asserts to guard possible wrong-code bugs.
645 2014-03-11  Richard Biener  <rguenther@suse.de>
647         PR tree-optimization/60429
648         PR tree-optimization/60485
649         * tree-ssa-structalias.c (set_union_with_increment): Properly
650         take into account all fields that overlap the shifted vars.
651         (do_sd_constraint): Likewise.
652         (do_ds_constraint): Likewise.
653         (get_constraint_for_ptr_offset): Likewise.
655 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
657         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
658         (nios2_compute_frame_layout):
659         Add calculation of cfun->machine->fp_save_offset.
660         (nios2_expand_prologue): Correct setting of frame pointer register
661         in prologue.
662         (nios2_expand_epilogue): Update recovery of stack pointer from
663         frame pointer accordingly.
664         (nios2_initial_elimination_offset): Update calculation of offset
665         for eliminating to HARD_FRAME_POINTER_REGNUM.
667 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
669         PR ipa/60457
670         * ipa.c (symtab_remove_unreachable_nodes): Don't call
671         cgraph_get_create_node on VAR_DECLs.
673 2014-03-10  Richard Biener  <rguenther@suse.de>
675         PR middle-end/60474
676         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
678 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
680         * config/vms/vms.opt (vms_float_format): New variable.
682 2014-03-08  Tobias Burnus  <burnus@net-b.de>
684         * doc/invoke.texi (-fcilkplus): Update implementation status.
686 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
687             Richard Biener  <rguenther@suse.de>
689         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
690         consistently accross all TUs.
691         (run_gcc): Enable -fshort-double automatically at link at link-time
692         and disallow override.
694 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
696         PR target/58271
697         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
698         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
699         if they can't be used.
701 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
703         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
704         for Solaris 11/x86 ld.
705         * configure: Regenerate.
707 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
709         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
710         (LIB_TLS_SPEC): Save as ld_tls_libs.
711         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
712         (HAVE_AS_IX86_TLSLDM): New test.
713         * configure, config.in: Regenerate.
714         * config/i386/i386.c (legitimize_tls_address): Fall back to
715         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
716         cannot support TLS_MODEL_LOCAL_DYNAMIC.
717         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
718         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
720 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
722         * common.opt (fira-loop-pressure): Mark as optimization.
724 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
726         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
727         an OpenMP mappable type.
729 2014-03-06  Matthias Klose  <doko@ubuntu.com>
731         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
732         MULTILIB_OSDIRNAMES is not defined.
734 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
735             Meador Inge  <meadori@codesourcery.com>
737         PR target/58595
738         * config/arm/arm.c (arm_tls_symbol_p): Remove.
739         (arm_legitimize_address): Call legitimize_tls_address for any
740         arm_tls_referenced_p expression, handle constant addend.  Call it
741         before testing for !TARGET_ARM.
742         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
744 2014-03-06  Richard Biener  <rguenther@suse.de>
746         PR middle-end/60445
747         PR lto/60424
748         PR lto/60427
749         Revert
750         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
752         * tree-streamer.c (record_common_node): Assert we don't record
753         nodes with type double.
754         (preload_common_node): Skip type double, complex double and double
755         pointer since it is now frontend dependent due to fshort-double option.
757 2014-03-06  Richard Biener  <rguenther@suse.de>
759         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
760         or -fno-lto is specified and the linker has full plugin support.
761         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
762         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
763         * lto-wrapper.c (merge_and_complain): Merge compile-time
764         optimization levels.
765         (run_gcc): And pass it through to the link options.
767 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
769         PR debug/60381
770         Revert:
771         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
772         PR debug/59992
773         * cselib.c (remove_useless_values): Skip to avoid quadratic
774         behavior if the condition moved from...
775         (cselib_process_insn): ... here holds.
777 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
779         PR plugins/59335
780         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
781         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
783         PR plugins/59335
784         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
785         (TM_H): Add x86-tune.def.
787 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
789         * config/aarch64/aarch64.c (generic_tunings):
790         Use cortexa57_extra_costs.
792 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
794         PR lto/60404
795         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
796         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
797         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
798         cost for in_lto_p.
800 2014-03-04  Heiher  <r@hev.cc>
802         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
803         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
805 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
807         * config/i386/predicates.md (const2356_operand): Change to ...
808         (const2367_operand): ... this.
809         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
810         const2367_operand.
811         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
812         (*avx512pf_scatterpf<mode>sf): Ditto.
813         (avx512pf_scatterpf<mode>df): Ditto.
814         (*avx512pf_scatterpf<mode>df_mask): Ditto.
815         (*avx512pf_scatterpf<mode>df): Ditto.
816         * config/i386/i386.c (ix86_expand_builtin): Update
817         incorrect hint operand error message.
819 2014-03-04  Richard Biener  <rguenther@suse.de>
821         * lto-section-in.c (lto_get_section_data): Fix const cast.
823 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
825         * tree-streamer.c (record_common_node): Assert we don't record
826         nodes with type double.
827         (preload_common_node): Skip type double, complex double and double
828         pointer since it is now frontend dependent due to fshort-double option.
830 2014-03-04  Richard Biener  <rguenther@suse.de>
832         PR lto/60405
833         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
834         (lto_input_toplevel_asms): Likewise.
835         * lto-section-in.c (lto_get_section_data): Instead do it here
836         for every section.
838 2014-03-04  Richard Biener  <rguenther@suse.de>
840         PR tree-optimization/60382
841         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
842         dead PHIs a reduction.
844 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
846         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
847         hint value.
848         (_mm_prefetch): Move out of GCC target("sse") pragma.
849         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
850         GCC target("prfchw") pragma.
851         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
852         for locality <= 2.
853         * config/i386/i386.c (ix86_option_override_internal): Enable
854         -mprfchw with -mprefetchwt1.
856 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
858         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
859         Mark as varying.
861 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
863         * opts.h (CL_PCH_IGNORE): Define.
864         * targhooks.c (option_affects_pch_p):
865         Return false for options that have CL_PCH_IGNORE set.
866         * opt-functions.awk: Process PchIgnore.
867         * doc/options.texi: Document PchIgnore.
869         * config/arc/arc.opt (misize): Add PchIgnore property.
871 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
873         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
874         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
875         constraint on constants to permit them being loaded into
876         GENERAL_REGS or BASE_REGS.
878 2014-03-03  Nick Clifton  <nickc@redhat.com>
880         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
881         anti-cacnonical alternatives.
882         (negandhi3_real): New pattern.
883         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
885 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
887         * config/avr/avr-mcus.def: Remove atxmega16x1.
888         * config/avr/avr-tables.opt: Regenerate.
889         * config/avr/t-multilib: Regenerate.
890         * doc/avr-mmcu.texi: Regenerate.
892 2014-03-03  Tobias Grosser  <tobias@grosser.es>
893             Mircea Namolaru  <mircea.namolaru@inria.fr>
895         PR tree-optimization/58028
896         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
897         scalar dimensions.
899 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
901         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
902         not handled by recognizers.
904 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
906         PR middle-end/60175
907         * function.c (expand_function_end): Don't emit
908         clobber_return_register sequence if clobber_after is a BARRIER.
909         * cfgexpand.c (construct_exit_block): Append instructions before
910         return_label to prev_bb.
912 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
914         * config/rs6000/constraints.md: Document reserved use of "wc".
916 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
918         PR ipa/60150
919         * ipa.c (function_and_variable_visibility): When dissolving comdat
920         group, also set all symbols to local.
922 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
924         PR ipa/60306
926         Revert:
927         2013-12-14   Jan Hubicka  <jh@suse.cz>
928         PR middle-end/58477
929         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
931 2014-03-02  Jon Beniston  <jon@beniston.com>
933         PR bootstrap/48230
934         PR bootstrap/50927
935         PR bootstrap/52466
936         PR target/46898
937         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
938           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
939         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
940         (simple_return, *simple_return): New patterns
941         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
942         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
944 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
946         * dwarf2out.c (gen_subprogram_die): Tidy.
948 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
950         PR target/60071
951         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
952         (*mov_t_msb_neg_negc): ... this new insn.
954 2014-02-28  Jason Merrill  <jason@redhat.com>
956         PR c++/58678
957         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
958         function.
960 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
962         PR c++/60314
963         * dwarf2out.c (decltype_auto_die): New static.
964         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
965         (gen_type_die_with_usage): Handle 'decltype(auto)'.
966         (is_cxx_auto): Likewise.
968 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
970         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
971         we are not using general regs only.
973 2014-02-28  Richard Biener  <rguenther@suse.de>
975         PR target/60280
976         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
977         previous fix and only allow to remove trivial pre-headers
978         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
979         (remove_forwarder_block): Properly update the latch of a loop.
981 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
983         PR debug/59992
984         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
985         (cselib_preserved_hash_table): New.
986         (preserve_constants_and_equivs): Move preserved vals to it.
987         (cselib_find_slot): Look it up first.
988         (cselib_init): Initialize it.
989         (cselib_finish): Release it.
990         (dump_cselib_table): Dump it.
992 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
994         PR debug/59992
995         * cselib.c (remove_useless_values): Skip to avoid quadratic
996         behavior if the condition moved from...
997         (cselib_process_insn): ... here holds.
999 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1001         PR debug/57232
1002         * var-tracking.c (vt_initialize): Apply the same condition to
1003         preserve the CFA base value.
1005 2014-02-28  Joey Ye  <joey.ye@arm.com>
1007         PR target/PR60169
1008         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1009         if reload in progress or completed.
1011 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1013         PR middle-end/60147
1014         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1015         NAMELIST_DECL.
1017 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1019         * doc/tm.texi.in (Condition Code Status): Update documention for
1020         relative locations of cc0-setter and cc0-user.
1022 2014-02-27  Jeff Law  <law@redhat.com>
1024         PR rtl-optimization/52714
1025         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1026         into two independent simple sets, if I3 is a jump, ensure the
1027         pattern we place into I3 is a (set (pc) ...).
1029 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1030             Jeff Law  <law@redhat.com>
1032         PR rtl-optimization/49847
1033         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1034         are in different blocks.
1035         * doc/tm.texi (Condition Code Status): Update documention for
1036         relative locations of cc0-setter and cc0-user.
1038 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1040         PR target/59222
1041         * lra.c (lra_emit_add): Check SUBREG too.
1043 2014-02-27  Andreas Schwab  <schwab@suse.de>
1045         * config/m68k/m68k.c (m68k_option_override): Disable
1046         -flive-range-shrinkage for classic m68k.
1047         (m68k_override_options_after_change): Likewise.
1049 2014-02-27  Marek Polacek  <polacek@redhat.com>
1051         PR middle-end/59223
1052         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1053         -Wmaybe-uninitialized.
1055 2014-02-27  Alan Modra  <amodra@gmail.com>
1057         PR target/57936
1058         * reload1.c (emit_input_reload_insns): When reload_override_in,
1059         set old to rl->in_reg when rl->in_reg is a subreg.
1061 2014-02-26  Richard Biener  <rguenther@suse.de>
1063         PR bootstrap/60343
1064         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1066 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1068         * common/config/i386/predicates.md (const1256_operand): Remove.
1069         (const2356_operand): New.
1070         (const_1_to_2_operand): Remove.
1071         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1072         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1073         (*avx512pf_gatherpf<mode>sf): Ditto.
1074         (avx512pf_gatherpf<mode>df): Ditto.
1075         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1076         (*avx512pf_gatherpf<mode>df): Ditto.
1077         (avx512pf_scatterpf<mode>sf): Ditto.
1078         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1079         (*avx512pf_scatterpf<mode>sf): Ditto.
1080         (avx512pf_scatterpf<mode>df): Ditto.
1081         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1082         (*avx512pf_scatterpf<mode>df): Ditto.
1083         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1085 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1087         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1088         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1089         (_mm512_mask_testn_epi64_mask): Move to ...
1090         * config/i386/avx512cdintrin.h: Here.
1091         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1092         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1093         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1094         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1095         TARGET_AVX512F from TARGET_AVX512CD.
1097 2014-02-26  Richard Biener  <rguenther@suse.de>
1099         PR ipa/60327
1100         * ipa.c (walk_polymorphic_call_targets): Properly guard
1101         call to inline_update_overall_summary.
1103 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1105         PR target/60280
1106         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1107         and latches only if requested.  Fix latch if it is removed.
1108         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1109         LOOPS_HAVE_PREHEADERS.
1111 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1113         * builtins.c (expand_builtin_thread_pointer): Create a new target
1114         when the target is NULL.
1116 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1118         PR rtl-optimization/60317
1119         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1120         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1121         * lra-assigns.c: Include params.h.
1122         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1123         other reload pseudos considerations.
1125 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1127         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1128         to use canonical form for nor<mode>3.
1130 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1132         PR target/55426
1133         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1134         conversions.
1136 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1138         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1139         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1140         (ix86_handle_option): Handle OPT_mprefetchwt1.
1141         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1142         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1143         PREFETCHWT1 CPUID.
1144         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1145         OPTION_MASK_ISA_PREFETCHWT1.
1146         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1147         (PTA_PREFETCHWT1): New.
1148         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1149         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1150         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1151         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1152         (*prefetch_avx512pf_<mode>_: Change into ...
1153         (*prefetch_prefetchwt1_<mode>: This.
1154         * config/i386/i386.opt (mprefetchwt1): New.
1155         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1156         (_mm_prefetch): Handle intent to write.
1157         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1159 2014-02-25  Richard Biener  <rguenther@suse.de>
1161         PR middle-end/60291
1162         * emit-rtl.c (mem_attrs_htab): Remove.
1163         (mem_attrs_htab_hash): Likewise.
1164         (mem_attrs_htab_eq): Likewise.
1165         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1166         (init_emit_once): Do not allocate mem_attrs_htab.
1168 2014-02-25  Richard Biener  <rguenther@suse.de>
1170         PR lto/60319
1171         * lto-opts.c (lto_write_options): Output non-explicit conservative
1172         -fwrapv, -fno-trapv and -fno-strict-overflow.
1173         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1174         (run_gcc): And pass them through.
1176 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1178         * sel-sched.c (calculate_new_fences): New parameter ptime.
1179         Calculate it as a maximum over all fence cycles.
1180         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1181         Print the final schedule timing when sched_verbose.
1183 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1185         PR rtl-optimization/60292
1186         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1187         bit fot the fence instruction.
1189 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1191         * calls.h: Fix typo in comment.
1193 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1195         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1196         adjusting offsetable addresses.
1198 2014-02-24  Guozhi Wei  <carrot@google.com>
1200         * sparseset.h (sparseset_pop): Fix the wrong index.
1202 2014-02-24  Walter Lee  <walt@tilera.com>
1204         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1205         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1206         triplet.
1207         * common/config/tilegx/tilegx-common.c
1208         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1209         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1210         (LINK_SPEC): Ditto.
1211         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1212         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1213         (tilegx_gimplify_va_arg_expr): Handle big endian.
1214         (tilegx_expand_unaligned_load): Ditto.
1215         (tilegx_expand_unaligned_store): Ditto.
1216         (TARGET_RETURN_IN_MSB): New.
1217         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1218         (TARGET_ENDIAN_DEFAULT): New.
1219         (TARGET_BIG_ENDIAN): Handle big endian.
1220         (BYTES_BIG_ENDIAN): Ditto.
1221         (WORDS_BIG_ENDIAN): Ditto.
1222         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1223         (ENDIAN_SPEC): New.
1224         (EXTRA_SPECS): New.
1225         * config/tilegx/tilegx.md (extv): Handle big endian.
1226         (extzv): Ditto.
1227         (insn_st<n>): Ditto.
1228         (insn_st<n>_add<bitsuffix>): Ditto.
