2014-03-25 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blobedb45ecc95dda5e33a52f6355d31f333dac51a0c
1 2014-03-25  Richard Biener  <rguenther@suse.de>
3         * lto-wrapper.c (merge_and_complain): Handle OPT_fPIE like
4         OPT_fpie.
5         (run_gcc): Likewise.
7 2014-03-25  Jakub Jelinek  <jakub@redhat.com>
9         * combine.c (simplify_compare_const): Add MODE argument.
10         Handle mode_width 0 as very large mode_width.
11         (try_combine, simplify_comparison): Adjust callers.
13         * cselib.c (cselib_hash_rtx): Perform addition in unsigned
14         type to avoid signed integer overflow.
15         * explow.c (plus_constant): Likewise.
17 2014-03-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
19         * doc/generic.texi: Correct typos.
21 2014-03-24  Tobias Burnus  <burnus@net-b.de>
23         * doc/invoke.texi (-flto): Expand section about
24         using static libraries with LTO.
26 2014-03-24  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
28         PR rtl-optimization/60501
29         * optabs.def (addptr3_optab): New optab.
30         * optabs.c (gen_addptr3_insn, have_addptr3_insn): New function.
31         * doc/md.texi ("addptrm3"): Document new RTL standard expander.
32         * expr.h (gen_addptr3_insn, have_addptr3_insn): Add prototypes.
34         * lra.c (emit_add3_insn): Use the addptr pattern if available.
36         * config/s390/s390.md ("addptrdi3", "addptrsi3"): New expanders.
38 2014-03-24  Ulrich Drepper  <drepper@gmail.com>
40         * config/i386/avx512fintrin.h: Define _mm512_set1_ps and
41         _mm512_set1_pd.
43         * config/i386/avxintrin.h (_mm256_undefined_si256): Define.
44         (_mm256_undefined_ps): Define.
45         (_mm256_undefined_pd): Define.
46         * config/i386/emmintrin.h (_mm_undefined_si128): Define.
47         (_mm_undefined_pd): Define.
48         * config/i386/xmmintrin.h (_mm_undefined_ps): Define.
49         * config/i386/avx512fintrin.h (_mm512_undefined_si512): Define.
50         (_mm512_undefined_ps): Define.
51         (_mm512_undefined_pd): Define.
52         Use _mm*_undefined_*.
53         * config/i386/avx2intrin.h: Use _mm*_undefined_*.
55 2014-03-24  Alex Velenko  <Alex.Velenko@arm.com>
57         * config/aarch64/aarch64-simd-builtins.def (lshr): DI mode excluded.
58         (lshr_simd): DI mode added.
59         * config/aarch64/aarch64-simd.md (aarch64_lshr_simddi): New pattern.
60         (aarch64_ushr_simddi): Likewise.
61         * config/aarch64/aarch64.md (UNSPEC_USHR64): New unspec.
62         * config/aarch64/arm_neon.h (vshr_n_u64): Intrinsic fixed.
63         (vshrd_n_u64): Likewise.
65 2014-03-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
67         * Makefile.in (s-macro_list): Depend on cc1.
69 2014-03-23  Teresa Johnson  <tejohnson@google.com>
71         * ipa-utils.c (ipa_print_order): Use specified dump file.
73 2014-03-23  Eric Botcazou  <ebotcazou@adacore.com>
75         PR rtl-optimization/60601
76         * bb-reorder.c (fix_up_fall_thru_edges): Test EDGE_FALLTHRU everywhere.
78         * gcc.c (eval_spec_function): Initialize save_growing_value.
80 2014-03-22  Jakub Jelinek  <jakub@redhat.com>
82         PR sanitizer/60613
83         * internal-fn.c (ubsan_expand_si_overflow_addsub_check): For
84         code == MINUS_EXPR, never swap op0 with op1.
86         * toplev.c (init_local_tick): Avoid signed integer multiplication
87         overflow.
88         * genautomata.c (reserv_sets_hash_value): Fix rotate idiom, avoid
89         shift by first operand's bitsize.
91 2014-03-21  Jakub Jelinek  <jakub@redhat.com>
93         PR target/60610
94         * config/i386/i386.h (TARGET_64BIT_P): If not TARGET_BI_ARCH,
95         redefine to 1 or 0.
96         * config/i386/darwin.h (TARGET_64BIT_P): Redefine to
97         TARGET_ISA_64BIT_P(x).
99 2014-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
101         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Generate a
102         pattern for vector nor instead of subtract from splat(-1).
103         (altivec_expand_vec_perm_const_le): Likewise.
105 2014-03-21  Richard Henderson  <rth@twiddle.net>
107         PR target/60598
108         * ifcvt.c (dead_or_predicable): Return FALSE if there are any frame
109         related insns after epilogue_completed.
111 2014-03-21  Martin Jambor  <mjambor@suse.cz>
113         PR ipa/59176
114         * cgraph.h (symtab_node): New flag body_removed.
115         * ipa.c (symtab_remove_unreachable_nodes): Set body_removed flag
116         when removing bodies.
117         * symtab.c (dump_symtab_base): Dump body_removed flag.
118         * cgraph.c (verify_edge_corresponds_to_fndecl): Skip nodes which
119         had their bodies removed.
121 2014-03-21  Martin Jambor  <mjambor@suse.cz>
123         PR ipa/60419
124         * ipa.c (symtab_remove_unreachable_nodes): Clear thunk flag of nodes
125         in the border.
127 2014-03-21  Richard Biener  <rguenther@suse.de>
129         PR tree-optimization/60577
130         * tree-core.h (struct tree_base): Document nothrow_flag use
131         in VAR_DECL_NONALIASED.
132         * tree.h (VAR_DECL_NONALIASED): New.
133         (may_be_aliased): Adjust.
134         * coverage.c (build_var): Set VAR_DECL_NONALIASED.
136 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
138         * expr.c (expand_expr_real_1): Remove outdated comment.
140 2014-03-20  Jakub Jelinek  <jakub@redhat.com>
142         PR middle-end/60597
143         * ira.c (adjust_cleared_regs): Call copy_rtx on
144         *reg_equiv[REGNO (loc)].src_p before passing it to
145         simplify_replace_fn_rtx.
147         PR target/60568
148         * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
149         into CONST, put pic register as first operand of PLUS.  Use
150         gen_const_mem for both 32-bit and 64-bit PIC got loads.
152 2014-03-20  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
154         * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
156 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
158         * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
159         around for store forwarding issue in the FPU on the UT699.
160         * config/sparc/sparc.md (in_branch_delay): Return false for single FP
161         loads and operations if -mfix-ut699 is specified.
162         (divtf3_hq): Tweak attribute.
163         (sqrttf2_hq): Likewise.
165 2014-03-20  Eric Botcazou  <ebotcazou@adacore.com>
167         * calls.c (store_one_arg): Remove incorrect const qualification on the
168         type of the temporary.
169         * cfgexpand.c (expand_return): Likewise.
170         * expr.c (expand_constructor): Likewise.
171         (expand_expr_real_1): Likewise.
173 2014-03-20  Zhenqiang Chen  <zhenqiang.chen@linaro.org>
175         * config/arm/arm.c (arm_dwarf_register_span): Update the element number
176         of parts.
178 2014-03-19  Kaz Kojima  <kkojima@gcc.gnu.org>
180         PR target/60039
181         * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
183 2014-03-19  James Greenhalgh  <james.greenhalgh@arm.com>
185         * config/arm/aarch-common-protos.h
186         (alu_cost_table): Fix spelling of "extend".
187         * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
189 2014-03-19  Richard Biener  <rguenther@suse.de>
191         PR middle-end/60553
192         * tree-core.h (tree_type_common): Re-order pointer members
193         to reduce recursion depth during GC walks.
195 2014-03-19  Marek Polacek  <polacek@redhat.com>
197         PR sanitizer/60569
198         * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
199         before accessing it.
201 2014-03-19  Richard Biener  <rguenther@suse.de>
203         PR lto/59543
204         * lto-streamer-in.c (input_function): In WPA stage do not drop
205         debug stmts.
207 2014-03-19  Jakub Jelinek  <jakub@redhat.com>
209         PR tree-optimization/60559
210         * vectorizable_mask_load_store): Replace scalar MASK_LOAD
211         with build_zero_cst assignment.
213 2014-03-18  Kai Tietz  <ktietz@redhat.com>
215         PR rtl-optimization/56356
216         * sdbout.c (sdbout_parms): Verify that parms'
217         incoming argument is valid.
218         (sdbout_reg_parms): Likewise.
220 2014-03-18  Richard Henderson  <rth@redhat.com>
222         PR target/60562
223         * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
224         be shadowed by *float<SWI48><MODEF>2_sse.  Test X87_ENABLE_FLOAT.
225         (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
227 2014-03-18  Basile Starynkevitch  <basile@starynkevitch.net>
229         * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
230         * doc/plugins.texi (Plugin callbacks): Mention
231         PLUGIN_INCLUDE_FILE.
232         Italicize plugin event names in description.  Explain that
233         PLUGIN_PRAGMAS has no sense for lto1. Explain
234         PLUGIN_INCLUDE_FILE.
235         Remind that no GCC functions should be called after
236         PLUGIN_FINISH.
237         Explain what pragmas with expansion are.
239 2014-03-18  Martin Liska  <mliska@suse.cz>
241         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
242         gimple call statement is update.
243         * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
244         GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
246 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
248         PR sanitizer/60557
249         * ubsan.c (ubsan_instrument_unreachable): Call
250         initialize_sanitizer_builtins.
251         (ubsan_pass): Likewise.
253         PR sanitizer/60535
254         * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
255         varpool_finalize_decl instead of rest_of_decl_compilation.
257 2014-03-18  Richard Biener  <rguenther@suse.de>
259         * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
260         by using bitmap_and_compl instead of bitmap_and_compl_into.
261         (df_rd_transfer_function): Likewise.
263 2014-03-18  Richard Biener  <rguenther@suse.de>
265         * doc/lto.texi (fresolution): Fix typo.
267 2014-03-18  Richard Biener  <rguenther@suse.de>
269         * doc/invoke.texi (flto): Update for changes in 4.9.
271 2014-03-18  Richard Biener  <rguenther@suse.de>
273         * doc/loop.texi: Remove section on the removed lambda framework.
274         Update loop docs with recent changes in preserving loop structure.
276 2014-03-18  Richard Biener  <rguenther@suse.de>
278         * doc/lto.texi (-fresolution): Document.
280 2014-03-18  Richard Biener  <rguenther@suse.de>
282         * doc/contrib.texi: Adjust my name.
284 2014-03-18  Jakub Jelinek  <jakub@redhat.com>
286         PR ipa/58721
287         * internal-fn.c: Include diagnostic-core.h.
288         (expand_BUILTIN_EXPECT): New function.
289         * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
290         (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
291         IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
292         * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
293         IFN_BUILTIN_EXPECT.
294         * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
295         Revert 3 argument __builtin_expect code.
296         (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
297         * gimple-fold.c (gimple_fold_call): Likewise.
298         * tree.h (fold_builtin_expect): New prototype.
299         * builtins.c (build_builtin_expect_predicate): Add predictor
300         argument, if non-NULL, create 3 argument __builtin_expect.
301         (fold_builtin_expect): No longer static.  Add ARG2 argument,
302         pass it through to build_builtin_expect_predicate.
303         (fold_builtin_2): Adjust caller.
304         (fold_builtin_3): Handle BUILT_IN_EXPECT.
305         * internal-fn.def (BUILTIN_EXPECT): New.
307 2014-03-18  Tobias Burnus  <burnus@net-b.de>
309         PR ipa/58721
310         * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
311         PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
312         PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
314 2014-03-18  Jan Hubicka  <hubicka@ucw.cz>
316         PR ipa/58721
317         * predict.c (combine_predictions_for_bb): Fix up formatting.
318         (expr_expected_value_1, expr_expected_value): Add predictor argument,
319         fill what it points to if non-NULL.
320         (tree_predict_by_opcode): Adjust caller, use the predictor.
321         * predict.def (PRED_COMPARE_AND_SWAP): Add.
323 2014-03-18  Eric Botcazou  <ebotcazou@adacore.com>
325         * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
326         proper constant for the store mode.
328 2014-03-18  Ilya Enkovich  <ilya.enkovich@intel.com>
330         * symtab.c (change_decl_assembler_name): Fix transparent alias
331         chain construction.
333 2014-03-16  Renlin Li  <Renlin.Li@arm.com>
335         * config/aarch64/aarch64.c: Correct the comments about the
336         aarch64 stack layout.
338 2014-03-18  Thomas Schwinge  <thomas@codesourcery.com>
340         * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
341         check for GF_OMP_FOR_KIND_FOR.
343 2013-03-18  Kirill Yukhin  <kirill.yukhin@intel.com>
345         * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
346         ymm and zmm register names.
348 2014-03-17  Jakub Jelinek  <jakub@redhat.com>
350         PR target/60516
351         * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
352         note creation for the 2010-08-31 changes.
354 2014-03-17  Marek Polacek  <polacek@redhat.com>
356         PR middle-end/60534
357         * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
358         as -fno-tree-loop-vectorize.
359         (expand_omp_simd): Likewise.
361 2014-03-15  Eric Botcazou  <ebotcazou@adacore.com>
363         * config/sparc/sparc-protos.h (tls_call_delay): Delete.
364         (eligible_for_call_delay): New prototype.
365         * config/sparc/sparc.c (tls_call_delay): Rename into...
366         (eligible_for_call_delay): ...this.  Return false if the instruction
367         cannot be put in the delay slot of a branch.
368         (eligible_for_restore_insn): Simplify.
369         (eligible_for_return_delay): Return false if the instruction cannot be
370         put in the delay slot of a branch and simplify.
371         (eligible_for_sibcall_delay): Return false if the instruction cannot be
372         put in the delay slot of a branch.
373         * config/sparc/sparc.md (fix_ut699): New attribute.
374         (tls_call_delay): Delete.
375         (in_call_delay): Reimplement.
376         (eligible_for_sibcall_delay): Rename into...
377         (in_sibcall_delay): ...this.
378         (eligible_for_return_delay): Rename into...
379         (in_return_delay): ...this.
380         (in_branch_delay): Reimplement.
381         (in_uncond_branch_delay): Delete.
382         (in_annul_branch_delay): Delete.
384 2014-03-14  Richard Henderson  <rth@redhat.com>
386         PR target/60525
387         * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
388         define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
389         (*floathi<X87MODEF>2_i387_with_temp): Remove.
390         (floathi splitters): Remove.
391         (float<SWI48x>xf2): New pattern.
392         (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2.  Drop
393         code that tried to handle DImode for 32-bit, but which was excluded
394         by the pattern's condition.  Drop allocation of stack temporary.
395         (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
396         (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
397         (*float<SWI48><MODEF>2_mixed_interunit): Remove.
398         (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
399         (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
400         (*float<SWI48><MODEF>2_sse_with_temp): Remove.
401         (*float<SWI48><MODEF>2_sse_interunit): Remove.
402         (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
403         (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
404         (*float<SWI48x><X87MODEF>2_i387): Remove.
405         (all float _with_temp splitters): Remove.
406         (*float<SWI48x><MODEF>2_i387): New pattern.
407         (*float<SWI48><MODEF>2_sse): New pattern.
408         (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
409         (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
411 2014-03-14  Jakub Jelinek  <jakub@redhat.com>
412             Marek Polacek  <polacek@redhat.com>
414         PR middle-end/60484
415         * common.opt (dump_base_name_prefixed): New Variable.
416         * opts.c (finish_options): Don't prepend directory to x_dump_base_name
417         if x_dump_base_name_prefixed is already set, set it at the end.
419 2014-03-14  Vladimir Makarov  <vmakarov@redhat.com>
421         PR rtl-optimization/60508
422         * lra-constraints.c (get_reload_reg): Add new parameter
423         in_subreg_p.
424         (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
425         Pass the new parameter values.
427 2014-03-14  Richard Biener  <rguenther@suse.de>
429         * common.opt: Revert unintented changes from r205065.
430         * opts.c: Likewise.
432 2014-03-14  Richard Biener  <rguenther@suse.de>
434         PR middle-end/60518
435         * cfghooks.c (split_block): Properly adjust all loops the
436         block was a latch of.
438 2014-03-14  Martin Jambor  <mjambor@suse.cz>
440         PR lto/60461
441         * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
442         and simplify it.
444 2014-03-14  Georg-Johann Lay  <avr@gjlay.de>
446         PR target/59396
447         * config/avr/avr.c (avr_set_current_function): Pass function name
448         through default_strip_name_encoding before sanity checking instead
449         of skipping the first char of the assembler name.
451 2014-03-13  Richard Henderson  <rth@redhat.com>
453         PR debug/60438
454         * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
455         (ix86_force_to_memory, ix86_free_from_memory): Remove.
456         * config/i386/i386-protos.h: Likewise.
457         * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
458         in the expander instead of a splitter.
459         (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
460         any possibility of requiring a memory.
461         (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
462         (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
463         (fp branch splitters): Update for ix86_split_fp_branch.
464         (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
465         (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
466         (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
467         (*fop_<MODEF>_2_i387): Remove f/r alternative.
468         (*fop_<MODEF>_3_i387): Likewise.
469         (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
470         (splitters for the fop_* register patterns): Remove.
471         (fscalexf4_i387): Rename from *fscalexf4_i387.
472         (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
473         
474 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
476         PR tree-optimization/59779
477         * tree-dfa.c (get_ref_base_and_extent): Use double_int
478         type for bitsize and maxsize instead of HOST_WIDE_INT.
480 2014-03-13  Steven Bosscher  <steven@gcc.gnu.org>
482         PR rtl-optimization/57320
483         * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
484         the CFG after thread_prologue_and_epilogue_insns.
486 2014-03-13  Vladimir Makarov  <vmakarov@redhat.com>
488         PR rtl-optimization/57189
489         * lra-constraints.c (process_alt_operands): Disfavor spilling
490         vector pseudos.
492 2014-03-13  Cesar Philippidis  <cesar@codesourcery.com>
494         * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
495         messages.
497 2014-03-13  Jakub Jelinek  <jakub@redhat.com>
499         PR tree-optimization/59025
500         PR middle-end/60418
501         * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
502         same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
504 2014-03-13  Georg-Johann Lay  <avr@gjlay.de>
506         PR target/60486
507         * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
508         calls of avr_out_plus_1.
510 2014-03-13  Bin Cheng  <bin.cheng@arm.com>
512         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
513         BB's single pred and update the father loop's latch info later.
515 2014-03-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
517         * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
518         (VEC_M): Likewise.
519         (VEC_N): Likewise.
520         (VEC_R): Likewise.
521         (VEC_base): Likewise.
522         (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
523         registers, we need to swap double words in little endian mode.
525         * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
526         to be a container mode for 128-bit integer operations added in ISA
527         2.07.  Unlike TImode and PTImode, the preferred register set is
528         the Altivec/VMX registers for the 128-bit operations.
530         * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
531         declarations.
532         (rs6000_split_128bit_ok_p): Likewise.
534         * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
535         macros for creating ISA 2.07 normal and overloaded builtin
536         functions with 3 arguments.
537         (BU_P8V_OVERLOAD_3): Likewise.
538         (VPERM_1T): Add support for V1TImode in 128-bit vector operations
539         for use as overloaded functions.
540         (VPERM_1TI_UNS): Likewise.
541         (VSEL_1TI): Likewise.
542         (VSEL_1TI_UNS): Likewise.
543         (ST_INTERNAL_1ti): Likewise.
544         (LD_INTERNAL_1ti): Likewise.
545         (XXSEL_1TI): Likewise.
546         (XXSEL_1TI_UNS): Likewise.
547         (VPERM_1TI): Likewise.
548         (VPERM_1TI_UNS): Likewise.
549         (XXPERMDI_1TI): Likewise.
550         (SET_1TI): Likewise.
551         (LXVD2X_V1TI): Likewise.
552         (STXVD2X_V1TI): Likewise.
553         (VEC_INIT_V1TI): Likewise.
554         (VEC_SET_V1TI): Likewise.
555         (VEC_EXT_V1TI): Likewise.
556         (EQV_V1TI): Likewise.