1229         (insn_stnt<n>): Ditto.
1230         (insn_stnt<n>_add<bitsuffix>):Ditto.
1231         (vec_interleave_highv8qi): Handle big endian.
1232         (vec_interleave_highv8qi_be): New.
1233         (vec_interleave_highv8qi_le): New.
1234         (insn_v1int_h): Handle big endian.
1235         (vec_interleave_lowv8qi): Handle big endian.
1236         (vec_interleave_lowv8qi_be): New.
1237         (vec_interleave_lowv8qi_le): New.
1238         (insn_v1int_l): Handle big endian.
1239         (vec_interleave_highv4hi): Handle big endian.
1240         (vec_interleave_highv4hi_be): New.
1241         (vec_interleave_highv4hi_le): New.
1242         (insn_v2int_h): Handle big endian.
1243         (vec_interleave_lowv4hi): Handle big endian.
1244         (vec_interleave_lowv4hi_be): New.
1245         (vec_interleave_lowv4hi_le): New.
1246         (insn_v2int_l): Handle big endian.
1247         (vec_interleave_highv2si): Handle big endian.
1248         (vec_interleave_highv2si_be): New.
1249         (vec_interleave_highv2si_le): New.
1250         (insn_v4int_h): Handle big endian.
1251         (vec_interleave_lowv2si): Handle big endian.
1252         (vec_interleave_lowv2si_be): New.
1253         (vec_interleave_lowv2si_le): New.
1254         (insn_v4int_l): Handle big endian.
1255         * config/tilegx/tilegx.opt (mbig-endian): New option.
1256         (mlittle-endian): New option.
1257         * doc/install.texi: Document tilegxbe-linux.
1258         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1260 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1262         PR ipa/60266
1263         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1264         there are no parameter descriptors.
1266 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1268         PR rtl-optimization/60268
1269         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1270         initialization to ...
1271         (sched_rgn_init): ... here.
1272         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1274 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1276         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1277         names.
1279 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1281         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1282         definition.
1284 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1286         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1287         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1289 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1291         * config/microblaze/predicates.md: Add cmp_op predicate.
1292         * config/microblaze/microblaze.md: Add branch_compare instruction
1293         which uses cmp_op predicate and emits cmp insn before branch.
1294         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1295         to microblaze_expand_conditional_branch and consolidate logic.
1296         (microblaze_expand_conditional_branch): emit branch_compare
1297         insn instead of handling cmp op separate from branch insn.
1299 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1301         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1302         to permit subregs.
1304 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1306         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1307         define_insn with define_expand and new define_insn
1308         *altivec_lve<VI_char>x_internal.
1309         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1310         and new define_insn *altivec_stve<VI_char>x_internal.
1311         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1312         prototype.
1313         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1314         lve*x built-ins.
1315         (altivec_expand_stvex_be): New function.
1317 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1319         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1320         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1321         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1322         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1324 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1326         PR target/60298
1327         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1328         instead of emit_move_insn.
1330 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1332         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1333         vspltw with vsldoi.
1334         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1335         gen_altivec_vsumsws.
1337 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1339         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1340         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1341         (altivec_lvxl_<mode>): New define_expand incorporating
1342         -maltivec=be semantics where needed.
1343         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1344         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1345         semantics where needed.
1346         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1347         (altivec_stvx_<mode>): New define_expand incorporating
1348         -maltivec=be semantics where needed.
1349         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1350         VM2 iterator instead of V4SI.
1351         (altivec_stvxl_<mode>): New define_expand incorporating
1352         -maltivec=be semantics where needed.
1353         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1354         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1355         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1356         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1357         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1358         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1359         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1360         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1361         ALTIVEC_BUILTIN_STVXL.
1362         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1363         (altivec_expand_stvx_be): Likewise.
1364         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1365         (altivec_expand_lvx_be): Likewise.
1366         (altivec_expand_stvx_be): Likewise.
1367         (altivec_expand_builtin): Add cases for
1368         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1369         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1370         (altivec_init_builtins): Add definitions for
1371         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1372         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1374 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1376         * doc/invoke.texi (mvirt, mno-virt): Document.
1377         * config/mips/mips.opt (mvirt): New option.
1378         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1380 2014-02-21  Richard Biener  <rguenther@suse.de>
1382         PR tree-optimization/60276
1383         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1384         (STMT_VINFO_MIN_NEG_DIST): New macro.
1385         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1386         STMT_VINFO_MIN_NEG_DIST.
1387         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1388         made for negative dependence distances still hold.
1390 2014-02-21  Richard Biener  <rguenther@suse.de>
1392         PR middle-end/60291
1393         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1394         DECL_INITIAL for globals not in the current function context.
1396 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1398         PR tree-optimization/56490
1399         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1400         * tree-ssa-uninit.c: Include params.h.
1401         (compute_control_dep_chain): Add num_calls argument, return false
1402         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1403         num_calls to recursive call.
1404         (find_predicates): Change dep_chain into normal array,
1405         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1406         variable and adjust compute_control_dep_chain caller.
1407         (find_def_preds): Likewise.
1409 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1411         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1412         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1414 2014-02-21  Nick Clifton  <nickc@redhat.com>
1416         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1417         (pushhi1): Likewise.
1418         (popqi1): Add mode to pre_dec.
1419         (pophi1): Likewise.
1421 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1423         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1424         mode for mask of V8SFmode permutation.
1426 2014-02-20  Richard Henderson  <rth@redhat.com>
1428         PR c++/60272
1429         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1430         a new pseudo for OLDVAL.
1432 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1434         PR target/57896
1435         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1436         gen_reg_rtx if d->testing_p.
1437         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1438         if d->testing_p and we will certainly return true.
1439         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1440         if d->testing_p.
1442 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1444         * emit-rtl.c (gen_reg_rtx): Assert that
1445         crtl->emit.regno_pointer_align_length is non-zero.
1447 2014-02-20  Richard Henderson  <rth@redhat.com>
1449         PR c++/60272
1450         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1451         on failure the store back into EXPECT.
1453 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1454             Sandra Loosemore  <sandra@codesourcery.com>
1456         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1457         * config/nios2/nios2.c (nios2_function_profiler): Add
1458         -fPIC (flag_pic == 2) support.
1459         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1460         (nios2_large_offset_p): New function.
1461         (nios2_unspec_reloc_p): Move up position, update to use
1462         nios2_large_offset_p.
1463         (nios2_unspec_address): Remove function.
1464         (nios2_unspec_offset): New function.
1465         (nios2_large_got_address): New function.
1466         (nios2_got_address): Add large offset support.
1467         (nios2_legitimize_tls_address): Update usage of removed and new
1468         functions.
1469         (nios2_symbol_binds_local_p): New function.
1470         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1471         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1472         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1473         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1474         processing for (const (unspec ...)).
1475         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1477 2014-02-20  Richard Biener  <rguenther@suse.de>
1479         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1480         doing the substitution.
1481         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1483 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1485         PR ipa/55260
1486         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1487         info when checking whether lattices are bottom.
1489 2014-02-20  Richard Biener  <rguenther@suse.de>
1491         PR middle-end/60221
1492         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1493         regions at -O0.
1495 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1497         PR ipa/58555
1498         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1499         parameter specifying the scaling.
1500         (inline_call): Update.
1501         (want_inline_recursively): Guard division by zero.
1502         (recursive_inlining): Update.
1503         * ipa-inline.h (clone_inlined_nodes): Update.
1505 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1507         PR target/60204
1508         * config/i386/i386.c (classify_argument): Pass structures of size
1509         64 bytes or less in register.
1511 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1512             Kirill Yukhin  <kirill.yukhin@intel.com>
1514         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1515         (_mm_rcp28_round_ss): Ditto.
1516         (_mm_rsqrt28_round_sd): Ditto.
1517         (_mm_rsqrt28_round_ss): Ditto.
1518         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1519         (_mm_rcp14_round_ss): Ditto.
1520         (_mm_rsqrt14_round_sd): Ditto.
1521         (_mm_rsqrt14_round_ss): Ditto.
1522         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1523         the first input operand, get rid of match_dup.
1524         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1525         attribute to sse.
1526         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1527         Ditto.
1528         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1529         operand as the first input operand, set type attribute.
1530         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1531         Set type attribute.
1532         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1533         operand as the first input operand, set type attribute.
1535 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1537         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1538         bit of zero.
1540 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1542         PR target/60207
1543         * config/i386/i386.c (construct_container): Remove TFmode check
1544         for X86_64_INTEGER_CLASS.
1546 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1548         PR target/59794
1549         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1550         only when -Wpsabi is enabled.
1552 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1554          PR target/59799
1555         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1556         passing arrays in registers are the same as for structs, so remove the
1557         special case for them.
1559 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1561         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1562         destination type, extract only the valid bits if the source type is not
1563         integral and has a different mode.
1565 2014-02-19  Richard Biener  <rguenther@suse.de>
1567         PR ipa/60243
1568         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1569         for all calls.
1571 2014-02-19  Richard Biener  <rguenther@suse.de>
1573         PR ipa/60243
1574         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1575         (ipa_modify_call_arguments): Emit an argument load explicitely and
1576         preserve virtual SSA form there and for the replacement call.
1577         Do not update SSA form nor free dominance info.
1579 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1581         * ipa.c (function_and_variable_visibility): Also clear WEAK
1582         flag when disolving COMDAT_GROUP.
1584 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1586         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1587         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1588         not devirtualizing.
1589         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1590         do more sanity checks.
1591         (detect_type_change): Return true when giving up early.
1592         (compute_complex_assign_jump_func): Fix type parameter of
1593         ipa_set_ancestor_jf.
1594         (compute_complex_ancestor_jump_func): Likewise.
1595         (update_jump_functions_after_inlining): Fix updating of
1596         ancestor function.
1597         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1599 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1601         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1602         inline clones when edge disappears.
1604 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1606         PR target/60203
1607         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1608         Split 64-bit moves into 2 patterns.  Do not allow the use of
1609         direct move for TDmode in little endian, since the decimal value
1610         has little endian bytes within a word, but the 64-bit pieces are
1611         ordered in a big endian fashion, and normal subreg's of TDmode are
1612         not allowed.
1613         (mov<mode>_64bit_dm): Likewise.
1614         (movtd_64bit_nodm): Likewise.
1616 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1618         PR tree-optimization/60174
1619         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1620         statement of an SSA_NAME that occurs in an abnormal PHI node.
1622 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1624         PR sanitizer/60142
1625         * final.c (SEEN_BB): Remove.
1626         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1627         (final_scan_insn): Don't force_source_line on second
1628         NOTE_INSN_BASIC_BLOCK.
1630 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1632         PR target/60205
1633         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1634         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1635         (type_natural_mode): Warn ABI change when %zmm register is not
1636         available for AVX512F vector value passing.
1638 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1640         PR target/60193
1641         * config/i386/i386.c (ix86_expand_prologue): Use value in
1642         rax register as displacement when restoring %r10 or %rax.
1643         Fix wrong offset when restoring both registers.
1645 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1647         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1648         assertion with conditional return.
1650 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1651             Uros Bizjak  <ubizjak@gmail.com>
1653         PR driver/60233
1654         * config/i386/driver-i386.c (host_detect_local_cpu): If
1655         YMM state is not saved by the OS, also clear has_f16c.  Move
1656         CPUID 0x80000001 handling before YMM state saving checking.
1658 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1660         PR rtl-optimization/58960
1661         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1662         factored out from ...
1663         (sched_init): ... here.
1664         (free_global_sched_pressure_data): New, factored out from ...
1665         (sched_finish): ... here.
1666         * sched-int.h (free_global_sched_pressure_data): Declare.
1667         * sched-rgn.c (nr_regions_initial): New static global.
1668         (haifa_find_rgns): Initialize it.
1669         (schedule_region): Disable sched-pressure for the newly
1670         generated regions.
1672 2014-02-17  Richard Biener  <rguenther@suse.de>
1674         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1675         release SSA defs of pattern stmts.
1677 2014-02-17  Richard Biener  <rguenther@suse.de>
1679         * tree-inline.c (expand_call_inline): Release the virtual
1680         operand defined by the call we are about to inline.
1682 2014-02-17  Richard Biener  <rguenther@suse.de>
1684         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1686 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1687             Ilya Tocar  <ilya.tocar@intel.com>
1689         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1690         arguments order in builtin.
1691         (_mm512_permutexvar_epi64): Ditto.
1692         (_mm512_mask_permutexvar_epi64): Ditto
1693         (_mm512_maskz_permutexvar_epi32): Ditto
1694         (_mm512_permutexvar_epi32): Ditto
1695         (_mm512_mask_permutexvar_epi32): Ditto
1697 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1699         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1700         (p8_vmrgow): Likewise.
1702 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1704         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1705         endian targets.
1707 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1709         PR target/60203
1710         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1711         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1712         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1713         using direct move instructions on ISA 2.07.  Also adjust
1714         instruction length for 64-bit.
1715         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1716         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1718 2014-02-15  Alan Modra  <amodra@gmail.com>
1720         PR target/58675
1721         PR target/57935
1722         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1723         find_replacement on parts of insn rtl that might be reloaded.
1725 2014-02-15  Richard Biener  <rguenther@suse.de>
1727         PR tree-optimization/60183
1728         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1729         (tree_ssa_phiprop): Calculate and free post-dominators.
1731 2014-02-14  Jeff Law  <law@redhat.com>
1733         PR rtl-optimization/60131
1734         * ree.c (get_extended_src_reg): New function.
1735         (combine_reaching_defs): Use it rather than assuming location of REG.
1736         (find_and_remove_re): Verify first operand of extension is
1737         a REG before adding the insns to the copy list.
1739 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1741         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1742         * configure: Regenerated.
1743         * config.in: Regenerated.
1744         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1745         instead of ASM_SHORT.
1747 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1748             Richard Earnshaw  <rearnsha@arm.com>
1750         PR rtl-optimization/59535
1751         * lra-constraints.c (process_alt_operands): Encourage alternative
1752         when unassigned pseudo class is superset of the alternative class.
1753         (inherit_reload_reg): Don't inherit when optimizing for code size.
1754         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1755         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1756         modes not less than 4 for Thumb1.
1758 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1760         PR pch/60010
1761         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1763 2014-02-14  Richard Biener  <rguenther@suse.de>
1765         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1766         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1767         Do not strip INDIRECT_REFs.
1769 2014-02-14  Richard Biener  <rguenther@suse.de>
1771         PR lto/60179
1772         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1773         DECL_FUNCTION_SPECIFIC_TARGET.
1774         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1775         * tree-streamer-out.c (pack_ts_target_option): Remove.
1776         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1777         (write_ts_function_decl_tree_pointers): Do not stream
1778         DECL_FUNCTION_SPECIFIC_TARGET.
1779         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1780         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1781         (lto_input_ts_function_decl_tree_pointers): Do not stream
1782         DECL_FUNCTION_SPECIFIC_TARGET.
1784 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1786         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1787         (get_initial_def_for_induction, vectorizable_induction): Ignore
1788         debug stmts when looking for exit_phi.
1789         (vectorizable_live_operation): Fix up condition.
1791 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1793         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1794         nreverse() because it changes the content of original tree list.
1796 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1798         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1799         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1801 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1803         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1804         GNU coding standards.
1806 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1808         PR debug/60152
1809         * dwarf2out.c (gen_subprogram_die): Don't call
1810         add_calling_convention_attribute if subr_die is old_die.
1812 2014-02-13  Sharad Singhai  <singhai@google.com>
1814         * doc/optinfo.texi: Fix order of nodes.
1816 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1818         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1819         operands[2], not operands[3].
1821 2014-02-13  Richard Biener  <rguenther@suse.de>
1823         PR bootstrap/59878
1824         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1825         mention the possibility of an in-tree build.
1826         (CLooG): Update recommended version to 0.18.1, mention the
1827         possibility of an in-tree build and clarify that the ISL
1828         bundled with CLooG does not work.
1830 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1832         PR target/43546
1833         * expr.c (compress_float_constant): If x is a hard register,
1834         extend into a pseudo and then move to x.
1836 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1838         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1839         caused by bad second argument to warning_at() with -mhotpatch and
1840         nested functions (e.g. with gfortran).
1842 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1844         * opts.c (option_name): Remove "enabled by default" rider.
1846 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1848         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1850 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1851             Uros Bizjak  <ubizjak@gmail.com>
1853         PR target/60151
1854         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1855         * configure: Regenerated.
1857 2014-02-12  Richard Biener  <rguenther@suse.de>
1859         * vec.c (vec_prefix::calculate_allocation): Move as
1860         inline variant to vec.h.
1861         (vec_prefix::calculate_allocation_1): New out-of-line version.
1862         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1863         (vec_prefix::m_has_auto_buf): Rename to ...
1864         (vec_prefix::m_using_auto_storage): ... this.
1865         (vec_prefix::calculate_allocation): Inline the easy cases
1866         and dispatch to calculate_allocation_1 which doesn't need the
1867         prefix address.
1868         (va_heap::reserve): Use gcc_checking_assert.
1869         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1870         m_using_auto_storage.
1871         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1872         member and adjust.
1873         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1874         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1875         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1877 2014-02-12  Richard Biener  <rguenther@suse.de>
1879         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1880         when we found a dependence.
1882 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1884         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1885         common code...
1886         (maybe_fold_stmt): ... into this new function.
1887         * omp-low.c (lower_omp): Update comment.
1889         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1890         last use.
1892         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1893         dereference.
1895 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1897         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1898         identifiers in comments.
1899         (cortexa53_extra_costs): Likewise.
1900         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1901         (cortexa7_extra_costs): Likewise.
1902         (cortexa12_extra_costs): Likewise.
1903         (cortexa15_extra_costs): Likewise.
1904         (v7m_extra_costs): Likewise.
1906 2014-02-12  Richard Biener  <rguenther@suse.de>
1908         PR middle-end/60092
1909         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1910         of posix_memalign being successful.
1911         (lower_stmt): Restrict lowering of posix_memalign to when
1912         -ftree-bit-ccp is enabled.
1914 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1916         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1917         arg_loc.
1918         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1920 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1922         PR rtl-optimization/60116
1923         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1924         other_insn once the combination has been validated.
1926 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1928         PR lto/59468
1929         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1930         and wrapper.
1931         * ipa-devirt.c: Include demangle.h
1932         (odr_violation_reported): New static variable.
1933         (add_type_duplicate): Update odr_violations.
1934         (maybe_record_node): Add completep parameter; update it.
1935         (record_target_from_binfo): Add COMPLETEP parameter;
1936         update it as needed.
1937         (possible_polymorphic_call_targets_1): Likewise.
1938         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1939         rename FINAL to COMPLETE.
1940         (record_targets_from_bases): Sanity check we found the binfo;
1941         fix COMPLETEP updating.
1942         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1943         parameter, fix computing of COMPLETEP.
1944         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1945         at LTO time do demangling.
1946         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1947         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1948         parameter.
1949         (gimple_get_virt_method_for_binfo): Likewise.
1950         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1951         gimple_get_virt_method_for_vtable): Update prototypes.
1953 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1955         PR target/49008
1956         * genautomata.c (add_presence_absence): Fix typo with
1957         {final_}presence_list.
1959 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1961         PR target/60137
1962         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1963         for VSX/Altivec vectors that land in GPR registers.
1965 2014-02-11  Richard Henderson  <rth@redhat.com>
1966             Jakub Jelinek  <jakub@redhat.com>
1968         PR debug/59776
1969         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1970         around drhs if type conversion to lacc->type is not useless.
1972 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1974         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1975         tuning struct.
1976         (cortex-a57.cortex-a53): Likewise.
1977         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1979 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1981         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1982         arm_restrict_it.
1984 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1986         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1987         add_options_for_arm_vfp3.
1989 2014-02-11  Jeff Law  <law@redhat.com>
1991         PR middle-end/54041
1992         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1993         object with an undesirable mode.
1995 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1997         PR libgomp/60107
1998         * config/i386/sol2-9.h: New file.
1999         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2000         *-*-solaris2.9*): Use it.
2002 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2004         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2005         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2007 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2009         * config/microblaze/microblaze.c: Extend mcpu version format
2011 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
2013         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2015 2014-02-10  Richard Henderson  <rth@redhat.com>
2017         PR target/59927
2018         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2019         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2020         ms-abi vs -mno-accumulate-outgoing-args.
2021         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2022         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2023         respect to ms-abi.
2025 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2027         PR middle-end/60080
2028         * cfgexpand.c (expand_asm_operands): Attach source location to
2029         ASM_INPUT rtx objects.
2030         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2032 2014-02-10  Nick Clifton  <nickc@redhat.com>
2034         * config/mn10300/mn10300.c (popcount): New function.
2035         (mn10300_expand_prologue): Include saved registers in stack usage
2036         count.
2038 2014-02-10  Jeff Law  <law@redhat.com>
2040         PR middle-end/52306
2041         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2042         when changing the SET_DEST of a prior insn to avoid an input reload.
2044 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2046         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2047         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2048         -mcall-openbsd, or -mcall-linux.
2049         (CC1_ENDIAN_BIG_SPEC): Remove.
2050         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2051         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2052         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2053         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2054         and %cc1_endian_default.
2055         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2057 2014-02-10  Richard Biener  <rguenther@suse.de>
2059         PR tree-optimization/60115
2060         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2061         MEM_REF handling.  Properly verify that the accesses are not
2062         out of the objects bound.
2064 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2066         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2067         coretex to cortex.
2069 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2071         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2072         proper constants and fix formatting.
2073         (possible_polymorphic_call_targets): Fix formatting.
2075 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2076             Ilya Tocar  <ilya.tocar@intel.com>
2078         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2079         (_mm512_loadu_epi32): Renamed into...
2080         (_mm512_loadu_si512): This.
2081         (_mm512_storeu_epi32): Renamed into...
2082         (_mm512_storeu_si512): This.
2083         (_mm512_maskz_ceil_ps): Removed.
2084         (_mm512_maskz_ceil_pd): Ditto.
2085         (_mm512_maskz_floor_ps): Ditto.
2086         (_mm512_maskz_floor_pd): Ditto.
2087         (_mm512_floor_round_ps): Ditto.
2088         (_mm512_floor_round_pd): Ditto.
2089         (_mm512_ceil_round_ps): Ditto.
2090         (_mm512_ceil_round_pd): Ditto.
2091         (_mm512_mask_floor_round_ps): Ditto.
2092         (_mm512_mask_floor_round_pd): Ditto.
2093         (_mm512_mask_ceil_round_ps): Ditto.
2094         (_mm512_mask_ceil_round_pd): Ditto.
2095         (_mm512_maskz_floor_round_ps): Ditto.
2096         (_mm512_maskz_floor_round_pd): Ditto.
2097         (_mm512_maskz_ceil_round_ps): Ditto.
2098         (_mm512_maskz_ceil_round_pd): Ditto.
2099         (_mm512_expand_pd): Ditto.
2100         (_mm512_expand_ps): Ditto.
2101         * config/i386/i386.c (ix86_builtins): Remove
2102         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2103         (bdesc_args): Ditto.
2104         * config/i386/predicates.md (const1256_operand): New.
2105         (const_1_to_2_operand): Ditto.
2106         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2107         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2108         (*avx512pf_gatherpf<mode>sf): Ditto.
2109         (avx512pf_gatherpf<mode>df): Ditto.
2110         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2111         (*avx512pf_gatherpf<mode>df): Ditto.
2112         (avx512pf_scatterpf<mode>sf): Ditto.
2113         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2114         (*avx512pf_scatterpf<mode>sf): Ditto.
2115         (avx512pf_scatterpf<mode>df): Ditto.
2116         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2117         (*avx512pf_scatterpf<mode>df): Ditto.
2118         (avx512f_expand<mode>): Removed.
2119         (<shift_insn><mode>3<mask_name>): Change predicate type.
2121 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2123         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2124         not at the end of datarefs vector use ordered_remove to avoid
2125         reordering datarefs vector.
2127         PR c/59984
2128         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2129         mark local addressable non-static vars as GOVD_PRIVATE
2130         instead of GOVD_LOCAL.
2131         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2132         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2133         than copying them.
2135         PR middle-end/60092
2136         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2137         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2138         assume_aligned or alloc_align attributes.
2139         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2140         arguments.  Handle also assume_aligned and alloc_align attributes.
2141         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2142         calls to functions with assume_aligned or alloc_align attributes.
2143         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2145 2014-02-08  Terry Guo  <terry.guo@arm.com>
2147         * doc/invoke.texi: Document ARM -march=armv7e-m.
2149 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2151         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2152         flag on __cilkrts_rethrow builtin.
2154         PR ipa/60026
2155         * ipa-cp.c (determine_versionability): Fail at -O0
2156         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2157         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2159         Revert:
2160         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2162         PR ipa/60026
2163         * tree-inline.c (copy_forbidden): Fail for
2164         __attribute__((optimize (0))) functions.
2166 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2168         * varpool.c: Include pointer-set.h.
2169         (varpool_remove_unreferenced_decls): Variables in other partitions
2170         will not be output; be however careful to not lose information
2171         about partitioning.
2173 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2175         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2176         lookup in the vtable constructor.
2178 2014-02-07  Jeff Law  <law@redhat.com>
2180         PR target/40977
2181         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2182         define_insn_and_split.
2184         * ipa-inline.c (inline_small_functions): Fix typos.
2186 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2188         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2189         (s390_can_use_return_insn): Declare.
2190         * config/s390/s390.h (EPILOGUE_USES): Define.
2191         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2192         instructions.
2193         (s390_chunkify_start): Handle return JUMP_LABELs.
2194         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2195         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2196         (s390_can_use_return_insn): New functions.
2197         (s390_fix_long_loop_prediction): Handle conditional returns.
2198         (TARGET_SET_UP_BY_PROLOGUE): Define.
2199         * config/s390/s390.md (ANY_RETURN): New code iterator.
2200         (*creturn, *csimple_return, return, simple_return): New patterns.
2202 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2204         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2205         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2206         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2207         REG_CFA_RESTORE list when deciding not to restore a register.
2209 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2211         * config/s390/s390.c: Include tree-pass.h and context.h.
2212         (s390_early_mach): New function, split out from...
2213         (s390_emit_prologue): ...here.
2214         (pass_data_s390_early_mach): New pass structure.
2215         (pass_s390_early_mach): New class.
2216         (s390_option_override): Create and register early_mach pass.
2217         Move to end of file.
2219 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2221         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2222         to match for the exit block.
2224 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2226         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2227         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2228         Reject misaligned operands.
2230 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2232         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2234 2014-02-07  Richard Biener  <rguenther@suse.de>
2236         PR middle-end/60092
2237         * gimple-low.c (lower_builtin_posix_memalign): New function.
2238         (lower_stmt): Call it to lower posix_memalign in a way
2239         to make alignment info accessible.
2241 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2243         PR c++/60082
2244         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2245         __builtin_setjmp_receiver.
2247 2014-02-07  Richard Biener  <rguenther@suse.de>
2249         PR middle-end/60092
2250         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2251         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2252         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2253         Handle BUILT_IN_POSIX_MEMALIGN.
2254         (find_func_clobbers): Likewise.
2255         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2256         (call_may_clobber_ref_p_1): Likewise.
2258 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2260         PR ipa/59918
2261         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2262         sanity check.
2264 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2266         PR ipa/59469
2267         * lto-cgraph.c (lto_output_node): Use
2268         symtab_get_symbol_partitioning_class.
2269         (lto_output_varpool_node): likewise.
2270         (symtab_get_symbol_partitioning_class): Move here from
2271         lto/lto-partition.c
2272         * cgraph.h (symbol_partitioning_class): Likewise.
2273         (symtab_get_symbol_partitioning_class): Declare.
2275 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2277         * ggc.h (ggc_internal_cleared_alloc): New macro.
2278         * vec.h (vec_safe_copy): Handle memory stats.
2279         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2280         * target-globals.c (save_target_globals): Likewise.
2282 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2284         PR target/60077
2285         * expr.c (emit_move_resolve_push): Export; be bit more selective
2286         on when to clear alias set.
2287         * expr.h (emit_move_resolve_push): Declare.
2288         * function.h (struct function): Add tail_call_marked.
2289         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2290         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2291         * config/i386/i386.md (TImode move expander): De not call
2292         ix86_expand_push.
2293         (FP push expanders): Preserve memory attributes.
2294         * config/i386/sse.md (push<mode>1): Remove.
2295         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2296         (ix86_expand_push): Remove.
2297         * config/i386/mmx.md (push<mode>1): Remove.
2299 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2301         PR rtl-optimization/60030
2302         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2303         lopart with paradoxical subreg before shifting it up by hprec.
2305 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2307         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2308         Remove extra newline at end of file.
2309         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2310         (arm_issue_rate): Handle cortexa57.
2311         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2312         (cortex-a57.cortex-a53): Likewise.
2314 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2316         PR target/59575
2317         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2318         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2319         bitmask.
2320         (arm_expand_prologue): Adjust all callers.
2321         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2322         info, registers also at the lowest numbered registers side.  Use
2323         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2324         XEXP.
2326         PR debug/59992
2327         * var-tracking.c (adjust_mems): Before adding a SET to
2328         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2330 2014-02-06  Alan Modra  <amodra@gmail.com>
2332         PR target/60032
2333         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2334         change SDmode to DDmode when lra_in_progress.
2336 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2338         PR middle-end/59150
2339         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2340         free_data_ref on the dr first, and before goto again also set dr
2341         to the next dr.  For simd_lane_access, free old datarefs[i] before
2342         overwriting it.  For get_vectype_for_scalar_type failure, don't
2343         free_data_ref if simd_lane_access.
2345         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2347         PR target/60062
2348         * tree.h (opts_for_fn): New inline function.
2349         (opt_for_fn): Define.
2350         * config/i386/i386.c (ix86_function_regparm): Use
2351         opt_for_fn (decl, optimize) instead of optimize.
2353 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2355         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2356         for SYMBOL_REF in large memory model.
2358 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2360         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2361         and crypto support.
2362         (cortex-a57): Likewise.
2363         (cortex-a57.cortex-a53): Likewise.
2365 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2366             Kugan Vivekanandarajah  <kuganv@linaro.org>
2368         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2369         unaligned_access.
2370         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2372 2014-02-06  Richard Biener  <rguenther@suse.de>
2374         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2375         set_loop_copy and initialize_original_copy_tables.
2377 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2379         * config/aarch64/aarch64-simd.md
2380         (aarch64_ashr_simddi): Change QI to SI.
2382 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2383             Jakub Jelinek  <jakub@redhat.com>
2385         PR middle-end/60013
2386         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2387         of the dataflow.
2389 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2391         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2392         CODE_FOR_altivec_vpku[hw]um to
2393         CODE_FOR_altivec_vpku[hw]um_direct.
2394         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2395         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2396         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2397         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2399 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2401         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2402         generation for -maltivec=be.
2403         (altivec_vsumsws): Simplify redundant test.
2405 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2407         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2408         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2409         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2410         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2411         gen_altivec_vpkuwum.
2412         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2413         BYTES_BIG_ENDIAN.
2414         (altivec_vpks<VI_char>ss): Likewise.