557         (NAND_V1TI): Likewise.
558         (ORC_V1TI): Likewise.
559         (VADDCUQ): Add support for 128-bit integer arithmetic instructions
560         added in ISA 2.07.  Add both normal 'altivec' builtins, and the
561         overloaded builtin.
562         (VADDUQM): Likewise.
563         (VSUBCUQ): Likewise.
564         (VADDEUQM): Likewise.
565         (VADDECUQ): Likewise.
566         (VSUBEUQM): Likewise.
567         (VSUBECUQ): Likewise.
569         * config/rs6000/rs6000-c.c (__int128_type): New static to hold
570         __int128_t and __uint128_t types.
571         (__uint128_type): Likewise.
572         (altivec_categorize_keyword): Add support for vector __int128_t,
573         vector __uint128_t, vector __int128, and vector unsigned __int128
574         as a container type for TImode operations that need to be done in
575         VSX/Altivec registers.
576         (rs6000_macro_to_expand): Likewise.
577         (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
578         to support 128-bit integer instructions vaddcuq, vadduqm,
579         vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
580         (altivec_resolve_overloaded_builtin): Add support for V1TImode.
582         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
583         for V1TImode, and set up preferences to use VSX/Altivec registers.
584         Setup VSX reload handlers.
585         (rs6000_debug_reg_global): Likewise.
586         (rs6000_init_hard_regno_mode_ok): Likewise.
587         (rs6000_preferred_simd_mode): Likewise.
588         (vspltis_constant): Do not allow V1TImode as easy altivec constants.
589         (easy_altivec_constant): Likewise.
590         (output_vec_const_move): Likewise.
591         (rs6000_expand_vector_set): Convert V1TImode set and extract to
592         simple move.
593         (rs6000_expand_vector_extract): Likewise.
594         (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
595         addressing.
596         (rs6000_const_vec): Add support for V1TImode.
597         (rs6000_emit_le_vsx_load): Swap double words when loading or
598         storing TImode/V1TImode.
599         (rs6000_emit_le_vsx_store): Likewise.
600         (rs6000_emit_le_vsx_move): Likewise.
601         (rs6000_emit_move): Add support for V1TImode.
602         (altivec_expand_ld_builtin): Likewise.
603         (altivec_expand_st_builtin): Likewise.
604         (altivec_expand_vec_init_builtin): Likewise.
605         (altivec_expand_builtin): Likewise.
606         (rs6000_init_builtins): Add support for V1TImode type.  Add
607         support for ISA 2.07 128-bit integer builtins.  Define type names
608         for the VSX/Altivec vector types.
609         (altivec_init_builtins): Add support for overloaded vector
610         functions with V1TImode type.
611         (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
612         (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
613         external function.
614         (rs6000_split_128bit_ok_p): Likewise.
615         (rs6000_handle_altivec_attribute): Create V1TImode from vector
616         __int128_t and vector __uint128_t.
618         * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
619         and mode attributes.
620         (VSX_M): Likewise.
621         (VSX_M2): Likewise.
622         (VSm): Likewise.
623         (VSs): Likewise.
624         (VSr): Likewise.
625         (VSv): Likewise.
626         (VS_scalar): Likewise.
627         (VS_double): Likewise.
628         (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
630         * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
631         we support the ISA 2.07 128-bit integer arithmetic instructions.
632         (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
633         (enum rs6000_builtin_type_index): Add fields to hold V1TImode
634         and TImode types for use with the builtin functions.
635         (V1TI_type_node): Likewise.
636         (unsigned_V1TI_type_node): Likewise.
637         (intTI_type_internal_node): Likewise.
638         (uintTI_type_internal_node): Likewise.
640         * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
641         128-bit builtin functions.
642         (UNSPEC_VADDEUQM): Likewise.
643         (UNSPEC_VADDECUQ): Likewise.
644         (UNSPEC_VSUBCUQ): Likewise.
645         (UNSPEC_VSUBEUQM): Likewise.
646         (UNSPEC_VSUBECUQ): Likewise.
647         (VM): Add V1TImode to vector mode iterators.
648         (VM2): Likewise.
649         (VI_unit): Likewise.
650         (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
651         (altivec_vaddcuq): Likewise.
652         (altivec_vsubuqm): Likewise.
653         (altivec_vsubcuq): Likewise.
654         (altivec_vaddeuqm): Likewise.
655         (altivec_vaddecuq): Likewise.
656         (altivec_vsubeuqm): Likewise.
657         (altivec_vsubecuq): Likewise.
659         * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
660         mode iterators.
661         (BOOL_128): Likewise.
662         (BOOL_REGS_OUTPUT): Likewise.
663         (BOOL_REGS_OP1): Likewise.
664         (BOOL_REGS_OP2): Likewise.
665         (BOOL_REGS_UNARY): Likewise.
666         (BOOL_REGS_AND_CR0): Likewise.
668         * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
669         128-bit integer builtin support.
670         (vec_vadduqm): Likewise.
671         (vec_vaddecuq): Likewise.
672         (vec_vaddeuqm): Likewise.
673         (vec_vsubecuq): Likewise.
674         (vec_vsubeuqm): Likewise.
675         (vec_vsubcuq): Likewise.
676         (vec_vsubuqm): Likewise.
678         * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
679         Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
680         vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
681         128-bit integer add/subtract to ISA 2.07.
683 2014-03-12  Joern Rennecke  <joern.rennecke@embecosm.com>
685         * config/arc/arc.c (arc_predicate_delay_insns):
686         Fix third argument passed to conditionalize_nonjump.
688 2014-03-12  Yufeng Zhang  <yufeng.zhang@arm.com>
690         * config/aarch64/aarch64-builtins.c
691         (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
692         BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
693         * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
694         instead of __builtin_lfloor.
695         (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
697 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
699         * tree-ssa-ifcombine.c (forwarder_block_to): New function.
700         (tree_ssa_ifcombine_bb_1): New function.
701         (tree_ssa_ifcombine_bb): Use it.  Handle also cases where else_bb
702         is an empty forwarder block to then_bb or vice versa and then_bb
703         and else_bb are effectively swapped.
705 2014-03-12  Christian Bruel  <christian.bruel@st.com>
707         PR target/60264
708         * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
709         REG_CFA_DEF_CFA note.
710         (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
711         (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
713 2014-03-12  Thomas Preud'homme  <thomas.preudhomme@arm.com>
715         PR tree-optimization/60454
716         * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
718 2014-03-12  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
720         * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
721         Do not define target_cpu_default2 to generic.
722         * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
723         * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
724         * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
726 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
727             Marc Glisse  <marc.glisse@inria.fr>
729         PR tree-optimization/60502
730         * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
731         instead of build_low_bits_mask.
733 2014-03-12  Jakub Jelinek  <jakub@redhat.com>
735         PR middle-end/60482
736         * tree-vrp.c (register_edge_assert_for_1): Don't add assert
737         if there are multiple uses, but op doesn't live on E edge.
738         * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
739         clobber stmts before __builtin_unreachable.
741 2014-03-11  Richard Sandiford  <rdsandiford@googlemail.com>
743         * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
744         hard_frame_pointer_rtx.
745         * cse.c (cse_insn): Remove volatile check.
746         * cselib.c (cselib_process_insn): Likewise.
747         * dse.c (scan_insn): Likewise.
749 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
751         * config/arc/arc.c (conditionalize_nonjump): New function,
752         broken out of ...
753         (arc_ifcvt): ... this.
754         (arc_predicate_delay_insns): Use it.
756 2014-03-11  Joern Rennecke  <joern.rennecke@embecosm.com>
758         * config/arc/predicates.md (extend_operand): During/after reload,
759         allow const_int_operand.
760         * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
761         (umulsidi3_700): Likewise.  Change operand 2 constraint back to "cL".
762         (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
763         to "i".
764         (umulsi3_highpart_i): Likewise.
766 2014-03-11  Richard Biener  <rguenther@suse.de>
768         * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
769         Add asserts to guard possible wrong-code bugs.
771 2014-03-11  Richard Biener  <rguenther@suse.de>
773         PR tree-optimization/60429
774         PR tree-optimization/60485
775         * tree-ssa-structalias.c (set_union_with_increment): Properly
776         take into account all fields that overlap the shifted vars.
777         (do_sd_constraint): Likewise.
778         (do_ds_constraint): Likewise.
779         (get_constraint_for_ptr_offset): Likewise.
781 2014-03-11  Chung-Lin Tang  <cltang@codesourcery.com>
783         * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
784         (nios2_compute_frame_layout):
785         Add calculation of cfun->machine->fp_save_offset.
786         (nios2_expand_prologue): Correct setting of frame pointer register
787         in prologue.
788         (nios2_expand_epilogue): Update recovery of stack pointer from
789         frame pointer accordingly.
790         (nios2_initial_elimination_offset): Update calculation of offset
791         for eliminating to HARD_FRAME_POINTER_REGNUM.
793 2014-03-10  Jakub Jelinek  <jakub@redhat.com>
795         PR ipa/60457
796         * ipa.c (symtab_remove_unreachable_nodes): Don't call
797         cgraph_get_create_node on VAR_DECLs.
799 2014-03-10  Richard Biener  <rguenther@suse.de>
801         PR middle-end/60474
802         * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
804 2014-03-08  Douglas B Rupp  <rupp@gnat.com>
806         * config/vms/vms.opt (vms_float_format): New variable.
808 2014-03-08  Tobias Burnus  <burnus@net-b.de>
810         * doc/invoke.texi (-fcilkplus): Update implementation status.
812 2014-03-08  Paulo Matos  <paulo@matos-sorge.com>
813             Richard Biener  <rguenther@suse.de>
815         * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
816         consistently accross all TUs.
817         (run_gcc): Enable -fshort-double automatically at link at link-time
818         and disallow override.
820 2014-03-08  Richard Sandiford  <rdsandiford@googlemail.com>
822         PR target/58271
823         * config/mips/mips.c (mips_option_override): Promote -mpaired-single
824         warning to an error.  Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
825         if they can't be used.
827 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
829         * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
830         for Solaris 11/x86 ld.
831         * configure: Regenerate.
833 2014-03-07  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
835         * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
836         (LIB_TLS_SPEC): Save as ld_tls_libs.
837         (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
838         (HAVE_AS_IX86_TLSLDM): New test.
839         * configure, config.in: Regenerate.
840         * config/i386/i386.c (legitimize_tls_address): Fall back to
841         TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
842         cannot support TLS_MODEL_LOCAL_DYNAMIC.
843         * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
844         instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
846 2014-03-07  Paulo Matos  <paulo@matos-sorge.com>
848         * common.opt (fira-loop-pressure): Mark as optimization.
850 2014-03-07  Thomas Schwinge  <thomas@codesourcery.com>
852         * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
853         an OpenMP mappable type.
855 2014-03-06  Matthias Klose  <doko@ubuntu.com>
857         * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
858         MULTILIB_OSDIRNAMES is not defined.
860 2014-03-06  Jakub Jelinek  <jakub@redhat.com>
861             Meador Inge  <meadori@codesourcery.com>
863         PR target/58595
864         * config/arm/arm.c (arm_tls_symbol_p): Remove.
865         (arm_legitimize_address): Call legitimize_tls_address for any
866         arm_tls_referenced_p expression, handle constant addend.  Call it
867         before testing for !TARGET_ARM.
868         (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
870 2014-03-06  Richard Biener  <rguenther@suse.de>
872         PR middle-end/60445
873         PR lto/60424
874         PR lto/60427
875         Revert
876         2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
878         * tree-streamer.c (record_common_node): Assert we don't record
879         nodes with type double.
880         (preload_common_node): Skip type double, complex double and double
881         pointer since it is now frontend dependent due to fshort-double option.
883 2014-03-06  Richard Biener  <rguenther@suse.de>
885         * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
886         or -fno-lto is specified and the linker has full plugin support.
887         * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
888         (main): Remove -flto processing, adjust lto_mode using use_plugin late.
889         * lto-wrapper.c (merge_and_complain): Merge compile-time
890         optimization levels.
891         (run_gcc): And pass it through to the link options.
893 2014-03-06  Alexandre Oliva <aoliva@redhat.com>
895         PR debug/60381
896         Revert:
897         2014-02-28  Alexandre Oliva <aoliva@redhat.com>
898         PR debug/59992
899         * cselib.c (remove_useless_values): Skip to avoid quadratic
900         behavior if the condition moved from...
901         (cselib_process_insn): ... here holds.
903 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
905         PR plugins/59335
906         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
907         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
909         PR plugins/59335
910         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
911         (TM_H): Add x86-tune.def.
913 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
915         * config/aarch64/aarch64.c (generic_tunings):
916         Use cortexa57_extra_costs.
918 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
920         PR lto/60404
921         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
922         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
923         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
924         cost for in_lto_p.
926 2014-03-04  Heiher  <r@hev.cc>
928         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
929         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
931 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
933         * config/i386/predicates.md (const2356_operand): Change to ...
934         (const2367_operand): ... this.
935         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
936         const2367_operand.
937         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
938         (*avx512pf_scatterpf<mode>sf): Ditto.
939         (avx512pf_scatterpf<mode>df): Ditto.
940         (*avx512pf_scatterpf<mode>df_mask): Ditto.
941         (*avx512pf_scatterpf<mode>df): Ditto.
942         * config/i386/i386.c (ix86_expand_builtin): Update
943         incorrect hint operand error message.
945 2014-03-04  Richard Biener  <rguenther@suse.de>
947         * lto-section-in.c (lto_get_section_data): Fix const cast.
949 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
951         * tree-streamer.c (record_common_node): Assert we don't record
952         nodes with type double.
953         (preload_common_node): Skip type double, complex double and double
954         pointer since it is now frontend dependent due to fshort-double option.
956 2014-03-04  Richard Biener  <rguenther@suse.de>
958         PR lto/60405
959         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
960         (lto_input_toplevel_asms): Likewise.
961         * lto-section-in.c (lto_get_section_data): Instead do it here
962         for every section.
964 2014-03-04  Richard Biener  <rguenther@suse.de>
966         PR tree-optimization/60382
967         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
968         dead PHIs a reduction.
970 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
972         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
973         hint value.
974         (_mm_prefetch): Move out of GCC target("sse") pragma.
975         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
976         GCC target("prfchw") pragma.
977         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
978         for locality <= 2.
979         * config/i386/i386.c (ix86_option_override_internal): Enable
980         -mprfchw with -mprefetchwt1.
982 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
984         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
985         Mark as varying.
987 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
989         * opts.h (CL_PCH_IGNORE): Define.
990         * targhooks.c (option_affects_pch_p):
991         Return false for options that have CL_PCH_IGNORE set.
992         * opt-functions.awk: Process PchIgnore.
993         * doc/options.texi: Document PchIgnore.
995         * config/arc/arc.opt (misize): Add PchIgnore property.
997 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
999         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
1000         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
1001         constraint on constants to permit them being loaded into
1002         GENERAL_REGS or BASE_REGS.
1004 2014-03-03  Nick Clifton  <nickc@redhat.com>
1006         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
1007         anti-cacnonical alternatives.
1008         (negandhi3_real): New pattern.
1009         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
1011 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1013         * config/avr/avr-mcus.def: Remove atxmega16x1.
1014         * config/avr/avr-tables.opt: Regenerate.
1015         * config/avr/t-multilib: Regenerate.
1016         * doc/avr-mmcu.texi: Regenerate.
1018 2014-03-03  Tobias Grosser  <tobias@grosser.es>
1019             Mircea Namolaru  <mircea.namolaru@inria.fr>
1021         PR tree-optimization/58028
1022         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
1023         scalar dimensions.
1025 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
1027         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
1028         not handled by recognizers.
1030 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
1032         PR middle-end/60175
1033         * function.c (expand_function_end): Don't emit
1034         clobber_return_register sequence if clobber_after is a BARRIER.
1035         * cfgexpand.c (construct_exit_block): Append instructions before
1036         return_label to prev_bb.
1038 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1040         * config/rs6000/constraints.md: Document reserved use of "wc".
1042 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1044         PR ipa/60150
1045         * ipa.c (function_and_variable_visibility): When dissolving comdat
1046         group, also set all symbols to local.
1048 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
1050         PR ipa/60306
1052         Revert:
1053         2013-12-14   Jan Hubicka  <jh@suse.cz>
1054         PR middle-end/58477
1055         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
1057 2014-03-02  Jon Beniston  <jon@beniston.com>
1059         PR bootstrap/48230
1060         PR bootstrap/50927
1061         PR bootstrap/52466
1062         PR target/46898
1063         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
1064           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
1065         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
1066         (simple_return, *simple_return): New patterns
1067         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
1068         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
1070 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
1072         * dwarf2out.c (gen_subprogram_die): Tidy.
1074 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
1076         PR target/60071
1077         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
1078         (*mov_t_msb_neg_negc): ... this new insn.
1080 2014-02-28  Jason Merrill  <jason@redhat.com>
1082         PR c++/58678
1083         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
1084         function.
1086 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
1088         PR c++/60314
1089         * dwarf2out.c (decltype_auto_die): New static.
1090         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
1091         (gen_type_die_with_usage): Handle 'decltype(auto)'.
1092         (is_cxx_auto): Likewise.
1094 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
1096         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
1097         we are not using general regs only.
1099 2014-02-28  Richard Biener  <rguenther@suse.de>
1101         PR target/60280
1102         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
1103         previous fix and only allow to remove trivial pre-headers
1104         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
1105         (remove_forwarder_block): Properly update the latch of a loop.
1107 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1109         PR debug/59992
1110         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
1111         (cselib_preserved_hash_table): New.
1112         (preserve_constants_and_equivs): Move preserved vals to it.
1113         (cselib_find_slot): Look it up first.
1114         (cselib_init): Initialize it.
1115         (cselib_finish): Release it.
1116         (dump_cselib_table): Dump it.
1118 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1120         PR debug/59992
1121         * cselib.c (remove_useless_values): Skip to avoid quadratic
1122         behavior if the condition moved from...
1123         (cselib_process_insn): ... here holds.
1125 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
1127         PR debug/57232
1128         * var-tracking.c (vt_initialize): Apply the same condition to
1129         preserve the CFA base value.
1131 2014-02-28  Joey Ye  <joey.ye@arm.com>
1133         PR target/PR60169
1134         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1135         if reload in progress or completed.
1137 2014-02-28  Tobias Burnus  <burnus@net-b.de>
1139         PR middle-end/60147
1140         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1141         NAMELIST_DECL.
1143 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
1145         * doc/tm.texi.in (Condition Code Status): Update documention for
1146         relative locations of cc0-setter and cc0-user.
1148 2014-02-27  Jeff Law  <law@redhat.com>
1150         PR rtl-optimization/52714
1151         * combine.c (try_combine): When splitting an unrecognized PARALLEL
1152         into two independent simple sets, if I3 is a jump, ensure the
1153         pattern we place into I3 is a (set (pc) ...).
1155 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
1156             Jeff Law  <law@redhat.com>
1158         PR rtl-optimization/49847
1159         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1160         are in different blocks.
1161         * doc/tm.texi (Condition Code Status): Update documention for
1162         relative locations of cc0-setter and cc0-user.
1164 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
1166         PR target/59222
1167         * lra.c (lra_emit_add): Check SUBREG too.
1169 2014-02-27  Andreas Schwab  <schwab@suse.de>
1171         * config/m68k/m68k.c (m68k_option_override): Disable
1172         -flive-range-shrinkage for classic m68k.
1173         (m68k_override_options_after_change): Likewise.
1175 2014-02-27  Marek Polacek  <polacek@redhat.com>
1177         PR middle-end/59223
1178         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1179         -Wmaybe-uninitialized.
1181 2014-02-27  Alan Modra  <amodra@gmail.com>
1183         PR target/57936
1184         * reload1.c (emit_input_reload_insns): When reload_override_in,
1185         set old to rl->in_reg when rl->in_reg is a subreg.
1187 2014-02-26  Richard Biener  <rguenther@suse.de>
1189         PR bootstrap/60343
1190         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1192 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1194         * common/config/i386/predicates.md (const1256_operand): Remove.
1195         (const2356_operand): New.
1196         (const_1_to_2_operand): Remove.