2415         (altivec_vpks<VI_char>us): Likewise.
2416         (altivec_vpku<VI_char>us): Likewise.
2417         (altivec_vpku<VI_char>um): Likewise.
2418         (altivec_vpku<VI_char>um_direct): New (copy of
2419         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2420         internal use).
2421         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2422         target is little endian and -maltivec=be is not specified.
2423         (*altivec_vupkhs<VU_char>_direct): New (copy of
2424         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2425         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2426         target is little endian and -maltivec=be is not specified.
2427         (*altivec_vupkls<VU_char>_direct): New (copy of
2428         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2429         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2430         little endian and -maltivec=be is not specified.
2431         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2432         little endian and -maltivec=be is not specified.
2434 2014-02-05  Richard Henderson  <rth@redhat.com>
2436         PR debug/52727
2437         * combine-stack-adj.c: Revert r206943.
2438         * sched-int.h (struct deps_desc): Add last_args_size.
2439         * sched-deps.c (init_deps): Initialize it.
2440         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2441         contain REG_ARGS_SIZE notes.
2443 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2445         * lto-cgraph.c (asm_nodes_output): Make global.
2446         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2447         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2448         (driver_handle_option): Handle OPT_fwpa.
2450 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2452         PR ipa/59947
2453         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2454         a comment typo and formatting issue.  If odr_hash hasn't been
2455         created, return vNULL and set *completep to false.
2457         PR middle-end/57499
2458         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2459         bb with no successors.
2461 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2463         PR target/59718
2464         * doc/invoke.texi (-march): Clarify documentation for ARM.
2465         (-mtune): Likewise.
2466         (-mcpu): Likewise.
2468 2014-02-05  Richard Biener  <rguenther@suse.de>
2470         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2471         when not vectorizing because of too many alias checks.
2472         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2473         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2475 2014-02-05  Nick Clifton  <nickc@redhat.com>
2477         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2478         accept extended registers in any mode when compiling for the MN10300.
2480 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2482         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2483         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2484         sanitization attributes.
2485         (can_inline_edge_p): Likewise.
2486         (sanitize_attrs_match_for_inline_p): New function.
2488 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2490         * ipa-prop.c (detect_type_change): Shor circuit testing of
2491         type changes on THIS pointer.
2493 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2495         PR target/59777
2496         * config/pa/pa.c (legitimize_tls_address): Return original address
2497         if not passed a SYMBOL_REF rtx.
2498         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2499         addresses.
2500         (pa_emit_move_sequence): Simplify TLS source operands.
2501         (pa_legitimate_constant_p): Reject all TLS constants.
2502         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2503         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2505 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2507         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2508         groups when we know they are controlled by LTO.
2509         * varasm.c (default_binds_local_p_1): If object is in other partition,
2510         it will be resolved locally.
2512 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2514         * config/host-linux.c (linux_gt_pch_use_address): Don't
2515         use SSIZE_MAX because it is not always defined.
2517 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2519         PR bootstrap/59913
2520         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2521         threshold for pseudo splitting.
2522         (update_ebb_live_info): Process call argument hard registers and
2523         hard registers from insn definition too.
2524         (max_small_class_regs_num): New constant.
2525         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2526         reloads_num only for small register classes.  Don't split for
2527         outputs of jumps.
2529 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2531         PR ipa/60058
2532         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2533         is non-null.
2535 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2537         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2538         visibility is safe.
2540 2014-02-04  Marek Polacek  <polacek@redhat.com>
2542         * gdbinit.in (pel): Define.
2544 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2546         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2547         behavior.
2549 2014-02-04  Richard Biener  <rguenther@suse.de>
2551         PR lto/59723
2552         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2553         in function context local.
2554         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2555         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2556         similar to LTO_imported_decl_ref.
2558 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2560         PR tree-optimization/60002
2561         * cgraphclones.c (build_function_decl_skip_args): Clear
2562         DECL_LANG_SPECIFIC.
2564         PR tree-optimization/60023
2565         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2566         false to gsi_replace.
2567         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2568         has been in some EH region and vec_stmt could throw, add
2569         vec_stmt into the same EH region.
2570         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2571         has no lhs, ignore it.
2572         * internal-fn.c (expand_MASK_LOAD): Likewise.
2574         PR ipa/60026
2575         * tree-inline.c (copy_forbidden): Fail for
2576         __attribute__((optimize (0))) functions.
2578         PR other/58712
2579         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2580         is set, copy one less argument.
2581         (expand_simd_clones): Don't subtract clone_info->inbranch
2582         from simd_clone_struct_alloc argument.
2584         PR rtl-optimization/57915
2585         * recog.c (simplify_while_replacing): If all unary/binary/relational
2586         operation arguments are constant, attempt to simplify those.
2588         PR middle-end/59261
2589         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2590         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2592 2014-02-04  Richard Biener  <rguenther@suse.de>
2594         PR tree-optimization/60012
2595         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2596         TBAA disambiguation to all DDRs.
2598 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2600         PR target/59788
2601         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2602         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2604 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2606         PR ipa/59882
2607         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2609 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2611         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2612         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2614 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2616         PR ipa/59831
2617         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2618         to figure out targets of polymorphic calls with known decl.
2619         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2620         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2621         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2622         (get_polymorphic_call_info): ... here.
2623         (get_polymorphic_call_info_from_invariant): New function.
2625 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2627         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2628         lookup via vtable pointer; check for type consistency
2629         and turn inconsitent facts into UNREACHABLE.
2630         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2631         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2632         type inconsistent querries; return UNREACHABLE instead.
2634 2014-02-03  Richard Henderson  <rth@twiddle.net>
2636         PR tree-opt/59924
2637         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2638         already processed this node.
2639         (normalize_one_pred_1): Pass along mark_set.
2640         (normalize_one_pred): Create and destroy a pointer_set_t.
2641         (normalize_one_pred_chain): Likewise.
2643 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2645         PR gcov-profile/58602
2646         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2648 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2650         PR ipa/59831
2651         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2652         -fno-devirtualize; try to devirtualize by the knowledge of
2653         virtual table pointer given by aggregate propagation.
2654         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2655         (ipa_print_node_jump_functions): Dump also offset that
2656         is relevant for polymorphic calls.
2657         (determine_known_aggregate_parts): Add arg_type parameter; use it
2658         instead of determining the type from pointer type.
2659         (ipa_compute_jump_functions_for_edge): Update call of
2660         determine_known_aggregate_parts.
2661         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2662         (gimple_get_virt_method_for_binfo): ... here; simplify using
2663         vtable_pointer_value_to_vtable.
2664         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2665         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2666         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2667         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2668         POINTER_PLUS_EXPR.
2669         (vtable_pointer_value_to_binfo): ... here.
2670         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2672 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2674         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2675         redef of outer loop index variable.
2677 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2679         PR c++/53017
2680         PR c++/59211
2681         * doc/extend.texi (Function Attributes): Typo.
2683 2014-02-03  Cong Hou  <congh@google.com>
2685         PR tree-optimization/60000
2686         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2687         if the vectorized statement is a store.  A store statement can only
2688         appear at the end of pattern statements.
2690 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2692         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2693         (ix86_option_override_internal): Default long double to 64-bit for
2694         32-bit Bionic and to 128-bit for 64-bit Bionic.
2696         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2697         TARGET_LONG_DOUBLE_128 is true.
2698         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2700         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2701         (mlong-double-64): Negate -mlong-double-128.
2702         (mlong-double-128): New option.
2704         * config/i386/i386-c.c (ix86_target_macros): Define
2705         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2707         * doc/invoke.texi: Document -mlong-double-128.
2709 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2711         PR rtl-optimization/60024
2712         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2714 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2716         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2718 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2720         PR rtl-optimization/57662
2721         * sel-sched.c (code_motion_path_driver): Do not mark already not
2722         existing blocks in the visiting bitmap.
2724 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2726         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2727         on the insn being emitted.
2729 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2730             Will Deacon  <will.deacon@arm.com>
2732         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2734 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2736         * config/arm/arm-tables.opt: Regenerate.
2738 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2740         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2741         for vector types other than V16QImode.
2742         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2743         define_expand, and call altivec_expand_vec_perm_le when producing
2744         code with little endian element order.
2745         (*altivec_vperm_<mode>_internal): New insn having previous
2746         behavior of altivec_vperm_<mode>.
2747         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2748         altivec_expand_vec_perm_le when producing code with little endian
2749         element order.
2750         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2751         behavior of altivec_vperm_<mode>_uns.
2753 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2755         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2756         (altivec_vsumsws): Add handling for -maltivec=be with a little
2757         endian target.
2758         (altivec_vsumsws_direct): New.
2759         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2760         gen_altivec_vsumsws.
2762 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2764         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2765         vtable_pointer_value_to_binfo): New functions.
2766         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2767         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2769 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2771         * config/nios2/nios2.md (load_got_register): Initialize GOT
2772         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2773         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2775 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2777         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2778         preserverd by passthrough, do not propagate the type.
2780 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2782         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2783         (mips_atomic_assign_expand_fenv): New function.
2784         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2786 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2788         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2789         (__builtin_mips_set_fcsr): Likewise.
2790         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2791         MIPS_USI_FTYPE_VOID.
2792         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2793         (mips16_expand_set_fcsr): Likewise.
2794         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2795         (mips16_set_fcsr_stub): Likewise.
2796         (mips16_get_fcsr_one_only_stub): New class.
2797         (mips16_set_fcsr_one_only_stub): Likewise.
2798         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2799         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2800         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2801         (hard_float): New availability predicate.
2802         (mips_builtins): Add get_fcsr and set_fcsr.
2803         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2804         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2805         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2806         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2807         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2808         patterns.
2810 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2812         * config/mips/mips.c (mips_one_only_stub): New class.
2813         (mips_need_mips16_rdhwr_p): Replace with...
2814         (mips16_rdhwr_stub): ...this new variable.
2815         (mips16_stub_call_address): New function.
2816         (mips16_rdhwr_one_only_stub): New class.
2817         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2818         (mips_output_mips16_rdhwr): Delete.
2819         (mips_finish_stub): New function.
2820         (mips_code_end): Use it to handle rdhwr stubs.
2822 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2824         PR target/60017
2825         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2826         when calculating size of integer atomic types.
2828 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2830         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2832 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2834         PR tree-optimization/60003
2835         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2836         * profile.c (branch_prob): Use gimple_call_builtin_p
2837         to check for BUILT_IN_SETJMP_RECEIVER.
2838         * tree-inline.c (copy_bb): Call notice_special_calls.
2840 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2842         PR bootstrap/59985
2843         * lra-constraints.c (process_alt_operands): Update reload_sum only
2844         on the first pass.
2846 2014-01-31  Richard Henderson  <rth@redhat.com>
2848         PR middle-end/60004
2849         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2850         until after else_eh is processed.
2852 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2854         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2855         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2856         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2857         in smmintrin.h, remove them.
2858         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2859         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2860         * config/i386/i386.md (ROUND_SAE): Fix value.
2861         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2862         (const48_operand): New.
2863         * config/i386/subst.md (round), (round_expand): Use
2864         const_4_or_8_to_11_operand.
2865         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2867 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2869         * config/i386/constraints.md (Yk): Swap meaning with k.
2870         * config/i386/i386.md (movhi_internal): Change Yk to k.
2871         (movqi_internal): Ditto.
2872         (*k<logic><mode>): Ditto.
2873         (*andhi_1): Ditto.
2874         (*andqi_1): Ditto.
2875         (kandn<mode>): Ditto.
2876         (*<code>hi_1): Ditto.
2877         (*<code>qi_1): Ditto.
2878         (kxnor<mode>): Ditto.
2879         (kortestzhi): Ditto.
2880         (kortestchi): Ditto.
2881         (kunpckhi): Ditto.
2882         (*one_cmplhi2_1): Ditto.
2883         (*one_cmplqi2_1): Ditto.
2884         * config/i386/sse.md (): Change k to Yk.
2885         (avx512f_load<mode>_mask): Ditto.
2886         (avx512f_blendm<mode>): Ditto.
2887         (avx512f_store<mode>_mask): Ditto.
2888         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2889         (avx512f_storedqu<mode>_mask): Ditto.
2890         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2891         Ditto.
2892         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2893         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2894         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2895         (avx512f_maskcmp<mode>3): Ditto.
2896         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2897         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2898         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2899         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2900         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2901         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2902         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2903         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2904         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2905         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2906         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2907         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2908         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2909         (vec_extract_lo_<mode>_maskm): Ditto.
2910         (vec_extract_hi_<mode>_maskm): Ditto.
2911         (avx512f_vternlog<mode>_mask): Ditto.
2912         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2913         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2914         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2915         (avx512f_<code>v8div16qi2_mask): Ditto.
2916         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2917         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2918         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2919         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2920         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2921         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2922         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2923         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2924         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2925         (avx512cd_maskb_vec_dupv8di): Ditto.
2926         (avx512cd_maskw_vec_dupv16si): Ditto.
2927         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2928         (avx512f_vpermi2var<mode>3_mask): Ditto.
2929         (avx512f_vpermi2var<mode>3_mask): Ditto.
2930         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2931         (*avx512f_gathersi<mode>): Ditto.
2932         (*avx512f_gathersi<mode>_2): Ditto.
2933         (*avx512f_gatherdi<mode>): Ditto.
2934         (*avx512f_gatherdi<mode>_2): Ditto.
2935         (*avx512f_scattersi<mode>): Ditto.
2936         (*avx512f_scatterdi<mode>): Ditto.
2937         (avx512f_compress<mode>_mask): Ditto.
2938         (avx512f_compressstore<mode>_mask): Ditto.
2939         (avx512f_expand<mode>_mask): Ditto.
2940         * config/i386/subst.md (mask): Change k to Yk.
2941         (mask_scalar_merge): Ditto.
2942         (sd): Ditto.
2944 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2946         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2948 2014-01-31  Richard Biener  <rguenther@suse.de>
2950         PR middle-end/59990
2951         * builtins.c (fold_builtin_memory_op): Make sure to not
2952         use a floating-point mode or a boolean or enumeral type for
2953         the copy operation.
2955 2014-01-30  DJ Delorie  <dj@redhat.com>
2957         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2958         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2959         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2960         whenever main() has an epilogue.
2962 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2964         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2965         unused variable "field".
2966         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2967         (vsx_mergeh_<mode>): Likewise.
2968         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2969         (altivec_vmrghh): Likewise.
2970         (altivec_vmrghw): Likewise.
2971         (altivec_vmrglb): Likewise.
2972         (altivec_vmrglh): Likewise.
2973         (altivec_vmrglw): Likewise.
2974         (altivec_vspltb): Add missing uses.
2975         (altivec_vsplth): Likewise.
2976         (altivec_vspltw): Likewise.
2977         (altivec_vspltsf): Likewise.
2979 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2981         PR target/59923
2982         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2983         frame related instructions.
2985 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2987         PR rtl-optimization/59959
2988         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2989         any reload of register whose subreg is invalid.
2991 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2993         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2994         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2995         Add missing return type - void.
2997 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2999         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3000         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3001         remove element index adjustment for endian (now handled in vsx.md
3002         and altivec.md).
3003         (altivec_expand_vec_perm_const): Use
3004         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3005         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3006         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3007         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3008         define_expand and a new define_insn *altivec_vspltb_internal;
3009         adjust for -maltivec=be on a little endian target.
3010         (altivec_vspltb_direct): New.
3011         (altivec_vsplth): Divide into a define_expand and a new
3012         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3013         little endian target.
3014         (altivec_vsplth_direct): New.
3015         (altivec_vspltw): Divide into a define_expand and a new
3016         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3017         little endian target.
3018         (altivec_vspltw_direct): New.
3019         (altivec_vspltsf): Divide into a define_expand and a new
3020         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3021         a little endian target.
3023 2014-01-30  Richard Biener  <rguenther@suse.de>
3025         PR tree-optimization/59993
3026         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3027         can propagate form the earlier stmt and avoid the transform
3028         when the intermediate result is needed.
3030 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3032         * README.Portability: Fix typo.
3034 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
3036         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3037         comparison_operator with ordered_comparison_operator.
3039 2014-01-30  Nick Clifton  <nickc@redhat.com>
3041         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3042         Rename to mn10300_store_multiple_regs.
3043         * config/mn10300/mn10300.c: Likewise.
3044         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3045         store_multiple_regs.
3046         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3047         Call mn10300_store_multiple_regs.
3049 2014-01-30  Nick Clifton  <nickc@redhat.com>
3050             DJ Delorie  <dj@redhat.com>
3052         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3053         %fp 2 to keep registers after it properly word-aligned.
3054         (rl78_alloc_physical_registers_umul): Handle the case where both
3055         input operands are the same.
3057 2014-01-30  Richard Biener  <rguenther@suse.de>
3059         PR tree-optimization/59903
3060         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3061         check properly.
3063 2014-01-30  Jason Merrill  <jason@redhat.com>
3065         PR c++/59633
3066         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3068         PR c++/59645
3069         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3071 2014-01-30  Richard Biener  <rguenther@suse.de>
3073         PR tree-optimization/59951
3074         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3076 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3078         PR target/59784
3079         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3080         SFmode to DFmode case.
3082 2014-01-29  DJ Delorie  <dj@redhat.com>
3084         * config/msp430/msp430.opt (-minrt): New.
3085         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3086         if -minrt given.
3087         (ENDFILE_SPEC): Likewise.
3089 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3091         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3092         (estimate_function_body_sizes): Use it.
3094 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3096         PR c++/58561
3097         * dwarf2out.c (is_cxx_auto): New.
3098         (is_base_type): Use it.
3099         (gen_type_die_with_usage): Likewise.
3101 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3103         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3104         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3105         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3106         -maltivec=be with LE targets.
3107         (vsx_mergeh_<mode>): Likewise.
3108         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3109         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3110         (altivec_vmrghb): Replace with define_expand and new
3111         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3112         (altivec_vmrghb_direct): New define_insn.
3113         (altivec_vmrghh): Replace with define_expand and new
3114         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3115         (altivec_vmrghh_direct): New define_insn.
3116         (altivec_vmrghw): Replace with define_expand and new
3117         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3118         (altivec_vmrghw_direct): New define_insn.
3119         (*altivec_vmrghsf): Adjust for endianness.
3120         (altivec_vmrglb): Replace with define_expand and new
3121         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3122         (altivec_vmrglb_direct): New define_insn.
3123         (altivec_vmrglh): Replace with define_expand and new
3124         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3125         (altivec_vmrglh_direct): New define_insn.
3126         (altivec_vmrglw): Replace with define_expand and new
3127         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3128         (altivec_vmrglw_direct): New define_insn.
3129         (*altivec_vmrglsf): Adjust for endianness.
3130         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3131         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3132         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3133         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3134         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3135         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3136         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3137         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3139 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3141         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3142         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3143         whitespace.
3145 2014-01-29  Richard Biener  <rguenther@suse.de>
3147         PR tree-optimization/58742
3148         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3149         associate_pointerplus_align.
3150         (associate_pointerplus_diff): New function.
3151         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3152         and associate_pointerplus_diff.
3154 2014-01-29  Richard Biener  <rguenther@suse.de>
3156         * lto-streamer.h (LTO_major_version): Bump to 3.
3157         (LTO_minor_version): Reset to 0.
3159 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3161         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3162         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3163         (arm_file_start): Generate correct asm header for armv7ve.
3164         * config/arm/bpabi.h: Add multilib support for armv7ve.
3165         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3166         and cortex-a15 to armv7ve.
3167         * config/arm/t-aprofile: Add multilib support for armv7ve.
3168         * doc/invoke.texi: Document -march=armv7ve.
3170 2014-01-29  Richard Biener  <rguenther@suse.de>
3172         PR tree-optimization/58742
3173         * tree-ssa-forwprop.c (associate_plusminus): Return true
3174         if we changed sth, defer EH cleanup to ...
3175         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3176         (simplify_mult): New function.
3178 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3180         PR middle-end/59917
3181         PR tree-optimization/59920
3182         * tree.c (build_common_builtin_nodes): Remove
3183         __builtin_setjmp_dispatcher initialization.
3184         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3185         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3186         instead of gsi_after_labels + manually skipping debug stmts.
3187         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3188         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3189         * tree-inline.c (copy_edges_for_bb): Remove
3190         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3191         argument.  Ignore computed_goto_p stmts.  Don't call
3192         make_abnormal_goto_edges.  If a call might need abnormal edges
3193         for non-local gotos, see if it already has an edge to
3194         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3195         with true argument, don't do anything then, otherwise add
3196         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3197         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3198         caller.
3199         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3200         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3201         (lower_stmt): Don't set data->calls_builtin_setjmp.
3202         (lower_builtin_setjmp): Adjust comment.
3203         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3204         * tree-cfg.c (found_computed_goto): Remove.
3205         (factor_computed_gotos): Remove.
3206         (make_goto_expr_edges): Return bool, true for computed gotos.
3207         Don't call make_abnormal_goto_edges.
3208         (build_gimple_cfg): Don't set found_computed_goto, don't call
3209         factor_computed_gotos.
3210         (computed_goto_p): No longer static.
3211         (make_blocks): Don't set found_computed_goto.
3212         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3213         (make_edges): If make_goto_expr_edges returns true, push bb
3214         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3215         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3216         vector.  Record mapping between bbs and OpenMP regions if there
3217         are any, adjust make_gimple_omp_edges caller.  Call
3218         handle_abnormal_edges.
3219         (make_abnormal_goto_edges): Remove.
3220         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3221         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3222         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3223         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3224         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3225         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3226         filling *region also set *region_idx to (*region)->entry->index.
3228         PR other/58712
3229         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3230         For REGs set ORIGINAL_REGNO.
3232 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3234         * doc/md.texi: Mention that a target shouldn't implement
3235         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3236         than hi/lo pair.
3238 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3240         PR tree-optimization/59594
3241         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3242         a copy of the datarefs vector rather than the vector itself.
3244 2014-01-28  Jason Merrill  <jason@redhat.com>
3246         PR c++/53756
3247         * dwarf2out.c (auto_die): New static.
3248         (gen_type_die_with_usage): Handle C++1y 'auto'.
3249         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3250         on definition.
3252 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3254         PR target/59672
3255         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3256         (SPEC_X32): Likewise.
3257         (SPEC_64): Likewise.
3258         * config/i386/i386.c (ix86_option_override_internal): Turn off
3259         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3260         for TARGET_16BIT.
3261         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3262         * config/i386/i386.h (TARGET_16BIT): New macro.
3263         (TARGET_16BIT_P): Likewise.
3264         * config/i386/i386.opt: Add m16.
3265         * doc/invoke.texi: Document -m16.
3267 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3269         PR preprocessor/59935
3270         * input.c (location_get_source_line): Bail out on when line number
3271         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3273 2014-01-28  Richard Biener  <rguenther@suse.de>
3275         PR tree-optimization/58742
3276         * tree-ssa-forwprop.c (associate_plusminus): Handle
3277         pointer subtraction of the form (T)(P + A) - (T)P.
3279 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3281         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3282         at const_int_cost.
3284 2014-01-28  Richard Biener  <rguenther@suse.de>
3286         Revert
3287         2014-01-28  Richard Biener  <rguenther@suse.de>
3289         PR rtl-optimization/45364
3290         PR rtl-optimization/59890
3291         * var-tracking.c (local_get_addr_clear_given_value): Handle
3292         already cleared slot.
3293         (val_reset): Handle not allocated local_get_addr_cache.
3294         (vt_find_locations): Use post-order on the inverted CFG.
3296 2014-01-28  Richard Biener  <rguenther@suse.de>
3298         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3300 2014-01-28  Richard Biener  <rguenther@suse.de>
3302         PR rtl-optimization/45364
3303         PR rtl-optimization/59890
3304         * var-tracking.c (local_get_addr_clear_given_value): Handle
3305         already cleared slot.
3306         (val_reset): Handle not allocated local_get_addr_cache.
3307         (vt_find_locations): Use post-order on the inverted CFG.
3309 2014-01-28  Alan Modra  <amodra@gmail.com>
3311         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3312         * configure.ac <recursive call for build != host>: Define
3313         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3314         and LD_FOR_BUILD too.
3315         * configure: Regenerate.
3317 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3319         * config/i386/i386.c (get_builtin_code_for_version): Separate
3320         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3321         Broadwell from Haswell.
3323 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3325         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3326         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3327         * config/mips/mips.c (mips_option_override): Change setting
3328         of TARGET_DSP.
3329         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3330         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3331         Change from Mask to Var.
3333 2014-01-27  Jeff Law  <law@redhat.com>
3335         * ipa-inline.c (inline_small_functions): Fix typo.
3337 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3339         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3340         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3341         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3342         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3343         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3344         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3345         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3346         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3347         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3348         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3349         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3350         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3351         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3352         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3353         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3354         (_mm512_storeu_epi64): Ditto.
3355         (_mm512_cmpge_epi32_mask): Ditto.
3356         (_mm512_cmpge_epu32_mask): Ditto.
3357         (_mm512_cmpge_epi64_mask): Ditto.
3358         (_mm512_cmpge_epu64_mask): Ditto.
3359         (_mm512_cmple_epi32_mask): Ditto.
3360         (_mm512_cmple_epu32_mask): Ditto.
3361         (_mm512_cmple_epi64_mask): Ditto.
3362         (_mm512_cmple_epu64_mask): Ditto.
3363         (_mm512_cmplt_epi32_mask): Ditto.
3364         (_mm512_cmplt_epu32_mask): Ditto.
3365         (_mm512_cmplt_epi64_mask): Ditto.
3366         (_mm512_cmplt_epu64_mask): Ditto.
3367         (_mm512_cmpneq_epi32_mask): Ditto.
3368         (_mm512_cmpneq_epu32_mask): Ditto.
3369         (_mm512_cmpneq_epi64_mask): Ditto.
3370         (_mm512_cmpneq_epu64_mask): Ditto.
3371         (_mm512_expand_pd): Ditto.
3372         (_mm512_expand_ps): Ditto.
3373         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3374         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3375         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3376         * config/i386/i386.c (ix86_builtins): Add
3377         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3378         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3379         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3380         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3381         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3382         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3383         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3384         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3385         IX86_BUILTIN_PMOVUSQW512_MEM.
3386         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3387         __builtin_ia32_pmovsqd512mem_mask,
3388         __builtin_ia32_pmovqd512mem_mask,
3389         __builtin_ia32_pmovusqw512mem_mask,
3390         __builtin_ia32_pmovsqw512mem_mask,
3391         __builtin_ia32_pmovqw512mem_mask,
3392         __builtin_ia32_pmovusdw512mem_mask,
3393         __builtin_ia32_pmovsdw512mem_mask,
3394         __builtin_ia32_pmovdw512mem_mask,
3395         __builtin_ia32_pmovqb512mem_mask,
3396         __builtin_ia32_pmovusqb512mem_mask,
3397         __builtin_ia32_pmovsqb512mem_mask,
3398         __builtin_ia32_pmovusdb512mem_mask,
3399         __builtin_ia32_pmovsdb512mem_mask,
3400         __builtin_ia32_pmovdb512mem_mask.
3401         (bdesc_args): Add __builtin_ia32_expanddf512,
3402         __builtin_ia32_expandsf512.
3403         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3404         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3405         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3406         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3407         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3408         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3409         (avx512f_<code>v8div16qi2_mask_store): This.
3410         (avx512f_expand<mode>): New.
3412 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3414         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3415         New.
3416         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3417         (_mm512_prefetch_i32scatter_pd): Ditto.
3418         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3419         (_mm512_prefetch_i64scatter_pd): Ditto.
3420         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3421         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3422         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3423         (_mm512_prefetch_i32scatter_ps): Ditto.
3424         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3425         (_mm512_prefetch_i64scatter_ps): Ditto.
3426         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3427         * config/i386/i386-builtin-types.def: Define
3428         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3429         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3430         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3431         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3432         IX86_BUILTIN_SCATTERPFQPD.
3433         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3434         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3435         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3436         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3437         __builtin_ia32_scatterpfqps.
3438         (ix86_expand_builtin): Expand new built-ins.
3439         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3440         fix memory access data type.
3441         (*avx512pf_gatherpf<mode>_mask): Ditto.
3442         (*avx512pf_gatherpf<mode>): Ditto.
3443         (avx512pf_scatterpf<mode>): Ditto.
3444         (*avx512pf_scatterpf<mode>_mask): Ditto.
3445         (*avx512pf_scatterpf<mode>): Ditto.
3446         (GATHER_SCATTER_SF_MEM_MODE): New.
3447         (avx512pf_gatherpf<mode>df): Ditto.
3448         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3449         (*avx512pf_scatterpf<mode>df): Ditto.
3451 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3453         PR bootstrap/59934
3454         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3455         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3456         reached.
3458 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3460         * common/config/arm/arm-common.c
3461         (arm_rewrite_mcpu): Handle multiple names.
3462         * config/arm/arm.h
3463         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3465 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3467         * gimple-builder.h (create_gimple_tmp): Delete.
3469 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3471         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3472         words comparisons.
3474 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3476         * config/pa/pa.md (call): Generate indirect long calls to non-local
3477         functions when outputing 32-bit code.
3478         (call_value): Likewise except for special call to buggy powf function.
3480         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3481         portable runtime and PIC indirect calls.
3482         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3483         and PIC call sequences.  Use ldo instead of blr to set return register
3484         in PIC call sequence.
3486 2014-01-25  Walter Lee  <walt@tilera.com>
3488         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3489         avoid clobbering a live register.
3491 2014-01-25  Walter Lee  <walt@tilera.com>
3493         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3494         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3495         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3496         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3498 2014-01-25  Walter Lee  <walt@tilera.com>
3500         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3501         arguments on even registers.
3502         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3503         STACK_BOUNDARY.
3504         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3505         (BIGGEST_ALIGNMENT): Ditto.
3506         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3508 2014-01-25  Walter Lee  <walt@tilera.com>
3510         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3511         insns before bundling.
3512         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3514 2014-01-25  Walter Lee  <walt@tilera.com>
3516         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3517         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3518         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3520 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3522         * config/mips/constraints.md (kl): Delete.
3523         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3524         define expands, using...
3525         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3526         instructions for MIPS16.
3527         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3528         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3530 2014-01-25  Walter Lee  <walt@tilera.com>
3532         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3533         (clzdi2): Ditto.
3534         (ffsdi2): Ditto.
3536 2014-01-25  Walter Lee  <walt@tilera.com>
3538         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3539         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3541 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3543         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3544         Handle XOR.
3546 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3548         * print-rtl.c (in_call_function_usage): New var.
3549         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3550         EXPR_LIST mode as mode and not as reg note name.
3552         PR middle-end/59561
3553         * cfgloopmanip.c (copy_loop_info): If
3554         loop->warned_aggressive_loop_optimizations, make sure
3555         the flag is set in target loop too.
3557 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3559         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3560         flag_cilkplus.
3561         * builtins.def: Likewise.
3562         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3563         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3564         * ira.c (ira_setup_eliminable_regset): Likewise.
3565         * omp-low.c (gate_expand_omp): Likewise.