1197         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1198         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1199         (*avx512pf_gatherpf<mode>sf): Ditto.
1200         (avx512pf_gatherpf<mode>df): Ditto.
1201         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1202         (*avx512pf_gatherpf<mode>df): Ditto.
1203         (avx512pf_scatterpf<mode>sf): Ditto.
1204         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1205         (*avx512pf_scatterpf<mode>sf): Ditto.
1206         (avx512pf_scatterpf<mode>df): Ditto.
1207         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1208         (*avx512pf_scatterpf<mode>df): Ditto.
1209         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1211 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
1213         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1214         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1215         (_mm512_mask_testn_epi64_mask): Move to ...
1216         * config/i386/avx512cdintrin.h: Here.
1217         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1218         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1219         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1220         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1221         TARGET_AVX512F from TARGET_AVX512CD.
1223 2014-02-26  Richard Biener  <rguenther@suse.de>
1225         PR ipa/60327
1226         * ipa.c (walk_polymorphic_call_targets): Properly guard
1227         call to inline_update_overall_summary.
1229 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
1231         PR target/60280
1232         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1233         and latches only if requested.  Fix latch if it is removed.
1234         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1235         LOOPS_HAVE_PREHEADERS.
1237 2014-02-25  Andrew Pinski  <apinski@cavium.com>
1239         * builtins.c (expand_builtin_thread_pointer): Create a new target
1240         when the target is NULL.
1242 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
1244         PR rtl-optimization/60317
1245         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1246         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1247         * lra-assigns.c: Include params.h.
1248         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1249         other reload pseudos considerations.
1251 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1253         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1254         to use canonical form for nor<mode>3.
1256 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1258         PR target/55426
1259         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1260         conversions.
1262 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
1264         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1265         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1266         (ix86_handle_option): Handle OPT_mprefetchwt1.
1267         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1268         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1269         PREFETCHWT1 CPUID.
1270         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1271         OPTION_MASK_ISA_PREFETCHWT1.
1272         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1273         (PTA_PREFETCHWT1): New.
1274         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1275         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1276         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1277         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1278         (*prefetch_avx512pf_<mode>_: Change into ...
1279         (*prefetch_prefetchwt1_<mode>: This.
1280         * config/i386/i386.opt (mprefetchwt1): New.
1281         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1282         (_mm_prefetch): Handle intent to write.
1283         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1285 2014-02-25  Richard Biener  <rguenther@suse.de>
1287         PR middle-end/60291
1288         * emit-rtl.c (mem_attrs_htab): Remove.
1289         (mem_attrs_htab_hash): Likewise.
1290         (mem_attrs_htab_eq): Likewise.
1291         (set_mem_attrs): Always allocate new mem-attrs when something changed.
1292         (init_emit_once): Do not allocate mem_attrs_htab.
1294 2014-02-25  Richard Biener  <rguenther@suse.de>
1296         PR lto/60319
1297         * lto-opts.c (lto_write_options): Output non-explicit conservative
1298         -fwrapv, -fno-trapv and -fno-strict-overflow.
1299         * lto-wrapper.c (merge_and_complain): Handle merging those options.
1300         (run_gcc): And pass them through.
1302 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1304         * sel-sched.c (calculate_new_fences): New parameter ptime.
1305         Calculate it as a maximum over all fence cycles.
1306         (sel_sched_region_2): Adjust the call to calculate_new_fences.
1307         Print the final schedule timing when sched_verbose.
1309 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
1311         PR rtl-optimization/60292
1312         * sel-sched.c (fill_vec_av_set): Do not reset target availability
1313         bit fot the fence instruction.
1315 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
1317         * calls.h: Fix typo in comment.
1319 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
1321         * config/pa/pa.c (pa_output_move_double): Don't valididate when
1322         adjusting offsetable addresses.
1324 2014-02-24  Guozhi Wei  <carrot@google.com>
1326         * sparseset.h (sparseset_pop): Fix the wrong index.
1328 2014-02-24  Walter Lee  <walt@tilera.com>
1330         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1331         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1332         triplet.
1333         * common/config/tilegx/tilegx-common.c
1334         (TARGET_DEFAULT_TARGET_FLAGS): Define.
1335         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1336         (LINK_SPEC): Ditto.
1337         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1338         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1339         (tilegx_gimplify_va_arg_expr): Handle big endian.
1340         (tilegx_expand_unaligned_load): Ditto.
1341         (tilegx_expand_unaligned_store): Ditto.
1342         (TARGET_RETURN_IN_MSB): New.
1343         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1344         (TARGET_ENDIAN_DEFAULT): New.
1345         (TARGET_BIG_ENDIAN): Handle big endian.
1346         (BYTES_BIG_ENDIAN): Ditto.
1347         (WORDS_BIG_ENDIAN): Ditto.
1348         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1349         (ENDIAN_SPEC): New.
1350         (EXTRA_SPECS): New.
1351         * config/tilegx/tilegx.md (extv): Handle big endian.
1352         (extzv): Ditto.
1353         (insn_st<n>): Ditto.
1354         (insn_st<n>_add<bitsuffix>): Ditto.
1355         (insn_stnt<n>): Ditto.
1356         (insn_stnt<n>_add<bitsuffix>):Ditto.
1357         (vec_interleave_highv8qi): Handle big endian.
1358         (vec_interleave_highv8qi_be): New.
1359         (vec_interleave_highv8qi_le): New.
1360         (insn_v1int_h): Handle big endian.
1361         (vec_interleave_lowv8qi): Handle big endian.
1362         (vec_interleave_lowv8qi_be): New.
1363         (vec_interleave_lowv8qi_le): New.
1364         (insn_v1int_l): Handle big endian.
1365         (vec_interleave_highv4hi): Handle big endian.
1366         (vec_interleave_highv4hi_be): New.
1367         (vec_interleave_highv4hi_le): New.
1368         (insn_v2int_h): Handle big endian.
1369         (vec_interleave_lowv4hi): Handle big endian.
1370         (vec_interleave_lowv4hi_be): New.
1371         (vec_interleave_lowv4hi_le): New.
1372         (insn_v2int_l): Handle big endian.
1373         (vec_interleave_highv2si): Handle big endian.
1374         (vec_interleave_highv2si_be): New.
1375         (vec_interleave_highv2si_le): New.
1376         (insn_v4int_h): Handle big endian.
1377         (vec_interleave_lowv2si): Handle big endian.
1378         (vec_interleave_lowv2si_be): New.
1379         (vec_interleave_lowv2si_le): New.
1380         (insn_v4int_l): Handle big endian.
1381         * config/tilegx/tilegx.opt (mbig-endian): New option.
1382         (mlittle-endian): New option.
1383         * doc/install.texi: Document tilegxbe-linux.
1384         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1386 2014-02-24  Martin Jambor  <mjambor@suse.cz>
1388         PR ipa/60266
1389         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1390         there are no parameter descriptors.
1392 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
1394         PR rtl-optimization/60268
1395         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1396         initialization to ...
1397         (sched_rgn_init): ... here.
1398         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1400 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1402         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1403         names.
1405 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1407         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1408         definition.
1410 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1412         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1413         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1415 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
1417         * config/microblaze/predicates.md: Add cmp_op predicate.
1418         * config/microblaze/microblaze.md: Add branch_compare instruction
1419         which uses cmp_op predicate and emits cmp insn before branch.
1420         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1421         to microblaze_expand_conditional_branch and consolidate logic.
1422         (microblaze_expand_conditional_branch): emit branch_compare
1423         insn instead of handling cmp op separate from branch insn.
1425 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1427         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1428         to permit subregs.
1430 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1432         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1433         define_insn with define_expand and new define_insn
1434         *altivec_lve<VI_char>x_internal.
1435         (altivec_stve<VI_char>x): Replace define_insn with define_expand
1436         and new define_insn *altivec_stve<VI_char>x_internal.
1437         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1438         prototype.
1439         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1440         lve*x built-ins.
1441         (altivec_expand_stvex_be): New function.
1443 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
1445         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1446         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1447         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1448         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1450 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
1452         PR target/60298
1453         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1454         instead of emit_move_insn.
1456 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1458         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1459         vspltw with vsldoi.
1460         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1461         gen_altivec_vsumsws.
1463 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1465         * config/rs6000/altivec.md (altivec_lvxl): Rename as
1466         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1467         (altivec_lvxl_<mode>): New define_expand incorporating
1468         -maltivec=be semantics where needed.
1469         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1470         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1471         semantics where needed.
1472         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1473         (altivec_stvx_<mode>): New define_expand incorporating
1474         -maltivec=be semantics where needed.
1475         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1476         VM2 iterator instead of V4SI.
1477         (altivec_stvxl_<mode>): New define_expand incorporating
1478         -maltivec=be semantics where needed.
1479         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1480         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1481         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1482         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1483         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1484         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1485         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1486         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1487         ALTIVEC_BUILTIN_STVXL.
1488         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1489         (altivec_expand_stvx_be): Likewise.
1490         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1491         (altivec_expand_lvx_be): Likewise.
1492         (altivec_expand_stvx_be): Likewise.
1493         (altivec_expand_builtin): Add cases for
1494         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1495         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1496         (altivec_init_builtins): Add definitions for
1497         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1498         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1500 2014-02-21  Catherine Moore  <clm@codesourcery.com>
1502         * doc/invoke.texi (mvirt, mno-virt): Document.
1503         * config/mips/mips.opt (mvirt): New option.
1504         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1506 2014-02-21  Richard Biener  <rguenther@suse.de>
1508         PR tree-optimization/60276
1509         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1510         (STMT_VINFO_MIN_NEG_DIST): New macro.
1511         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1512         STMT_VINFO_MIN_NEG_DIST.
1513         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1514         made for negative dependence distances still hold.
1516 2014-02-21  Richard Biener  <rguenther@suse.de>
1518         PR middle-end/60291
1519         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1520         DECL_INITIAL for globals not in the current function context.
1522 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1524         PR tree-optimization/56490
1525         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1526         * tree-ssa-uninit.c: Include params.h.
1527         (compute_control_dep_chain): Add num_calls argument, return false
1528         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1529         num_calls to recursive call.
1530         (find_predicates): Change dep_chain into normal array,
1531         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1532         variable and adjust compute_control_dep_chain caller.
1533         (find_def_preds): Likewise.
1535 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
1537         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1538         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1540 2014-02-21  Nick Clifton  <nickc@redhat.com>
1542         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1543         (pushhi1): Likewise.
1544         (popqi1): Add mode to pre_dec.
1545         (pophi1): Likewise.
1547 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
1549         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1550         mode for mask of V8SFmode permutation.
1552 2014-02-20  Richard Henderson  <rth@redhat.com>
1554         PR c++/60272
1555         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1556         a new pseudo for OLDVAL.
1558 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
1560         PR target/57896
1561         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1562         gen_reg_rtx if d->testing_p.
1563         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1564         if d->testing_p and we will certainly return true.
1565         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
1566         if d->testing_p.
1568 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
1570         * emit-rtl.c (gen_reg_rtx): Assert that
1571         crtl->emit.regno_pointer_align_length is non-zero.
1573 2014-02-20  Richard Henderson  <rth@redhat.com>
1575         PR c++/60272
1576         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1577         on failure the store back into EXPECT.
1579 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
1580             Sandra Loosemore  <sandra@codesourcery.com>
1582         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1583         * config/nios2/nios2.c (nios2_function_profiler): Add
1584         -fPIC (flag_pic == 2) support.
1585         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1586         (nios2_large_offset_p): New function.
1587         (nios2_unspec_reloc_p): Move up position, update to use
1588         nios2_large_offset_p.
1589         (nios2_unspec_address): Remove function.
1590         (nios2_unspec_offset): New function.
1591         (nios2_large_got_address): New function.
1592         (nios2_got_address): Add large offset support.
1593         (nios2_legitimize_tls_address): Update usage of removed and new
1594         functions.
1595         (nios2_symbol_binds_local_p): New function.
1596         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1597         (nios2_legitimize_address): Update to use nios2_large_offset_p.
1598         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1599         (nios2_print_operand): Merge H/L processing, add hiadj/lo
1600         processing for (const (unspec ...)).
1601         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1603 2014-02-20  Richard Biener  <rguenther@suse.de>
1605         * tree-cfg.c (replace_uses_by): Mark altered BBs before
1606         doing the substitution.
1607         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1609 2014-02-20  Martin Jambor  <mjambor@suse.cz>
1611         PR ipa/55260
1612         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1613         info when checking whether lattices are bottom.
1615 2014-02-20  Richard Biener  <rguenther@suse.de>
1617         PR middle-end/60221
1618         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1619         regions at -O0.
1621 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
1623         PR ipa/58555
1624         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1625         parameter specifying the scaling.
1626         (inline_call): Update.
1627         (want_inline_recursively): Guard division by zero.
1628         (recursive_inlining): Update.
1629         * ipa-inline.h (clone_inlined_nodes): Update.
1631 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1633         PR target/60204
1634         * config/i386/i386.c (classify_argument): Pass structures of size
1635         64 bytes or less in register.
1637 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
1638             Kirill Yukhin  <kirill.yukhin@intel.com>
1640         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1641         (_mm_rcp28_round_ss): Ditto.
1642         (_mm_rsqrt28_round_sd): Ditto.
1643         (_mm_rsqrt28_round_ss): Ditto.
1644         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1645         (_mm_rcp14_round_ss): Ditto.
1646         (_mm_rsqrt14_round_sd): Ditto.
1647         (_mm_rsqrt14_round_ss): Ditto.
1648         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1649         the first input operand, get rid of match_dup.
1650         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1651         attribute to sse.
1652         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1653         Ditto.
1654         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1655         operand as the first input operand, set type attribute.
1656         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1657         Set type attribute.
1658         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1659         operand as the first input operand, set type attribute.
1661 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1663         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1664         bit of zero.
1666 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
1668         PR target/60207
1669         * config/i386/i386.c (construct_container): Remove TFmode check
1670         for X86_64_INTEGER_CLASS.
1672 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
1674         PR target/59794
1675         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1676         only when -Wpsabi is enabled.
1678 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
1680          PR target/59799
1681         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1682         passing arrays in registers are the same as for structs, so remove the
1683         special case for them.
1685 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
1687         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1688         destination type, extract only the valid bits if the source type is not
1689         integral and has a different mode.
1691 2014-02-19  Richard Biener  <rguenther@suse.de>
1693         PR ipa/60243
1694         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1695         for all calls.
1697 2014-02-19  Richard Biener  <rguenther@suse.de>
1699         PR ipa/60243
1700         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1701         (ipa_modify_call_arguments): Emit an argument load explicitely and
1702         preserve virtual SSA form there and for the replacement call.
1703         Do not update SSA form nor free dominance info.
1705 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1707         * ipa.c (function_and_variable_visibility): Also clear WEAK
1708         flag when disolving COMDAT_GROUP.
1710 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1712         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1713         * ipa-prop.c (ipa_set_jf_known_type): Return early when
1714         not devirtualizing.
1715         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1716         do more sanity checks.
1717         (detect_type_change): Return true when giving up early.
1718         (compute_complex_assign_jump_func): Fix type parameter of
1719         ipa_set_ancestor_jf.
1720         (compute_complex_ancestor_jump_func): Likewise.
1721         (update_jump_functions_after_inlining): Fix updating of
1722         ancestor function.
1723         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1725 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
1727         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1728         inline clones when edge disappears.
1730 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
1732         PR target/60203
1733         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1734         Split 64-bit moves into 2 patterns.  Do not allow the use of
1735         direct move for TDmode in little endian, since the decimal value
1736         has little endian bytes within a word, but the 64-bit pieces are
1737         ordered in a big endian fashion, and normal subreg's of TDmode are
1738         not allowed.
1739         (mov<mode>_64bit_dm): Likewise.
1740         (movtd_64bit_nodm): Likewise.
1742 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1744         PR tree-optimization/60174
1745         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1746         statement of an SSA_NAME that occurs in an abnormal PHI node.
1748 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1750         PR sanitizer/60142
1751         * final.c (SEEN_BB): Remove.
1752         (SEEN_NOTE, SEEN_EMITTED): Renumber.
1753         (final_scan_insn): Don't force_source_line on second
1754         NOTE_INSN_BASIC_BLOCK.
1756 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
1758         PR target/60205
1759         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1760         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1761         (type_natural_mode): Warn ABI change when %zmm register is not
1762         available for AVX512F vector value passing.
1764 2014-02-18  Kai Tietz  <ktietz@redhat.com>
1766         PR target/60193
1767         * config/i386/i386.c (ix86_expand_prologue): Use value in
1768         rax register as displacement when restoring %r10 or %rax.
1769         Fix wrong offset when restoring both registers.
1771 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
1773         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1774         assertion with conditional return.
1776 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
1777             Uros Bizjak  <ubizjak@gmail.com>
1779         PR driver/60233
1780         * config/i386/driver-i386.c (host_detect_local_cpu): If
1781         YMM state is not saved by the OS, also clear has_f16c.  Move
1782         CPUID 0x80000001 handling before YMM state saving checking.
1784 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
1786         PR rtl-optimization/58960
1787         * haifa-sched.c (alloc_global_sched_pressure_data): New,
1788         factored out from ...
1789         (sched_init): ... here.
1790         (free_global_sched_pressure_data): New, factored out from ...
1791         (sched_finish): ... here.
1792         * sched-int.h (free_global_sched_pressure_data): Declare.
1793         * sched-rgn.c (nr_regions_initial): New static global.
1794         (haifa_find_rgns): Initialize it.
1795         (schedule_region): Disable sched-pressure for the newly
1796         generated regions.
1798 2014-02-17  Richard Biener  <rguenther@suse.de>
1800         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1801         release SSA defs of pattern stmts.
1803 2014-02-17  Richard Biener  <rguenther@suse.de>
1805         * tree-inline.c (expand_call_inline): Release the virtual
1806         operand defined by the call we are about to inline.
1808 2014-02-17  Richard Biener  <rguenther@suse.de>
1810         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1812 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
1813             Ilya Tocar  <ilya.tocar@intel.com>
1815         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1816         arguments order in builtin.
1817         (_mm512_permutexvar_epi64): Ditto.
1818         (_mm512_mask_permutexvar_epi64): Ditto
1819         (_mm512_maskz_permutexvar_epi32): Ditto
1820         (_mm512_permutexvar_epi32): Ditto
1821         (_mm512_mask_permutexvar_epi32): Ditto
1823 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1825         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1826         (p8_vmrgow): Likewise.
1828 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1830         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1831         endian targets.
1833 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
1835         PR target/60203
1836         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1837         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1838         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
1839         using direct move instructions on ISA 2.07.  Also adjust
1840         instruction length for 64-bit.
1841         (mov<mode>_64bit, TFmode/TDmode): Likewise.
1842         (mov<mode>_32bit, TFmode/TDmode): Likewise.
1844 2014-02-15  Alan Modra  <amodra@gmail.com>
1846         PR target/58675
1847         PR target/57935
1848         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1849         find_replacement on parts of insn rtl that might be reloaded.
1851 2014-02-15  Richard Biener  <rguenther@suse.de>
1853         PR tree-optimization/60183
1854         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1855         (tree_ssa_phiprop): Calculate and free post-dominators.
1857 2014-02-14  Jeff Law  <law@redhat.com>
1859         PR rtl-optimization/60131
1860         * ree.c (get_extended_src_reg): New function.
1861         (combine_reaching_defs): Use it rather than assuming location of REG.
1862         (find_and_remove_re): Verify first operand of extension is
1863         a REG before adding the insns to the copy list.
1865 2014-02-14  Roland McGrath  <mcgrathr@google.com>
1867         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1868         * configure: Regenerated.
1869         * config.in: Regenerated.
1870         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1871         instead of ASM_SHORT.
1873 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
1874             Richard Earnshaw  <rearnsha@arm.com>
1876         PR rtl-optimization/59535
1877         * lra-constraints.c (process_alt_operands): Encourage alternative
1878         when unassigned pseudo class is superset of the alternative class.
1879         (inherit_reload_reg): Don't inherit when optimizing for code size.
1880         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1881         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1882         modes not less than 4 for Thumb1.