3566         (execute_lower_omp): Likewise.
3567         (diagnose_sb_0): Likewise.
3568         (gate_diagnose_omp_blocks): Likewise.
3569         (simd_clone_clauses_extract): Likewise.
3570         (gate): Likewise.
3572 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3574         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3575         correction for little endian...
3576         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3577         here.
3579 2014-01-24  Jeff Law  <law@redhat.com>
3581         PR tree-optimization/59919
3582         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3583         for non-returning calls.
3585 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3587         * common/config/aarch64/aarch64-common.c
3588         (aarch64_rewrite_mcpu): Handle multiple names.
3589         * config/aarch64/aarch64.h
3590         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3592 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3594         * input.c (add_file_to_cache_tab): Handle the case where fopen
3595         returns NULL.
3597 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3599         PR target/59929
3600         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3601         from push operand if code of push isn't PRE_DEC.
3603 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3605         PR target/59909
3606         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3607         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3608         it is only used for non-atomic loads/stores.
3610         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3611         -mquad-memory or -mquad-memory-atomic switches.
3613         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3614         -mquad-memory-atomic to ISA 2.07 support.
3616         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3617         to separate support of normal quad word memory operations (ldq, stq)
3618         from the atomic quad word memory operations.
3620         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3621         support to separate non-atomic quad word operations from atomic
3622         quad word operations.  Disable non-atomic quad word operations in
3623         little endian mode so that we don't have to swap words after the
3624         load and before the store.
3625         (quad_load_store_p): Add comment about atomic quad word support.
3626         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3627         options printed with -mdebug=reg.
3629         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3630         -mquad-memory-atomic as the test for whether we have quad word
3631         atomic instructions.
3632         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3633         or -mp8-vector are used, allow byte/half-word atomic operations.
3635         * config/rs6000/sync.md (load_lockedti): Insure that the address
3636         is a proper indexed or indirect address for the lqarx instruction.
3637         On little endian systems, swap the hi/lo registers after the lqarx
3638         instruction.
3639         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3640         insure the address is valid for the lqarx instruction.
3641         (store_conditionalti): Insure that the address is a proper indexed
3642         or indirect address for the stqcrx. instruction.  On little endian
3643         systems, swap the hi/lo registers before doing the stqcrx.
3644         instruction.
3645         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3646         insure the address is valid for the stqcrx. instruction.
3648         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3649         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3650         type of quad memory support is available.
3652 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3654         PR regression/59915
3655         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3656         there is a danger of looping.
3658 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3660         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3661         force flag_ira_loop_pressure if set via command line.
3663 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3665         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3666         (ashr_simd): New builtin handling DI mode.
3667         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3668         (aarch64_sshr_simddi): New match pattern.
3669         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3670         (vshrd_n_s64): Likewise.
3671         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3673 2014-01-23  Nick Clifton  <nickc@redhat.com>
3675         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3676         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3677         favour of mcu specific scripts.
3678         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3679         430x multilibs.
3681 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3682             Alex Velenko  <Alex.Velenko@arm.com>
3684         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3685         (vaddv_s16): Likewise.
3686         (vaddv_s32): Likewise.
3687         (vaddv_u8): Likewise.
3688         (vaddv_u16): Likewise.
3689         (vaddv_u32): Likewise.
3690         (vaddvq_s8): Likewise.
3691         (vaddvq_s16): Likewise.
3692         (vaddvq_s32): Likewise.
3693         (vaddvq_s64): Likewise.
3694         (vaddvq_u8): Likewise.
3695         (vaddvq_u16): Likewise.
3696         (vaddvq_u32): Likewise.
3697         (vaddvq_u64): Likewise.
3698         (vaddv_f32): Likewise.
3699         (vaddvq_f32): Likewise.
3700         (vaddvq_f64): Likewise.
3701         (vmaxv_f32): Likewise.
3702         (vmaxv_s8): Likewise.
3703         (vmaxv_s16): Likewise.
3704         (vmaxv_s32): Likewise.
3705         (vmaxv_u8): Likewise.
3706         (vmaxv_u16): Likewise.
3707         (vmaxv_u32): Likewise.
3708         (vmaxvq_f32): Likewise.
3709         (vmaxvq_f64): Likewise.
3710         (vmaxvq_s8): Likewise.
3711         (vmaxvq_s16): Likewise.
3712         (vmaxvq_s32): Likewise.
3713         (vmaxvq_u8): Likewise.
3714         (vmaxvq_u16): Likewise.
3715         (vmaxvq_u32): Likewise.
3716         (vmaxnmv_f32): Likewise.
3717         (vmaxnmvq_f32): Likewise.
3718         (vmaxnmvq_f64): Likewise.
3719         (vminv_f32): Likewise.
3720         (vminv_s8): Likewise.
3721         (vminv_s16): Likewise.
3722         (vminv_s32): Likewise.
3723         (vminv_u8): Likewise.
3724         (vminv_u16): Likewise.
3725         (vminv_u32): Likewise.
3726         (vminvq_f32): Likewise.
3727         (vminvq_f64): Likewise.
3728         (vminvq_s8): Likewise.
3729         (vminvq_s16): Likewise.
3730         (vminvq_s32): Likewise.
3731         (vminvq_u8): Likewise.
3732         (vminvq_u16): Likewise.
3733         (vminvq_u32): Likewise.
3734         (vminnmv_f32): Likewise.
3735         (vminnmvq_f32): Likewise.
3736         (vminnmvq_f64): Likewise.
3738 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3740         * config/aarch64/aarch64-simd.md
3741         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3742         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3743         (*aarch64_mul3_elt<mode>): Likewise.
3744         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3745         (*aarch64_mul3_elt_to_64v2df): Likewise.
3746         (*aarch64_mla_elt<mode>): Likewise.
3747         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3748         (*aarch64_mls_elt<mode>): Likewise.
3749         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3750         (*aarch64_fma4_elt<mode>): Likewise.
3751         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3752         (*aarch64_fma4_elt_to_64v2df): Likewise.
3753         (*aarch64_fnma4_elt<mode>): Likewise.
3754         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3755         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3756         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3757         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3758         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3759         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3760         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3761         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3762         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3764 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3766         * config/aarch64/aarch64-simd.md
3767         (aarch64_be_checked_get_lane<mode>): New define_expand.
3768         * config/aarch64/aarch64-simd-builtins.def
3769         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3770         New builtin definition.
3771         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3772         Use new safe be builtin.
3774 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3776         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3777         New define_insn.
3778         (aarch64_be_st1<mode>): Likewise.
3779         (aarch_ld1<VALL:mode>): Define_expand modified.
3780         (aarch_st1<VALL:mode>): Likewise.
3781         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3782         (UNSPEC_ST1): Likewise.
3784 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3786         * config/microblaze/microblaze.md: Add trap insn and attribute
3788 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3790         PR preprocessor/58580
3791         * input.h (location_get_source_line): Take an additional line_size
3792         parameter.
3793         (void diagnostics_file_cache_fini): Declare new function.
3794         * input.c (struct fcache): New type.
3795         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3796         New static constants.
3797         (diagnostic_file_cache_init, total_lines_num)
3798         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3799         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3800         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3801         (get_next_line, read_next_line, goto_next_line, read_line_num):
3802         New static function definitions.
3803         (diagnostic_file_cache_fini): New function.
3804         (location_get_source_line): Take an additional output line_len
3805         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3806         read_line_num.
3807         * diagnostic.c (diagnostic_finish): Call
3808         diagnostic_file_cache_fini.
3809         (adjust_line): Take an additional input parameter for the length
3810         of the line, rather than calculating it with strlen.
3811         (diagnostic_show_locus): Adjust the use of
3812         location_get_source_line and adjust_line with respect to their new
3813         signature.  While displaying a line now, do not stop at the first
3814         null byte.  Rather, display the zero byte as a space and keep
3815         going until we reach the size of the line.
3816         * Makefile.in: Add vec.o to OBJS-libcommon
3818 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3819             Ilya Tocar     <ilya.tocar@intel.com>
3821         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3822         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3823         (__builtin_ia32_kmov16): Ditto.
3824         * config/i386/i386.md (UNSPEC_KMOV): New.
3825         (kmovw): Ditto.
3827 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3829         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3830         (_mm512_storeu_si512): Ditto.
3832 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3834         PR target/52125
3835         * rtl.h (get_referenced_operands): Declare.
3836         * recog.c (get_referenced_operands): New function.
3837         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3838         operands have been referenced when recording LO_SUM references.
3840 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3842         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3844 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3846         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3847         Enable for generic and recent AMD targets.
3849 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3851         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3852         ARG_SIZE note when adjustment was eliminated.
3854 2014-01-22  Jeff Law  <law@redhat.com>
3856         PR tree-optimization/59597
3857         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3858         in file.  Accept new argument REGISTERING and use it to modify
3859         dump output appropriately.
3860         (register_jump_thread): Corresponding changes.
3861         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3862         thread paths involving joiner blocks.  Add code to dump cancelled
3863         jump threading paths.
3865 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3867         PR rtl-optimization/59477
3868         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3869         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3871 2014-01-22  Tom Tromey  <tromey@redhat.com>
3873         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3874         PARAMS.
3875         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3877 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3879         PR rtl-optimization/59896
3880         * lra-constraints.c (process_alt_operands): Check unused note for
3881         matched operands of insn with no output reloads.
3883 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3885         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3886         (mips_move_from_gpr_cost): Likewise.
3888 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3890         PR rtl-optimization/59858
3891         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3892         ira_class_hard_regs_num.
3893         (process_alt_operands): Increase reject for dying matched operand.
3895 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3897         PR target/59003
3898         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3899         smaller than size, perform several stores or loads and stores
3900         at dst + count - size to store or copy all of size bytes, rather
3901         than just last modesize bytes.
3903 2014-01-20  DJ Delorie  <dj@redhat.com>
3905         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3906         that CLOBBERs are REGs before propogating their values.
3908 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3910         PR middle-end/59789
3911         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3912         (cgraph_inline_failed_type): New function.
3913         * cgraph.h (DEFCIFCODE): Add type.
3914         (cgraph_inline_failed_type_t): New enum.
3915         (cgraph_inline_failed_type): New prototype.
3916         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3917         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3918         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3919         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3920         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3921         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3922         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3923         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3924         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3925         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3926         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3927         OPTIMIZATION_MISMATCH.
3928         * tree-inline.c (expand_call_inline): Emit errors during
3929         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3931 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3933         PR target/59685
3934         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3935         mode attribute in insn output.
3937 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3939         * output.h (output_constant): Delete.
3940         * varasm.c (output_constant): Make private.
3942 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3944         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3946 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3948         PR middle-end/59860
3949         * tree.h (fold_builtin_strcat): New prototype.
3950         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3951         argument, if non-NULL, don't call c_strlen.  Optimize
3952         directly into __builtin_memcpy instead of __builtin_strcpy.
3953         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3954         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3956 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3958         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3959         for SImode_address_operand operands, having only a REG argument.
3961 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3963         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3964         loader name using mbig-endian.
3965         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3967 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3969         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3970         (-mtune): Likewise.
3971         (-mcpu): Likewise.
3973 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3975         * config/aarch64/aarch64-protos.h
3976         (aarch64_cannot_change_mode_class_ptr): Declare.
3977         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3978         aarch64_cannot_change_mode_class_ptr): New.
3979         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3980         backend hook aarch64_cannot_change_mode_class.
3982 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3984         * common/config/aarch64/aarch64-common.c
3985         (aarch64_handle_option): Don't handle any option order logic here.
3986         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3987         selected_cpu, warn on architecture version mismatch.
3988         (aarch64_override_options): Fix parsing order for option strings.
3990 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3991             Iain Sandoe  <iain@codesourcery.com>
3993         PR bootstrap/59496
3994         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3995         warning.  Amend comment to reflect current functionality.
3997 2014-01-20  Richard Biener  <rguenther@suse.de>
3999         PR middle-end/59860
4000         * builtins.c (fold_builtin_strcat): Remove case better handled
4001         by tree-ssa-strlen.c.
4003 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4005         * config/aarch64/aarch64.opt
4006         (mcpu, march, mtune): Make case-insensitive.
4008 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4010         PR target/59880
4011         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4012         if operands[1] is a REG or ZERO_EXTEND of a REG.
4014 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4016         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4018 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4020         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4021         long non-pic millicode calls.
4023 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4025         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4027 2014-01-19  Kito Cheng  <kito@0xlab.org>
4029         * builtins.c (expand_movstr): Check movstr expand done or fail.
4031 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4032             H.J. Lu  <hongjiu.lu@intel.com>
4034         PR target/59379
4035         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4036         to DImode for zero-extended addresses.
4038 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4040         PR rtl-optimization/57763
4041         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4042         on the new indirect jump_insn and increment LABEL_NUSES (label).
4044 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4046         PR bootstrap/59580
4047         PR bootstrap/59583
4048         * config.gcc (x86_archs): New variable.
4049         (x86_64_archs): Likewise.
4050         (x86_cpus): Likewise.
4051         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4052         --with-arch/--with-cpu= options.
4053         Support --with-arch=/--with-cpu={nehalem,westmere,
4054         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4056 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4058         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4059         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4061 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4063         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4065 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4067         PR target/58944
4068         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4069         clear cpp_get_options (parse_in)->warn_unused_macros for
4070         ix86_target_macros_internal with cpp_define.
4072 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4074         * jump.c (delete_related_insns): Keep (use (insn))s.
4075         * reorg.c (redundant_insn): Check for barriers too.
4077 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4079         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4081 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4083         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4084         call to $$dyncall when TARGET_LONG_CALLS is true.
4086 2014-01-17  Jeff Law  <law@redhat.com>
4088         * ree.c (combine_set_extension): Temporarily disable test for
4089         changing number of hard registers.
4091 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4093         PR middle-end/58125
4094         * ipa-inline-analysis.c (inline_free_summary):
4095         Do not free summary of aliases.
4097 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4099         PR middle-end/59706
4100         * gimplify.c (gimplify_expr): Use create_tmp_var
4101         instead of create_tmp_var_raw.  If cond doesn't have
4102         integral type, don't add the IFN_ANNOTATE builtin at all.
4104 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4106         PR ipa/59736
4107         * ipa-cp.c (prev_edge_clone): New variable.
4108         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4109         Also resize prev_edge_clone vector.
4110         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4111         (ipcp_edge_removal_hook): New function.
4112         (ipcp_driver): Register ipcp_edge_removal_hook.
4114 2014-01-17  Andrew Pinski <apinski@cavium.com>
4115             Steve Ellcey  <sellcey@mips.com>
4117         PR target/59462
4118         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4119         of operator mode.
4121 2014-01-17  Jeff Law  <law@redhat.com>
4123         PR middle-end/57904
4124         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4125         so that pass_ccp runs first.
4127 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4129         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4130         (ix86_adjust_cost): Use !TARGET_XXX.
4131         (do_reorder_for_imul): Likewise.
4132         (swap_top_of_ready_list): Likewise.
4133         (ix86_sched_reorder): Likewise.
4135 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4137         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4138         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4139         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4140         (intel_memset): New.  Duplicate slm_memset.
4141         (intel_cost): New.  Duplicate slm_cost.
4142         (m_INTEL): New macro.
4143         (processor_target_table): Add "intel".
4144         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4145         with PROCESSOR_INTEL for "intel".
4146         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4147         PROCESSOR_SILVERMONT.
4148         (ix86_issue_rate): Likewise.
4149         (ix86_adjust_cost): Likewise.
4150         (ia32_multipass_dfa_lookahead): Likewise.
4151         (swap_top_of_ready_list): Likewise.
4152         (ix86_sched_reorder): Likewise.
4153         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4154         instead of TARGET_OPT_AGU.
4155         * config/i386/i386.h (TARGET_INTEL): New.
4156         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4157         (processor_type): Add PROCESSOR_INTEL.
4158         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4159         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4161 2014-01-17  Marek Polacek  <polacek@redhat.com>
4163         PR c/58346
4164         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4165         size is zero.
4167 2014-01-17  Richard Biener  <rguenther@suse.de>
4169         PR tree-optimization/46590
4170         * opts.c (default_options_table): Add entries for
4171         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4172         all enabled at -O1 but not for -Og.
4173         * common.opt (fbranch-count-reg): Remove Init(1).
4174         (fmove-loop-invariants): Likewise.
4175         (ftree-pta): Likewise.
4177 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4179         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4180         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4181         decls to at least the GCC 4.8 used alignments.
4183         PR fortran/59440
4184         * tree-nested.c (convert_nonlocal_reference_stmt,
4185         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4186         of GIMPLE_BIND stmts, adjust associated decls.
4188 2014-01-17  Richard Biener  <rguenther@suse.de>
4190         PR tree-optimization/46590
4191         * vec.h (vec<>::bseach): New member function implementing
4192         binary search according to C89 bsearch.
4193         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4194         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4195         bitmap pointer again.  Make accesses_in_loop a flat array.
4196         (mem_ref_obstack): New global.
4197         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4198         (mark_ref_stored): Likewise.
4199         (ref_indep_loop_p_2): Likewise.
4200         (set_ref_stored_in_loop): New helper function.
4201         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4202         (memref_free): Adjust.
4203         (record_mem_ref_loc): Simplify.
4204         (gather_mem_refs_stmt): Adjust.
4205         (sort_locs_in_loop_postorder_cmp): New function.
4206         (analyze_memory_references): Sort accesses_in_loop after
4207         loop postorder number.
4208         (find_ref_loc_in_loop_cmp): New function.
4209         (for_all_locs_in_loop): Find relevant cluster of locs in
4210         accesses_in_loop and iterate without recursion.
4211         (execute_sm): Avoid uninit warning.
4212         (struct ref_always_accessed): Simplify.
4213         (ref_always_accessed::operator ()): Likewise.
4214         (ref_always_accessed_p): Likewise.
4215         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4216         loop postorder numbers here.
4217         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4218         numbers.
4220 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4222         PR c++/57945
4223         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4224         on decls for which assemble_alias has been called.
4226 2014-01-17  Nick Clifton  <nickc@redhat.com>
4228         * config/msp430/msp430.opt: (mcpu): New option.
4229         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4230         (msp430_option_override): Parse target_cpu.  If the MCU name
4231         matches a generic string, clear target_mcu.
4232         (msp430_attr): Allow numeric interrupt values up to 63.
4233         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4234         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4235         option.
4236         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4237         Add mcpu matches.
4238         * config/msp430/msp430.md (popm): Use %J rather than %I.
4239         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4240         (addhi_cy_i): Use immediate_operand for operand 2.
4241         * doc/invoke.texi: Document -mcpu option.
4243 2014-01-17  Richard Biener  <rguenther@suse.de>
4245         PR rtl-optimization/38518
4246         * df.h (df_analyze_loop): Declare.
4247         * df-core.c: Include cfgloop.h.
4248         (df_analyze_1): Split out main part of df_analyze.
4249         (df_analyze): Adjust.
4250         (loop_inverted_post_order_compute): New function.
4251         (loop_post_order_compute): Likewise.
4252         (df_analyze_loop): New function avoiding whole-function
4253         postorder computes.
4254         * loop-invariant.c (find_defs): Use df_analyze_loop.
4255         (find_invariants): Adjust.
4256         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4258 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4260         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4261         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4263 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4265         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4266         traversal when removing references.
4268 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4270         PR ipa/59775
4271         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4273 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4275         PR middle-end/56791
4276         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4277         pushing a reload for an autoinc when we had previously reloaded an
4278         inner part of the address.
4280 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4282         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4283         field.
4284         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4285         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4286         when not giving up or versioning for alias only because of
4287         loop->safelen.
4288         (vect_analyze_data_ref_dependences): Set to true.
4289         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4290         is a GIMPLE_PHI.
4291         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4292         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4293         to the condition.
4295         PR middle-end/58344
4296         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4298         PR target/59839
4299         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4300         operand 0 predicate for gathers, use a new pseudo as subtarget.
4302 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4304         PR middle-end/59609
4305         * lra-constraints.c (process_alt_operands): Add printing debug info.
4306         Check absence of input/output reloads for matched operands too.
4308 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4310         PR rtl-optimization/59835
4311         * ira.c (ira_init_register_move_cost): Increase cost for
4312         impossible modes.
4314 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4316         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4318 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4320         PR target/59780
4321         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4322         non-register objects.  Use gen_(high/low)part more consistently.
4323         Fix assertions.
4325 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4327         PR target/59844
4328         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4329         endian support, remove tests for WORDS_BIG_ENDIAN.
4330         (p8_mfvsrd_3_<mode>): Likewise.
4331         (reload_gpr_from_vsx<mode>): Likewise.
4332         (reload_gpr_from_vsxsf): Likewise.
4333         (p8_mfvsrd_4_disf): Likewise.
4335 2014-01-16  Richard Biener  <rguenther@suse.de>
4337         PR rtl-optimization/46590
4338         * lcm.c (compute_antinout_edge): Use postorder iteration.
4339         (compute_laterin): Use inverted postorder iteration.
4341 2014-01-16  Nick Clifton  <nickc@redhat.com>
4343         PR middle-end/28865
4344         * varasm.c (output_constant): Return the number of bytes actually
4345         emitted.
4346         (output_constructor_array_range): Update the field size with the
4347         number of bytes emitted by output_constant.
4348         (output_constructor_regular_field): Likewise.  Also do not
4349         complain if the total number of bytes emitted is now greater
4350         than the expected fieldpos.
4351         * output.h (output_constant): Update prototype and descriptive comment.
4353 2014-01-16  Marek Polacek  <polacek@redhat.com>
4355         PR middle-end/59827
4356         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4357         it is error_mark_node.
4359 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4361         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4362         VALID_AVX256_REG_OR_OI_MODE.
4364 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4366         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4367         current procedure should be profiled.
4369 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4371         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4372         of moving from/to the STACK_REG register class.
4374 2014-01-15  Richard Henderson  <rth@redhat.com>
4376         PR debug/54694
4377         * reginfo.c (global_regs_decl): Globalize.
4378         * rtl.h (global_regs_decl): Declare.
4379         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4380         reserved via global_regs.
4382 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4384         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4386 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4388         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4389         and vmulosh rather than call gen_vec_widen_smult_*.
4390         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4391         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4392         (vec_widen_smult_even_v16qi): Likewise.
4393         (vec_widen_umult_even_v8hi): Likewise.
4394         (vec_widen_smult_even_v8hi): Likewise.
4395         (vec_widen_umult_odd_v16qi): Likewise.
4396         (vec_widen_smult_odd_v16qi): Likewise.
4397         (vec_widen_umult_odd_v8hi): Likewise.
4398         (vec_widen_smult_odd_v8hi): Likewise.
4399         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4400         vmuloub rather than call gen_vec_widen_umult_*.
4401         (vec_widen_umult_lo_v16qi): Likewise.
4402         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4403         vmulosb rather than call gen_vec_widen_smult_*.
4404         (vec_widen_smult_lo_v16qi): Likewise.
4405         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4406         rather than call gen_vec_widen_umult_*.
4407         (vec_widen_umult_lo_v8hi): Likewise.
4408         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4409         rather than call gen_vec_widen_smult_*.
4410         (vec_widen_smult_lo_v8hi): Likewise.
4412 2014-01-15  Jeff Law  <law@redhat.com>
4414         PR tree-optimization/59747
4415         * ree.c (find_and_remove_re): Properly handle case where a second
4416         eliminated extension requires widening a copy created for elimination
4417         of a prior extension.
4418         (combine_set_extension): Ensure that the number of hard regs needed
4419         for a destination register does not change when we widen it.
4421 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4423         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4424         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4425         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4426         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4427         (avr-*-rtems*): Likewise.
4428         (bfin*-rtems*): Likewise.
4429         (moxie-*-rtems*): Likewise.
4430         (h8300-*-rtems*): Likewise.
4431         (i[34567]86-*-rtems*): Likewise.
4432         (lm32-*-rtems*): Likewise.
4433         (m32r-*-rtems*): Likewise.
4434         (m68k-*-rtems*): Likewise.
4435         (microblaze*-*-rtems*): Likewise.
4436         (mips*-*-rtems*): Likewise.
4437         (powerpc-*-rtems*): Likewise.
4438         (sh-*-rtems*): Likewise.
4439         (sparc-*-rtems*): Likewise.
4440         (sparc64-*-rtems*): Likewise.
4441         (v850-*-rtems*): Likewise.
4442         (m32c-*-rtems*): Likewise.
4444 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4446         PR rtl-optimization/59511
4447         * ira.c (ira_init_register_move_cost): Use memory costs for some
4448         cases of register move cost calculations.
4449         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4450         instead of BB frequency.
4451         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4452         * lra-assigns.c (find_hard_regno_for): Ditto.
4454 2014-01-15  Richard Biener  <rguenther@suse.de>
4456         PR tree-optimization/59822
4457         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4458         (vectorizable_load): Use it to hoist defs of uses of invariant
4459         loads out of the loop.
4461 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4462             Kugan Vivekanandarajah  <kuganv@linaro.org>
4464         PR target/59695
4465         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4466         truncation.
4468 2014-01-15  Richard Biener  <rguenther@suse.de>
4470         PR rtl-optimization/59802
4471         * lcm.c (compute_available): Use inverted postorder to seed
4472         the initial worklist.
4474 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4476         PR target/59803
4477         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4478         ADDR_REGS for invalid symrefs in non-PIC code.
4480 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4482         PR other/58712
4483         * builtins.c (determine_block_size): Initialize *probable_max_size
4484         even if len_rtx is CONST_INT.
4486 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4488         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4489         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4490         (cortexa53_tunings): Likewise.
4491         (aarch64_sched_issue_rate): New function.
4492         (TARGET_SCHED_ISSUE_RATE): Define.
4494 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4496         * ira-costs.c (find_costs_and_classes): Add missed
4497         ira_init_register_move_cost_if_necessary.
4499 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4501         PR target/59787
4502         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4504 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4506         PR target/59794
4507         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4508         to indicate if type is used for function return value.  Warn ABI
4509         change if the vector mode isn't available for function return value.
4510         (ix86_function_arg_advance): Pass false to type_natural_mode.
4511         (ix86_function_arg): Likewise.
4512         (ix86_gimplify_va_arg): Likewise.
4513         (function_arg_32): Don't warn ABI change.
4514         (ix86_function_value): Pass true to type_natural_mode.
4515         (ix86_return_in_memory): Likewise.
4516         (ix86_struct_value_rtx): Removed.
4517         (TARGET_STRUCT_VALUE_RTX): Likewise.
4519 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4521         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4522         converting a conditional jump into a conditional return.
4524 2014-01-14  Richard Biener  <rguenther@suse.de>
4526         PR tree-optimization/58921
4527         PR tree-optimization/59006
4528         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4529         hoisting invariant stmts.
4530         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4531         invariant loads on the preheader edge if possible.
4533 2014-01-14  Joey Ye  <joey.ye@arm.com>
4535         * doc/plugin.texi (Building GCC plugins): Update to C++.
4537 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4539         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4540         (_mm_rcp28_round_ss): Ditto.
4541         (_mm_rsqrt28_round_sd): Ditto.
4542         (_mm_rsqrt28_round_ss): Ditto.
4543         (_mm_rcp28_sd): Ditto.
4544         (_mm_rcp28_ss): Ditto.
4545         (_mm_rsqrt28_sd): Ditto.
4546         (_mm_rsqrt28_ss): Ditto.
4547         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4548         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4549         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4550         (IX86_BUILTIN_RCP28SD): Ditto.
4551         (IX86_BUILTIN_RCP28SS): Ditto.
4552         (IX86_BUILTIN_RSQRT28SD): Ditto.
4553         (IX86_BUILTIN_RSQRT28SS): Ditto.
4554         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4555         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4556         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4557         (ix86_expand_special_args_builtin): Expand new FTYPE.
4558         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4559         (srcp14<mode>): Make insn unary.
4560         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4561         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4562         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4563         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4564         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4565         Fix rounding: make it SAE only.
4566         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4567         Ditto.
4568         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4569         Ditto.
4570         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4571         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4572         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4573         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4574         (round_saeonly_mask_scalar_operand4): Ditto.
4575         (round_saeonly_mask_scalar_op3): Ditto.
4576         (round_saeonly_mask_scalar_op4): Ditto.
4578 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4580         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4581         Implement -maltivec=be for vec_insert and vec_extract.
4583 2014-01-10  DJ Delorie  <dj@redhat.com>
4585         * config/msp430/msp430.md (call_internal): Don't allow memory
4586         references with SP as the base register.
4587         (call_value_internal): Likewise.
4588         * config/msp430/constraints.md (Yc): New.  For memory references
4589         that don't use SP as a base register.
4591         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4592         "an integer without a # prefix"
4593         * config/msp430/msp430.md (epilogue_helper): Use it.
4595 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4597         PR target/59617
4598         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4599         AVX512F gather builtins.
4600         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4601         on gather decls with INTEGER_TYPE masktype.
4602         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4603         directly into the builtin rather than hoisting it before loop.
4605         PR tree-optimization/59387
4606         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4607         (scev_const_prop): If folded_casts and type has undefined overflow,
4608         use force_gimple_operand instead of force_gimple_operand_gsi and
4609         for each added stmt if it is assign with
4610         arith_code_with_undefined_signed_overflow, call
4611         rewrite_to_defined_overflow.
4612         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4613         gimple-fold.h instead.
4614         (arith_code_with_undefined_signed_overflow,
4615         rewrite_to_defined_overflow): Moved to ...
4616         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4617         rewrite_to_defined_overflow): ... here.  No longer static.
4618         Include gimplify-me.h.
4619         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4620         rewrite_to_defined_overflow): New prototypes.
4622 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4624         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4626 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4628         * builtins.c (get_object_alignment_2): Minor tweak.
4629         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4631 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4633         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4634         optimized non constant lengths.
4636 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4638         PR libgomp/59194
4639         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4640         load as __atomic_load_N if possible.
4642 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4644         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4645         target parameter.
4646         (rs6000_expand_builtin): Adjust call.
4648 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4650         PR target/58115
4651         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4652         * config/rs6000/rs6000.c: Include target-globals.h.
4653         (rs6000_set_current_function): Instead of doing target_reinit
4654         unconditionally, use save_target_globals_default_opts and
4655         restore_target_globals.
4657         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4658         FPSCR.
4659         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4660         (rs6000_expand_builtin): Handle mffs and mtfsf.
4661         (rs6000_init_builtins): Define mffs and mtfsf.
4662         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4663         (rs6000_mffs): New pattern.
4664         (rs6000_mtfsf): New pattern.
4666 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4668         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4669         Start narrowing with START.  Apply candidate-use pair
4670         and check overall cost in narrowing.
4671         (iv_ca_prune): Pass new argument.
4673 2014-01-10  Jeff Law  <law@redhat.com>
4675         PR middle-end/59743
4676         * ree.c (combine_reaching_defs): Ensure the defining statement
4677         occurs before the extension when optimizing extensions with
4678         different source and destination hard registers.
4680 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4682         PR ipa/58585
4683         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4684         vtables into the type inheritance graph.