1884 2014-02-14  Kyle McMartin  <kyle@redhat.com>
1886         PR pch/60010
1887         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1889 2014-02-14  Richard Biener  <rguenther@suse.de>
1891         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1892         (get_frame_arg): Drop the assert with langhook types_compatible_p.
1893         Do not strip INDIRECT_REFs.
1895 2014-02-14  Richard Biener  <rguenther@suse.de>
1897         PR lto/60179
1898         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1899         DECL_FUNCTION_SPECIFIC_TARGET.
1900         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1901         * tree-streamer-out.c (pack_ts_target_option): Remove.
1902         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1903         (write_ts_function_decl_tree_pointers): Do not stream
1904         DECL_FUNCTION_SPECIFIC_TARGET.
1905         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1906         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1907         (lto_input_ts_function_decl_tree_pointers): Do not stream
1908         DECL_FUNCTION_SPECIFIC_TARGET.
1910 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1912         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1913         (get_initial_def_for_induction, vectorizable_induction): Ignore
1914         debug stmts when looking for exit_phi.
1915         (vectorizable_live_operation): Fix up condition.
1917 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1919         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1920         nreverse() because it changes the content of original tree list.
1922 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1924         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1925         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1927 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1929         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1930         GNU coding standards.
1932 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1934         PR debug/60152
1935         * dwarf2out.c (gen_subprogram_die): Don't call
1936         add_calling_convention_attribute if subr_die is old_die.
1938 2014-02-13  Sharad Singhai  <singhai@google.com>
1940         * doc/optinfo.texi: Fix order of nodes.
1942 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1944         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1945         operands[2], not operands[3].
1947 2014-02-13  Richard Biener  <rguenther@suse.de>
1949         PR bootstrap/59878
1950         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1951         mention the possibility of an in-tree build.
1952         (CLooG): Update recommended version to 0.18.1, mention the
1953         possibility of an in-tree build and clarify that the ISL
1954         bundled with CLooG does not work.
1956 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1958         PR target/43546
1959         * expr.c (compress_float_constant): If x is a hard register,
1960         extend into a pseudo and then move to x.
1962 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1964         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1965         caused by bad second argument to warning_at() with -mhotpatch and
1966         nested functions (e.g. with gfortran).
1968 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1970         * opts.c (option_name): Remove "enabled by default" rider.
1972 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1974         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1976 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1977             Uros Bizjak  <ubizjak@gmail.com>
1979         PR target/60151
1980         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1981         * configure: Regenerated.
1983 2014-02-12  Richard Biener  <rguenther@suse.de>
1985         * vec.c (vec_prefix::calculate_allocation): Move as
1986         inline variant to vec.h.
1987         (vec_prefix::calculate_allocation_1): New out-of-line version.
1988         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1989         (vec_prefix::m_has_auto_buf): Rename to ...
1990         (vec_prefix::m_using_auto_storage): ... this.
1991         (vec_prefix::calculate_allocation): Inline the easy cases
1992         and dispatch to calculate_allocation_1 which doesn't need the
1993         prefix address.
1994         (va_heap::reserve): Use gcc_checking_assert.
1995         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1996         m_using_auto_storage.
1997         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1998         member and adjust.
1999         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
2000         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
2001         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
2003 2014-02-12  Richard Biener  <rguenther@suse.de>
2005         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
2006         when we found a dependence.
2008 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
2010         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
2011         common code...
2012         (maybe_fold_stmt): ... into this new function.
2013         * omp-low.c (lower_omp): Update comment.
2015         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
2016         last use.
2018         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
2019         dereference.
2021 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
2023         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
2024         identifiers in comments.
2025         (cortexa53_extra_costs): Likewise.
2026         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
2027         (cortexa7_extra_costs): Likewise.
2028         (cortexa12_extra_costs): Likewise.
2029         (cortexa15_extra_costs): Likewise.
2030         (v7m_extra_costs): Likewise.
2032 2014-02-12  Richard Biener  <rguenther@suse.de>
2034         PR middle-end/60092
2035         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
2036         of posix_memalign being successful.
2037         (lower_stmt): Restrict lowering of posix_memalign to when
2038         -ftree-bit-ccp is enabled.
2040 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
2042         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
2043         arg_loc.
2044         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
2046 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
2048         PR rtl-optimization/60116
2049         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
2050         other_insn once the combination has been validated.
2052 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
2054         PR lto/59468
2055         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
2056         and wrapper.
2057         * ipa-devirt.c: Include demangle.h
2058         (odr_violation_reported): New static variable.
2059         (add_type_duplicate): Update odr_violations.
2060         (maybe_record_node): Add completep parameter; update it.
2061         (record_target_from_binfo): Add COMPLETEP parameter;
2062         update it as needed.
2063         (possible_polymorphic_call_targets_1): Likewise.
2064         (struct polymorphic_call_target_d): Add nonconstruction_targets;
2065         rename FINAL to COMPLETE.
2066         (record_targets_from_bases): Sanity check we found the binfo;
2067         fix COMPLETEP updating.
2068         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
2069         parameter, fix computing of COMPLETEP.
2070         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
2071         at LTO time do demangling.
2072         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
2073         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
2074         parameter.
2075         (gimple_get_virt_method_for_binfo): Likewise.
2076         * gimple-fold.h (gimple_get_virt_method_for_binfo,
2077         gimple_get_virt_method_for_vtable): Update prototypes.
2079 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
2081         PR target/49008
2082         * genautomata.c (add_presence_absence): Fix typo with
2083         {final_}presence_list.
2085 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
2087         PR target/60137
2088         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
2089         for VSX/Altivec vectors that land in GPR registers.
2091 2014-02-11  Richard Henderson  <rth@redhat.com>
2092             Jakub Jelinek  <jakub@redhat.com>
2094         PR debug/59776
2095         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
2096         around drhs if type conversion to lacc->type is not useless.
2098 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2100         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
2101         tuning struct.
2102         (cortex-a57.cortex-a53): Likewise.
2103         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
2105 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2107         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
2108         arm_restrict_it.
2110 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
2112         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
2113         add_options_for_arm_vfp3.
2115 2014-02-11  Jeff Law  <law@redhat.com>
2117         PR middle-end/54041
2118         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
2119         object with an undesirable mode.
2121 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2123         PR libgomp/60107
2124         * config/i386/sol2-9.h: New file.
2125         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2126         *-*-solaris2.9*): Use it.
2128 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2130         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2131         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2133 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2135         * config/microblaze/microblaze.c: Extend mcpu version format
2137 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
2139         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2141 2014-02-10  Richard Henderson  <rth@redhat.com>
2143         PR target/59927
2144         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2145         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2146         ms-abi vs -mno-accumulate-outgoing-args.
2147         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2148         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2149         respect to ms-abi.
2151 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2153         PR middle-end/60080
2154         * cfgexpand.c (expand_asm_operands): Attach source location to
2155         ASM_INPUT rtx objects.
2156         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2158 2014-02-10  Nick Clifton  <nickc@redhat.com>
2160         * config/mn10300/mn10300.c (popcount): New function.
2161         (mn10300_expand_prologue): Include saved registers in stack usage
2162         count.
2164 2014-02-10  Jeff Law  <law@redhat.com>
2166         PR middle-end/52306
2167         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2168         when changing the SET_DEST of a prior insn to avoid an input reload.
2170 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
2172         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2173         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2174         -mcall-openbsd, or -mcall-linux.
2175         (CC1_ENDIAN_BIG_SPEC): Remove.
2176         (CC1_ENDIAN_LITTLE_SPEC): Remove.
2177         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2178         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2179         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2180         and %cc1_endian_default.
2181         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2183 2014-02-10  Richard Biener  <rguenther@suse.de>
2185         PR tree-optimization/60115
2186         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2187         MEM_REF handling.  Properly verify that the accesses are not
2188         out of the objects bound.
2190 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2192         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2193         coretex to cortex.
2195 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
2197         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2198         proper constants and fix formatting.
2199         (possible_polymorphic_call_targets): Fix formatting.
2201 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
2202             Ilya Tocar  <ilya.tocar@intel.com>
2204         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2205         (_mm512_loadu_epi32): Renamed into...
2206         (_mm512_loadu_si512): This.
2207         (_mm512_storeu_epi32): Renamed into...
2208         (_mm512_storeu_si512): This.
2209         (_mm512_maskz_ceil_ps): Removed.
2210         (_mm512_maskz_ceil_pd): Ditto.
2211         (_mm512_maskz_floor_ps): Ditto.
2212         (_mm512_maskz_floor_pd): Ditto.
2213         (_mm512_floor_round_ps): Ditto.
2214         (_mm512_floor_round_pd): Ditto.
2215         (_mm512_ceil_round_ps): Ditto.
2216         (_mm512_ceil_round_pd): Ditto.
2217         (_mm512_mask_floor_round_ps): Ditto.
2218         (_mm512_mask_floor_round_pd): Ditto.
2219         (_mm512_mask_ceil_round_ps): Ditto.
2220         (_mm512_mask_ceil_round_pd): Ditto.
2221         (_mm512_maskz_floor_round_ps): Ditto.
2222         (_mm512_maskz_floor_round_pd): Ditto.
2223         (_mm512_maskz_ceil_round_ps): Ditto.
2224         (_mm512_maskz_ceil_round_pd): Ditto.
2225         (_mm512_expand_pd): Ditto.
2226         (_mm512_expand_ps): Ditto.
2227         * config/i386/i386.c (ix86_builtins): Remove
2228         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2229         (bdesc_args): Ditto.
2230         * config/i386/predicates.md (const1256_operand): New.
2231         (const_1_to_2_operand): Ditto.
2232         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2233         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2234         (*avx512pf_gatherpf<mode>sf): Ditto.
2235         (avx512pf_gatherpf<mode>df): Ditto.
2236         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2237         (*avx512pf_gatherpf<mode>df): Ditto.
2238         (avx512pf_scatterpf<mode>sf): Ditto.
2239         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2240         (*avx512pf_scatterpf<mode>sf): Ditto.
2241         (avx512pf_scatterpf<mode>df): Ditto.
2242         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2243         (*avx512pf_scatterpf<mode>df): Ditto.
2244         (avx512f_expand<mode>): Removed.
2245         (<shift_insn><mode>3<mask_name>): Change predicate type.
2247 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2249         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2250         not at the end of datarefs vector use ordered_remove to avoid
2251         reordering datarefs vector.
2253         PR c/59984
2254         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2255         mark local addressable non-static vars as GOVD_PRIVATE
2256         instead of GOVD_LOCAL.
2257         * omp-low.c (lower_omp_for): Move gimple_bind_vars
2258         and BLOCK_VARS of gimple_bind_block to new_stmt rather
2259         than copying them.
2261         PR middle-end/60092
2262         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2263         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2264         assume_aligned or alloc_align attributes.
2265         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2266         arguments.  Handle also assume_aligned and alloc_align attributes.
2267         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
2268         calls to functions with assume_aligned or alloc_align attributes.
2269         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2271 2014-02-08  Terry Guo  <terry.guo@arm.com>
2273         * doc/invoke.texi: Document ARM -march=armv7e-m.
2275 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
2277         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2278         flag on __cilkrts_rethrow builtin.
2280         PR ipa/60026
2281         * ipa-cp.c (determine_versionability): Fail at -O0
2282         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2283         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2285         Revert:
2286         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2288         PR ipa/60026
2289         * tree-inline.c (copy_forbidden): Fail for
2290         __attribute__((optimize (0))) functions.
2292 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2294         * varpool.c: Include pointer-set.h.
2295         (varpool_remove_unreferenced_decls): Variables in other partitions
2296         will not be output; be however careful to not lose information
2297         about partitioning.
2299 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
2301         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2302         lookup in the vtable constructor.
2304 2014-02-07  Jeff Law  <law@redhat.com>
2306         PR target/40977
2307         * config/m68k/m68k.md (ashldi_extsi): Turn into a
2308         define_insn_and_split.
2310         * ipa-inline.c (inline_small_functions): Fix typos.
2312 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2314         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2315         (s390_can_use_return_insn): Declare.
2316         * config/s390/s390.h (EPILOGUE_USES): Define.
2317         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2318         instructions.
2319         (s390_chunkify_start): Handle return JUMP_LABELs.
2320         (s390_early_mach): Emit a main_pool instruction on the entry edge.
2321         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2322         (s390_can_use_return_insn): New functions.
2323         (s390_fix_long_loop_prediction): Handle conditional returns.
2324         (TARGET_SET_UP_BY_PROLOGUE): Define.
2325         * config/s390/s390.md (ANY_RETURN): New code iterator.
2326         (*creturn, *csimple_return, return, simple_return): New patterns.
2328 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2330         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2331         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
2332         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
2333         REG_CFA_RESTORE list when deciding not to restore a register.
2335 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2337         * config/s390/s390.c: Include tree-pass.h and context.h.
2338         (s390_early_mach): New function, split out from...
2339         (s390_emit_prologue): ...here.
2340         (pass_data_s390_early_mach): New pass structure.
2341         (pass_s390_early_mach): New class.
2342         (s390_option_override): Create and register early_mach pass.
2343         Move to end of file.
2345 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
2347         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2348         to match for the exit block.
2350 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2352         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2353         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2354         Reject misaligned operands.
2356 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
2358         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2360 2014-02-07  Richard Biener  <rguenther@suse.de>
2362         PR middle-end/60092
2363         * gimple-low.c (lower_builtin_posix_memalign): New function.
2364         (lower_stmt): Call it to lower posix_memalign in a way
2365         to make alignment info accessible.
2367 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
2369         PR c++/60082
2370         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2371         __builtin_setjmp_receiver.
2373 2014-02-07  Richard Biener  <rguenther@suse.de>
2375         PR middle-end/60092
2376         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2377         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2378         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2379         Handle BUILT_IN_POSIX_MEMALIGN.
2380         (find_func_clobbers): Likewise.
2381         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2382         (call_may_clobber_ref_p_1): Likewise.
2384 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2386         PR ipa/59918
2387         * ipa-devirt.c (record_target_from_binfo): Remove overactive
2388         sanity check.
2390 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2392         PR ipa/59469
2393         * lto-cgraph.c (lto_output_node): Use
2394         symtab_get_symbol_partitioning_class.
2395         (lto_output_varpool_node): likewise.
2396         (symtab_get_symbol_partitioning_class): Move here from
2397         lto/lto-partition.c
2398         * cgraph.h (symbol_partitioning_class): Likewise.
2399         (symtab_get_symbol_partitioning_class): Declare.
2401 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2403         * ggc.h (ggc_internal_cleared_alloc): New macro.
2404         * vec.h (vec_safe_copy): Handle memory stats.
2405         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2406         * target-globals.c (save_target_globals): Likewise.
2408 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
2410         PR target/60077
2411         * expr.c (emit_move_resolve_push): Export; be bit more selective
2412         on when to clear alias set.
2413         * expr.h (emit_move_resolve_push): Declare.
2414         * function.h (struct function): Add tail_call_marked.
2415         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2416         * config/i386/i386-protos.h (ix86_expand_push): Remove.
2417         * config/i386/i386.md (TImode move expander): De not call
2418         ix86_expand_push.
2419         (FP push expanders): Preserve memory attributes.
2420         * config/i386/sse.md (push<mode>1): Remove.
2421         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2422         (ix86_expand_push): Remove.
2423         * config/i386/mmx.md (push<mode>1): Remove.
2425 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2427         PR rtl-optimization/60030
2428         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2429         lopart with paradoxical subreg before shifting it up by hprec.
2431 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2433         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2434         Remove extra newline at end of file.
2435         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2436         (arm_issue_rate): Handle cortexa57.
2437         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2438         (cortex-a57.cortex-a53): Likewise.
2440 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2442         PR target/59575
2443         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2444         don't record in REG_FRAME_RELATED_EXPR registers not set in that
2445         bitmask.
2446         (arm_expand_prologue): Adjust all callers.
2447         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2448         info, registers also at the lowest numbered registers side.  Use
2449         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2450         XEXP.
2452         PR debug/59992
2453         * var-tracking.c (adjust_mems): Before adding a SET to
2454         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2456 2014-02-06  Alan Modra  <amodra@gmail.com>
2458         PR target/60032
2459         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2460         change SDmode to DDmode when lra_in_progress.
2462 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
2464         PR middle-end/59150
2465         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2466         free_data_ref on the dr first, and before goto again also set dr
2467         to the next dr.  For simd_lane_access, free old datarefs[i] before
2468         overwriting it.  For get_vectype_for_scalar_type failure, don't
2469         free_data_ref if simd_lane_access.
2471         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2473         PR target/60062
2474         * tree.h (opts_for_fn): New inline function.
2475         (opt_for_fn): Define.
2476         * config/i386/i386.c (ix86_function_regparm): Use
2477         opt_for_fn (decl, optimize) instead of optimize.
2479 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2481         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2482         for SYMBOL_REF in large memory model.
2484 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2486         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2487         and crypto support.
2488         (cortex-a57): Likewise.
2489         (cortex-a57.cortex-a53): Likewise.
2491 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
2492             Kugan Vivekanandarajah  <kuganv@linaro.org>
2494         * config/arm/arm.c (arm_vector_alignment_reachable): Check
2495         unaligned_access.
2496         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2498 2014-02-06  Richard Biener  <rguenther@suse.de>
2500         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2501         set_loop_copy and initialize_original_copy_tables.
2503 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
2505         * config/aarch64/aarch64-simd.md
2506         (aarch64_ashr_simddi): Change QI to SI.
2508 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2509             Jakub Jelinek  <jakub@redhat.com>
2511         PR middle-end/60013
2512         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2513         of the dataflow.
2515 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2517         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2518         CODE_FOR_altivec_vpku[hw]um to
2519         CODE_FOR_altivec_vpku[hw]um_direct.
2520         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2521         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2522         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2523         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2525 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2527         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2528         generation for -maltivec=be.
2529         (altivec_vsumsws): Simplify redundant test.
2531 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2533         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2534         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2535         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2536         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2537         gen_altivec_vpkuwum.
2538         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2539         BYTES_BIG_ENDIAN.
2540         (altivec_vpks<VI_char>ss): Likewise.
2541         (altivec_vpks<VI_char>us): Likewise.
2542         (altivec_vpku<VI_char>us): Likewise.
2543         (altivec_vpku<VI_char>um): Likewise.
2544         (altivec_vpku<VI_char>um_direct): New (copy of
2545         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2546         internal use).
2547         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2548         target is little endian and -maltivec=be is not specified.
2549         (*altivec_vupkhs<VU_char>_direct): New (copy of
2550         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2551         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2552         target is little endian and -maltivec=be is not specified.
2553         (*altivec_vupkls<VU_char>_direct): New (copy of
2554         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2555         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2556         little endian and -maltivec=be is not specified.
2557         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2558         little endian and -maltivec=be is not specified.
2560 2014-02-05  Richard Henderson  <rth@redhat.com>
2562         PR debug/52727
2563         * combine-stack-adj.c: Revert r206943.
2564         * sched-int.h (struct deps_desc): Add last_args_size.
2565         * sched-deps.c (init_deps): Initialize it.
2566         (sched_analyze_insn): Add OUTPUT dependencies between insns that
2567         contain REG_ARGS_SIZE notes.
2569 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
2571         * lto-cgraph.c (asm_nodes_output): Make global.
2572         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2573         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2574         (driver_handle_option): Handle OPT_fwpa.
2576 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
2578         PR ipa/59947
2579         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2580         a comment typo and formatting issue.  If odr_hash hasn't been
2581         created, return vNULL and set *completep to false.
2583         PR middle-end/57499
2584         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2585         bb with no successors.
2587 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
2589         PR target/59718
2590         * doc/invoke.texi (-march): Clarify documentation for ARM.
2591         (-mtune): Likewise.
2592         (-mcpu): Likewise.
2594 2014-02-05  Richard Biener  <rguenther@suse.de>
2596         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2597         when not vectorizing because of too many alias checks.
2598         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2599         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2601 2014-02-05  Nick Clifton  <nickc@redhat.com>
2603         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2604         accept extended registers in any mode when compiling for the MN10300.
2606 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
2608         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2609         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2610         sanitization attributes.