4686 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4688         PR rtl-optimization/59754
4689         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4690         modes in the REGNO != REGNO case.
4692 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4694         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4696 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4698         PR tree-optimization/59745
4699         * tree-predcom.c (tree_predictive_commoning_loop): Call
4700         free_affine_expand_cache if giving up because components is NULL.
4702         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4703         GC in payload of target_globals struct instead of allocating them on
4704         the heap and the larger structs separately using GC.
4705         * target-globals.h (struct target_globals): Make regs, hard_regs,
4706         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4707         of GTY((skip)) and change type to void *.
4708         (reset_target_globals): Cast loads from those fields to corresponding
4709         types.
4711 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4713         PR plugins/59335
4714         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4715         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4716         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4718 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4720         PR target/59744
4721         * aarch64-modes.def (CC_Zmode): New flags mode.
4722         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4723         represents an equality.
4724         (aarch64_get_condition_code): Handle CC_Zmode.
4725         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4727 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4729         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4730         extraction in good case.
4732 2014-01-10  Richard Biener  <rguenther@suse.de>
4734         PR tree-optimization/59374
4735         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4736         checking after SLP discovery.  Mark stmts not participating
4737         in any SLP instance properly.
4739 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4741         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4742         when handling a SET rtx.
4744 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4746         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4747         (cortex-a57): Likewise.
4748         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4750 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4752         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4753         non-iwmmxt builtins.
4755 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4757         PR ipa/58252
4758         PR ipa/59226
4759         * ipa-devirt.c record_target_from_binfo): Take as argument
4760         stack of binfos and lookup matching one for virtual inheritance.
4761         (possible_polymorphic_call_targets_1): Update.
4763 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4765         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4766         kernel strings for Loongson-2E/2F/3A.
4768 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4770         PR middle-end/59670
4771         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4772         is_gimple_call before calling gimple_call_internal_p.
4774 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4776         * Makefile.in (TREE_FLOW_H): Remove.
4777         (TREE_SSA_H): Add file names from tree-flow.h.
4778         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4779         * tree.h: Remove tree-flow.h reference.
4780         * hash-table.h: Remove tree-flow.h reference.
4781         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4782         reference with tree-ssa-loop.h.
4784 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4786         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4787         default element-order behavior for -maltivec.
4788         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4789         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4790         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4791         when targeting big endian, at least for now.
4792         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4794 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4796         PR middle-end/47735
4797         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4798         var satisfies use_register_for_decl, just take into account type
4799         alignment, rather than decl alignment.
4801         PR tree-optimization/59622
4802         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4803         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4804         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4805         Don't devirtualize for inplace at all.  For targets.length () == 1,
4806         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4808 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4810         * config/i386/i386.md (cpu): Remove the unused btver1.
4812 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4814         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4816 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4818         PR target/58115
4819         * tree-core.h (struct target_globals): New forward declaration.
4820         (struct tree_target_option): Add globals field.
4821         * tree.h (TREE_TARGET_GLOBALS): Define.
4822         (prepare_target_option_nodes_for_pch): New prototype.
4823         * target-globals.h (struct target_globals): Define even if
4824         !SWITCHABLE_TARGET.
4825         * tree.c (prepare_target_option_node_for_pch,
4826         prepare_target_option_nodes_for_pch): New functions.
4827         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4828         * config/i386/i386.c: Include target-globals.h.
4829         (ix86_set_current_function): Instead of doing target_reinit
4830         unconditionally, use save_target_globals_default_opts and
4831         restore_target_globals.
4833 2014-01-09  Richard Biener  <rguenther@suse.de>
4835         PR tree-optimization/59715
4836         * tree-cfg.h (split_critical_edges): Declare.
4837         * tree-cfg.c (split_critical_edges): Export.
4838         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4840 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4842         * cfgexpand.c (expand_stack_vars): Optionally disable
4843         asan stack protection.
4844         (expand_used_vars): Likewise.
4845         (partition_stack_vars): Likewise.
4846         * asan.c (asan_emit_stack_protection): Optionally disable
4847         after return stack usage.
4848         (instrument_derefs): Optionally disable memory access instrumentation.
4849         (instrument_builtin_call): Likewise.
4850         (instrument_strlen_call): Likewise.
4851         (asan_protect_global): Optionally disable global variables protection.
4852         * doc/invoke.texi: Added doc for new options.
4853         * params.def: Added new options.
4854         * params.h: Likewise.
4856 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4858         PR rtl-optimization/59724
4859         * ifcvt.c (cond_exec_process_if_block): Don't call
4860         flow_find_head_matching_sequence with 0 longest_match.
4861         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4862         non-active insns if !stop_after.
4863         (try_head_merge_bb): Revert 2014-01-07 changes.
4865 2014-01-08  Jeff Law  <law@redhat.com>
4867         * ree.c (get_sub_rtx): New function, extracted from...
4868         (merge_def_and_ext): Here.
4869         (combine_reaching_defs): Use get_sub_rtx.
4871 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4873         * cgraph.h (varpool_variable_node): Do not choke on null node.
4875 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4877         * config/mips/mips.md (simple_return): Attempt to use JRC
4878         for microMIPS.
4879         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4881 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4883         PR rtl-optimization/59137
4884         * reorg.c (steal_delay_list_from_target): Call update_block for
4885         elided insns.
4886         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4888 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4890         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4891         two duplicate entries.
4893 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4895         Revert:
4896         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4898         * config/mips/mips.c (mips_truncated_op_cost): New function.
4899         (mips_rtx_costs): Adjust test for BADDU.
4900         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4902         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4904         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4905         (*baddu_si): ...this new pattern.
4907 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4909         PR ipa/59722
4910         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4912 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4914         PR middle-end/57748
4915         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4916         inner_reference_p.
4917         (expand_expr, expand_normal): Adjust.
4918         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4919         inner_reference_p. Use inner_reference_p to expand inner references.
4920         (store_expr): Adjust.
4921         * cfgexpand.c (expand_call_stmt): Adjust.
4923 2014-01-08  Rong Xu  <xur@google.com>
4925         * gcov-io.c (gcov_var): Move from gcov-io.h.
4926         (gcov_position): Ditto.
4927         (gcov_is_error): Ditto.
4928         (gcov_rewrite): Ditto.
4929         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4930         only part to libgcc/libgcov.h.
4932 2014-01-08  Marek Polacek  <polacek@redhat.com>
4934         PR middle-end/59669
4935         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4937 2014-01-08  Marek Polacek  <polacek@redhat.com>
4939         PR sanitizer/59667
4940         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4942 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4944         PR rtl-optimization/59649
4945         * stor-layout.c (get_mode_bounds): For BImode return
4946         0 and STORE_FLAG_VALUE.
4948 2014-01-08  Richard Biener  <rguenther@suse.de>
4950         PR middle-end/59630
4951         * gimple.h (is_gimple_builtin_call): Remove.
4952         (gimple_builtin_call_types_compatible_p): New.
4953         (gimple_call_builtin_p): New overload.
4954         * gimple.c (is_gimple_builtin_call): Remove.
4955         (validate_call): Rename to ...
4956         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4957         check return types.
4958         (validate_type): New static function.
4959         (gimple_call_builtin_p): New overload and adjust.
4960         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4961         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4962         (gimple_fold_stmt_to_constant_1): Likewise.
4963         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4965 2014-01-08  Richard Biener  <rguenther@suse.de>
4967         PR middle-end/59471
4968         * gimplify.c (gimplify_expr): Gimplify register-register type
4969         VIEW_CONVERT_EXPRs to separate stmts.
4971 2014-01-07  Jeff Law  <law@redhat.com>
4973         PR middle-end/53623
4974         * ree.c (combine_set_extension): Handle case where source
4975         and destination registers in an extension insn are different.
4976         (combine_reaching_defs): Allow source and destination registers
4977         in extension to be different under limited circumstances.
4978         (add_removable_extension): Remove restriction that the
4979         source and destination registers in the extension are the same.
4980         (find_and_remove_re): Emit a copy from the extension's
4981         destination to its source after the defining insn if
4982         the source and destination registers are different.
4984         PR middle-end/59285
4985         * ifcvt.c (merge_if_block): If we are merging a block with more than
4986         one successor with a block with no successors, remove any BARRIER
4987         after the second block.
4989 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4991         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4993 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4995         PR target/59652
4996         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4997         for 14-bit register offsets when INT14_OK_STRICT is false.
4999 2014-01-07  Roland Stigge  <stigge@antcom.de>
5000             Michael Meissner  <meissner@linux.vnet.ibm.com>
5002         PR 57386/target
5003         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5004         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5006 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5008         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5009         -mcpu.
5011 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5013         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5014         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5015         rtx is const0_rtx or not.
5017 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5019         PR target/58115
5020         * target-globals.c (save_target_globals): Remove this_fn_optab
5021         handling.
5022         * toplev.c: Include optabs.h.
5023         (target_reinit): Temporarily restore the global options if another
5024         set of options are in force.
5026 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5028         PR rtl-optimization/58668
5029         * cfgcleanup.c (flow_find_cross_jump): Don't count
5030         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5031         to determine what is counted.
5032         (flow_find_head_matching_sequence): Use active_insn_p to determine
5033         what is counted.
5034         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5035         counting change.
5036         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5037         determine what is counted.
5039         PR tree-optimization/59643
5040         * tree-predcom.c (split_data_refs_to_components): If one dr is
5041         read and one write, determine_offset fails and the write isn't
5042         in the bad component, just put the read into the bad component.
5044 2014-01-07  Mike Stump  <mikestump@comcast.net>
5045             Jakub Jelinek  <jakub@redhat.com>
5047         PR pch/59436
5048         * tree-core.h (struct tree_optimization_option): Change optabs
5049         type from unsigned char * to void *.
5050         * optabs.c (init_tree_optimization_optabs): Adjust
5051         TREE_OPTIMIZATION_OPTABS initialization.
5053 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5055         PR target/59644
5056         * config/i386/i386.h (struct machine_function): Add
5057         no_drap_save_restore field.
5058         * config/i386/i386.c (ix86_save_reg): Use
5059         !cfun->machine->no_drap_save_restore instead of
5060         crtl->stack_realign_needed.
5061         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5062         this function clears frame_pointer_needed.  Set
5063         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5064         and DRAP reg is needed.
5066 2014-01-06  Marek Polacek  <polacek@redhat.com>
5068         PR c/57773
5069         * doc/implement-c.texi: Mention that other integer types are
5070         permitted as bit-field types in strictly conforming mode.
5072 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5074         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5075         is newly allocated.
5077 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5079         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5081 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5083         PR ipa/59008
5084         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5085         to int.
5086         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5088 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5090         PR debug/59350
5091         PR debug/59510
5092         * var-tracking.c (add_stores): Preserve the value of the source even if
5093         we don't record the store.
5095 2014-01-06  Terry Guo  <terry.guo@arm.com>
5097         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5099 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5101         PR bootstrap/59541
5102         * config/darwin.c (darwin_function_section): Adjust return values to
5103         correspond to optimisation changes made in r206070.
5105 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5107         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5108         from prefetch_block tune setting.
5109         (nocona_cost): Correct size of prefetch block to 64.
5111 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5113         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5114         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5115         used to save the static chain register in the computation of the offset
5116         from which the FP registers need to be restored.
5118 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5120         PR tree-optimization/59519
5121         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5122         ICE if get_current_def (current_new_name) is already non-NULL, as long
5123         as it is a phi result of some other phi in *new_exit_bb that has
5124         the same argument.
5126         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5127         or vmovdqu* for misaligned_operand.
5128         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5129         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5130         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5131         aligned_mem for AVX512F masked aligned load and store builtins and for
5132         non-temporal moves.
5134 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5136         PR tree-optimization/59651
5137         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5138         Address range for negative step should be added by TYPE_SIZE_UNIT.
5140 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5142         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5143         overlapping registers also for registers other than the stack pointer.
5145 2014-01-03  Marek Polacek  <polacek@redhat.com>
5147         PR other/59661
5148         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5149         __builtin_FILE.
5151 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5153         PR target/59625
5154         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5155         asm goto as jump.
5157         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5158         (push splitter): Use <P:MODE_SIZE> instead of
5159         GET_MODE_SIZE (<P:MODE>mode).
5160         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5161         (mov -1, reg peephole2): Likewise.
5162         * config/i386/sse.md (*mov<mode>_internal,
5163         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5164         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5165         *<code><mode>3, *andnot<mode>3<mask_name>,
5166         <mask_codefor><code><mode>3<mask_name>): Likewise.
5167         * config/i386/subst.md (mask_mode512bit_condition,
5168         sd_mask_mode512bit_condition): Likewise.
5170 2014-01-02  Xinliang David Li  <davidxl@google.com>
5172         PR tree-optimization/59303
5173         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5174         (dump_predicates): Better output format.
5175         (pred_equal_p): New function.
5176         (is_neq_relop_p): Ditto.
5177         (is_neq_zero_form_p): Ditto.
5178         (pred_expr_equal_p): Ditto.
5179         (pred_neg_p): Ditto.
5180         (simplify_pred): Ditto.
5181         (simplify_preds_2): Ditto.
5182         (simplify_preds_3): Ditto.
5183         (simplify_preds_4): Ditto.
5184         (simplify_preds): Ditto.
5185         (push_pred): Ditto.
5186         (push_to_worklist): Ditto.
5187         (get_pred_info_from_cmp): Ditto.
5188         (is_degenerated_phi): Ditto.
5189         (normalize_one_pred_1): Ditto.
5190         (normalize_one_pred): Ditto.
5191         (normalize_one_pred_chain): Ditto.
5192         (normalize_preds): Ditto.
5193         (normalize_cond_1): Remove function.
5194         (normalize_cond): Ditto.
5195         (is_gcond_subset_of): Ditto.
5196         (is_subset_of_any): Ditto.
5197         (is_or_set_subset_of): Ditto.
5198         (is_and_set_subset_of): Ditto.
5199         (is_norm_cond_subset_of): Ditto.
5200         (pred_chain_length_cmp): Ditto.
5201         (convert_control_dep_chain_into_preds): Type change.
5202         (find_predicates): Ditto.
5203         (find_def_preds): Ditto.
5204         (destroy_predicates_vecs): Ditto.
5205         (find_matching_predicates_in_rest_chains): Ditto.
5206         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5207         (is_pred_expr_subset): Ditto.
5208         (is_pred_chain_subset_of): Ditto.
5209         (is_included_in): Ditto.
5210         (is_superset_of): Ditto.
5212 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5214         Update copyright years.
5216 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5218         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5219         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5220         config/arc/arc.md, config/arc/arc.opt,
5221         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5222         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5223         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5224         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5225         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5226         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5227         vtable-verify.c, vtable-verify.h: Use the standard form for the
5228         copyright notice.
5230 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5232         * gcc.c (process_command): Update copyright notice dates.
5233         * gcov-dump.c: Ditto.
5234         * gcov.c: Ditto.
5235         * doc/cpp.texi: Bump @copying's copyright year.
5236         * doc/cppinternals.texi: Ditto.
5237         * doc/gcc.texi: Ditto.
5238         * doc/gccint.texi: Ditto.
5239         * doc/gcov.texi: Ditto.
5240         * doc/install.texi: Ditto.
5241         * doc/invoke.texi: Ditto.
5243 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5245         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5247 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5249         * config/i386/sse.md (*mov<mode>_internal): Guard
5250         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5252         PR rtl-optimization/59647
5253         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5254         new_rtx into UNSIGNED_FLOAT rtxes.
5256 Copyright (C) 2014 Free Software Foundation, Inc.
5258 Copying and distribution of this file, with or without modification,
5259 are permitted in any medium without royalty provided the copyright
5260 notice and this notice are preserved.