2611         (can_inline_edge_p): Likewise.
2612         (sanitize_attrs_match_for_inline_p): New function.
2614 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2616         * ipa-prop.c (detect_type_change): Shor circuit testing of
2617         type changes on THIS pointer.
2619 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
2621         PR target/59777
2622         * config/pa/pa.c (legitimize_tls_address): Return original address
2623         if not passed a SYMBOL_REF rtx.
2624         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2625         addresses.
2626         (pa_emit_move_sequence): Simplify TLS source operands.
2627         (pa_legitimate_constant_p): Reject all TLS constants.
2628         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2629         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2631 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2633         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2634         groups when we know they are controlled by LTO.
2635         * varasm.c (default_binds_local_p_1): If object is in other partition,
2636         it will be resolved locally.
2638 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2640         * config/host-linux.c (linux_gt_pch_use_address): Don't
2641         use SSIZE_MAX because it is not always defined.
2643 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
2645         PR bootstrap/59913
2646         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2647         threshold for pseudo splitting.
2648         (update_ebb_live_info): Process call argument hard registers and
2649         hard registers from insn definition too.
2650         (max_small_class_regs_num): New constant.
2651         (inherit_in_ebb): Update live hard regs through EBBs.  Update
2652         reloads_num only for small register classes.  Don't split for
2653         outputs of jumps.
2655 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
2657         PR ipa/60058
2658         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2659         is non-null.
2661 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
2663         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2664         visibility is safe.
2666 2014-02-04  Marek Polacek  <polacek@redhat.com>
2668         * gdbinit.in (pel): Define.
2670 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2672         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2673         behavior.
2675 2014-02-04  Richard Biener  <rguenther@suse.de>
2677         PR lto/59723
2678         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2679         in function context local.
2680         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2681         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2682         similar to LTO_imported_decl_ref.
2684 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
2686         PR tree-optimization/60002
2687         * cgraphclones.c (build_function_decl_skip_args): Clear
2688         DECL_LANG_SPECIFIC.
2690         PR tree-optimization/60023
2691         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2692         false to gsi_replace.
2693         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2694         has been in some EH region and vec_stmt could throw, add
2695         vec_stmt into the same EH region.
2696         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2697         has no lhs, ignore it.
2698         * internal-fn.c (expand_MASK_LOAD): Likewise.
2700         PR ipa/60026
2701         * tree-inline.c (copy_forbidden): Fail for
2702         __attribute__((optimize (0))) functions.
2704         PR other/58712
2705         * omp-low.c (simd_clone_struct_copy): If from->inbranch
2706         is set, copy one less argument.
2707         (expand_simd_clones): Don't subtract clone_info->inbranch
2708         from simd_clone_struct_alloc argument.
2710         PR rtl-optimization/57915
2711         * recog.c (simplify_while_replacing): If all unary/binary/relational
2712         operation arguments are constant, attempt to simplify those.
2714         PR middle-end/59261
2715         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2716         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2718 2014-02-04  Richard Biener  <rguenther@suse.de>
2720         PR tree-optimization/60012
2721         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2722         TBAA disambiguation to all DDRs.
2724 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2726         PR target/59788
2727         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2728         (LINK_SPEC): Use it for -shared, -shared-libgcc.
2730 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2732         PR ipa/59882
2733         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2735 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2737         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2738         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2740 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2742         PR ipa/59831
2743         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2744         to figure out targets of polymorphic calls with known decl.
2745         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2746         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2747         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2748         (get_polymorphic_call_info): ... here.
2749         (get_polymorphic_call_info_from_invariant): New function.
2751 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2753         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2754         lookup via vtable pointer; check for type consistency
2755         and turn inconsitent facts into UNREACHABLE.
2756         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2757         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2758         type inconsistent querries; return UNREACHABLE instead.
2760 2014-02-03  Richard Henderson  <rth@twiddle.net>
2762         PR tree-opt/59924
2763         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2764         already processed this node.
2765         (normalize_one_pred_1): Pass along mark_set.
2766         (normalize_one_pred): Create and destroy a pointer_set_t.
2767         (normalize_one_pred_chain): Likewise.
2769 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
2771         PR gcov-profile/58602
2772         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2774 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
2776         PR ipa/59831
2777         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2778         -fno-devirtualize; try to devirtualize by the knowledge of
2779         virtual table pointer given by aggregate propagation.
2780         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2781         (ipa_print_node_jump_functions): Dump also offset that
2782         is relevant for polymorphic calls.
2783         (determine_known_aggregate_parts): Add arg_type parameter; use it
2784         instead of determining the type from pointer type.
2785         (ipa_compute_jump_functions_for_edge): Update call of
2786         determine_known_aggregate_parts.
2787         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2788         (gimple_get_virt_method_for_binfo): ... here; simplify using
2789         vtable_pointer_value_to_vtable.
2790         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2791         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2792         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2793         (vtable_pointer_value_to_vtable): Break out from ...; handle also
2794         POINTER_PLUS_EXPR.
2795         (vtable_pointer_value_to_binfo): ... here.
2796         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2798 2014-02-03  Teresa Johnson  <tejohnson@google.com>
2800         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2801         redef of outer loop index variable.
2803 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
2805         PR c++/53017
2806         PR c++/59211
2807         * doc/extend.texi (Function Attributes): Typo.
2809 2014-02-03  Cong Hou  <congh@google.com>
2811         PR tree-optimization/60000
2812         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2813         if the vectorized statement is a store.  A store statement can only
2814         appear at the end of pattern statements.
2816 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2818         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2819         (ix86_option_override_internal): Default long double to 64-bit for
2820         32-bit Bionic and to 128-bit for 64-bit Bionic.
2822         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2823         TARGET_LONG_DOUBLE_128 is true.
2824         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2826         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2827         (mlong-double-64): Negate -mlong-double-128.
2828         (mlong-double-128): New option.
2830         * config/i386/i386-c.c (ix86_target_macros): Define
2831         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2833         * doc/invoke.texi: Document -mlong-double-128.
2835 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
2837         PR rtl-optimization/60024
2838         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2840 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
2842         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2844 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2846         PR rtl-optimization/57662
2847         * sel-sched.c (code_motion_path_driver): Do not mark already not
2848         existing blocks in the visiting bitmap.
2850 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
2852         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2853         on the insn being emitted.
2855 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
2856             Will Deacon  <will.deacon@arm.com>
2858         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2860 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2862         * config/arm/arm-tables.opt: Regenerate.
2864 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2866         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2867         for vector types other than V16QImode.
2868         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2869         define_expand, and call altivec_expand_vec_perm_le when producing
2870         code with little endian element order.
2871         (*altivec_vperm_<mode>_internal): New insn having previous
2872         behavior of altivec_vperm_<mode>.
2873         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2874         altivec_expand_vec_perm_le when producing code with little endian
2875         element order.
2876         (*altivec_vperm_<mode>_uns_internal): New insn having previous
2877         behavior of altivec_vperm_<mode>_uns.
2879 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2881         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2882         (altivec_vsumsws): Add handling for -maltivec=be with a little
2883         endian target.
2884         (altivec_vsumsws_direct): New.
2885         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2886         gen_altivec_vsumsws.
2888 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2890         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2891         vtable_pointer_value_to_binfo): New functions.
2892         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2893         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2895 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
2897         * config/nios2/nios2.md (load_got_register): Initialize GOT
2898         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2899         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2901 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2903         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2904         preserverd by passthrough, do not propagate the type.
2906 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2908         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2909         (mips_atomic_assign_expand_fenv): New function.
2910         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2912 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2914         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2915         (__builtin_mips_set_fcsr): Likewise.
2916         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2917         MIPS_USI_FTYPE_VOID.
2918         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2919         (mips16_expand_set_fcsr): Likewise.
2920         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2921         (mips16_set_fcsr_stub): Likewise.
2922         (mips16_get_fcsr_one_only_stub): New class.
2923         (mips16_set_fcsr_one_only_stub): Likewise.
2924         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2925         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2926         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2927         (hard_float): New availability predicate.
2928         (mips_builtins): Add get_fcsr and set_fcsr.
2929         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2930         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2931         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2932         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2933         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2934         patterns.
2936 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2938         * config/mips/mips.c (mips_one_only_stub): New class.
2939         (mips_need_mips16_rdhwr_p): Replace with...
2940         (mips16_rdhwr_stub): ...this new variable.
2941         (mips16_stub_call_address): New function.
2942         (mips16_rdhwr_one_only_stub): New class.
2943         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2944         (mips_output_mips16_rdhwr): Delete.
2945         (mips_finish_stub): New function.
2946         (mips_code_end): Use it to handle rdhwr stubs.
2948 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2950         PR target/60017
2951         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2952         when calculating size of integer atomic types.
2954 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2956         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2958 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2960         PR tree-optimization/60003
2961         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2962         * profile.c (branch_prob): Use gimple_call_builtin_p
2963         to check for BUILT_IN_SETJMP_RECEIVER.
2964         * tree-inline.c (copy_bb): Call notice_special_calls.
2966 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2968         PR bootstrap/59985
2969         * lra-constraints.c (process_alt_operands): Update reload_sum only
2970         on the first pass.
2972 2014-01-31  Richard Henderson  <rth@redhat.com>
2974         PR middle-end/60004
2975         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2976         until after else_eh is processed.
2978 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2980         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2981         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2982         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2983         in smmintrin.h, remove them.
2984         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2985         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2986         * config/i386/i386.md (ROUND_SAE): Fix value.
2987         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2988         (const48_operand): New.
2989         * config/i386/subst.md (round), (round_expand): Use
2990         const_4_or_8_to_11_operand.
2991         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2993 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2995         * config/i386/constraints.md (Yk): Swap meaning with k.
2996         * config/i386/i386.md (movhi_internal): Change Yk to k.
2997         (movqi_internal): Ditto.
2998         (*k<logic><mode>): Ditto.
2999         (*andhi_1): Ditto.
3000         (*andqi_1): Ditto.
3001         (kandn<mode>): Ditto.
3002         (*<code>hi_1): Ditto.
3003         (*<code>qi_1): Ditto.
3004         (kxnor<mode>): Ditto.
3005         (kortestzhi): Ditto.
3006         (kortestchi): Ditto.
3007         (kunpckhi): Ditto.
3008         (*one_cmplhi2_1): Ditto.
3009         (*one_cmplqi2_1): Ditto.
3010         * config/i386/sse.md (): Change k to Yk.
3011         (avx512f_load<mode>_mask): Ditto.
3012         (avx512f_blendm<mode>): Ditto.
3013         (avx512f_store<mode>_mask): Ditto.
3014         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
3015         (avx512f_storedqu<mode>_mask): Ditto.
3016         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3017         Ditto.
3018         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
3019         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
3020         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
3021         (avx512f_maskcmp<mode>3): Ditto.
3022         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
3023         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
3024         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
3025         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
3026         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
3027         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
3028         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
3029         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
3030         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
3031         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
3032         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
3033         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
3034         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
3035         (vec_extract_lo_<mode>_maskm): Ditto.
3036         (vec_extract_hi_<mode>_maskm): Ditto.
3037         (avx512f_vternlog<mode>_mask): Ditto.
3038         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
3039         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
3040         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
3041         (avx512f_<code>v8div16qi2_mask): Ditto.
3042         (avx512f_<code>v8div16qi2_mask_store): Ditto.
3043         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
3044         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
3045         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
3046         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
3047         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
3048         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3049         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
3050         (*avx512pf_scatterpf<mode>df_mask): Ditto.
3051         (avx512cd_maskb_vec_dupv8di): Ditto.
3052         (avx512cd_maskw_vec_dupv16si): Ditto.
3053         (avx512f_vpermi2var<mode>3_maskz): Ditto.
3054         (avx512f_vpermi2var<mode>3_mask): Ditto.
3055         (avx512f_vpermi2var<mode>3_mask): Ditto.
3056         (avx512f_vpermt2var<mode>3_maskz): Ditto.
3057         (*avx512f_gathersi<mode>): Ditto.
3058         (*avx512f_gathersi<mode>_2): Ditto.
3059         (*avx512f_gatherdi<mode>): Ditto.
3060         (*avx512f_gatherdi<mode>_2): Ditto.
3061         (*avx512f_scattersi<mode>): Ditto.
3062         (*avx512f_scatterdi<mode>): Ditto.
3063         (avx512f_compress<mode>_mask): Ditto.
3064         (avx512f_compressstore<mode>_mask): Ditto.
3065         (avx512f_expand<mode>_mask): Ditto.
3066         * config/i386/subst.md (mask): Change k to Yk.
3067         (mask_scalar_merge): Ditto.
3068         (sd): Ditto.
3070 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
3072         * doc/extend.texi (Vector Extensions): Document ?: in C++.
3074 2014-01-31  Richard Biener  <rguenther@suse.de>
3076         PR middle-end/59990
3077         * builtins.c (fold_builtin_memory_op): Make sure to not
3078         use a floating-point mode or a boolean or enumeral type for
3079         the copy operation.
3081 2014-01-30  DJ Delorie  <dj@redhat.com>
3083         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
3084         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
3085         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
3086         whenever main() has an epilogue.
3088 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3090         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
3091         unused variable "field".
3092         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
3093         (vsx_mergeh_<mode>): Likewise.
3094         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
3095         (altivec_vmrghh): Likewise.
3096         (altivec_vmrghw): Likewise.
3097         (altivec_vmrglb): Likewise.
3098         (altivec_vmrglh): Likewise.
3099         (altivec_vmrglw): Likewise.
3100         (altivec_vspltb): Add missing uses.
3101         (altivec_vsplth): Likewise.
3102         (altivec_vspltw): Likewise.
3103         (altivec_vspltsf): Likewise.
3105 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3107         PR target/59923
3108         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
3109         frame related instructions.
3111 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
3113         PR rtl-optimization/59959
3114         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
3115         any reload of register whose subreg is invalid.
3117 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
3119         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
3120         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
3121         Add missing return type - void.
3123 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3125         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3126         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3127         remove element index adjustment for endian (now handled in vsx.md
3128         and altivec.md).
3129         (altivec_expand_vec_perm_const): Use
3130         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3131         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3132         (vsx_xxspltw_<mode>): Adjust element index for little endian.
3133         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3134         define_expand and a new define_insn *altivec_vspltb_internal;
3135         adjust for -maltivec=be on a little endian target.
3136         (altivec_vspltb_direct): New.
3137         (altivec_vsplth): Divide into a define_expand and a new
3138         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3139         little endian target.
3140         (altivec_vsplth_direct): New.
3141         (altivec_vspltw): Divide into a define_expand and a new
3142         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3143         little endian target.
3144         (altivec_vspltw_direct): New.
3145         (altivec_vspltsf): Divide into a define_expand and a new
3146         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3147         a little endian target.
3149 2014-01-30  Richard Biener  <rguenther@suse.de>
3151         PR tree-optimization/59993
3152         * tree-ssa-forwprop.c (associate_pointerplus): Check we
3153         can propagate form the earlier stmt and avoid the transform
3154         when the intermediate result is needed.
3156 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
3158         * README.Portability: Fix typo.
3160 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
3162         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3163         comparison_operator with ordered_comparison_operator.
3165 2014-01-30  Nick Clifton  <nickc@redhat.com>
3167         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3168         Rename to mn10300_store_multiple_regs.
3169         * config/mn10300/mn10300.c: Likewise.
3170         * config/mn10300/mn10300.md (store_movm): Fix typo: call
3171         store_multiple_regs.
3172         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3173         Call mn10300_store_multiple_regs.
3175 2014-01-30  Nick Clifton  <nickc@redhat.com>
3176             DJ Delorie  <dj@redhat.com>
3178         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3179         %fp 2 to keep registers after it properly word-aligned.
3180         (rl78_alloc_physical_registers_umul): Handle the case where both
3181         input operands are the same.
3183 2014-01-30  Richard Biener  <rguenther@suse.de>
3185         PR tree-optimization/59903
3186         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3187         check properly.
3189 2014-01-30  Jason Merrill  <jason@redhat.com>
3191         PR c++/59633
3192         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3194         PR c++/59645
3195         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3197 2014-01-30  Richard Biener  <rguenther@suse.de>
3199         PR tree-optimization/59951
3200         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3202 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
3204         PR target/59784
3205         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3206         SFmode to DFmode case.
3208 2014-01-29  DJ Delorie  <dj@redhat.com>
3210         * config/msp430/msp430.opt (-minrt): New.
3211         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3212         if -minrt given.
3213         (ENDFILE_SPEC): Likewise.
3215 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
3217         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3218         (estimate_function_body_sizes): Use it.
3220 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
3222         PR c++/58561
3223         * dwarf2out.c (is_cxx_auto): New.
3224         (is_base_type): Use it.
3225         (gen_type_die_with_usage): Likewise.
3227 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3229         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
3230         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3231         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3232         -maltivec=be with LE targets.
3233         (vsx_mergeh_<mode>): Likewise.
3234         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3235         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3236         (altivec_vmrghb): Replace with define_expand and new
3237         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3238         (altivec_vmrghb_direct): New define_insn.
3239         (altivec_vmrghh): Replace with define_expand and new
3240         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3241         (altivec_vmrghh_direct): New define_insn.
3242         (altivec_vmrghw): Replace with define_expand and new
3243         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3244         (altivec_vmrghw_direct): New define_insn.
3245         (*altivec_vmrghsf): Adjust for endianness.
3246         (altivec_vmrglb): Replace with define_expand and new
3247         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3248         (altivec_vmrglb_direct): New define_insn.
3249         (altivec_vmrglh): Replace with define_expand and new
3250         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3251         (altivec_vmrglh_direct): New define_insn.
3252         (altivec_vmrglw): Replace with define_expand and new
3253         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3254         (altivec_vmrglw_direct): New define_insn.
3255         (*altivec_vmrglsf): Adjust for endianness.
3256         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3257         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3258         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3259         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3260         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3261         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3262         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3263         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3265 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3267         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3268         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3269         whitespace.
3271 2014-01-29  Richard Biener  <rguenther@suse.de>
3273         PR tree-optimization/58742
3274         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3275         associate_pointerplus_align.
3276         (associate_pointerplus_diff): New function.
3277         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
3278         and associate_pointerplus_diff.
3280 2014-01-29  Richard Biener  <rguenther@suse.de>
3282         * lto-streamer.h (LTO_major_version): Bump to 3.
3283         (LTO_minor_version): Reset to 0.
3285 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
3287         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3288         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3289         (arm_file_start): Generate correct asm header for armv7ve.
3290         * config/arm/bpabi.h: Add multilib support for armv7ve.
3291         * config/arm/driver-arm.c: Change the architectures of cortex-a7
3292         and cortex-a15 to armv7ve.
3293         * config/arm/t-aprofile: Add multilib support for armv7ve.
3294         * doc/invoke.texi: Document -march=armv7ve.
3296 2014-01-29  Richard Biener  <rguenther@suse.de>
3298         PR tree-optimization/58742
3299         * tree-ssa-forwprop.c (associate_plusminus): Return true
3300         if we changed sth, defer EH cleanup to ...
3301         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
3302         (simplify_mult): New function.
3304 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3306         PR middle-end/59917
3307         PR tree-optimization/59920
3308         * tree.c (build_common_builtin_nodes): Remove
3309         __builtin_setjmp_dispatcher initialization.
3310         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3311         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3312         instead of gsi_after_labels + manually skipping debug stmts.
3313         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3314         ignore bbs with IFN_ABNORMAL_DISPATCHER.
3315         * tree-inline.c (copy_edges_for_bb): Remove
3316         can_make_abnormal_goto argument, instead add abnormal_goto_dest
3317         argument.  Ignore computed_goto_p stmts.  Don't call
3318         make_abnormal_goto_edges.  If a call might need abnormal edges
3319         for non-local gotos, see if it already has an edge to
3320         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3321         with true argument, don't do anything then, otherwise add
3322         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3323         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3324         caller.
3325         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3326         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3327         (lower_stmt): Don't set data->calls_builtin_setjmp.
3328         (lower_builtin_setjmp): Adjust comment.
3329         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3330         * tree-cfg.c (found_computed_goto): Remove.
3331         (factor_computed_gotos): Remove.
3332         (make_goto_expr_edges): Return bool, true for computed gotos.
3333         Don't call make_abnormal_goto_edges.
3334         (build_gimple_cfg): Don't set found_computed_goto, don't call
3335         factor_computed_gotos.
3336         (computed_goto_p): No longer static.
3337         (make_blocks): Don't set found_computed_goto.
3338         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3339         (make_edges): If make_goto_expr_edges returns true, push bb
3340         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3341         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3342         vector.  Record mapping between bbs and OpenMP regions if there
3343         are any, adjust make_gimple_omp_edges caller.  Call
3344         handle_abnormal_edges.
3345         (make_abnormal_goto_edges): Remove.
3346         * tree-cfg.h (make_abnormal_goto_edges): Remove.
3347         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3348         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3349         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3350         * internal-fn.def (ABNORMAL_DISPATCHER): New.
3351         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3352         filling *region also set *region_idx to (*region)->entry->index.
3354         PR other/58712
3355         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3356         For REGs set ORIGINAL_REGNO.
3358 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
3360         * doc/md.texi: Mention that a target shouldn't implement
3361         vec_widen_(s|u)mul_even/odd pair if it is less efficient
3362         than hi/lo pair.
3364 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
3366         PR tree-optimization/59594
3367         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3368         a copy of the datarefs vector rather than the vector itself.
3370 2014-01-28  Jason Merrill  <jason@redhat.com>
3372         PR c++/53756
3373         * dwarf2out.c (auto_die): New static.
3374         (gen_type_die_with_usage): Handle C++1y 'auto'.
3375         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3376         on definition.
3378 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
3380         PR target/59672
3381         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3382         (SPEC_X32): Likewise.
3383         (SPEC_64): Likewise.
3384         * config/i386/i386.c (ix86_option_override_internal): Turn off
3385         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3386         for TARGET_16BIT.
3387         (x86_file_start): Output .code16gcc for TARGET_16BIT.
3388         * config/i386/i386.h (TARGET_16BIT): New macro.
3389         (TARGET_16BIT_P): Likewise.
3390         * config/i386/i386.opt: Add m16.
3391         * doc/invoke.texi: Document -m16.
3393 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
3395         PR preprocessor/59935
3396         * input.c (location_get_source_line): Bail out on when line number
3397         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3399 2014-01-28  Richard Biener  <rguenther@suse.de>
3401         PR tree-optimization/58742
3402         * tree-ssa-forwprop.c (associate_plusminus): Handle
3403         pointer subtraction of the form (T)(P + A) - (T)P.
3405 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3407         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3408         at const_int_cost.
3410 2014-01-28  Richard Biener  <rguenther@suse.de>
3412         Revert
3413         2014-01-28  Richard Biener  <rguenther@suse.de>
3415         PR rtl-optimization/45364
3416         PR rtl-optimization/59890
3417         * var-tracking.c (local_get_addr_clear_given_value): Handle
3418         already cleared slot.
3419         (val_reset): Handle not allocated local_get_addr_cache.
3420         (vt_find_locations): Use post-order on the inverted CFG.
3422 2014-01-28  Richard Biener  <rguenther@suse.de>
3424         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3426 2014-01-28  Richard Biener  <rguenther@suse.de>
3428         PR rtl-optimization/45364
3429         PR rtl-optimization/59890
3430         * var-tracking.c (local_get_addr_clear_given_value): Handle
3431         already cleared slot.
3432         (val_reset): Handle not allocated local_get_addr_cache.
3433         (vt_find_locations): Use post-order on the inverted CFG.
3435 2014-01-28  Alan Modra  <amodra@gmail.com>
3437         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3438         * configure.ac <recursive call for build != host>: Define
3439         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3440         and LD_FOR_BUILD too.
3441         * configure: Regenerate.
3443 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
3445         * config/i386/i386.c (get_builtin_code_for_version): Separate
3446         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3447         Broadwell from Haswell.
3449 2014-01-27  Steve Ellcey  <sellcey@mips.com>
3451         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3452         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3453         * config/mips/mips.c (mips_option_override): Change setting
3454         of TARGET_DSP.
3455         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3456         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3457         Change from Mask to Var.
3459 2014-01-27  Jeff Law  <law@redhat.com>
3461         * ipa-inline.c (inline_small_functions): Fix typo.
3463 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
3465         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3466         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3467         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3468         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3469         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3470         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3471         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3472         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3473         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3474         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3475         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3476         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3477         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3478         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3479         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3480         (_mm512_storeu_epi64): Ditto.
3481         (_mm512_cmpge_epi32_mask): Ditto.
3482         (_mm512_cmpge_epu32_mask): Ditto.
3483         (_mm512_cmpge_epi64_mask): Ditto.
3484         (_mm512_cmpge_epu64_mask): Ditto.
3485         (_mm512_cmple_epi32_mask): Ditto.
3486         (_mm512_cmple_epu32_mask): Ditto.
3487         (_mm512_cmple_epi64_mask): Ditto.
3488         (_mm512_cmple_epu64_mask): Ditto.
3489         (_mm512_cmplt_epi32_mask): Ditto.
3490         (_mm512_cmplt_epu32_mask): Ditto.
3491         (_mm512_cmplt_epi64_mask): Ditto.
3492         (_mm512_cmplt_epu64_mask): Ditto.
3493         (_mm512_cmpneq_epi32_mask): Ditto.
3494         (_mm512_cmpneq_epu32_mask): Ditto.
3495         (_mm512_cmpneq_epi64_mask): Ditto.
3496         (_mm512_cmpneq_epu64_mask): Ditto.
3497         (_mm512_expand_pd): Ditto.
3498         (_mm512_expand_ps): Ditto.
3499         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3500         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3501         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3502         * config/i386/i386.c (ix86_builtins): Add
3503         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3504         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3505         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3506         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3507         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3508         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3509         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3510         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3511         IX86_BUILTIN_PMOVUSQW512_MEM.
3512         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3513         __builtin_ia32_pmovsqd512mem_mask,
3514         __builtin_ia32_pmovqd512mem_mask,
3515         __builtin_ia32_pmovusqw512mem_mask,
3516         __builtin_ia32_pmovsqw512mem_mask,
3517         __builtin_ia32_pmovqw512mem_mask,
3518         __builtin_ia32_pmovusdw512mem_mask,
3519         __builtin_ia32_pmovsdw512mem_mask,
3520         __builtin_ia32_pmovdw512mem_mask,
3521         __builtin_ia32_pmovqb512mem_mask,
3522         __builtin_ia32_pmovusqb512mem_mask,
3523         __builtin_ia32_pmovsqb512mem_mask,
3524         __builtin_ia32_pmovusdb512mem_mask,
3525         __builtin_ia32_pmovsdb512mem_mask,
3526         __builtin_ia32_pmovdb512mem_mask.
3527         (bdesc_args): Add __builtin_ia32_expanddf512,
3528         __builtin_ia32_expandsf512.
3529         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3530         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3531         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3532         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3533         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3534         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3535         (avx512f_<code>v8div16qi2_mask_store): This.
3536         (avx512f_expand<mode>): New.
3538 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3540         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3541         New.
3542         (_mm512_mask_prefetch_i64gather_pd): Ditto.
3543         (_mm512_prefetch_i32scatter_pd): Ditto.
3544         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3545         (_mm512_prefetch_i64scatter_pd): Ditto.
3546         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3547         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3548         (_mm512_mask_prefetch_i64gather_ps): Ditto.
3549         (_mm512_prefetch_i32scatter_ps): Ditto.
3550         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3551         (_mm512_prefetch_i64scatter_ps): Ditto.
3552         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3553         * config/i386/i386-builtin-types.def: Define
3554         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3555         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3556         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3557         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3558         IX86_BUILTIN_SCATTERPFQPD.
3559         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3560         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3561         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3562         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3563         __builtin_ia32_scatterpfqps.
3564         (ix86_expand_builtin): Expand new built-ins.
3565         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3566         fix memory access data type.
3567         (*avx512pf_gatherpf<mode>_mask): Ditto.
3568         (*avx512pf_gatherpf<mode>): Ditto.
3569         (avx512pf_scatterpf<mode>): Ditto.
3570         (*avx512pf_scatterpf<mode>_mask): Ditto.
3571         (*avx512pf_scatterpf<mode>): Ditto.
3572         (GATHER_SCATTER_SF_MEM_MODE): New.
3573         (avx512pf_gatherpf<mode>df): Ditto.
3574         (*avx512pf_gatherpf<mode>df_mask): Ditto.
3575         (*avx512pf_scatterpf<mode>df): Ditto.
3577 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
3579         PR bootstrap/59934
3580         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3581         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3582         reached.
3584 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3586         * common/config/arm/arm-common.c
3587         (arm_rewrite_mcpu): Handle multiple names.
3588         * config/arm/arm.h
3589         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3591 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
3593         * gimple-builder.h (create_gimple_tmp): Delete.
3595 2014-01-27  Christian Bruel  <christian.bruel@st.com>
3597         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3598         words comparisons.
3600 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
3602         * config/pa/pa.md (call): Generate indirect long calls to non-local
3603         functions when outputing 32-bit code.
3604         (call_value): Likewise except for special call to buggy powf function.
3606         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3607         portable runtime and PIC indirect calls.
3608         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3609         and PIC call sequences.  Use ldo instead of blr to set return register
3610         in PIC call sequence.
3612 2014-01-25  Walter Lee  <walt@tilera.com>
3614         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3615         avoid clobbering a live register.
3617 2014-01-25  Walter Lee  <walt@tilera.com>
3619         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3620         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3621         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3622         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3624 2014-01-25  Walter Lee  <walt@tilera.com>
3626         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3627         arguments on even registers.
3628         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3629         STACK_BOUNDARY.
3630         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3631         (BIGGEST_ALIGNMENT): Ditto.
3632         (BIGGEST_FIELD_ALIGNMENT): Ditto.
3634 2014-01-25  Walter Lee  <walt@tilera.com>
3636         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3637         insns before bundling.
3638         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3640 2014-01-25  Walter Lee  <walt@tilera.com>
3642         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3643         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3644         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3646 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3648         * config/mips/constraints.md (kl): Delete.
3649         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3650         define expands, using...
3651         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3652         instructions for MIPS16.
3653         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3654         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3656 2014-01-25  Walter Lee  <walt@tilera.com>
3658         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3659         (clzdi2): Ditto.
3660         (ffsdi2): Ditto.
3662 2014-01-25  Walter Lee  <walt@tilera.com>
3664         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3665         (TARGET_EXPAND_TO_RTL_HOOK): Define.
3667 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
3669         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3670         Handle XOR.
3672 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
3674         * print-rtl.c (in_call_function_usage): New var.
3675         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3676         EXPR_LIST mode as mode and not as reg note name.
3678         PR middle-end/59561
3679         * cfgloopmanip.c (copy_loop_info): If
3680         loop->warned_aggressive_loop_optimizations, make sure
3681         the flag is set in target loop too.
3683 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
3685         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3686         flag_cilkplus.
3687         * builtins.def: Likewise.
3688         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3689         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3690         * ira.c (ira_setup_eliminable_regset): Likewise.
3691         * omp-low.c (gate_expand_omp): Likewise.
3692         (execute_lower_omp): Likewise.
3693         (diagnose_sb_0): Likewise.
3694         (gate_diagnose_omp_blocks): Likewise.
3695         (simd_clone_clauses_extract): Likewise.
3696         (gate): Likewise.
3698 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3700         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3701         correction for little endian...
3702         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3703         here.
3705 2014-01-24  Jeff Law  <law@redhat.com>
3707         PR tree-optimization/59919
3708         * tree-vrp.c (find_assert_locations_1): Do not register asserts
3709         for non-returning calls.
3711 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
3713         * common/config/aarch64/aarch64-common.c
3714         (aarch64_rewrite_mcpu): Handle multiple names.
3715         * config/aarch64/aarch64.h
3716         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3718 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
3720         * input.c (add_file_to_cache_tab): Handle the case where fopen
3721         returns NULL.
3723 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
3725         PR target/59929
3726         * config/i386/i386.md (pushsf splitter): Get stack adjustment
3727         from push operand if code of push isn't PRE_DEC.
3729 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
3731         PR target/59909
3732         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3733         -mquad-memory-atomic.  Update -mquad-memory documentation to say
3734         it is only used for non-atomic loads/stores.
3736         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3737         -mquad-memory or -mquad-memory-atomic switches.
3739         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3740         -mquad-memory-atomic to ISA 2.07 support.
3742         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3743         to separate support of normal quad word memory operations (ldq, stq)
3744         from the atomic quad word memory operations.
3746         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3747         support to separate non-atomic quad word operations from atomic
3748         quad word operations.  Disable non-atomic quad word operations in
3749         little endian mode so that we don't have to swap words after the
3750         load and before the store.
3751         (quad_load_store_p): Add comment about atomic quad word support.
3752         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3753         options printed with -mdebug=reg.
3755         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3756         -mquad-memory-atomic as the test for whether we have quad word
3757         atomic instructions.
3758         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3759         or -mp8-vector are used, allow byte/half-word atomic operations.
3761         * config/rs6000/sync.md (load_lockedti): Insure that the address
3762         is a proper indexed or indirect address for the lqarx instruction.
3763         On little endian systems, swap the hi/lo registers after the lqarx
3764         instruction.
3765         (load_lockedpti): Use indexed_or_indirect_operand predicate to
3766         insure the address is valid for the lqarx instruction.
3767         (store_conditionalti): Insure that the address is a proper indexed
3768         or indirect address for the stqcrx. instruction.  On little endian
3769         systems, swap the hi/lo registers before doing the stqcrx.
3770         instruction.
3771         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3772         insure the address is valid for the stqcrx. instruction.
3774         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3775         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3776         type of quad memory support is available.
3778 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
3780         PR regression/59915
3781         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3782         there is a danger of looping.
3784 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
3786         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3787         force flag_ira_loop_pressure if set via command line.
3789 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3791         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3792         (ashr_simd): New builtin handling DI mode.
3793         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3794         (aarch64_sshr_simddi): New match pattern.
3795         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3796         (vshrd_n_s64): Likewise.
3797         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3799 2014-01-23  Nick Clifton  <nickc@redhat.com>
3801         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3802         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3803         favour of mcu specific scripts.
3804         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3805         430x multilibs.
3807 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3808             Alex Velenko  <Alex.Velenko@arm.com>
3810         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3811         (vaddv_s16): Likewise.
3812         (vaddv_s32): Likewise.
3813         (vaddv_u8): Likewise.
3814         (vaddv_u16): Likewise.
3815         (vaddv_u32): Likewise.
3816         (vaddvq_s8): Likewise.
3817         (vaddvq_s16): Likewise.
3818         (vaddvq_s32): Likewise.
3819         (vaddvq_s64): Likewise.
3820         (vaddvq_u8): Likewise.
3821         (vaddvq_u16): Likewise.
3822         (vaddvq_u32): Likewise.
3823         (vaddvq_u64): Likewise.
3824         (vaddv_f32): Likewise.
3825         (vaddvq_f32): Likewise.
3826         (vaddvq_f64): Likewise.
3827         (vmaxv_f32): Likewise.
3828         (vmaxv_s8): Likewise.
3829         (vmaxv_s16): Likewise.
3830         (vmaxv_s32): Likewise.
3831         (vmaxv_u8): Likewise.
3832         (vmaxv_u16): Likewise.
3833         (vmaxv_u32): Likewise.
3834         (vmaxvq_f32): Likewise.
3835         (vmaxvq_f64): Likewise.
3836         (vmaxvq_s8): Likewise.
3837         (vmaxvq_s16): Likewise.
3838         (vmaxvq_s32): Likewise.
3839         (vmaxvq_u8): Likewise.
3840         (vmaxvq_u16): Likewise.
3841         (vmaxvq_u32): Likewise.
3842         (vmaxnmv_f32): Likewise.
3843         (vmaxnmvq_f32): Likewise.
3844         (vmaxnmvq_f64): Likewise.
3845         (vminv_f32): Likewise.
3846         (vminv_s8): Likewise.
3847         (vminv_s16): Likewise.
3848         (vminv_s32): Likewise.
3849         (vminv_u8): Likewise.
3850         (vminv_u16): Likewise.
3851         (vminv_u32): Likewise.
3852         (vminvq_f32): Likewise.
3853         (vminvq_f64): Likewise.
3854         (vminvq_s8): Likewise.
3855         (vminvq_s16): Likewise.
3856         (vminvq_s32): Likewise.
3857         (vminvq_u8): Likewise.
3858         (vminvq_u16): Likewise.
3859         (vminvq_u32): Likewise.
3860         (vminnmv_f32): Likewise.
3861         (vminnmvq_f32): Likewise.
3862         (vminnmvq_f64): Likewise.
3864 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
3866         * config/aarch64/aarch64-simd.md
3867         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3868         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3869         (*aarch64_mul3_elt<mode>): Likewise.
3870         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3871         (*aarch64_mul3_elt_to_64v2df): Likewise.
3872         (*aarch64_mla_elt<mode>): Likewise.
3873         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3874         (*aarch64_mls_elt<mode>): Likewise.
3875         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3876         (*aarch64_fma4_elt<mode>): Likewise.
3877         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3878         (*aarch64_fma4_elt_to_64v2df): Likewise.
3879         (*aarch64_fnma4_elt<mode>): Likewise.
3880         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3881         (*aarch64_fnma4_elt_to_64v2df): Likewise.
3882         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3883         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3884         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3885         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3886         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3887         (aarch64_sqdmull_lane<mode>_internal): Likewise.
3888         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3890 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3892         * config/aarch64/aarch64-simd.md
3893         (aarch64_be_checked_get_lane<mode>): New define_expand.
3894         * config/aarch64/aarch64-simd-builtins.def
3895         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3896         New builtin definition.
3897         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3898         Use new safe be builtin.
3900 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3902         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3903         New define_insn.
3904         (aarch64_be_st1<mode>): Likewise.
3905         (aarch_ld1<VALL:mode>): Define_expand modified.
3906         (aarch_st1<VALL:mode>): Likewise.
3907         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3908         (UNSPEC_ST1): Likewise.
3910 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3912         * config/microblaze/microblaze.md: Add trap insn and attribute
3914 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3916         PR preprocessor/58580
3917         * input.h (location_get_source_line): Take an additional line_size
3918         parameter.
3919         (void diagnostics_file_cache_fini): Declare new function.
3920         * input.c (struct fcache): New type.
3921         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3922         New static constants.
3923         (diagnostic_file_cache_init, total_lines_num)
3924         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3925         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3926         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3927         (get_next_line, read_next_line, goto_next_line, read_line_num):
3928         New static function definitions.
3929         (diagnostic_file_cache_fini): New function.
3930         (location_get_source_line): Take an additional output line_len
3931         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3932         read_line_num.
3933         * diagnostic.c (diagnostic_finish): Call
3934         diagnostic_file_cache_fini.
3935         (adjust_line): Take an additional input parameter for the length
3936         of the line, rather than calculating it with strlen.
3937         (diagnostic_show_locus): Adjust the use of
3938         location_get_source_line and adjust_line with respect to their new
3939         signature.  While displaying a line now, do not stop at the first
3940         null byte.  Rather, display the zero byte as a space and keep
3941         going until we reach the size of the line.
3942         * Makefile.in: Add vec.o to OBJS-libcommon
3944 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3945             Ilya Tocar     <ilya.tocar@intel.com>
3947         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3948         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3949         (__builtin_ia32_kmov16): Ditto.
3950         * config/i386/i386.md (UNSPEC_KMOV): New.
3951         (kmovw): Ditto.
3953 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3955         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3956         (_mm512_storeu_si512): Ditto.
3958 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3960         PR target/52125
3961         * rtl.h (get_referenced_operands): Declare.
3962         * recog.c (get_referenced_operands): New function.
3963         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3964         operands have been referenced when recording LO_SUM references.
3966 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3968         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3970 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3972         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3973         Enable for generic and recent AMD targets.
3975 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3977         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3978         ARG_SIZE note when adjustment was eliminated.
3980 2014-01-22  Jeff Law  <law@redhat.com>
3982         PR tree-optimization/59597
3983         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3984         in file.  Accept new argument REGISTERING and use it to modify
3985         dump output appropriately.
3986         (register_jump_thread): Corresponding changes.
3987         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3988         thread paths involving joiner blocks.  Add code to dump cancelled
3989         jump threading paths.
3991 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3993         PR rtl-optimization/59477
3994         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3995         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3997 2014-01-22  Tom Tromey  <tromey@redhat.com>
3999         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
4000         PARAMS.
4001         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
4003 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4005         PR rtl-optimization/59896
4006         * lra-constraints.c (process_alt_operands): Check unused note for
4007         matched operands of insn with no output reloads.
4009 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
4011         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
4012         (mips_move_from_gpr_cost): Likewise.
4014 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
4016         PR rtl-optimization/59858
4017         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
4018         ira_class_hard_regs_num.
4019         (process_alt_operands): Increase reject for dying matched operand.
4021 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
4023         PR target/59003
4024         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
4025         smaller than size, perform several stores or loads and stores
4026         at dst + count - size to store or copy all of size bytes, rather
4027         than just last modesize bytes.
4029 2014-01-20  DJ Delorie  <dj@redhat.com>
4031         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
4032         that CLOBBERs are REGs before propogating their values.
4034 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
4036         PR middle-end/59789
4037         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
4038         (cgraph_inline_failed_type): New function.
4039         * cgraph.h (DEFCIFCODE): Add type.
4040         (cgraph_inline_failed_type_t): New enum.
4041         (cgraph_inline_failed_type): New prototype.
4042         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
4043         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
4044         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
4045         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
4046         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
4047         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
4048         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
4049         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
4050         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
4051         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
4052         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
4053         OPTIMIZATION_MISMATCH.
4054         * tree-inline.c (expand_call_inline): Emit errors during
4055         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
4057 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4059         PR target/59685
4060         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
4061         mode attribute in insn output.
4063 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
4065         * output.h (output_constant): Delete.
4066         * varasm.c (output_constant): Make private.
4068 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
4070         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
4072 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4074         PR middle-end/59860
4075         * tree.h (fold_builtin_strcat): New prototype.
4076         * builtins.c (fold_builtin_strcat): No longer static.  Add len
4077         argument, if non-NULL, don't call c_strlen.  Optimize
4078         directly into __builtin_memcpy instead of __builtin_strcpy.
4079         (fold_builtin_2): Adjust fold_builtin_strcat caller.
4080         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
4082 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
4084         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4085         for SImode_address_operand operands, having only a REG argument.
4087 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
4089         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
4090         loader name using mbig-endian.
4091         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
4093 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4095         * doc/invoke.texi (-march): Clarify documentation for AArch64.
4096         (-mtune): Likewise.
4097         (-mcpu): Likewise.
4099 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
4101         * config/aarch64/aarch64-protos.h
4102         (aarch64_cannot_change_mode_class_ptr): Declare.
4103         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
4104         aarch64_cannot_change_mode_class_ptr): New.
4105         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
4106         backend hook aarch64_cannot_change_mode_class.
4108 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
4110         * common/config/aarch64/aarch64-common.c
4111         (aarch64_handle_option): Don't handle any option order logic here.
4112         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
4113         selected_cpu, warn on architecture version mismatch.
4114         (aarch64_override_options): Fix parsing order for option strings.
4116 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4117             Iain Sandoe  <iain@codesourcery.com>
4119         PR bootstrap/59496
4120         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
4121         warning.  Amend comment to reflect current functionality.
4123 2014-01-20  Richard Biener  <rguenther@suse.de>
4125         PR middle-end/59860
4126         * builtins.c (fold_builtin_strcat): Remove case better handled
4127         by tree-ssa-strlen.c.
4129 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
4131         * config/aarch64/aarch64.opt
4132         (mcpu, march, mtune): Make case-insensitive.
4134 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
4136         PR target/59880
4137         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4138         if operands[1] is a REG or ZERO_EXTEND of a REG.
4140 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
4142         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4144 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
4146         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4147         long non-pic millicode calls.
4149 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4151         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4153 2014-01-19  Kito Cheng  <kito@0xlab.org>
4155         * builtins.c (expand_movstr): Check movstr expand done or fail.
4157 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4158             H.J. Lu  <hongjiu.lu@intel.com>
4160         PR target/59379
4161         * config/i386/i386.md (*lea<mode>): Zero-extend return register
4162         to DImode for zero-extended addresses.
4164 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
4166         PR rtl-optimization/57763
4167         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4168         on the new indirect jump_insn and increment LABEL_NUSES (label).
4170 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
4172         PR bootstrap/59580
4173         PR bootstrap/59583
4174         * config.gcc (x86_archs): New variable.
4175         (x86_64_archs): Likewise.
4176         (x86_cpus): Likewise.
4177         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4178         --with-arch/--with-cpu= options.
4179         Support --with-arch=/--with-cpu={nehalem,westmere,
4180         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4182 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4184         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4185         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
4187 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
4189         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4191 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
4193         PR target/58944
4194         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4195         clear cpp_get_options (parse_in)->warn_unused_macros for
4196         ix86_target_macros_internal with cpp_define.
4198 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
4200         * jump.c (delete_related_insns): Keep (use (insn))s.
4201         * reorg.c (redundant_insn): Check for barriers too.
4203 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4205         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4207 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
4209         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4210         call to $$dyncall when TARGET_LONG_CALLS is true.
4212 2014-01-17  Jeff Law  <law@redhat.com>
4214         * ree.c (combine_set_extension): Temporarily disable test for
4215         changing number of hard registers.
4217 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4219         PR middle-end/58125
4220         * ipa-inline-analysis.c (inline_free_summary):
4221         Do not free summary of aliases.
4223 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4225         PR middle-end/59706
4226         * gimplify.c (gimplify_expr): Use create_tmp_var
4227         instead of create_tmp_var_raw.  If cond doesn't have
4228         integral type, don't add the IFN_ANNOTATE builtin at all.
4230 2014-01-17  Martin Jambor  <mjambor@suse.cz>
4232         PR ipa/59736
4233         * ipa-cp.c (prev_edge_clone): New variable.
4234         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4235         Also resize prev_edge_clone vector.
4236         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4237         (ipcp_edge_removal_hook): New function.
4238         (ipcp_driver): Register ipcp_edge_removal_hook.
4240 2014-01-17  Andrew Pinski <apinski@cavium.com>
4241             Steve Ellcey  <sellcey@mips.com>
4243         PR target/59462
4244         * config/mips/mips.c (mips_print_operand): Check operand mode instead
4245         of operator mode.
4247 2014-01-17  Jeff Law  <law@redhat.com>
4249         PR middle-end/57904
4250         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4251         so that pass_ccp runs first.
4253 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4255         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4256         (ix86_adjust_cost): Use !TARGET_XXX.
4257         (do_reorder_for_imul): Likewise.
4258         (swap_top_of_ready_list): Likewise.
4259         (ix86_sched_reorder): Likewise.
4261 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
4263         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4264         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
4265         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
4266         (intel_memset): New.  Duplicate slm_memset.
4267         (intel_cost): New.  Duplicate slm_cost.
4268         (m_INTEL): New macro.
4269         (processor_target_table): Add "intel".
4270         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4271         with PROCESSOR_INTEL for "intel".
4272         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
4273         PROCESSOR_SILVERMONT.
4274         (ix86_issue_rate): Likewise.
4275         (ix86_adjust_cost): Likewise.
4276         (ia32_multipass_dfa_lookahead): Likewise.
4277         (swap_top_of_ready_list): Likewise.
4278         (ix86_sched_reorder): Likewise.
4279         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4280         instead of TARGET_OPT_AGU.
4281         * config/i386/i386.h (TARGET_INTEL): New.
4282         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4283         (processor_type): Add PROCESSOR_INTEL.
4284         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4285         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4287 2014-01-17  Marek Polacek  <polacek@redhat.com>
4289         PR c/58346
4290         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4291         size is zero.
4293 2014-01-17  Richard Biener  <rguenther@suse.de>
4295         PR tree-optimization/46590
4296         * opts.c (default_options_table): Add entries for
4297         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4298         all enabled at -O1 but not for -Og.
4299         * common.opt (fbranch-count-reg): Remove Init(1).
4300         (fmove-loop-invariants): Likewise.
4301         (ftree-pta): Likewise.
4303 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
4305         * config/i386/i386.c (ix86_data_alignment): For compatibility with
4306         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4307         decls to at least the GCC 4.8 used alignments.
4309         PR fortran/59440
4310         * tree-nested.c (convert_nonlocal_reference_stmt,
4311         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4312         of GIMPLE_BIND stmts, adjust associated decls.
4314 2014-01-17  Richard Biener  <rguenther@suse.de>
4316         PR tree-optimization/46590
4317         * vec.h (vec<>::bseach): New member function implementing
4318         binary search according to C89 bsearch.
4319         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4320         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4321         bitmap pointer again.  Make accesses_in_loop a flat array.
4322         (mem_ref_obstack): New global.
4323         (outermost_indep_loop): Adjust for mem_ref->stored changes.
4324         (mark_ref_stored): Likewise.
4325         (ref_indep_loop_p_2): Likewise.
4326         (set_ref_stored_in_loop): New helper function.
4327         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4328         (memref_free): Adjust.
4329         (record_mem_ref_loc): Simplify.
4330         (gather_mem_refs_stmt): Adjust.
4331         (sort_locs_in_loop_postorder_cmp): New function.
4332         (analyze_memory_references): Sort accesses_in_loop after
4333         loop postorder number.
4334         (find_ref_loc_in_loop_cmp): New function.
4335         (for_all_locs_in_loop): Find relevant cluster of locs in
4336         accesses_in_loop and iterate without recursion.
4337         (execute_sm): Avoid uninit warning.
4338         (struct ref_always_accessed): Simplify.
4339         (ref_always_accessed::operator ()): Likewise.
4340         (ref_always_accessed_p): Likewise.
4341         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4342         loop postorder numbers here.
4343         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4344         numbers.
4346 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
4348         PR c++/57945
4349         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4350         on decls for which assemble_alias has been called.
4352 2014-01-17  Nick Clifton  <nickc@redhat.com>
4354         * config/msp430/msp430.opt: (mcpu): New option.
4355         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4356         (msp430_option_override): Parse target_cpu.  If the MCU name
4357         matches a generic string, clear target_mcu.
4358         (msp430_attr): Allow numeric interrupt values up to 63.
4359         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4360         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4361         option.
4362         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4363         Add mcpu matches.
4364         * config/msp430/msp430.md (popm): Use %J rather than %I.
4365         (addsi3): Use msp430_nonimmediate_operand for operand 2.
4366         (addhi_cy_i): Use immediate_operand for operand 2.
4367         * doc/invoke.texi: Document -mcpu option.
4369 2014-01-17  Richard Biener  <rguenther@suse.de>
4371         PR rtl-optimization/38518
4372         * df.h (df_analyze_loop): Declare.
4373         * df-core.c: Include cfgloop.h.
4374         (df_analyze_1): Split out main part of df_analyze.
4375         (df_analyze): Adjust.
4376         (loop_inverted_post_order_compute): New function.
4377         (loop_post_order_compute): Likewise.
4378         (df_analyze_loop): New function avoiding whole-function
4379         postorder computes.
4380         * loop-invariant.c (find_defs): Use df_analyze_loop.
4381         (find_invariants): Adjust.
4382         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4384 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
4386         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4387         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4389 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
4391         * ipa-ref.c (ipa_remove_stmt_references): Fix references
4392         traversal when removing references.
4394 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
4396         PR ipa/59775
4397         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4399 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
4401         PR middle-end/56791
4402         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4403         pushing a reload for an autoinc when we had previously reloaded an
4404         inner part of the address.
4406 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
4408         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4409         field.
4410         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4411         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4412         when not giving up or versioning for alias only because of
4413         loop->safelen.
4414         (vect_analyze_data_ref_dependences): Set to true.
4415         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4416         is a GIMPLE_PHI.
4417         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4418         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4419         to the condition.
4421         PR middle-end/58344
4422         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4424         PR target/59839
4425         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4426         operand 0 predicate for gathers, use a new pseudo as subtarget.
4428 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4430         PR middle-end/59609
4431         * lra-constraints.c (process_alt_operands): Add printing debug info.
4432         Check absence of input/output reloads for matched operands too.
4434 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
4436         PR rtl-optimization/59835
4437         * ira.c (ira_init_register_move_cost): Increase cost for
4438         impossible modes.
4440 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
4442         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4444 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
4446         PR target/59780
4447         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4448         non-register objects.  Use gen_(high/low)part more consistently.
4449         Fix assertions.
4451 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
4453         PR target/59844
4454         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4455         endian support, remove tests for WORDS_BIG_ENDIAN.
4456         (p8_mfvsrd_3_<mode>): Likewise.
4457         (reload_gpr_from_vsx<mode>): Likewise.
4458         (reload_gpr_from_vsxsf): Likewise.
4459         (p8_mfvsrd_4_disf): Likewise.
4461 2014-01-16  Richard Biener  <rguenther@suse.de>
4463         PR rtl-optimization/46590
4464         * lcm.c (compute_antinout_edge): Use postorder iteration.
4465         (compute_laterin): Use inverted postorder iteration.
4467 2014-01-16  Nick Clifton  <nickc@redhat.com>
4469         PR middle-end/28865
4470         * varasm.c (output_constant): Return the number of bytes actually
4471         emitted.
4472         (output_constructor_array_range): Update the field size with the
4473         number of bytes emitted by output_constant.
4474         (output_constructor_regular_field): Likewise.  Also do not
4475         complain if the total number of bytes emitted is now greater
4476         than the expected fieldpos.
4477         * output.h (output_constant): Update prototype and descriptive comment.
4479 2014-01-16  Marek Polacek  <polacek@redhat.com>
4481         PR middle-end/59827
4482         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4483         it is error_mark_node.
4485 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
4487         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4488         VALID_AVX256_REG_OR_OI_MODE.
4490 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
4492         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4493         current procedure should be profiled.
4495 2014-01-15  Andrew Pinski  <apinski@cavium.com>
4497         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4498         of moving from/to the STACK_REG register class.
4500 2014-01-15  Richard Henderson  <rth@redhat.com>
4502         PR debug/54694
4503         * reginfo.c (global_regs_decl): Globalize.
4504         * rtl.h (global_regs_decl): Declare.
4505         * ira.c (do_reload): Diagnose frame_pointer_needed and it
4506         reserved via global_regs.
4508 2014-01-15  Teresa Johnson  <tejohnson@google.com>
4510         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4512 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
4514         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4515         and vmulosh rather than call gen_vec_widen_smult_*.
4516         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4517         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4518         (vec_widen_smult_even_v16qi): Likewise.
4519         (vec_widen_umult_even_v8hi): Likewise.
4520         (vec_widen_smult_even_v8hi): Likewise.
4521         (vec_widen_umult_odd_v16qi): Likewise.
4522         (vec_widen_smult_odd_v16qi): Likewise.
4523         (vec_widen_umult_odd_v8hi): Likewise.
4524         (vec_widen_smult_odd_v8hi): Likewise.
4525         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4526         vmuloub rather than call gen_vec_widen_umult_*.
4527         (vec_widen_umult_lo_v16qi): Likewise.
4528         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4529         vmulosb rather than call gen_vec_widen_smult_*.
4530         (vec_widen_smult_lo_v16qi): Likewise.
4531         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4532         rather than call gen_vec_widen_umult_*.
4533         (vec_widen_umult_lo_v8hi): Likewise.
4534         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4535         rather than call gen_vec_widen_smult_*.
4536         (vec_widen_smult_lo_v8hi): Likewise.
4538 2014-01-15  Jeff Law  <law@redhat.com>
4540         PR tree-optimization/59747
4541         * ree.c (find_and_remove_re): Properly handle case where a second
4542         eliminated extension requires widening a copy created for elimination
4543         of a prior extension.
4544         (combine_set_extension): Ensure that the number of hard regs needed
4545         for a destination register does not change when we widen it.
4547 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
4549         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4550         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4551         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4552         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4553         (avr-*-rtems*): Likewise.
4554         (bfin*-rtems*): Likewise.
4555         (moxie-*-rtems*): Likewise.
4556         (h8300-*-rtems*): Likewise.
4557         (i[34567]86-*-rtems*): Likewise.
4558         (lm32-*-rtems*): Likewise.
4559         (m32r-*-rtems*): Likewise.
4560         (m68k-*-rtems*): Likewise.
4561         (microblaze*-*-rtems*): Likewise.
4562         (mips*-*-rtems*): Likewise.
4563         (powerpc-*-rtems*): Likewise.
4564         (sh-*-rtems*): Likewise.
4565         (sparc-*-rtems*): Likewise.
4566         (sparc64-*-rtems*): Likewise.
4567         (v850-*-rtems*): Likewise.
4568         (m32c-*-rtems*): Likewise.
4570 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
4572         PR rtl-optimization/59511
4573         * ira.c (ira_init_register_move_cost): Use memory costs for some
4574         cases of register move cost calculations.
4575         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4576         instead of BB frequency.
4577         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4578         * lra-assigns.c (find_hard_regno_for): Ditto.
4580 2014-01-15  Richard Biener  <rguenther@suse.de>
4582         PR tree-optimization/59822
4583         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4584         (vectorizable_load): Use it to hoist defs of uses of invariant
4585         loads out of the loop.
4587 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
4588             Kugan Vivekanandarajah  <kuganv@linaro.org>
4590         PR target/59695
4591         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4592         truncation.
4594 2014-01-15  Richard Biener  <rguenther@suse.de>
4596         PR rtl-optimization/59802
4597         * lcm.c (compute_available): Use inverted postorder to seed
4598         the initial worklist.
4600 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4602         PR target/59803
4603         * config/s390/s390.c (s390_preferred_reload_class): Don't return
4604         ADDR_REGS for invalid symrefs in non-PIC code.
4606 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
4608         PR other/58712
4609         * builtins.c (determine_block_size): Initialize *probable_max_size
4610         even if len_rtx is CONST_INT.
4612 2014-01-14  Andrew Pinski  <apinski@cavium.com>
4614         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4615         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4616         (cortexa53_tunings): Likewise.
4617         (aarch64_sched_issue_rate): New function.
4618         (TARGET_SCHED_ISSUE_RATE): Define.
4620 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4622         * ira-costs.c (find_costs_and_classes): Add missed
4623         ira_init_register_move_cost_if_necessary.
4625 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
4627         PR target/59787
4628         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4630 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
4632         PR target/59794
4633         * config/i386/i386.c (type_natural_mode): Add a bool parameter
4634         to indicate if type is used for function return value.  Warn ABI
4635         change if the vector mode isn't available for function return value.
4636         (ix86_function_arg_advance): Pass false to type_natural_mode.
4637         (ix86_function_arg): Likewise.
4638         (ix86_gimplify_va_arg): Likewise.
4639         (function_arg_32): Don't warn ABI change.
4640         (ix86_function_value): Pass true to type_natural_mode.
4641         (ix86_return_in_memory): Likewise.
4642         (ix86_struct_value_rtx): Removed.
4643         (TARGET_STRUCT_VALUE_RTX): Likewise.
4645 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
4647         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4648         converting a conditional jump into a conditional return.
4650 2014-01-14  Richard Biener  <rguenther@suse.de>
4652         PR tree-optimization/58921
4653         PR tree-optimization/59006
4654         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4655         hoisting invariant stmts.
4656         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4657         invariant loads on the preheader edge if possible.
4659 2014-01-14  Joey Ye  <joey.ye@arm.com>
4661         * doc/plugin.texi (Building GCC plugins): Update to C++.
4663 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
4665         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4666         (_mm_rcp28_round_ss): Ditto.
4667         (_mm_rsqrt28_round_sd): Ditto.
4668         (_mm_rsqrt28_round_ss): Ditto.
4669         (_mm_rcp28_sd): Ditto.
4670         (_mm_rcp28_ss): Ditto.
4671         (_mm_rsqrt28_sd): Ditto.
4672         (_mm_rsqrt28_ss): Ditto.
4673         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4674         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4675         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4676         (IX86_BUILTIN_RCP28SD): Ditto.
4677         (IX86_BUILTIN_RCP28SS): Ditto.
4678         (IX86_BUILTIN_RSQRT28SD): Ditto.
4679         (IX86_BUILTIN_RSQRT28SS): Ditto.
4680         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4681         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4682         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4683         (ix86_expand_special_args_builtin): Expand new FTYPE.
4684         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4685         (srcp14<mode>): Make insn unary.
4686         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4687         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4688         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4689         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4690         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4691         Fix rounding: make it SAE only.
4692         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4693         Ditto.
4694         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4695         Ditto.
4696         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4697         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4698         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4699         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4700         (round_saeonly_mask_scalar_operand4): Ditto.
4701         (round_saeonly_mask_scalar_op3): Ditto.
4702         (round_saeonly_mask_scalar_op4): Ditto.
4704 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4706         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4707         Implement -maltivec=be for vec_insert and vec_extract.
4709 2014-01-10  DJ Delorie  <dj@redhat.com>
4711         * config/msp430/msp430.md (call_internal): Don't allow memory
4712         references with SP as the base register.
4713         (call_value_internal): Likewise.
4714         * config/msp430/constraints.md (Yc): New.  For memory references
4715         that don't use SP as a base register.
4717         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4718         "an integer without a # prefix"
4719         * config/msp430/msp430.md (epilogue_helper): Use it.
4721 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4723         PR target/59617
4724         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4725         AVX512F gather builtins.
4726         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4727         on gather decls with INTEGER_TYPE masktype.
4728         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4729         directly into the builtin rather than hoisting it before loop.
4731         PR tree-optimization/59387
4732         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4733         (scev_const_prop): If folded_casts and type has undefined overflow,
4734         use force_gimple_operand instead of force_gimple_operand_gsi and
4735         for each added stmt if it is assign with
4736         arith_code_with_undefined_signed_overflow, call
4737         rewrite_to_defined_overflow.
4738         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4739         gimple-fold.h instead.
4740         (arith_code_with_undefined_signed_overflow,
4741         rewrite_to_defined_overflow): Moved to ...
4742         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4743         rewrite_to_defined_overflow): ... here.  No longer static.
4744         Include gimplify-me.h.
4745         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4746         rewrite_to_defined_overflow): New prototypes.
4748 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4750         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4752 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
4754         * builtins.c (get_object_alignment_2): Minor tweak.
4755         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4757 2014-01-13  Christian Bruel  <christian.bruel@st.com>
4759         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4760         optimized non constant lengths.
4762 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
4764         PR libgomp/59194
4765         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4766         load as __atomic_load_N if possible.
4768 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4770         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4771         target parameter.
4772         (rs6000_expand_builtin): Adjust call.
4774 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
4776         PR target/58115
4777         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4778         * config/rs6000/rs6000.c: Include target-globals.h.
4779         (rs6000_set_current_function): Instead of doing target_reinit
4780         unconditionally, use save_target_globals_default_opts and
4781         restore_target_globals.
4783         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4784         FPSCR.
4785         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4786         (rs6000_expand_builtin): Handle mffs and mtfsf.
4787         (rs6000_init_builtins): Define mffs and mtfsf.
4788         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4789         (rs6000_mffs): New pattern.
4790         (rs6000_mtfsf): New pattern.
4792 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
4794         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4795         Start narrowing with START.  Apply candidate-use pair
4796         and check overall cost in narrowing.
4797         (iv_ca_prune): Pass new argument.
4799 2014-01-10  Jeff Law  <law@redhat.com>
4801         PR middle-end/59743
4802         * ree.c (combine_reaching_defs): Ensure the defining statement
4803         occurs before the extension when optimizing extensions with
4804         different source and destination hard registers.
4806 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4808         PR ipa/58585
4809         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4810         vtables into the type inheritance graph.
4812 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4814         PR rtl-optimization/59754
4815         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4816         modes in the REGNO != REGNO case.
4818 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4820         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4822 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4824         PR tree-optimization/59745
4825         * tree-predcom.c (tree_predictive_commoning_loop): Call
4826         free_affine_expand_cache if giving up because components is NULL.
4828         * target-globals.c (save_target_globals): Allocate < 4KB structs using
4829         GC in payload of target_globals struct instead of allocating them on
4830         the heap and the larger structs separately using GC.
4831         * target-globals.h (struct target_globals): Make regs, hard_regs,
4832         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4833         of GTY((skip)) and change type to void *.
4834         (reset_target_globals): Cast loads from those fields to corresponding
4835         types.
4837 2014-01-10  Steve Ellcey  <sellcey@mips.com>
4839         PR plugins/59335
4840         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4841         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4842         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4844 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
4846         PR target/59744
4847         * aarch64-modes.def (CC_Zmode): New flags mode.
4848         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4849         represents an equality.
4850         (aarch64_get_condition_code): Handle CC_Zmode.
4851         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4853 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
4855         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4856         extraction in good case.
4858 2014-01-10  Richard Biener  <rguenther@suse.de>
4860         PR tree-optimization/59374
4861         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4862         checking after SLP discovery.  Mark stmts not participating
4863         in any SLP instance properly.
4865 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4867         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4868         when handling a SET rtx.
4870 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4872         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4873         (cortex-a57): Likewise.
4874         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4876 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4878         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4879         non-iwmmxt builtins.
4881 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
4883         PR ipa/58252
4884         PR ipa/59226
4885         * ipa-devirt.c record_target_from_binfo): Take as argument
4886         stack of binfos and lookup matching one for virtual inheritance.
4887         (possible_polymorphic_call_targets_1): Update.
4889 2014-01-10  Huacai Chen  <chenhc@lemote.com>
4891         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4892         kernel strings for Loongson-2E/2F/3A.
4894 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
4896         PR middle-end/59670
4897         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4898         is_gimple_call before calling gimple_call_internal_p.
4900 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4902         * Makefile.in (TREE_FLOW_H): Remove.
4903         (TREE_SSA_H): Add file names from tree-flow.h.
4904         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4905         * tree.h: Remove tree-flow.h reference.
4906         * hash-table.h: Remove tree-flow.h reference.
4907         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4908         reference with tree-ssa-loop.h.
4910 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4912         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4913         default element-order behavior for -maltivec.
4914         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4915         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4916         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4917         when targeting big endian, at least for now.
4918         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4920 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4922         PR middle-end/47735
4923         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4924         var satisfies use_register_for_decl, just take into account type
4925         alignment, rather than decl alignment.
4927         PR tree-optimization/59622
4928         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4929         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4930         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4931         Don't devirtualize for inplace at all.  For targets.length () == 1,
4932         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4934 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4936         * config/i386/i386.md (cpu): Remove the unused btver1.
4938 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4940         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4942 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4944         PR target/58115
4945         * tree-core.h (struct target_globals): New forward declaration.
4946         (struct tree_target_option): Add globals field.
4947         * tree.h (TREE_TARGET_GLOBALS): Define.
4948         (prepare_target_option_nodes_for_pch): New prototype.
4949         * target-globals.h (struct target_globals): Define even if
4950         !SWITCHABLE_TARGET.
4951         * tree.c (prepare_target_option_node_for_pch,
4952         prepare_target_option_nodes_for_pch): New functions.
4953         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4954         * config/i386/i386.c: Include target-globals.h.
4955         (ix86_set_current_function): Instead of doing target_reinit
4956         unconditionally, use save_target_globals_default_opts and
4957         restore_target_globals.
4959 2014-01-09  Richard Biener  <rguenther@suse.de>
4961         PR tree-optimization/59715
4962         * tree-cfg.h (split_critical_edges): Declare.
4963         * tree-cfg.c (split_critical_edges): Export.
4964         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4966 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4968         * cfgexpand.c (expand_stack_vars): Optionally disable
4969         asan stack protection.
4970         (expand_used_vars): Likewise.
4971         (partition_stack_vars): Likewise.
4972         * asan.c (asan_emit_stack_protection): Optionally disable
4973         after return stack usage.
4974         (instrument_derefs): Optionally disable memory access instrumentation.
4975         (instrument_builtin_call): Likewise.
4976         (instrument_strlen_call): Likewise.
4977         (asan_protect_global): Optionally disable global variables protection.
4978         * doc/invoke.texi: Added doc for new options.
4979         * params.def: Added new options.
4980         * params.h: Likewise.
4982 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4984         PR rtl-optimization/59724
4985         * ifcvt.c (cond_exec_process_if_block): Don't call
4986         flow_find_head_matching_sequence with 0 longest_match.
4987         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4988         non-active insns if !stop_after.
4989         (try_head_merge_bb): Revert 2014-01-07 changes.
4991 2014-01-08  Jeff Law  <law@redhat.com>
4993         * ree.c (get_sub_rtx): New function, extracted from...
4994         (merge_def_and_ext): Here.
4995         (combine_reaching_defs): Use get_sub_rtx.
4997 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4999         * cgraph.h (varpool_variable_node): Do not choke on null node.
5001 2014-01-08  Catherine Moore  <clm@codesourcery.com>
5003         * config/mips/mips.md (simple_return): Attempt to use JRC
5004         for microMIPS.
5005         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
5007 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5009         PR rtl-optimization/59137
5010         * reorg.c (steal_delay_list_from_target): Call update_block for
5011         elided insns.
5012         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
5014 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
5016         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
5017         two duplicate entries.
5019 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
5021         Revert:
5022         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
5024         * config/mips/mips.c (mips_truncated_op_cost): New function.
5025         (mips_rtx_costs): Adjust test for BADDU.
5026         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
5028         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
5030         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
5031         (*baddu_si): ...this new pattern.
5033 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5035         PR ipa/59722
5036         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
5038 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5040         PR middle-end/57748
5041         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
5042         inner_reference_p.
5043         (expand_expr, expand_normal): Adjust.
5044         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
5045         inner_reference_p. Use inner_reference_p to expand inner references.
5046         (store_expr): Adjust.
5047         * cfgexpand.c (expand_call_stmt): Adjust.
5049 2014-01-08  Rong Xu  <xur@google.com>
5051         * gcov-io.c (gcov_var): Move from gcov-io.h.
5052         (gcov_position): Ditto.
5053         (gcov_is_error): Ditto.
5054         (gcov_rewrite): Ditto.
5055         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
5056         only part to libgcc/libgcov.h.
5058 2014-01-08  Marek Polacek  <polacek@redhat.com>
5060         PR middle-end/59669
5061         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
5063 2014-01-08  Marek Polacek  <polacek@redhat.com>
5065         PR sanitizer/59667
5066         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
5068 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
5070         PR rtl-optimization/59649
5071         * stor-layout.c (get_mode_bounds): For BImode return
5072         0 and STORE_FLAG_VALUE.
5074 2014-01-08  Richard Biener  <rguenther@suse.de>
5076         PR middle-end/59630
5077         * gimple.h (is_gimple_builtin_call): Remove.
5078         (gimple_builtin_call_types_compatible_p): New.
5079         (gimple_call_builtin_p): New overload.
5080         * gimple.c (is_gimple_builtin_call): Remove.
5081         (validate_call): Rename to ...
5082         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
5083         check return types.
5084         (validate_type): New static function.
5085         (gimple_call_builtin_p): New overload and adjust.
5086         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
5087         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
5088         (gimple_fold_stmt_to_constant_1): Likewise.
5089         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
5091 2014-01-08  Richard Biener  <rguenther@suse.de>
5093         PR middle-end/59471
5094         * gimplify.c (gimplify_expr): Gimplify register-register type
5095         VIEW_CONVERT_EXPRs to separate stmts.
5097 2014-01-07  Jeff Law  <law@redhat.com>
5099         PR middle-end/53623
5100         * ree.c (combine_set_extension): Handle case where source
5101         and destination registers in an extension insn are different.
5102         (combine_reaching_defs): Allow source and destination registers
5103         in extension to be different under limited circumstances.
5104         (add_removable_extension): Remove restriction that the
5105         source and destination registers in the extension are the same.
5106         (find_and_remove_re): Emit a copy from the extension's
5107         destination to its source after the defining insn if
5108         the source and destination registers are different.
5110         PR middle-end/59285
5111         * ifcvt.c (merge_if_block): If we are merging a block with more than
5112         one successor with a block with no successors, remove any BARRIER
5113         after the second block.
5115 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
5117         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
5119 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
5121         PR target/59652
5122         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
5123         for 14-bit register offsets when INT14_OK_STRICT is false.
5125 2014-01-07  Roland Stigge  <stigge@antcom.de>
5126             Michael Meissner  <meissner@linux.vnet.ibm.com>
5128         PR 57386/target
5129         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5130         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
5132 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
5134         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5135         -mcpu.
5137 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
5139         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5140         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5141         rtx is const0_rtx or not.
5143 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
5145         PR target/58115
5146         * target-globals.c (save_target_globals): Remove this_fn_optab
5147         handling.
5148         * toplev.c: Include optabs.h.
5149         (target_reinit): Temporarily restore the global options if another
5150         set of options are in force.
5152 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
5154         PR rtl-optimization/58668
5155         * cfgcleanup.c (flow_find_cross_jump): Don't count
5156         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
5157         to determine what is counted.
5158         (flow_find_head_matching_sequence): Use active_insn_p to determine
5159         what is counted.
5160         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5161         counting change.
5162         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5163         determine what is counted.
5165         PR tree-optimization/59643
5166         * tree-predcom.c (split_data_refs_to_components): If one dr is
5167         read and one write, determine_offset fails and the write isn't
5168         in the bad component, just put the read into the bad component.
5170 2014-01-07  Mike Stump  <mikestump@comcast.net>
5171             Jakub Jelinek  <jakub@redhat.com>
5173         PR pch/59436
5174         * tree-core.h (struct tree_optimization_option): Change optabs
5175         type from unsigned char * to void *.
5176         * optabs.c (init_tree_optimization_optabs): Adjust
5177         TREE_OPTIMIZATION_OPTABS initialization.
5179 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
5181         PR target/59644
5182         * config/i386/i386.h (struct machine_function): Add
5183         no_drap_save_restore field.
5184         * config/i386/i386.c (ix86_save_reg): Use
5185         !cfun->machine->no_drap_save_restore instead of
5186         crtl->stack_realign_needed.
5187         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5188         this function clears frame_pointer_needed.  Set
5189         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5190         and DRAP reg is needed.
5192 2014-01-06  Marek Polacek  <polacek@redhat.com>
5194         PR c/57773
5195         * doc/implement-c.texi: Mention that other integer types are
5196         permitted as bit-field types in strictly conforming mode.
5198 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
5200         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5201         is newly allocated.
5203 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
5205         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5207 2014-01-06  Martin Jambor  <mjambor@suse.cz>
5209         PR ipa/59008
5210         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5211         to int.
5212         * ipa-prop.c (ipa_print_node_params): Fix indentation.
5214 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
5216         PR debug/59350
5217         PR debug/59510
5218         * var-tracking.c (add_stores): Preserve the value of the source even if
5219         we don't record the store.
5221 2014-01-06  Terry Guo  <terry.guo@arm.com>
5223         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5225 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
5227         PR bootstrap/59541
5228         * config/darwin.c (darwin_function_section): Adjust return values to
5229         correspond to optimisation changes made in r206070.
5231 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
5233         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5234         from prefetch_block tune setting.
5235         (nocona_cost): Correct size of prefetch block to 64.
5237 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
5239         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5240         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5241         used to save the static chain register in the computation of the offset
5242         from which the FP registers need to be restored.
5244 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
5246         PR tree-optimization/59519
5247         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5248         ICE if get_current_def (current_new_name) is already non-NULL, as long
5249         as it is a phi result of some other phi in *new_exit_bb that has
5250         the same argument.
5252         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5253         or vmovdqu* for misaligned_operand.
5254         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5255         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5256         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5257         aligned_mem for AVX512F masked aligned load and store builtins and for
5258         non-temporal moves.
5260 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
5262         PR tree-optimization/59651
5263         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5264         Address range for negative step should be added by TYPE_SIZE_UNIT.
5266 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
5268         * config/m68k/m68k.c (handle_move_double): Handle pushes with
5269         overlapping registers also for registers other than the stack pointer.
5271 2014-01-03  Marek Polacek  <polacek@redhat.com>
5273         PR other/59661
5274         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5275         __builtin_FILE.
5277 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
5279         PR target/59625
5280         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5281         asm goto as jump.
5283         * config/i386/i386.md (MODE_SIZE): New mode attribute.
5284         (push splitter): Use <P:MODE_SIZE> instead of
5285         GET_MODE_SIZE (<P:MODE>mode).
5286         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5287         (mov -1, reg peephole2): Likewise.
5288         * config/i386/sse.md (*mov<mode>_internal,
5289         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5290         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5291         *<code><mode>3, *andnot<mode>3<mask_name>,
5292         <mask_codefor><code><mode>3<mask_name>): Likewise.
5293         * config/i386/subst.md (mask_mode512bit_condition,
5294         sd_mask_mode512bit_condition): Likewise.
5296 2014-01-02  Xinliang David Li  <davidxl@google.com>
5298         PR tree-optimization/59303
5299         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5300         (dump_predicates): Better output format.
5301         (pred_equal_p): New function.
5302         (is_neq_relop_p): Ditto.
5303         (is_neq_zero_form_p): Ditto.
5304         (pred_expr_equal_p): Ditto.
5305         (pred_neg_p): Ditto.
5306         (simplify_pred): Ditto.
5307         (simplify_preds_2): Ditto.
5308         (simplify_preds_3): Ditto.
5309         (simplify_preds_4): Ditto.
5310         (simplify_preds): Ditto.
5311         (push_pred): Ditto.
5312         (push_to_worklist): Ditto.
5313         (get_pred_info_from_cmp): Ditto.
5314         (is_degenerated_phi): Ditto.
5315         (normalize_one_pred_1): Ditto.
5316         (normalize_one_pred): Ditto.
5317         (normalize_one_pred_chain): Ditto.
5318         (normalize_preds): Ditto.
5319         (normalize_cond_1): Remove function.
5320         (normalize_cond): Ditto.
5321         (is_gcond_subset_of): Ditto.
5322         (is_subset_of_any): Ditto.
5323         (is_or_set_subset_of): Ditto.
5324         (is_and_set_subset_of): Ditto.
5325         (is_norm_cond_subset_of): Ditto.
5326         (pred_chain_length_cmp): Ditto.
5327         (convert_control_dep_chain_into_preds): Type change.
5328         (find_predicates): Ditto.
5329         (find_def_preds): Ditto.
5330         (destroy_predicates_vecs): Ditto.
5331         (find_matching_predicates_in_rest_chains): Ditto.
5332         (use_pred_not_overlap_with_undef_path_pred): Ditto.
5333         (is_pred_expr_subset): Ditto.
5334         (is_pred_chain_subset_of): Ditto.
5335         (is_included_in): Ditto.
5336         (is_superset_of): Ditto.
5338 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5340         Update copyright years.
5342 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
5344         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5345         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5346         config/arc/arc.md, config/arc/arc.opt,
5347         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5348         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5349         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5350         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5351         config/linux-protos.h, config/linux.c, config/winnt-c.c,
5352         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5353         vtable-verify.c, vtable-verify.h: Use the standard form for the
5354         copyright notice.
5356 2014-01-02  Tobias Burnus  <burnus@net-b.de>
5358         * gcc.c (process_command): Update copyright notice dates.
5359         * gcov-dump.c: Ditto.
5360         * gcov.c: Ditto.
5361         * doc/cpp.texi: Bump @copying's copyright year.
5362         * doc/cppinternals.texi: Ditto.
5363         * doc/gcc.texi: Ditto.
5364         * doc/gccint.texi: Ditto.
5365         * doc/gcov.texi: Ditto.
5366         * doc/install.texi: Ditto.
5367         * doc/invoke.texi: Ditto.
5369 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
5371         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5373 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
5375         * config/i386/sse.md (*mov<mode>_internal): Guard
5376         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5378         PR rtl-optimization/59647
5379         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5380         new_rtx into UNSIGNED_FLOAT rtxes.
5382 Copyright (C) 2014 Free Software Foundation, Inc.
5384 Copying and distribution of this file, with or without modification,
5385 are permitted in any medium without royalty provided the copyright
5386 notice and this notice are preserved.